nfs41: lock sequence setup/done support
[linux-2.6.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/utsname.h>
40 #include <linux/delay.h>
41 #include <linux/errno.h>
42 #include <linux/string.h>
43 #include <linux/sunrpc/clnt.h>
44 #include <linux/nfs.h>
45 #include <linux/nfs4.h>
46 #include <linux/nfs_fs.h>
47 #include <linux/nfs_page.h>
48 #include <linux/smp_lock.h>
49 #include <linux/namei.h>
50 #include <linux/mount.h>
51
52 #include "nfs4_fs.h"
53 #include "delegation.h"
54 #include "internal.h"
55 #include "iostat.h"
56
57 #define NFSDBG_FACILITY         NFSDBG_PROC
58
59 #define NFS4_POLL_RETRY_MIN     (HZ/10)
60 #define NFS4_POLL_RETRY_MAX     (15*HZ)
61
62 struct nfs4_opendata;
63 static int _nfs4_proc_open(struct nfs4_opendata *data);
64 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
65 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
66 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
67 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
68
69 /* Prevent leaks of NFSv4 errors into userland */
70 static int nfs4_map_errors(int err)
71 {
72         if (err < -1000) {
73                 dprintk("%s could not handle NFSv4 error %d\n",
74                                 __func__, -err);
75                 return -EIO;
76         }
77         return err;
78 }
79
80 /*
81  * This is our standard bitmap for GETATTR requests.
82  */
83 const u32 nfs4_fattr_bitmap[2] = {
84         FATTR4_WORD0_TYPE
85         | FATTR4_WORD0_CHANGE
86         | FATTR4_WORD0_SIZE
87         | FATTR4_WORD0_FSID
88         | FATTR4_WORD0_FILEID,
89         FATTR4_WORD1_MODE
90         | FATTR4_WORD1_NUMLINKS
91         | FATTR4_WORD1_OWNER
92         | FATTR4_WORD1_OWNER_GROUP
93         | FATTR4_WORD1_RAWDEV
94         | FATTR4_WORD1_SPACE_USED
95         | FATTR4_WORD1_TIME_ACCESS
96         | FATTR4_WORD1_TIME_METADATA
97         | FATTR4_WORD1_TIME_MODIFY
98 };
99
100 const u32 nfs4_statfs_bitmap[2] = {
101         FATTR4_WORD0_FILES_AVAIL
102         | FATTR4_WORD0_FILES_FREE
103         | FATTR4_WORD0_FILES_TOTAL,
104         FATTR4_WORD1_SPACE_AVAIL
105         | FATTR4_WORD1_SPACE_FREE
106         | FATTR4_WORD1_SPACE_TOTAL
107 };
108
109 const u32 nfs4_pathconf_bitmap[2] = {
110         FATTR4_WORD0_MAXLINK
111         | FATTR4_WORD0_MAXNAME,
112         0
113 };
114
115 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
116                         | FATTR4_WORD0_MAXREAD
117                         | FATTR4_WORD0_MAXWRITE
118                         | FATTR4_WORD0_LEASE_TIME,
119                         0
120 };
121
122 const u32 nfs4_fs_locations_bitmap[2] = {
123         FATTR4_WORD0_TYPE
124         | FATTR4_WORD0_CHANGE
125         | FATTR4_WORD0_SIZE
126         | FATTR4_WORD0_FSID
127         | FATTR4_WORD0_FILEID
128         | FATTR4_WORD0_FS_LOCATIONS,
129         FATTR4_WORD1_MODE
130         | FATTR4_WORD1_NUMLINKS
131         | FATTR4_WORD1_OWNER
132         | FATTR4_WORD1_OWNER_GROUP
133         | FATTR4_WORD1_RAWDEV
134         | FATTR4_WORD1_SPACE_USED
135         | FATTR4_WORD1_TIME_ACCESS
136         | FATTR4_WORD1_TIME_METADATA
137         | FATTR4_WORD1_TIME_MODIFY
138         | FATTR4_WORD1_MOUNTED_ON_FILEID
139 };
140
141 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
142                 struct nfs4_readdir_arg *readdir)
143 {
144         __be32 *start, *p;
145
146         BUG_ON(readdir->count < 80);
147         if (cookie > 2) {
148                 readdir->cookie = cookie;
149                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
150                 return;
151         }
152
153         readdir->cookie = 0;
154         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
155         if (cookie == 2)
156                 return;
157         
158         /*
159          * NFSv4 servers do not return entries for '.' and '..'
160          * Therefore, we fake these entries here.  We let '.'
161          * have cookie 0 and '..' have cookie 1.  Note that
162          * when talking to the server, we always send cookie 0
163          * instead of 1 or 2.
164          */
165         start = p = kmap_atomic(*readdir->pages, KM_USER0);
166         
167         if (cookie == 0) {
168                 *p++ = xdr_one;                                  /* next */
169                 *p++ = xdr_zero;                   /* cookie, first word */
170                 *p++ = xdr_one;                   /* cookie, second word */
171                 *p++ = xdr_one;                             /* entry len */
172                 memcpy(p, ".\0\0\0", 4);                        /* entry */
173                 p++;
174                 *p++ = xdr_one;                         /* bitmap length */
175                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
176                 *p++ = htonl(8);              /* attribute buffer length */
177                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
178         }
179         
180         *p++ = xdr_one;                                  /* next */
181         *p++ = xdr_zero;                   /* cookie, first word */
182         *p++ = xdr_two;                   /* cookie, second word */
183         *p++ = xdr_two;                             /* entry len */
184         memcpy(p, "..\0\0", 4);                         /* entry */
185         p++;
186         *p++ = xdr_one;                         /* bitmap length */
187         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
188         *p++ = htonl(8);              /* attribute buffer length */
189         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
190
191         readdir->pgbase = (char *)p - (char *)start;
192         readdir->count -= readdir->pgbase;
193         kunmap_atomic(start, KM_USER0);
194 }
195
196 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
197 {
198         int res;
199
200         might_sleep();
201
202         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
203                         nfs_wait_bit_killable, TASK_KILLABLE);
204         return res;
205 }
206
207 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
208 {
209         int res = 0;
210
211         might_sleep();
212
213         if (*timeout <= 0)
214                 *timeout = NFS4_POLL_RETRY_MIN;
215         if (*timeout > NFS4_POLL_RETRY_MAX)
216                 *timeout = NFS4_POLL_RETRY_MAX;
217         schedule_timeout_killable(*timeout);
218         if (fatal_signal_pending(current))
219                 res = -ERESTARTSYS;
220         *timeout <<= 1;
221         return res;
222 }
223
224 /* This is the error handling routine for processes that are allowed
225  * to sleep.
226  */
227 static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
228 {
229         struct nfs_client *clp = server->nfs_client;
230         struct nfs4_state *state = exception->state;
231         int ret = errorcode;
232
233         exception->retry = 0;
234         switch(errorcode) {
235                 case 0:
236                         return 0;
237                 case -NFS4ERR_ADMIN_REVOKED:
238                 case -NFS4ERR_BAD_STATEID:
239                 case -NFS4ERR_OPENMODE:
240                         if (state == NULL)
241                                 break;
242                         nfs4_state_mark_reclaim_nograce(clp, state);
243                 case -NFS4ERR_STALE_CLIENTID:
244                 case -NFS4ERR_STALE_STATEID:
245                 case -NFS4ERR_EXPIRED:
246                         nfs4_schedule_state_recovery(clp);
247                         ret = nfs4_wait_clnt_recover(clp);
248                         if (ret == 0)
249                                 exception->retry = 1;
250                         break;
251                 case -NFS4ERR_FILE_OPEN:
252                 case -NFS4ERR_GRACE:
253                 case -NFS4ERR_DELAY:
254                         ret = nfs4_delay(server->client, &exception->timeout);
255                         if (ret != 0)
256                                 break;
257                 case -NFS4ERR_OLD_STATEID:
258                         exception->retry = 1;
259         }
260         /* We failed to handle the error */
261         return nfs4_map_errors(ret);
262 }
263
264
265 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
266 {
267         struct nfs_client *clp = server->nfs_client;
268         spin_lock(&clp->cl_lock);
269         if (time_before(clp->cl_last_renewal,timestamp))
270                 clp->cl_last_renewal = timestamp;
271         spin_unlock(&clp->cl_lock);
272 }
273
274 #if defined(CONFIG_NFS_V4_1)
275
276 /*
277  * nfs4_free_slot - free a slot and efficiently update slot table.
278  *
279  * freeing a slot is trivially done by clearing its respective bit
280  * in the bitmap.
281  * If the freed slotid equals highest_used_slotid we want to update it
282  * so that the server would be able to size down the slot table if needed,
283  * otherwise we know that the highest_used_slotid is still in use.
284  * When updating highest_used_slotid there may be "holes" in the bitmap
285  * so we need to scan down from highest_used_slotid to 0 looking for the now
286  * highest slotid in use.
287  * If none found, highest_used_slotid is set to -1.
288  */
289 static void
290 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
291 {
292         int slotid = free_slotid;
293
294         spin_lock(&tbl->slot_tbl_lock);
295         /* clear used bit in bitmap */
296         __clear_bit(slotid, tbl->used_slots);
297
298         /* update highest_used_slotid when it is freed */
299         if (slotid == tbl->highest_used_slotid) {
300                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
301                 if (slotid >= 0 && slotid < tbl->max_slots)
302                         tbl->highest_used_slotid = slotid;
303                 else
304                         tbl->highest_used_slotid = -1;
305         }
306         rpc_wake_up_next(&tbl->slot_tbl_waitq);
307         spin_unlock(&tbl->slot_tbl_lock);
308         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
309                 free_slotid, tbl->highest_used_slotid);
310 }
311
312 void nfs41_sequence_free_slot(const struct nfs_client *clp,
313                               struct nfs4_sequence_res *res)
314 {
315         struct nfs4_slot_table *tbl;
316
317         if (!nfs4_has_session(clp)) {
318                 dprintk("%s: No session\n", __func__);
319                 return;
320         }
321         tbl = &clp->cl_session->fc_slot_table;
322         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) {
323                 dprintk("%s: No slot\n", __func__);
324                 /* just wake up the next guy waiting since
325                  * we may have not consumed a slot after all */
326                 rpc_wake_up_next(&tbl->slot_tbl_waitq);
327                 return;
328         }
329         nfs4_free_slot(tbl, res->sr_slotid);
330         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
331 }
332
333 static void nfs41_sequence_done(struct nfs_client *clp,
334                                 struct nfs4_sequence_res *res,
335                                 int rpc_status)
336 {
337         unsigned long timestamp;
338         struct nfs4_slot_table *tbl;
339         struct nfs4_slot *slot;
340
341         /*
342          * sr_status remains 1 if an RPC level error occurred. The server
343          * may or may not have processed the sequence operation..
344          * Proceed as if the server received and processed the sequence
345          * operation.
346          */
347         if (res->sr_status == 1)
348                 res->sr_status = NFS_OK;
349
350         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
351         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE)
352                 goto out;
353
354         tbl = &clp->cl_session->fc_slot_table;
355         slot = tbl->slots + res->sr_slotid;
356
357         if (res->sr_status == 0) {
358                 /* Update the slot's sequence and clientid lease timer */
359                 ++slot->seq_nr;
360                 timestamp = res->sr_renewal_time;
361                 spin_lock(&clp->cl_lock);
362                 if (time_before(clp->cl_last_renewal, timestamp))
363                         clp->cl_last_renewal = timestamp;
364                 spin_unlock(&clp->cl_lock);
365                 return;
366         }
367 out:
368         /* The session may be reset by one of the error handlers. */
369         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
370         nfs41_sequence_free_slot(clp, res);
371 }
372
373 /*
374  * nfs4_find_slot - efficiently look for a free slot
375  *
376  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
377  * If found, we mark the slot as used, update the highest_used_slotid,
378  * and respectively set up the sequence operation args.
379  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
380  *
381  * Note: must be called with under the slot_tbl_lock.
382  */
383 static u8
384 nfs4_find_slot(struct nfs4_slot_table *tbl, struct rpc_task *task)
385 {
386         int slotid;
387         u8 ret_id = NFS4_MAX_SLOT_TABLE;
388         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
389
390         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
391                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
392                 tbl->max_slots);
393         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
394         if (slotid >= tbl->max_slots)
395                 goto out;
396         __set_bit(slotid, tbl->used_slots);
397         if (slotid > tbl->highest_used_slotid)
398                 tbl->highest_used_slotid = slotid;
399         ret_id = slotid;
400 out:
401         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
402                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
403         return ret_id;
404 }
405
406 static int nfs41_setup_sequence(struct nfs4_session *session,
407                                 struct nfs4_sequence_args *args,
408                                 struct nfs4_sequence_res *res,
409                                 int cache_reply,
410                                 struct rpc_task *task)
411 {
412         struct nfs4_slot *slot;
413         struct nfs4_slot_table *tbl;
414         u8 slotid;
415
416         dprintk("--> %s\n", __func__);
417         /* slot already allocated? */
418         if (res->sr_slotid != NFS4_MAX_SLOT_TABLE)
419                 return 0;
420
421         memset(res, 0, sizeof(*res));
422         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
423         tbl = &session->fc_slot_table;
424
425         spin_lock(&tbl->slot_tbl_lock);
426         slotid = nfs4_find_slot(tbl, task);
427         if (slotid == NFS4_MAX_SLOT_TABLE) {
428                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
429                 spin_unlock(&tbl->slot_tbl_lock);
430                 dprintk("<-- %s: no free slots\n", __func__);
431                 return -EAGAIN;
432         }
433         spin_unlock(&tbl->slot_tbl_lock);
434
435         slot = tbl->slots + slotid;
436         args->sa_slotid = slotid;
437         args->sa_cache_this = cache_reply;
438
439         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
440
441         res->sr_slotid = slotid;
442         res->sr_renewal_time = jiffies;
443         /*
444          * sr_status is only set in decode_sequence, and so will remain
445          * set to 1 if an rpc level failure occurs.
446          */
447         res->sr_status = 1;
448         return 0;
449 }
450
451 int nfs4_setup_sequence(struct nfs_client *clp,
452                         struct nfs4_sequence_args *args,
453                         struct nfs4_sequence_res *res,
454                         int cache_reply,
455                         struct rpc_task *task)
456 {
457         int ret = 0;
458
459         dprintk("--> %s clp %p session %p sr_slotid %d\n",
460                 __func__, clp, clp->cl_session, res->sr_slotid);
461
462         if (!nfs4_has_session(clp))
463                 goto out;
464         ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply,
465                                    task);
466         if (ret != -EAGAIN) {
467                 /* terminate rpc task */
468                 task->tk_status = ret;
469                 task->tk_action = NULL;
470         }
471 out:
472         dprintk("<-- %s status=%d\n", __func__, ret);
473         return ret;
474 }
475
476 struct nfs41_call_sync_data {
477         struct nfs_client *clp;
478         struct nfs4_sequence_args *seq_args;
479         struct nfs4_sequence_res *seq_res;
480         int cache_reply;
481 };
482
483 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
484 {
485         struct nfs41_call_sync_data *data = calldata;
486
487         dprintk("--> %s data->clp->cl_session %p\n", __func__,
488                 data->clp->cl_session);
489         if (nfs4_setup_sequence(data->clp, data->seq_args,
490                                 data->seq_res, data->cache_reply, task))
491                 return;
492         rpc_call_start(task);
493 }
494
495 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
496 {
497         struct nfs41_call_sync_data *data = calldata;
498
499         nfs41_sequence_done(data->clp, data->seq_res, task->tk_status);
500         nfs41_sequence_free_slot(data->clp, data->seq_res);
501 }
502
503 struct rpc_call_ops nfs41_call_sync_ops = {
504         .rpc_call_prepare = nfs41_call_sync_prepare,
505         .rpc_call_done = nfs41_call_sync_done,
506 };
507
508 static int nfs4_call_sync_sequence(struct nfs_client *clp,
509                                    struct rpc_clnt *clnt,
510                                    struct rpc_message *msg,
511                                    struct nfs4_sequence_args *args,
512                                    struct nfs4_sequence_res *res,
513                                    int cache_reply)
514 {
515         int ret;
516         struct rpc_task *task;
517         struct nfs41_call_sync_data data = {
518                 .clp = clp,
519                 .seq_args = args,
520                 .seq_res = res,
521                 .cache_reply = cache_reply,
522         };
523         struct rpc_task_setup task_setup = {
524                 .rpc_client = clnt,
525                 .rpc_message = msg,
526                 .callback_ops = &nfs41_call_sync_ops,
527                 .callback_data = &data
528         };
529
530         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
531         task = rpc_run_task(&task_setup);
532         if (IS_ERR(task))
533                 ret = PTR_ERR(task);
534         else {
535                 ret = task->tk_status;
536                 rpc_put_task(task);
537         }
538         return ret;
539 }
540
541 int _nfs4_call_sync_session(struct nfs_server *server,
542                             struct rpc_message *msg,
543                             struct nfs4_sequence_args *args,
544                             struct nfs4_sequence_res *res,
545                             int cache_reply)
546 {
547         return nfs4_call_sync_sequence(server->nfs_client, server->client,
548                                        msg, args, res, cache_reply);
549 }
550
551 #endif /* CONFIG_NFS_V4_1 */
552
553 int _nfs4_call_sync(struct nfs_server *server,
554                     struct rpc_message *msg,
555                     struct nfs4_sequence_args *args,
556                     struct nfs4_sequence_res *res,
557                     int cache_reply)
558 {
559         args->sa_session = res->sr_session = NULL;
560         return rpc_call_sync(server->client, msg, 0);
561 }
562
563 #define nfs4_call_sync(server, msg, args, res, cache_reply) \
564         (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \
565                         &(res)->seq_res, (cache_reply))
566
567 static void nfs4_sequence_done(const struct nfs_server *server,
568                                struct nfs4_sequence_res *res, int rpc_status)
569 {
570 #ifdef CONFIG_NFS_V4_1
571         if (nfs4_has_session(server->nfs_client))
572                 nfs41_sequence_done(server->nfs_client, res, rpc_status);
573 #endif /* CONFIG_NFS_V4_1 */
574 }
575
576 /* no restart, therefore free slot here */
577 static void nfs4_sequence_done_free_slot(const struct nfs_server *server,
578                                          struct nfs4_sequence_res *res,
579                                          int rpc_status)
580 {
581         nfs4_sequence_done(server, res, rpc_status);
582         nfs4_sequence_free_slot(server->nfs_client, res);
583 }
584
585 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
586 {
587         struct nfs_inode *nfsi = NFS_I(dir);
588
589         spin_lock(&dir->i_lock);
590         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
591         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
592                 nfs_force_lookup_revalidate(dir);
593         nfsi->change_attr = cinfo->after;
594         spin_unlock(&dir->i_lock);
595 }
596
597 struct nfs4_opendata {
598         struct kref kref;
599         struct nfs_openargs o_arg;
600         struct nfs_openres o_res;
601         struct nfs_open_confirmargs c_arg;
602         struct nfs_open_confirmres c_res;
603         struct nfs_fattr f_attr;
604         struct nfs_fattr dir_attr;
605         struct path path;
606         struct dentry *dir;
607         struct nfs4_state_owner *owner;
608         struct nfs4_state *state;
609         struct iattr attrs;
610         unsigned long timestamp;
611         unsigned int rpc_done : 1;
612         int rpc_status;
613         int cancelled;
614 };
615
616
617 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
618 {
619         p->o_res.f_attr = &p->f_attr;
620         p->o_res.dir_attr = &p->dir_attr;
621         p->o_res.seqid = p->o_arg.seqid;
622         p->c_res.seqid = p->c_arg.seqid;
623         p->o_res.server = p->o_arg.server;
624         nfs_fattr_init(&p->f_attr);
625         nfs_fattr_init(&p->dir_attr);
626 }
627
628 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
629                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
630                 const struct iattr *attrs)
631 {
632         struct dentry *parent = dget_parent(path->dentry);
633         struct inode *dir = parent->d_inode;
634         struct nfs_server *server = NFS_SERVER(dir);
635         struct nfs4_opendata *p;
636
637         p = kzalloc(sizeof(*p), GFP_KERNEL);
638         if (p == NULL)
639                 goto err;
640         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
641         if (p->o_arg.seqid == NULL)
642                 goto err_free;
643         p->path.mnt = mntget(path->mnt);
644         p->path.dentry = dget(path->dentry);
645         p->dir = parent;
646         p->owner = sp;
647         atomic_inc(&sp->so_count);
648         p->o_arg.fh = NFS_FH(dir);
649         p->o_arg.open_flags = flags;
650         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
651         p->o_arg.clientid = server->nfs_client->cl_clientid;
652         p->o_arg.id = sp->so_owner_id.id;
653         p->o_arg.name = &p->path.dentry->d_name;
654         p->o_arg.server = server;
655         p->o_arg.bitmask = server->attr_bitmask;
656         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
657         p->o_res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
658         if (flags & O_EXCL) {
659                 u32 *s = (u32 *) p->o_arg.u.verifier.data;
660                 s[0] = jiffies;
661                 s[1] = current->pid;
662         } else if (flags & O_CREAT) {
663                 p->o_arg.u.attrs = &p->attrs;
664                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
665         }
666         p->c_arg.fh = &p->o_res.fh;
667         p->c_arg.stateid = &p->o_res.stateid;
668         p->c_arg.seqid = p->o_arg.seqid;
669         nfs4_init_opendata_res(p);
670         kref_init(&p->kref);
671         return p;
672 err_free:
673         kfree(p);
674 err:
675         dput(parent);
676         return NULL;
677 }
678
679 static void nfs4_opendata_free(struct kref *kref)
680 {
681         struct nfs4_opendata *p = container_of(kref,
682                         struct nfs4_opendata, kref);
683
684         nfs_free_seqid(p->o_arg.seqid);
685         if (p->state != NULL)
686                 nfs4_put_open_state(p->state);
687         nfs4_put_state_owner(p->owner);
688         dput(p->dir);
689         path_put(&p->path);
690         kfree(p);
691 }
692
693 static void nfs4_opendata_put(struct nfs4_opendata *p)
694 {
695         if (p != NULL)
696                 kref_put(&p->kref, nfs4_opendata_free);
697 }
698
699 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
700 {
701         int ret;
702
703         ret = rpc_wait_for_completion_task(task);
704         return ret;
705 }
706
707 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
708 {
709         int ret = 0;
710
711         if (open_mode & O_EXCL)
712                 goto out;
713         switch (mode & (FMODE_READ|FMODE_WRITE)) {
714                 case FMODE_READ:
715                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0;
716                         break;
717                 case FMODE_WRITE:
718                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0;
719                         break;
720                 case FMODE_READ|FMODE_WRITE:
721                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0;
722         }
723 out:
724         return ret;
725 }
726
727 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
728 {
729         if ((delegation->type & fmode) != fmode)
730                 return 0;
731         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
732                 return 0;
733         nfs_mark_delegation_referenced(delegation);
734         return 1;
735 }
736
737 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
738 {
739         switch (fmode) {
740                 case FMODE_WRITE:
741                         state->n_wronly++;
742                         break;
743                 case FMODE_READ:
744                         state->n_rdonly++;
745                         break;
746                 case FMODE_READ|FMODE_WRITE:
747                         state->n_rdwr++;
748         }
749         nfs4_state_set_mode_locked(state, state->state | fmode);
750 }
751
752 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
753 {
754         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
755                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
756         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
757         switch (fmode) {
758                 case FMODE_READ:
759                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
760                         break;
761                 case FMODE_WRITE:
762                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
763                         break;
764                 case FMODE_READ|FMODE_WRITE:
765                         set_bit(NFS_O_RDWR_STATE, &state->flags);
766         }
767 }
768
769 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
770 {
771         write_seqlock(&state->seqlock);
772         nfs_set_open_stateid_locked(state, stateid, fmode);
773         write_sequnlock(&state->seqlock);
774 }
775
776 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
777 {
778         /*
779          * Protect the call to nfs4_state_set_mode_locked and
780          * serialise the stateid update
781          */
782         write_seqlock(&state->seqlock);
783         if (deleg_stateid != NULL) {
784                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
785                 set_bit(NFS_DELEGATED_STATE, &state->flags);
786         }
787         if (open_stateid != NULL)
788                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
789         write_sequnlock(&state->seqlock);
790         spin_lock(&state->owner->so_lock);
791         update_open_stateflags(state, fmode);
792         spin_unlock(&state->owner->so_lock);
793 }
794
795 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
796 {
797         struct nfs_inode *nfsi = NFS_I(state->inode);
798         struct nfs_delegation *deleg_cur;
799         int ret = 0;
800
801         fmode &= (FMODE_READ|FMODE_WRITE);
802
803         rcu_read_lock();
804         deleg_cur = rcu_dereference(nfsi->delegation);
805         if (deleg_cur == NULL)
806                 goto no_delegation;
807
808         spin_lock(&deleg_cur->lock);
809         if (nfsi->delegation != deleg_cur ||
810             (deleg_cur->type & fmode) != fmode)
811                 goto no_delegation_unlock;
812
813         if (delegation == NULL)
814                 delegation = &deleg_cur->stateid;
815         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
816                 goto no_delegation_unlock;
817
818         nfs_mark_delegation_referenced(deleg_cur);
819         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
820         ret = 1;
821 no_delegation_unlock:
822         spin_unlock(&deleg_cur->lock);
823 no_delegation:
824         rcu_read_unlock();
825
826         if (!ret && open_stateid != NULL) {
827                 __update_open_stateid(state, open_stateid, NULL, fmode);
828                 ret = 1;
829         }
830
831         return ret;
832 }
833
834
835 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
836 {
837         struct nfs_delegation *delegation;
838
839         rcu_read_lock();
840         delegation = rcu_dereference(NFS_I(inode)->delegation);
841         if (delegation == NULL || (delegation->type & fmode) == fmode) {
842                 rcu_read_unlock();
843                 return;
844         }
845         rcu_read_unlock();
846         nfs_inode_return_delegation(inode);
847 }
848
849 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
850 {
851         struct nfs4_state *state = opendata->state;
852         struct nfs_inode *nfsi = NFS_I(state->inode);
853         struct nfs_delegation *delegation;
854         int open_mode = opendata->o_arg.open_flags & O_EXCL;
855         fmode_t fmode = opendata->o_arg.fmode;
856         nfs4_stateid stateid;
857         int ret = -EAGAIN;
858
859         for (;;) {
860                 if (can_open_cached(state, fmode, open_mode)) {
861                         spin_lock(&state->owner->so_lock);
862                         if (can_open_cached(state, fmode, open_mode)) {
863                                 update_open_stateflags(state, fmode);
864                                 spin_unlock(&state->owner->so_lock);
865                                 goto out_return_state;
866                         }
867                         spin_unlock(&state->owner->so_lock);
868                 }
869                 rcu_read_lock();
870                 delegation = rcu_dereference(nfsi->delegation);
871                 if (delegation == NULL ||
872                     !can_open_delegated(delegation, fmode)) {
873                         rcu_read_unlock();
874                         break;
875                 }
876                 /* Save the delegation */
877                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
878                 rcu_read_unlock();
879                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
880                 if (ret != 0)
881                         goto out;
882                 ret = -EAGAIN;
883
884                 /* Try to update the stateid using the delegation */
885                 if (update_open_stateid(state, NULL, &stateid, fmode))
886                         goto out_return_state;
887         }
888 out:
889         return ERR_PTR(ret);
890 out_return_state:
891         atomic_inc(&state->count);
892         return state;
893 }
894
895 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
896 {
897         struct inode *inode;
898         struct nfs4_state *state = NULL;
899         struct nfs_delegation *delegation;
900         int ret;
901
902         if (!data->rpc_done) {
903                 state = nfs4_try_open_cached(data);
904                 goto out;
905         }
906
907         ret = -EAGAIN;
908         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
909                 goto err;
910         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
911         ret = PTR_ERR(inode);
912         if (IS_ERR(inode))
913                 goto err;
914         ret = -ENOMEM;
915         state = nfs4_get_open_state(inode, data->owner);
916         if (state == NULL)
917                 goto err_put_inode;
918         if (data->o_res.delegation_type != 0) {
919                 int delegation_flags = 0;
920
921                 rcu_read_lock();
922                 delegation = rcu_dereference(NFS_I(inode)->delegation);
923                 if (delegation)
924                         delegation_flags = delegation->flags;
925                 rcu_read_unlock();
926                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
927                         nfs_inode_set_delegation(state->inode,
928                                         data->owner->so_cred,
929                                         &data->o_res);
930                 else
931                         nfs_inode_reclaim_delegation(state->inode,
932                                         data->owner->so_cred,
933                                         &data->o_res);
934         }
935
936         update_open_stateid(state, &data->o_res.stateid, NULL,
937                         data->o_arg.fmode);
938         iput(inode);
939 out:
940         return state;
941 err_put_inode:
942         iput(inode);
943 err:
944         return ERR_PTR(ret);
945 }
946
947 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
948 {
949         struct nfs_inode *nfsi = NFS_I(state->inode);
950         struct nfs_open_context *ctx;
951
952         spin_lock(&state->inode->i_lock);
953         list_for_each_entry(ctx, &nfsi->open_files, list) {
954                 if (ctx->state != state)
955                         continue;
956                 get_nfs_open_context(ctx);
957                 spin_unlock(&state->inode->i_lock);
958                 return ctx;
959         }
960         spin_unlock(&state->inode->i_lock);
961         return ERR_PTR(-ENOENT);
962 }
963
964 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
965 {
966         struct nfs4_opendata *opendata;
967
968         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL);
969         if (opendata == NULL)
970                 return ERR_PTR(-ENOMEM);
971         opendata->state = state;
972         atomic_inc(&state->count);
973         return opendata;
974 }
975
976 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
977 {
978         struct nfs4_state *newstate;
979         int ret;
980
981         opendata->o_arg.open_flags = 0;
982         opendata->o_arg.fmode = fmode;
983         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
984         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
985         nfs4_init_opendata_res(opendata);
986         ret = _nfs4_proc_open(opendata);
987         if (ret != 0)
988                 return ret; 
989         newstate = nfs4_opendata_to_nfs4_state(opendata);
990         if (IS_ERR(newstate))
991                 return PTR_ERR(newstate);
992         nfs4_close_state(&opendata->path, newstate, fmode);
993         *res = newstate;
994         return 0;
995 }
996
997 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
998 {
999         struct nfs4_state *newstate;
1000         int ret;
1001
1002         /* memory barrier prior to reading state->n_* */
1003         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1004         smp_rmb();
1005         if (state->n_rdwr != 0) {
1006                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1007                 if (ret != 0)
1008                         return ret;
1009                 if (newstate != state)
1010                         return -ESTALE;
1011         }
1012         if (state->n_wronly != 0) {
1013                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1014                 if (ret != 0)
1015                         return ret;
1016                 if (newstate != state)
1017                         return -ESTALE;
1018         }
1019         if (state->n_rdonly != 0) {
1020                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1021                 if (ret != 0)
1022                         return ret;
1023                 if (newstate != state)
1024                         return -ESTALE;
1025         }
1026         /*
1027          * We may have performed cached opens for all three recoveries.
1028          * Check if we need to update the current stateid.
1029          */
1030         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1031             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1032                 write_seqlock(&state->seqlock);
1033                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1034                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1035                 write_sequnlock(&state->seqlock);
1036         }
1037         return 0;
1038 }
1039
1040 /*
1041  * OPEN_RECLAIM:
1042  *      reclaim state on the server after a reboot.
1043  */
1044 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1045 {
1046         struct nfs_delegation *delegation;
1047         struct nfs4_opendata *opendata;
1048         fmode_t delegation_type = 0;
1049         int status;
1050
1051         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1052         if (IS_ERR(opendata))
1053                 return PTR_ERR(opendata);
1054         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1055         opendata->o_arg.fh = NFS_FH(state->inode);
1056         rcu_read_lock();
1057         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1058         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1059                 delegation_type = delegation->type;
1060         rcu_read_unlock();
1061         opendata->o_arg.u.delegation_type = delegation_type;
1062         status = nfs4_open_recover(opendata, state);
1063         nfs4_opendata_put(opendata);
1064         return status;
1065 }
1066
1067 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1068 {
1069         struct nfs_server *server = NFS_SERVER(state->inode);
1070         struct nfs4_exception exception = { };
1071         int err;
1072         do {
1073                 err = _nfs4_do_open_reclaim(ctx, state);
1074                 if (err != -NFS4ERR_DELAY)
1075                         break;
1076                 nfs4_handle_exception(server, err, &exception);
1077         } while (exception.retry);
1078         return err;
1079 }
1080
1081 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1082 {
1083         struct nfs_open_context *ctx;
1084         int ret;
1085
1086         ctx = nfs4_state_find_open_context(state);
1087         if (IS_ERR(ctx))
1088                 return PTR_ERR(ctx);
1089         ret = nfs4_do_open_reclaim(ctx, state);
1090         put_nfs_open_context(ctx);
1091         return ret;
1092 }
1093
1094 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1095 {
1096         struct nfs4_opendata *opendata;
1097         int ret;
1098
1099         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1100         if (IS_ERR(opendata))
1101                 return PTR_ERR(opendata);
1102         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1103         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1104                         sizeof(opendata->o_arg.u.delegation.data));
1105         ret = nfs4_open_recover(opendata, state);
1106         nfs4_opendata_put(opendata);
1107         return ret;
1108 }
1109
1110 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1111 {
1112         struct nfs4_exception exception = { };
1113         struct nfs_server *server = NFS_SERVER(state->inode);
1114         int err;
1115         do {
1116                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1117                 switch (err) {
1118                         case 0:
1119                                 return err;
1120                         case -NFS4ERR_STALE_CLIENTID:
1121                         case -NFS4ERR_STALE_STATEID:
1122                         case -NFS4ERR_EXPIRED:
1123                                 /* Don't recall a delegation if it was lost */
1124                                 nfs4_schedule_state_recovery(server->nfs_client);
1125                                 return err;
1126                 }
1127                 err = nfs4_handle_exception(server, err, &exception);
1128         } while (exception.retry);
1129         return err;
1130 }
1131
1132 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1133 {
1134         struct nfs4_opendata *data = calldata;
1135
1136         data->rpc_status = task->tk_status;
1137         if (RPC_ASSASSINATED(task))
1138                 return;
1139         if (data->rpc_status == 0) {
1140                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1141                                 sizeof(data->o_res.stateid.data));
1142                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1143                 renew_lease(data->o_res.server, data->timestamp);
1144                 data->rpc_done = 1;
1145         }
1146 }
1147
1148 static void nfs4_open_confirm_release(void *calldata)
1149 {
1150         struct nfs4_opendata *data = calldata;
1151         struct nfs4_state *state = NULL;
1152
1153         /* If this request hasn't been cancelled, do nothing */
1154         if (data->cancelled == 0)
1155                 goto out_free;
1156         /* In case of error, no cleanup! */
1157         if (!data->rpc_done)
1158                 goto out_free;
1159         state = nfs4_opendata_to_nfs4_state(data);
1160         if (!IS_ERR(state))
1161                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1162 out_free:
1163         nfs4_opendata_put(data);
1164 }
1165
1166 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1167         .rpc_call_done = nfs4_open_confirm_done,
1168         .rpc_release = nfs4_open_confirm_release,
1169 };
1170
1171 /*
1172  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1173  */
1174 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1175 {
1176         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1177         struct rpc_task *task;
1178         struct  rpc_message msg = {
1179                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1180                 .rpc_argp = &data->c_arg,
1181                 .rpc_resp = &data->c_res,
1182                 .rpc_cred = data->owner->so_cred,
1183         };
1184         struct rpc_task_setup task_setup_data = {
1185                 .rpc_client = server->client,
1186                 .rpc_message = &msg,
1187                 .callback_ops = &nfs4_open_confirm_ops,
1188                 .callback_data = data,
1189                 .workqueue = nfsiod_workqueue,
1190                 .flags = RPC_TASK_ASYNC,
1191         };
1192         int status;
1193
1194         kref_get(&data->kref);
1195         data->rpc_done = 0;
1196         data->rpc_status = 0;
1197         data->timestamp = jiffies;
1198         task = rpc_run_task(&task_setup_data);
1199         if (IS_ERR(task))
1200                 return PTR_ERR(task);
1201         status = nfs4_wait_for_completion_rpc_task(task);
1202         if (status != 0) {
1203                 data->cancelled = 1;
1204                 smp_wmb();
1205         } else
1206                 status = data->rpc_status;
1207         rpc_put_task(task);
1208         return status;
1209 }
1210
1211 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1212 {
1213         struct nfs4_opendata *data = calldata;
1214         struct nfs4_state_owner *sp = data->owner;
1215
1216         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1217                 return;
1218         /*
1219          * Check if we still need to send an OPEN call, or if we can use
1220          * a delegation instead.
1221          */
1222         if (data->state != NULL) {
1223                 struct nfs_delegation *delegation;
1224
1225                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1226                         goto out_no_action;
1227                 rcu_read_lock();
1228                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1229                 if (delegation != NULL &&
1230                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1231                         rcu_read_unlock();
1232                         goto out_no_action;
1233                 }
1234                 rcu_read_unlock();
1235         }
1236         /* Update sequence id. */
1237         data->o_arg.id = sp->so_owner_id.id;
1238         data->o_arg.clientid = sp->so_client->cl_clientid;
1239         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1240                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1241                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1242         }
1243         data->timestamp = jiffies;
1244         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
1245                                 &data->o_arg.seq_args,
1246                                 &data->o_res.seq_res, 1, task))
1247                 return;
1248         rpc_call_start(task);
1249         return;
1250 out_no_action:
1251         task->tk_action = NULL;
1252
1253 }
1254
1255 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1256 {
1257         struct nfs4_opendata *data = calldata;
1258
1259         data->rpc_status = task->tk_status;
1260
1261         nfs4_sequence_done_free_slot(data->o_arg.server, &data->o_res.seq_res,
1262                                      task->tk_status);
1263
1264         if (RPC_ASSASSINATED(task))
1265                 return;
1266         if (task->tk_status == 0) {
1267                 switch (data->o_res.f_attr->mode & S_IFMT) {
1268                         case S_IFREG:
1269                                 break;
1270                         case S_IFLNK:
1271                                 data->rpc_status = -ELOOP;
1272                                 break;
1273                         case S_IFDIR:
1274                                 data->rpc_status = -EISDIR;
1275                                 break;
1276                         default:
1277                                 data->rpc_status = -ENOTDIR;
1278                 }
1279                 renew_lease(data->o_res.server, data->timestamp);
1280                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1281                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1282         }
1283         data->rpc_done = 1;
1284 }
1285
1286 static void nfs4_open_release(void *calldata)
1287 {
1288         struct nfs4_opendata *data = calldata;
1289         struct nfs4_state *state = NULL;
1290
1291         /* If this request hasn't been cancelled, do nothing */
1292         if (data->cancelled == 0)
1293                 goto out_free;
1294         /* In case of error, no cleanup! */
1295         if (data->rpc_status != 0 || !data->rpc_done)
1296                 goto out_free;
1297         /* In case we need an open_confirm, no cleanup! */
1298         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1299                 goto out_free;
1300         state = nfs4_opendata_to_nfs4_state(data);
1301         if (!IS_ERR(state))
1302                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1303 out_free:
1304         nfs4_opendata_put(data);
1305 }
1306
1307 static const struct rpc_call_ops nfs4_open_ops = {
1308         .rpc_call_prepare = nfs4_open_prepare,
1309         .rpc_call_done = nfs4_open_done,
1310         .rpc_release = nfs4_open_release,
1311 };
1312
1313 /*
1314  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1315  */
1316 static int _nfs4_proc_open(struct nfs4_opendata *data)
1317 {
1318         struct inode *dir = data->dir->d_inode;
1319         struct nfs_server *server = NFS_SERVER(dir);
1320         struct nfs_openargs *o_arg = &data->o_arg;
1321         struct nfs_openres *o_res = &data->o_res;
1322         struct rpc_task *task;
1323         struct rpc_message msg = {
1324                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1325                 .rpc_argp = o_arg,
1326                 .rpc_resp = o_res,
1327                 .rpc_cred = data->owner->so_cred,
1328         };
1329         struct rpc_task_setup task_setup_data = {
1330                 .rpc_client = server->client,
1331                 .rpc_message = &msg,
1332                 .callback_ops = &nfs4_open_ops,
1333                 .callback_data = data,
1334                 .workqueue = nfsiod_workqueue,
1335                 .flags = RPC_TASK_ASYNC,
1336         };
1337         int status;
1338
1339         kref_get(&data->kref);
1340         data->rpc_done = 0;
1341         data->rpc_status = 0;
1342         data->cancelled = 0;
1343         task = rpc_run_task(&task_setup_data);
1344         if (IS_ERR(task))
1345                 return PTR_ERR(task);
1346         status = nfs4_wait_for_completion_rpc_task(task);
1347         if (status != 0) {
1348                 data->cancelled = 1;
1349                 smp_wmb();
1350         } else
1351                 status = data->rpc_status;
1352         rpc_put_task(task);
1353         if (status != 0 || !data->rpc_done)
1354                 return status;
1355
1356         if (o_res->fh.size == 0)
1357                 _nfs4_proc_lookup(dir, o_arg->name, &o_res->fh, o_res->f_attr);
1358
1359         if (o_arg->open_flags & O_CREAT) {
1360                 update_changeattr(dir, &o_res->cinfo);
1361                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1362         } else
1363                 nfs_refresh_inode(dir, o_res->dir_attr);
1364         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1365                 status = _nfs4_proc_open_confirm(data);
1366                 if (status != 0)
1367                         return status;
1368         }
1369         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1370                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1371         return 0;
1372 }
1373
1374 static int nfs4_recover_expired_lease(struct nfs_server *server)
1375 {
1376         struct nfs_client *clp = server->nfs_client;
1377         int ret;
1378
1379         for (;;) {
1380                 ret = nfs4_wait_clnt_recover(clp);
1381                 if (ret != 0)
1382                         return ret;
1383                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1384                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1385                         break;
1386                 nfs4_schedule_state_recovery(clp);
1387         }
1388         return 0;
1389 }
1390
1391 /*
1392  * OPEN_EXPIRED:
1393  *      reclaim state on the server after a network partition.
1394  *      Assumes caller holds the appropriate lock
1395  */
1396 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1397 {
1398         struct nfs4_opendata *opendata;
1399         int ret;
1400
1401         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1402         if (IS_ERR(opendata))
1403                 return PTR_ERR(opendata);
1404         ret = nfs4_open_recover(opendata, state);
1405         if (ret == -ESTALE)
1406                 d_drop(ctx->path.dentry);
1407         nfs4_opendata_put(opendata);
1408         return ret;
1409 }
1410
1411 static inline int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1412 {
1413         struct nfs_server *server = NFS_SERVER(state->inode);
1414         struct nfs4_exception exception = { };
1415         int err;
1416
1417         do {
1418                 err = _nfs4_open_expired(ctx, state);
1419                 if (err != -NFS4ERR_DELAY)
1420                         break;
1421                 nfs4_handle_exception(server, err, &exception);
1422         } while (exception.retry);
1423         return err;
1424 }
1425
1426 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1427 {
1428         struct nfs_open_context *ctx;
1429         int ret;
1430
1431         ctx = nfs4_state_find_open_context(state);
1432         if (IS_ERR(ctx))
1433                 return PTR_ERR(ctx);
1434         ret = nfs4_do_open_expired(ctx, state);
1435         put_nfs_open_context(ctx);
1436         return ret;
1437 }
1438
1439 /*
1440  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1441  * fields corresponding to attributes that were used to store the verifier.
1442  * Make sure we clobber those fields in the later setattr call
1443  */
1444 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1445 {
1446         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1447             !(sattr->ia_valid & ATTR_ATIME_SET))
1448                 sattr->ia_valid |= ATTR_ATIME;
1449
1450         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1451             !(sattr->ia_valid & ATTR_MTIME_SET))
1452                 sattr->ia_valid |= ATTR_MTIME;
1453 }
1454
1455 /*
1456  * Returns a referenced nfs4_state
1457  */
1458 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1459 {
1460         struct nfs4_state_owner  *sp;
1461         struct nfs4_state     *state = NULL;
1462         struct nfs_server       *server = NFS_SERVER(dir);
1463         struct nfs4_opendata *opendata;
1464         int status;
1465
1466         /* Protect against reboot recovery conflicts */
1467         status = -ENOMEM;
1468         if (!(sp = nfs4_get_state_owner(server, cred))) {
1469                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1470                 goto out_err;
1471         }
1472         status = nfs4_recover_expired_lease(server);
1473         if (status != 0)
1474                 goto err_put_state_owner;
1475         if (path->dentry->d_inode != NULL)
1476                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1477         status = -ENOMEM;
1478         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr);
1479         if (opendata == NULL)
1480                 goto err_put_state_owner;
1481
1482         if (path->dentry->d_inode != NULL)
1483                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1484
1485         status = _nfs4_proc_open(opendata);
1486         if (status != 0)
1487                 goto err_opendata_put;
1488
1489         if (opendata->o_arg.open_flags & O_EXCL)
1490                 nfs4_exclusive_attrset(opendata, sattr);
1491
1492         state = nfs4_opendata_to_nfs4_state(opendata);
1493         status = PTR_ERR(state);
1494         if (IS_ERR(state))
1495                 goto err_opendata_put;
1496         nfs4_opendata_put(opendata);
1497         nfs4_put_state_owner(sp);
1498         *res = state;
1499         return 0;
1500 err_opendata_put:
1501         nfs4_opendata_put(opendata);
1502 err_put_state_owner:
1503         nfs4_put_state_owner(sp);
1504 out_err:
1505         *res = NULL;
1506         return status;
1507 }
1508
1509
1510 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1511 {
1512         struct nfs4_exception exception = { };
1513         struct nfs4_state *res;
1514         int status;
1515
1516         do {
1517                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1518                 if (status == 0)
1519                         break;
1520                 /* NOTE: BAD_SEQID means the server and client disagree about the
1521                  * book-keeping w.r.t. state-changing operations
1522                  * (OPEN/CLOSE/LOCK/LOCKU...)
1523                  * It is actually a sign of a bug on the client or on the server.
1524                  *
1525                  * If we receive a BAD_SEQID error in the particular case of
1526                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1527                  * have unhashed the old state_owner for us, and that we can
1528                  * therefore safely retry using a new one. We should still warn
1529                  * the user though...
1530                  */
1531                 if (status == -NFS4ERR_BAD_SEQID) {
1532                         printk(KERN_WARNING "NFS: v4 server %s "
1533                                         " returned a bad sequence-id error!\n",
1534                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1535                         exception.retry = 1;
1536                         continue;
1537                 }
1538                 /*
1539                  * BAD_STATEID on OPEN means that the server cancelled our
1540                  * state before it received the OPEN_CONFIRM.
1541                  * Recover by retrying the request as per the discussion
1542                  * on Page 181 of RFC3530.
1543                  */
1544                 if (status == -NFS4ERR_BAD_STATEID) {
1545                         exception.retry = 1;
1546                         continue;
1547                 }
1548                 if (status == -EAGAIN) {
1549                         /* We must have found a delegation */
1550                         exception.retry = 1;
1551                         continue;
1552                 }
1553                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1554                                         status, &exception));
1555         } while (exception.retry);
1556         return res;
1557 }
1558
1559 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1560                             struct nfs_fattr *fattr, struct iattr *sattr,
1561                             struct nfs4_state *state)
1562 {
1563         struct nfs_server *server = NFS_SERVER(inode);
1564         struct nfs_setattrargs  arg = {
1565                 .fh             = NFS_FH(inode),
1566                 .iap            = sattr,
1567                 .server         = server,
1568                 .bitmask = server->attr_bitmask,
1569         };
1570         struct nfs_setattrres  res = {
1571                 .fattr          = fattr,
1572                 .server         = server,
1573         };
1574         struct rpc_message msg = {
1575                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1576                 .rpc_argp       = &arg,
1577                 .rpc_resp       = &res,
1578                 .rpc_cred       = cred,
1579         };
1580         unsigned long timestamp = jiffies;
1581         int status;
1582
1583         nfs_fattr_init(fattr);
1584
1585         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1586                 /* Use that stateid */
1587         } else if (state != NULL) {
1588                 nfs4_copy_stateid(&arg.stateid, state, current->files);
1589         } else
1590                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1591
1592         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
1593         if (status == 0 && state != NULL)
1594                 renew_lease(server, timestamp);
1595         return status;
1596 }
1597
1598 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1599                            struct nfs_fattr *fattr, struct iattr *sattr,
1600                            struct nfs4_state *state)
1601 {
1602         struct nfs_server *server = NFS_SERVER(inode);
1603         struct nfs4_exception exception = { };
1604         int err;
1605         do {
1606                 err = nfs4_handle_exception(server,
1607                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1608                                 &exception);
1609         } while (exception.retry);
1610         return err;
1611 }
1612
1613 struct nfs4_closedata {
1614         struct path path;
1615         struct inode *inode;
1616         struct nfs4_state *state;
1617         struct nfs_closeargs arg;
1618         struct nfs_closeres res;
1619         struct nfs_fattr fattr;
1620         unsigned long timestamp;
1621 };
1622
1623 static void nfs4_free_closedata(void *data)
1624 {
1625         struct nfs4_closedata *calldata = data;
1626         struct nfs4_state_owner *sp = calldata->state->owner;
1627
1628         nfs4_put_open_state(calldata->state);
1629         nfs_free_seqid(calldata->arg.seqid);
1630         nfs4_put_state_owner(sp);
1631         path_put(&calldata->path);
1632         kfree(calldata);
1633 }
1634
1635 static void nfs4_close_done(struct rpc_task *task, void *data)
1636 {
1637         struct nfs4_closedata *calldata = data;
1638         struct nfs4_state *state = calldata->state;
1639         struct nfs_server *server = NFS_SERVER(calldata->inode);
1640
1641         nfs4_sequence_done(server, &calldata->res.seq_res, task->tk_status);
1642         if (RPC_ASSASSINATED(task))
1643                 return;
1644         /* hmm. we are done with the inode, and in the process of freeing
1645          * the state_owner. we keep this around to process errors
1646          */
1647         switch (task->tk_status) {
1648                 case 0:
1649                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1650                         renew_lease(server, calldata->timestamp);
1651                         break;
1652                 case -NFS4ERR_STALE_STATEID:
1653                 case -NFS4ERR_OLD_STATEID:
1654                 case -NFS4ERR_BAD_STATEID:
1655                 case -NFS4ERR_EXPIRED:
1656                         if (calldata->arg.fmode == 0)
1657                                 break;
1658                 default:
1659                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
1660                                 rpc_restart_call(task);
1661                                 return;
1662                         }
1663         }
1664         nfs4_sequence_free_slot(server->nfs_client, &calldata->res.seq_res);
1665         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1666 }
1667
1668 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1669 {
1670         struct nfs4_closedata *calldata = data;
1671         struct nfs4_state *state = calldata->state;
1672         int clear_rd, clear_wr, clear_rdwr;
1673
1674         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1675                 return;
1676
1677         clear_rd = clear_wr = clear_rdwr = 0;
1678         spin_lock(&state->owner->so_lock);
1679         /* Calculate the change in open mode */
1680         if (state->n_rdwr == 0) {
1681                 if (state->n_rdonly == 0) {
1682                         clear_rd |= test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1683                         clear_rdwr |= test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags);
1684                 }
1685                 if (state->n_wronly == 0) {
1686                         clear_wr |= test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1687                         clear_rdwr |= test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags);
1688                 }
1689         }
1690         spin_unlock(&state->owner->so_lock);
1691         if (!clear_rd && !clear_wr && !clear_rdwr) {
1692                 /* Note: exit _without_ calling nfs4_close_done */
1693                 task->tk_action = NULL;
1694                 return;
1695         }
1696         nfs_fattr_init(calldata->res.fattr);
1697         if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0) {
1698                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1699                 calldata->arg.fmode = FMODE_READ;
1700         } else if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0) {
1701                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1702                 calldata->arg.fmode = FMODE_WRITE;
1703         }
1704         calldata->timestamp = jiffies;
1705         if (nfs4_setup_sequence((NFS_SERVER(calldata->inode))->nfs_client,
1706                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1707                                 1, task))
1708                 return;
1709         rpc_call_start(task);
1710 }
1711
1712 static const struct rpc_call_ops nfs4_close_ops = {
1713         .rpc_call_prepare = nfs4_close_prepare,
1714         .rpc_call_done = nfs4_close_done,
1715         .rpc_release = nfs4_free_closedata,
1716 };
1717
1718 /* 
1719  * It is possible for data to be read/written from a mem-mapped file 
1720  * after the sys_close call (which hits the vfs layer as a flush).
1721  * This means that we can't safely call nfsv4 close on a file until 
1722  * the inode is cleared. This in turn means that we are not good
1723  * NFSv4 citizens - we do not indicate to the server to update the file's 
1724  * share state even when we are done with one of the three share 
1725  * stateid's in the inode.
1726  *
1727  * NOTE: Caller must be holding the sp->so_owner semaphore!
1728  */
1729 int nfs4_do_close(struct path *path, struct nfs4_state *state, int wait)
1730 {
1731         struct nfs_server *server = NFS_SERVER(state->inode);
1732         struct nfs4_closedata *calldata;
1733         struct nfs4_state_owner *sp = state->owner;
1734         struct rpc_task *task;
1735         struct rpc_message msg = {
1736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
1737                 .rpc_cred = state->owner->so_cred,
1738         };
1739         struct rpc_task_setup task_setup_data = {
1740                 .rpc_client = server->client,
1741                 .rpc_message = &msg,
1742                 .callback_ops = &nfs4_close_ops,
1743                 .workqueue = nfsiod_workqueue,
1744                 .flags = RPC_TASK_ASYNC,
1745         };
1746         int status = -ENOMEM;
1747
1748         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
1749         if (calldata == NULL)
1750                 goto out;
1751         calldata->inode = state->inode;
1752         calldata->state = state;
1753         calldata->arg.fh = NFS_FH(state->inode);
1754         calldata->arg.stateid = &state->open_stateid;
1755         /* Serialization for the sequence id */
1756         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
1757         if (calldata->arg.seqid == NULL)
1758                 goto out_free_calldata;
1759         calldata->arg.fmode = 0;
1760         calldata->arg.bitmask = server->cache_consistency_bitmask;
1761         calldata->res.fattr = &calldata->fattr;
1762         calldata->res.seqid = calldata->arg.seqid;
1763         calldata->res.server = server;
1764         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
1765         calldata->path.mnt = mntget(path->mnt);
1766         calldata->path.dentry = dget(path->dentry);
1767
1768         msg.rpc_argp = &calldata->arg,
1769         msg.rpc_resp = &calldata->res,
1770         task_setup_data.callback_data = calldata;
1771         task = rpc_run_task(&task_setup_data);
1772         if (IS_ERR(task))
1773                 return PTR_ERR(task);
1774         status = 0;
1775         if (wait)
1776                 status = rpc_wait_for_completion_task(task);
1777         rpc_put_task(task);
1778         return status;
1779 out_free_calldata:
1780         kfree(calldata);
1781 out:
1782         nfs4_put_open_state(state);
1783         nfs4_put_state_owner(sp);
1784         return status;
1785 }
1786
1787 static int nfs4_intent_set_file(struct nameidata *nd, struct path *path, struct nfs4_state *state, fmode_t fmode)
1788 {
1789         struct file *filp;
1790         int ret;
1791
1792         /* If the open_intent is for execute, we have an extra check to make */
1793         if (fmode & FMODE_EXEC) {
1794                 ret = nfs_may_open(state->inode,
1795                                 state->owner->so_cred,
1796                                 nd->intent.open.flags);
1797                 if (ret < 0)
1798                         goto out_close;
1799         }
1800         filp = lookup_instantiate_filp(nd, path->dentry, NULL);
1801         if (!IS_ERR(filp)) {
1802                 struct nfs_open_context *ctx;
1803                 ctx = nfs_file_open_context(filp);
1804                 ctx->state = state;
1805                 return 0;
1806         }
1807         ret = PTR_ERR(filp);
1808 out_close:
1809         nfs4_close_sync(path, state, fmode & (FMODE_READ|FMODE_WRITE));
1810         return ret;
1811 }
1812
1813 struct dentry *
1814 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1815 {
1816         struct path path = {
1817                 .mnt = nd->path.mnt,
1818                 .dentry = dentry,
1819         };
1820         struct dentry *parent;
1821         struct iattr attr;
1822         struct rpc_cred *cred;
1823         struct nfs4_state *state;
1824         struct dentry *res;
1825         fmode_t fmode = nd->intent.open.flags & (FMODE_READ | FMODE_WRITE | FMODE_EXEC);
1826
1827         if (nd->flags & LOOKUP_CREATE) {
1828                 attr.ia_mode = nd->intent.open.create_mode;
1829                 attr.ia_valid = ATTR_MODE;
1830                 if (!IS_POSIXACL(dir))
1831                         attr.ia_mode &= ~current_umask();
1832         } else {
1833                 attr.ia_valid = 0;
1834                 BUG_ON(nd->intent.open.flags & O_CREAT);
1835         }
1836
1837         cred = rpc_lookup_cred();
1838         if (IS_ERR(cred))
1839                 return (struct dentry *)cred;
1840         parent = dentry->d_parent;
1841         /* Protect against concurrent sillydeletes */
1842         nfs_block_sillyrename(parent);
1843         state = nfs4_do_open(dir, &path, fmode, nd->intent.open.flags, &attr, cred);
1844         put_rpccred(cred);
1845         if (IS_ERR(state)) {
1846                 if (PTR_ERR(state) == -ENOENT) {
1847                         d_add(dentry, NULL);
1848                         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
1849                 }
1850                 nfs_unblock_sillyrename(parent);
1851                 return (struct dentry *)state;
1852         }
1853         res = d_add_unique(dentry, igrab(state->inode));
1854         if (res != NULL)
1855                 path.dentry = res;
1856         nfs_set_verifier(path.dentry, nfs_save_change_attribute(dir));
1857         nfs_unblock_sillyrename(parent);
1858         nfs4_intent_set_file(nd, &path, state, fmode);
1859         return res;
1860 }
1861
1862 int
1863 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
1864 {
1865         struct path path = {
1866                 .mnt = nd->path.mnt,
1867                 .dentry = dentry,
1868         };
1869         struct rpc_cred *cred;
1870         struct nfs4_state *state;
1871         fmode_t fmode = openflags & (FMODE_READ | FMODE_WRITE);
1872
1873         cred = rpc_lookup_cred();
1874         if (IS_ERR(cred))
1875                 return PTR_ERR(cred);
1876         state = nfs4_do_open(dir, &path, fmode, openflags, NULL, cred);
1877         put_rpccred(cred);
1878         if (IS_ERR(state)) {
1879                 switch (PTR_ERR(state)) {
1880                         case -EPERM:
1881                         case -EACCES:
1882                         case -EDQUOT:
1883                         case -ENOSPC:
1884                         case -EROFS:
1885                                 lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
1886                                 return 1;
1887                         default:
1888                                 goto out_drop;
1889                 }
1890         }
1891         if (state->inode == dentry->d_inode) {
1892                 nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
1893                 nfs4_intent_set_file(nd, &path, state, fmode);
1894                 return 1;
1895         }
1896         nfs4_close_sync(&path, state, fmode);
1897 out_drop:
1898         d_drop(dentry);
1899         return 0;
1900 }
1901
1902 void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
1903 {
1904         if (ctx->state == NULL)
1905                 return;
1906         if (is_sync)
1907                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
1908         else
1909                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
1910 }
1911
1912 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
1913 {
1914         struct nfs4_server_caps_arg args = {
1915                 .fhandle = fhandle,
1916         };
1917         struct nfs4_server_caps_res res = {};
1918         struct rpc_message msg = {
1919                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
1920                 .rpc_argp = &args,
1921                 .rpc_resp = &res,
1922         };
1923         int status;
1924
1925         status = nfs4_call_sync(server, &msg, &args, &res, 0);
1926         if (status == 0) {
1927                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
1928                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
1929                         server->caps |= NFS_CAP_ACLS;
1930                 if (res.has_links != 0)
1931                         server->caps |= NFS_CAP_HARDLINKS;
1932                 if (res.has_symlinks != 0)
1933                         server->caps |= NFS_CAP_SYMLINKS;
1934                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
1935                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
1936                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
1937                 server->acl_bitmask = res.acl_bitmask;
1938         }
1939
1940         return status;
1941 }
1942
1943 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
1944 {
1945         struct nfs4_exception exception = { };
1946         int err;
1947         do {
1948                 err = nfs4_handle_exception(server,
1949                                 _nfs4_server_capabilities(server, fhandle),
1950                                 &exception);
1951         } while (exception.retry);
1952         return err;
1953 }
1954
1955 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
1956                 struct nfs_fsinfo *info)
1957 {
1958         struct nfs4_lookup_root_arg args = {
1959                 .bitmask = nfs4_fattr_bitmap,
1960         };
1961         struct nfs4_lookup_res res = {
1962                 .server = server,
1963                 .fattr = info->fattr,
1964                 .fh = fhandle,
1965         };
1966         struct rpc_message msg = {
1967                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
1968                 .rpc_argp = &args,
1969                 .rpc_resp = &res,
1970         };
1971         nfs_fattr_init(info->fattr);
1972         return nfs4_call_sync(server, &msg, &args, &res, 0);
1973 }
1974
1975 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
1976                 struct nfs_fsinfo *info)
1977 {
1978         struct nfs4_exception exception = { };
1979         int err;
1980         do {
1981                 err = nfs4_handle_exception(server,
1982                                 _nfs4_lookup_root(server, fhandle, info),
1983                                 &exception);
1984         } while (exception.retry);
1985         return err;
1986 }
1987
1988 /*
1989  * get the file handle for the "/" directory on the server
1990  */
1991 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
1992                               struct nfs_fsinfo *info)
1993 {
1994         int status;
1995
1996         status = nfs4_lookup_root(server, fhandle, info);
1997         if (status == 0)
1998                 status = nfs4_server_capabilities(server, fhandle);
1999         if (status == 0)
2000                 status = nfs4_do_fsinfo(server, fhandle, info);
2001         return nfs4_map_errors(status);
2002 }
2003
2004 /*
2005  * Get locations and (maybe) other attributes of a referral.
2006  * Note that we'll actually follow the referral later when
2007  * we detect fsid mismatch in inode revalidation
2008  */
2009 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2010 {
2011         int status = -ENOMEM;
2012         struct page *page = NULL;
2013         struct nfs4_fs_locations *locations = NULL;
2014
2015         page = alloc_page(GFP_KERNEL);
2016         if (page == NULL)
2017                 goto out;
2018         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2019         if (locations == NULL)
2020                 goto out;
2021
2022         status = nfs4_proc_fs_locations(dir, name, locations, page);
2023         if (status != 0)
2024                 goto out;
2025         /* Make sure server returned a different fsid for the referral */
2026         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2027                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2028                 status = -EIO;
2029                 goto out;
2030         }
2031
2032         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2033         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2034         if (!fattr->mode)
2035                 fattr->mode = S_IFDIR;
2036         memset(fhandle, 0, sizeof(struct nfs_fh));
2037 out:
2038         if (page)
2039                 __free_page(page);
2040         if (locations)
2041                 kfree(locations);
2042         return status;
2043 }
2044
2045 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2046 {
2047         struct nfs4_getattr_arg args = {
2048                 .fh = fhandle,
2049                 .bitmask = server->attr_bitmask,
2050         };
2051         struct nfs4_getattr_res res = {
2052                 .fattr = fattr,
2053                 .server = server,
2054         };
2055         struct rpc_message msg = {
2056                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2057                 .rpc_argp = &args,
2058                 .rpc_resp = &res,
2059         };
2060         
2061         nfs_fattr_init(fattr);
2062         return nfs4_call_sync(server, &msg, &args, &res, 0);
2063 }
2064
2065 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2066 {
2067         struct nfs4_exception exception = { };
2068         int err;
2069         do {
2070                 err = nfs4_handle_exception(server,
2071                                 _nfs4_proc_getattr(server, fhandle, fattr),
2072                                 &exception);
2073         } while (exception.retry);
2074         return err;
2075 }
2076
2077 /* 
2078  * The file is not closed if it is opened due to the a request to change
2079  * the size of the file. The open call will not be needed once the
2080  * VFS layer lookup-intents are implemented.
2081  *
2082  * Close is called when the inode is destroyed.
2083  * If we haven't opened the file for O_WRONLY, we
2084  * need to in the size_change case to obtain a stateid.
2085  *
2086  * Got race?
2087  * Because OPEN is always done by name in nfsv4, it is
2088  * possible that we opened a different file by the same
2089  * name.  We can recognize this race condition, but we
2090  * can't do anything about it besides returning an error.
2091  *
2092  * This will be fixed with VFS changes (lookup-intent).
2093  */
2094 static int
2095 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2096                   struct iattr *sattr)
2097 {
2098         struct inode *inode = dentry->d_inode;
2099         struct rpc_cred *cred = NULL;
2100         struct nfs4_state *state = NULL;
2101         int status;
2102
2103         nfs_fattr_init(fattr);
2104         
2105         /* Search for an existing open(O_WRITE) file */
2106         if (sattr->ia_valid & ATTR_FILE) {
2107                 struct nfs_open_context *ctx;
2108
2109                 ctx = nfs_file_open_context(sattr->ia_file);
2110                 if (ctx) {
2111                         cred = ctx->cred;
2112                         state = ctx->state;
2113                 }
2114         }
2115
2116         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2117         if (status == 0)
2118                 nfs_setattr_update_inode(inode, sattr);
2119         return status;
2120 }
2121
2122 static int _nfs4_proc_lookupfh(struct nfs_server *server, const struct nfs_fh *dirfh,
2123                 const struct qstr *name, struct nfs_fh *fhandle,
2124                 struct nfs_fattr *fattr)
2125 {
2126         int                    status;
2127         struct nfs4_lookup_arg args = {
2128                 .bitmask = server->attr_bitmask,
2129                 .dir_fh = dirfh,
2130                 .name = name,
2131         };
2132         struct nfs4_lookup_res res = {
2133                 .server = server,
2134                 .fattr = fattr,
2135                 .fh = fhandle,
2136         };
2137         struct rpc_message msg = {
2138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2139                 .rpc_argp = &args,
2140                 .rpc_resp = &res,
2141         };
2142
2143         nfs_fattr_init(fattr);
2144
2145         dprintk("NFS call  lookupfh %s\n", name->name);
2146         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2147         dprintk("NFS reply lookupfh: %d\n", status);
2148         return status;
2149 }
2150
2151 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2152                               struct qstr *name, struct nfs_fh *fhandle,
2153                               struct nfs_fattr *fattr)
2154 {
2155         struct nfs4_exception exception = { };
2156         int err;
2157         do {
2158                 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
2159                 /* FIXME: !!!! */
2160                 if (err == -NFS4ERR_MOVED) {
2161                         err = -EREMOTE;
2162                         break;
2163                 }
2164                 err = nfs4_handle_exception(server, err, &exception);
2165         } while (exception.retry);
2166         return err;
2167 }
2168
2169 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
2170                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2171 {
2172         int status;
2173         
2174         dprintk("NFS call  lookup %s\n", name->name);
2175         status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2176         if (status == -NFS4ERR_MOVED)
2177                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2178         dprintk("NFS reply lookup: %d\n", status);
2179         return status;
2180 }
2181
2182 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2183 {
2184         struct nfs4_exception exception = { };
2185         int err;
2186         do {
2187                 err = nfs4_handle_exception(NFS_SERVER(dir),
2188                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
2189                                 &exception);
2190         } while (exception.retry);
2191         return err;
2192 }
2193
2194 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2195 {
2196         struct nfs_server *server = NFS_SERVER(inode);
2197         struct nfs_fattr fattr;
2198         struct nfs4_accessargs args = {
2199                 .fh = NFS_FH(inode),
2200                 .bitmask = server->attr_bitmask,
2201         };
2202         struct nfs4_accessres res = {
2203                 .server = server,
2204                 .fattr = &fattr,
2205         };
2206         struct rpc_message msg = {
2207                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2208                 .rpc_argp = &args,
2209                 .rpc_resp = &res,
2210                 .rpc_cred = entry->cred,
2211         };
2212         int mode = entry->mask;
2213         int status;
2214
2215         /*
2216          * Determine which access bits we want to ask for...
2217          */
2218         if (mode & MAY_READ)
2219                 args.access |= NFS4_ACCESS_READ;
2220         if (S_ISDIR(inode->i_mode)) {
2221                 if (mode & MAY_WRITE)
2222                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2223                 if (mode & MAY_EXEC)
2224                         args.access |= NFS4_ACCESS_LOOKUP;
2225         } else {
2226                 if (mode & MAY_WRITE)
2227                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2228                 if (mode & MAY_EXEC)
2229                         args.access |= NFS4_ACCESS_EXECUTE;
2230         }
2231         nfs_fattr_init(&fattr);
2232         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2233         if (!status) {
2234                 entry->mask = 0;
2235                 if (res.access & NFS4_ACCESS_READ)
2236                         entry->mask |= MAY_READ;
2237                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2238                         entry->mask |= MAY_WRITE;
2239                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2240                         entry->mask |= MAY_EXEC;
2241                 nfs_refresh_inode(inode, &fattr);
2242         }
2243         return status;
2244 }
2245
2246 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2247 {
2248         struct nfs4_exception exception = { };
2249         int err;
2250         do {
2251                 err = nfs4_handle_exception(NFS_SERVER(inode),
2252                                 _nfs4_proc_access(inode, entry),
2253                                 &exception);
2254         } while (exception.retry);
2255         return err;
2256 }
2257
2258 /*
2259  * TODO: For the time being, we don't try to get any attributes
2260  * along with any of the zero-copy operations READ, READDIR,
2261  * READLINK, WRITE.
2262  *
2263  * In the case of the first three, we want to put the GETATTR
2264  * after the read-type operation -- this is because it is hard
2265  * to predict the length of a GETATTR response in v4, and thus
2266  * align the READ data correctly.  This means that the GETATTR
2267  * may end up partially falling into the page cache, and we should
2268  * shift it into the 'tail' of the xdr_buf before processing.
2269  * To do this efficiently, we need to know the total length
2270  * of data received, which doesn't seem to be available outside
2271  * of the RPC layer.
2272  *
2273  * In the case of WRITE, we also want to put the GETATTR after
2274  * the operation -- in this case because we want to make sure
2275  * we get the post-operation mtime and size.  This means that
2276  * we can't use xdr_encode_pages() as written: we need a variant
2277  * of it which would leave room in the 'tail' iovec.
2278  *
2279  * Both of these changes to the XDR layer would in fact be quite
2280  * minor, but I decided to leave them for a subsequent patch.
2281  */
2282 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2283                 unsigned int pgbase, unsigned int pglen)
2284 {
2285         struct nfs4_readlink args = {
2286                 .fh       = NFS_FH(inode),
2287                 .pgbase   = pgbase,
2288                 .pglen    = pglen,
2289                 .pages    = &page,
2290         };
2291         struct nfs4_readlink_res res;
2292         struct rpc_message msg = {
2293                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2294                 .rpc_argp = &args,
2295                 .rpc_resp = &res,
2296         };
2297
2298         return nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
2299 }
2300
2301 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2302                 unsigned int pgbase, unsigned int pglen)
2303 {
2304         struct nfs4_exception exception = { };
2305         int err;
2306         do {
2307                 err = nfs4_handle_exception(NFS_SERVER(inode),
2308                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2309                                 &exception);
2310         } while (exception.retry);
2311         return err;
2312 }
2313
2314 /*
2315  * Got race?
2316  * We will need to arrange for the VFS layer to provide an atomic open.
2317  * Until then, this create/open method is prone to inefficiency and race
2318  * conditions due to the lookup, create, and open VFS calls from sys_open()
2319  * placed on the wire.
2320  *
2321  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2322  * The file will be opened again in the subsequent VFS open call
2323  * (nfs4_proc_file_open).
2324  *
2325  * The open for read will just hang around to be used by any process that
2326  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2327  */
2328
2329 static int
2330 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2331                  int flags, struct nameidata *nd)
2332 {
2333         struct path path = {
2334                 .mnt = nd->path.mnt,
2335                 .dentry = dentry,
2336         };
2337         struct nfs4_state *state;
2338         struct rpc_cred *cred;
2339         fmode_t fmode = flags & (FMODE_READ | FMODE_WRITE);
2340         int status = 0;
2341
2342         cred = rpc_lookup_cred();
2343         if (IS_ERR(cred)) {
2344                 status = PTR_ERR(cred);
2345                 goto out;
2346         }
2347         state = nfs4_do_open(dir, &path, fmode, flags, sattr, cred);
2348         d_drop(dentry);
2349         if (IS_ERR(state)) {
2350                 status = PTR_ERR(state);
2351                 goto out_putcred;
2352         }
2353         d_add(dentry, igrab(state->inode));
2354         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2355         if (flags & O_EXCL) {
2356                 struct nfs_fattr fattr;
2357                 status = nfs4_do_setattr(state->inode, cred, &fattr, sattr, state);
2358                 if (status == 0)
2359                         nfs_setattr_update_inode(state->inode, sattr);
2360                 nfs_post_op_update_inode(state->inode, &fattr);
2361         }
2362         if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0)
2363                 status = nfs4_intent_set_file(nd, &path, state, fmode);
2364         else
2365                 nfs4_close_sync(&path, state, fmode);
2366 out_putcred:
2367         put_rpccred(cred);
2368 out:
2369         return status;
2370 }
2371
2372 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2373 {
2374         struct nfs_server *server = NFS_SERVER(dir);
2375         struct nfs_removeargs args = {
2376                 .fh = NFS_FH(dir),
2377                 .name.len = name->len,
2378                 .name.name = name->name,
2379                 .bitmask = server->attr_bitmask,
2380         };
2381         struct nfs_removeres res = {
2382                 .server = server,
2383         };
2384         struct rpc_message msg = {
2385                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2386                 .rpc_argp = &args,
2387                 .rpc_resp = &res,
2388         };
2389         int                     status;
2390
2391         nfs_fattr_init(&res.dir_attr);
2392         status = nfs4_call_sync(server, &msg, &args, &res, 1);
2393         if (status == 0) {
2394                 update_changeattr(dir, &res.cinfo);
2395                 nfs_post_op_update_inode(dir, &res.dir_attr);
2396         }
2397         return status;
2398 }
2399
2400 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2401 {
2402         struct nfs4_exception exception = { };
2403         int err;
2404         do {
2405                 err = nfs4_handle_exception(NFS_SERVER(dir),
2406                                 _nfs4_proc_remove(dir, name),
2407                                 &exception);
2408         } while (exception.retry);
2409         return err;
2410 }
2411
2412 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2413 {
2414         struct nfs_server *server = NFS_SERVER(dir);
2415         struct nfs_removeargs *args = msg->rpc_argp;
2416         struct nfs_removeres *res = msg->rpc_resp;
2417
2418         args->bitmask = server->cache_consistency_bitmask;
2419         res->server = server;
2420         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2421 }
2422
2423 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2424 {
2425         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2426
2427         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2428                 return 0;
2429         update_changeattr(dir, &res->cinfo);
2430         nfs_post_op_update_inode(dir, &res->dir_attr);
2431         return 1;
2432 }
2433
2434 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2435                 struct inode *new_dir, struct qstr *new_name)
2436 {
2437         struct nfs_server *server = NFS_SERVER(old_dir);
2438         struct nfs4_rename_arg arg = {
2439                 .old_dir = NFS_FH(old_dir),
2440                 .new_dir = NFS_FH(new_dir),
2441                 .old_name = old_name,
2442                 .new_name = new_name,
2443                 .bitmask = server->attr_bitmask,
2444         };
2445         struct nfs_fattr old_fattr, new_fattr;
2446         struct nfs4_rename_res res = {
2447                 .server = server,
2448                 .old_fattr = &old_fattr,
2449                 .new_fattr = &new_fattr,
2450         };
2451         struct rpc_message msg = {
2452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2453                 .rpc_argp = &arg,
2454                 .rpc_resp = &res,
2455         };
2456         int                     status;
2457         
2458         nfs_fattr_init(res.old_fattr);
2459         nfs_fattr_init(res.new_fattr);
2460         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2461
2462         if (!status) {
2463                 update_changeattr(old_dir, &res.old_cinfo);
2464                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2465                 update_changeattr(new_dir, &res.new_cinfo);
2466                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2467         }
2468         return status;
2469 }
2470
2471 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2472                 struct inode *new_dir, struct qstr *new_name)
2473 {
2474         struct nfs4_exception exception = { };
2475         int err;
2476         do {
2477                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2478                                 _nfs4_proc_rename(old_dir, old_name,
2479                                         new_dir, new_name),
2480                                 &exception);
2481         } while (exception.retry);
2482         return err;
2483 }
2484
2485 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2486 {
2487         struct nfs_server *server = NFS_SERVER(inode);
2488         struct nfs4_link_arg arg = {
2489                 .fh     = NFS_FH(inode),
2490                 .dir_fh = NFS_FH(dir),
2491                 .name   = name,
2492                 .bitmask = server->attr_bitmask,
2493         };
2494         struct nfs_fattr fattr, dir_attr;
2495         struct nfs4_link_res res = {
2496                 .server = server,
2497                 .fattr = &fattr,
2498                 .dir_attr = &dir_attr,
2499         };
2500         struct rpc_message msg = {
2501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2502                 .rpc_argp = &arg,
2503                 .rpc_resp = &res,
2504         };
2505         int                     status;
2506
2507         nfs_fattr_init(res.fattr);
2508         nfs_fattr_init(res.dir_attr);
2509         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2510         if (!status) {
2511                 update_changeattr(dir, &res.cinfo);
2512                 nfs_post_op_update_inode(dir, res.dir_attr);
2513                 nfs_post_op_update_inode(inode, res.fattr);
2514         }
2515
2516         return status;
2517 }
2518
2519 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2520 {
2521         struct nfs4_exception exception = { };
2522         int err;
2523         do {
2524                 err = nfs4_handle_exception(NFS_SERVER(inode),
2525                                 _nfs4_proc_link(inode, dir, name),
2526                                 &exception);
2527         } while (exception.retry);
2528         return err;
2529 }
2530
2531 struct nfs4_createdata {
2532         struct rpc_message msg;
2533         struct nfs4_create_arg arg;
2534         struct nfs4_create_res res;
2535         struct nfs_fh fh;
2536         struct nfs_fattr fattr;
2537         struct nfs_fattr dir_fattr;
2538 };
2539
2540 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2541                 struct qstr *name, struct iattr *sattr, u32 ftype)
2542 {
2543         struct nfs4_createdata *data;
2544
2545         data = kzalloc(sizeof(*data), GFP_KERNEL);
2546         if (data != NULL) {
2547                 struct nfs_server *server = NFS_SERVER(dir);
2548
2549                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2550                 data->msg.rpc_argp = &data->arg;
2551                 data->msg.rpc_resp = &data->res;
2552                 data->arg.dir_fh = NFS_FH(dir);
2553                 data->arg.server = server;
2554                 data->arg.name = name;
2555                 data->arg.attrs = sattr;
2556                 data->arg.ftype = ftype;
2557                 data->arg.bitmask = server->attr_bitmask;
2558                 data->res.server = server;
2559                 data->res.fh = &data->fh;
2560                 data->res.fattr = &data->fattr;
2561                 data->res.dir_fattr = &data->dir_fattr;
2562                 nfs_fattr_init(data->res.fattr);
2563                 nfs_fattr_init(data->res.dir_fattr);
2564         }
2565         return data;
2566 }
2567
2568 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2569 {
2570         int status = nfs4_call_sync(NFS_SERVER(dir), &data->msg,
2571                                     &data->arg, &data->res, 1);
2572         if (status == 0) {
2573                 update_changeattr(dir, &data->res.dir_cinfo);
2574                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2575                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2576         }
2577         return status;
2578 }
2579
2580 static void nfs4_free_createdata(struct nfs4_createdata *data)
2581 {
2582         kfree(data);
2583 }
2584
2585 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2586                 struct page *page, unsigned int len, struct iattr *sattr)
2587 {
2588         struct nfs4_createdata *data;
2589         int status = -ENAMETOOLONG;
2590
2591         if (len > NFS4_MAXPATHLEN)
2592                 goto out;
2593
2594         status = -ENOMEM;
2595         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2596         if (data == NULL)
2597                 goto out;
2598
2599         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2600         data->arg.u.symlink.pages = &page;
2601         data->arg.u.symlink.len = len;
2602         
2603         status = nfs4_do_create(dir, dentry, data);
2604
2605         nfs4_free_createdata(data);
2606 out:
2607         return status;
2608 }
2609
2610 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2611                 struct page *page, unsigned int len, struct iattr *sattr)
2612 {
2613         struct nfs4_exception exception = { };
2614         int err;
2615         do {
2616                 err = nfs4_handle_exception(NFS_SERVER(dir),
2617                                 _nfs4_proc_symlink(dir, dentry, page,
2618                                                         len, sattr),
2619                                 &exception);
2620         } while (exception.retry);
2621         return err;
2622 }
2623
2624 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2625                 struct iattr *sattr)
2626 {
2627         struct nfs4_createdata *data;
2628         int status = -ENOMEM;
2629
2630         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2631         if (data == NULL)
2632                 goto out;
2633
2634         status = nfs4_do_create(dir, dentry, data);
2635
2636         nfs4_free_createdata(data);
2637 out:
2638         return status;
2639 }
2640
2641 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2642                 struct iattr *sattr)
2643 {
2644         struct nfs4_exception exception = { };
2645         int err;
2646         do {
2647                 err = nfs4_handle_exception(NFS_SERVER(dir),
2648                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2649                                 &exception);
2650         } while (exception.retry);
2651         return err;
2652 }
2653
2654 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2655                   u64 cookie, struct page *page, unsigned int count, int plus)
2656 {
2657         struct inode            *dir = dentry->d_inode;
2658         struct nfs4_readdir_arg args = {
2659                 .fh = NFS_FH(dir),
2660                 .pages = &page,
2661                 .pgbase = 0,
2662                 .count = count,
2663                 .bitmask = NFS_SERVER(dentry->d_inode)->cache_consistency_bitmask,
2664         };
2665         struct nfs4_readdir_res res;
2666         struct rpc_message msg = {
2667                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2668                 .rpc_argp = &args,
2669                 .rpc_resp = &res,
2670                 .rpc_cred = cred,
2671         };
2672         int                     status;
2673
2674         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2675                         dentry->d_parent->d_name.name,
2676                         dentry->d_name.name,
2677                         (unsigned long long)cookie);
2678         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2679         res.pgbase = args.pgbase;
2680         status = nfs4_call_sync(NFS_SERVER(dir), &msg, &args, &res, 0);
2681         if (status == 0)
2682                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2683
2684         nfs_invalidate_atime(dir);
2685
2686         dprintk("%s: returns %d\n", __func__, status);
2687         return status;
2688 }
2689
2690 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2691                   u64 cookie, struct page *page, unsigned int count, int plus)
2692 {
2693         struct nfs4_exception exception = { };
2694         int err;
2695         do {
2696                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2697                                 _nfs4_proc_readdir(dentry, cred, cookie,
2698                                         page, count, plus),
2699                                 &exception);
2700         } while (exception.retry);
2701         return err;
2702 }
2703
2704 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2705                 struct iattr *sattr, dev_t rdev)
2706 {
2707         struct nfs4_createdata *data;
2708         int mode = sattr->ia_mode;
2709         int status = -ENOMEM;
2710
2711         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2712         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2713
2714         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2715         if (data == NULL)
2716                 goto out;
2717
2718         if (S_ISFIFO(mode))
2719                 data->arg.ftype = NF4FIFO;
2720         else if (S_ISBLK(mode)) {
2721                 data->arg.ftype = NF4BLK;
2722                 data->arg.u.device.specdata1 = MAJOR(rdev);
2723                 data->arg.u.device.specdata2 = MINOR(rdev);
2724         }
2725         else if (S_ISCHR(mode)) {
2726                 data->arg.ftype = NF4CHR;
2727                 data->arg.u.device.specdata1 = MAJOR(rdev);
2728                 data->arg.u.device.specdata2 = MINOR(rdev);
2729         }
2730         
2731         status = nfs4_do_create(dir, dentry, data);
2732
2733         nfs4_free_createdata(data);
2734 out:
2735         return status;
2736 }
2737
2738 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2739                 struct iattr *sattr, dev_t rdev)
2740 {
2741         struct nfs4_exception exception = { };
2742         int err;
2743         do {
2744                 err = nfs4_handle_exception(NFS_SERVER(dir),
2745                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
2746                                 &exception);
2747         } while (exception.retry);
2748         return err;
2749 }
2750
2751 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
2752                  struct nfs_fsstat *fsstat)
2753 {
2754         struct nfs4_statfs_arg args = {
2755                 .fh = fhandle,
2756                 .bitmask = server->attr_bitmask,
2757         };
2758         struct nfs4_statfs_res res = {
2759                 .fsstat = fsstat,
2760         };
2761         struct rpc_message msg = {
2762                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
2763                 .rpc_argp = &args,
2764                 .rpc_resp = &res,
2765         };
2766
2767         nfs_fattr_init(fsstat->fattr);
2768         return  nfs4_call_sync(server, &msg, &args, &res, 0);
2769 }
2770
2771 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
2772 {
2773         struct nfs4_exception exception = { };
2774         int err;
2775         do {
2776                 err = nfs4_handle_exception(server,
2777                                 _nfs4_proc_statfs(server, fhandle, fsstat),
2778                                 &exception);
2779         } while (exception.retry);
2780         return err;
2781 }
2782
2783 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
2784                 struct nfs_fsinfo *fsinfo)
2785 {
2786         struct nfs4_fsinfo_arg args = {
2787                 .fh = fhandle,
2788                 .bitmask = server->attr_bitmask,
2789         };
2790         struct nfs4_fsinfo_res res = {
2791                 .fsinfo = fsinfo,
2792         };
2793         struct rpc_message msg = {
2794                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
2795                 .rpc_argp = &args,
2796                 .rpc_resp = &res,
2797         };
2798
2799         return nfs4_call_sync(server, &msg, &args, &res, 0);
2800 }
2801
2802 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
2803 {
2804         struct nfs4_exception exception = { };
2805         int err;
2806
2807         do {
2808                 err = nfs4_handle_exception(server,
2809                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
2810                                 &exception);
2811         } while (exception.retry);
2812         return err;
2813 }
2814
2815 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
2816 {
2817         nfs_fattr_init(fsinfo->fattr);
2818         return nfs4_do_fsinfo(server, fhandle, fsinfo);
2819 }
2820
2821 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
2822                 struct nfs_pathconf *pathconf)
2823 {
2824         struct nfs4_pathconf_arg args = {
2825                 .fh = fhandle,
2826                 .bitmask = server->attr_bitmask,
2827         };
2828         struct nfs4_pathconf_res res = {
2829                 .pathconf = pathconf,
2830         };
2831         struct rpc_message msg = {
2832                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
2833                 .rpc_argp = &args,
2834                 .rpc_resp = &res,
2835         };
2836
2837         /* None of the pathconf attributes are mandatory to implement */
2838         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
2839                 memset(pathconf, 0, sizeof(*pathconf));
2840                 return 0;
2841         }
2842
2843         nfs_fattr_init(pathconf->fattr);
2844         return nfs4_call_sync(server, &msg, &args, &res, 0);
2845 }
2846
2847 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
2848                 struct nfs_pathconf *pathconf)
2849 {
2850         struct nfs4_exception exception = { };
2851         int err;
2852
2853         do {
2854                 err = nfs4_handle_exception(server,
2855                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
2856                                 &exception);
2857         } while (exception.retry);
2858         return err;
2859 }
2860
2861 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
2862 {
2863         struct nfs_server *server = NFS_SERVER(data->inode);
2864
2865         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
2866                 rpc_restart_call(task);
2867                 return -EAGAIN;
2868         }
2869
2870         nfs_invalidate_atime(data->inode);
2871         if (task->tk_status > 0)
2872                 renew_lease(server, data->timestamp);
2873         return 0;
2874 }
2875
2876 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
2877 {
2878         data->timestamp   = jiffies;
2879         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
2880 }
2881
2882 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
2883 {
2884         struct inode *inode = data->inode;
2885         
2886         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
2887                 rpc_restart_call(task);
2888                 return -EAGAIN;
2889         }
2890         if (task->tk_status >= 0) {
2891                 renew_lease(NFS_SERVER(inode), data->timestamp);
2892                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
2893         }
2894         return 0;
2895 }
2896
2897 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
2898 {
2899         struct nfs_server *server = NFS_SERVER(data->inode);
2900
2901         data->args.bitmask = server->cache_consistency_bitmask;
2902         data->res.server = server;
2903         data->timestamp   = jiffies;
2904
2905         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
2906 }
2907
2908 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
2909 {
2910         struct inode *inode = data->inode;
2911         
2912         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
2913                 rpc_restart_call(task);
2914                 return -EAGAIN;
2915         }
2916         nfs_refresh_inode(inode, data->res.fattr);
2917         return 0;
2918 }
2919
2920 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
2921 {
2922         struct nfs_server *server = NFS_SERVER(data->inode);
2923         
2924         data->args.bitmask = server->cache_consistency_bitmask;
2925         data->res.server = server;
2926         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
2927 }
2928
2929 /*
2930  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
2931  * standalone procedure for queueing an asynchronous RENEW.
2932  */
2933 static void nfs4_renew_done(struct rpc_task *task, void *data)
2934 {
2935         struct nfs_client *clp = (struct nfs_client *)task->tk_msg.rpc_argp;
2936         unsigned long timestamp = (unsigned long)data;
2937
2938         if (task->tk_status < 0) {
2939                 /* Unless we're shutting down, schedule state recovery! */
2940                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
2941                         nfs4_schedule_state_recovery(clp);
2942                 return;
2943         }
2944         spin_lock(&clp->cl_lock);
2945         if (time_before(clp->cl_last_renewal,timestamp))
2946                 clp->cl_last_renewal = timestamp;
2947         spin_unlock(&clp->cl_lock);
2948 }
2949
2950 static const struct rpc_call_ops nfs4_renew_ops = {
2951         .rpc_call_done = nfs4_renew_done,
2952 };
2953
2954 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
2955 {
2956         struct rpc_message msg = {
2957                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2958                 .rpc_argp       = clp,
2959                 .rpc_cred       = cred,
2960         };
2961
2962         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
2963                         &nfs4_renew_ops, (void *)jiffies);
2964 }
2965
2966 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
2967 {
2968         struct rpc_message msg = {
2969                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2970                 .rpc_argp       = clp,
2971                 .rpc_cred       = cred,
2972         };
2973         unsigned long now = jiffies;
2974         int status;
2975
2976         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2977         if (status < 0)
2978                 return status;
2979         spin_lock(&clp->cl_lock);
2980         if (time_before(clp->cl_last_renewal,now))
2981                 clp->cl_last_renewal = now;
2982         spin_unlock(&clp->cl_lock);
2983         return 0;
2984 }
2985
2986 static inline int nfs4_server_supports_acls(struct nfs_server *server)
2987 {
2988         return (server->caps & NFS_CAP_ACLS)
2989                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2990                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
2991 }
2992
2993 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
2994  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
2995  * the stack.
2996  */
2997 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
2998
2999 static void buf_to_pages(const void *buf, size_t buflen,
3000                 struct page **pages, unsigned int *pgbase)
3001 {
3002         const void *p = buf;
3003
3004         *pgbase = offset_in_page(buf);
3005         p -= *pgbase;
3006         while (p < buf + buflen) {
3007                 *(pages++) = virt_to_page(p);
3008                 p += PAGE_CACHE_SIZE;
3009         }
3010 }
3011
3012 struct nfs4_cached_acl {
3013         int cached;
3014         size_t len;
3015         char data[0];
3016 };
3017
3018 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3019 {
3020         struct nfs_inode *nfsi = NFS_I(inode);
3021
3022         spin_lock(&inode->i_lock);
3023         kfree(nfsi->nfs4_acl);
3024         nfsi->nfs4_acl = acl;
3025         spin_unlock(&inode->i_lock);
3026 }
3027
3028 static void nfs4_zap_acl_attr(struct inode *inode)
3029 {
3030         nfs4_set_cached_acl(inode, NULL);
3031 }
3032
3033 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3034 {
3035         struct nfs_inode *nfsi = NFS_I(inode);
3036         struct nfs4_cached_acl *acl;
3037         int ret = -ENOENT;
3038
3039         spin_lock(&inode->i_lock);
3040         acl = nfsi->nfs4_acl;
3041         if (acl == NULL)
3042                 goto out;
3043         if (buf == NULL) /* user is just asking for length */
3044                 goto out_len;
3045         if (acl->cached == 0)
3046                 goto out;
3047         ret = -ERANGE; /* see getxattr(2) man page */
3048         if (acl->len > buflen)
3049                 goto out;
3050         memcpy(buf, acl->data, acl->len);
3051 out_len:
3052         ret = acl->len;
3053 out:
3054         spin_unlock(&inode->i_lock);
3055         return ret;
3056 }
3057
3058 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3059 {
3060         struct nfs4_cached_acl *acl;
3061
3062         if (buf && acl_len <= PAGE_SIZE) {
3063                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3064                 if (acl == NULL)
3065                         goto out;
3066                 acl->cached = 1;
3067                 memcpy(acl->data, buf, acl_len);
3068         } else {
3069                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3070                 if (acl == NULL)
3071                         goto out;
3072                 acl->cached = 0;
3073         }
3074         acl->len = acl_len;
3075 out:
3076         nfs4_set_cached_acl(inode, acl);
3077 }
3078
3079 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3080 {
3081         struct page *pages[NFS4ACL_MAXPAGES];
3082         struct nfs_getaclargs args = {
3083                 .fh = NFS_FH(inode),
3084                 .acl_pages = pages,
3085                 .acl_len = buflen,
3086         };
3087         struct nfs_getaclres res = {
3088                 .acl_len = buflen,
3089         };
3090         void *resp_buf;
3091         struct rpc_message msg = {
3092                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3093                 .rpc_argp = &args,
3094                 .rpc_resp = &res,
3095         };
3096         struct page *localpage = NULL;
3097         int ret;
3098
3099         if (buflen < PAGE_SIZE) {
3100                 /* As long as we're doing a round trip to the server anyway,
3101                  * let's be prepared for a page of acl data. */
3102                 localpage = alloc_page(GFP_KERNEL);
3103                 resp_buf = page_address(localpage);
3104                 if (localpage == NULL)
3105                         return -ENOMEM;
3106                 args.acl_pages[0] = localpage;
3107                 args.acl_pgbase = 0;
3108                 args.acl_len = PAGE_SIZE;
3109         } else {
3110                 resp_buf = buf;
3111                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3112         }
3113         ret = nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
3114         if (ret)
3115                 goto out_free;
3116         if (res.acl_len > args.acl_len)
3117                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3118         else
3119                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3120         if (buf) {
3121                 ret = -ERANGE;
3122                 if (res.acl_len > buflen)
3123                         goto out_free;
3124                 if (localpage)
3125                         memcpy(buf, resp_buf, res.acl_len);
3126         }
3127         ret = res.acl_len;
3128 out_free:
3129         if (localpage)
3130                 __free_page(localpage);
3131         return ret;
3132 }
3133
3134 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3135 {
3136         struct nfs4_exception exception = { };
3137         ssize_t ret;
3138         do {
3139                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3140                 if (ret >= 0)
3141                         break;
3142                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3143         } while (exception.retry);
3144         return ret;
3145 }
3146
3147 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3148 {
3149         struct nfs_server *server = NFS_SERVER(inode);
3150         int ret;
3151
3152         if (!nfs4_server_supports_acls(server))
3153                 return -EOPNOTSUPP;
3154         ret = nfs_revalidate_inode(server, inode);
3155         if (ret < 0)
3156                 return ret;
3157         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3158                 nfs_zap_acl_cache(inode);
3159         ret = nfs4_read_cached_acl(inode, buf, buflen);
3160         if (ret != -ENOENT)
3161                 return ret;
3162         return nfs4_get_acl_uncached(inode, buf, buflen);
3163 }
3164
3165 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3166 {
3167         struct nfs_server *server = NFS_SERVER(inode);
3168         struct page *pages[NFS4ACL_MAXPAGES];
3169         struct nfs_setaclargs arg = {
3170                 .fh             = NFS_FH(inode),
3171                 .acl_pages      = pages,
3172                 .acl_len        = buflen,
3173         };
3174         struct nfs_setaclres res;
3175         struct rpc_message msg = {
3176                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3177                 .rpc_argp       = &arg,
3178                 .rpc_resp       = &res,
3179         };
3180         int ret;
3181
3182         if (!nfs4_server_supports_acls(server))
3183                 return -EOPNOTSUPP;
3184         nfs_inode_return_delegation(inode);
3185         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3186         ret = nfs4_call_sync(server, &msg, &arg, &res, 1);
3187         nfs_access_zap_cache(inode);
3188         nfs_zap_acl_cache(inode);
3189         return ret;
3190 }
3191
3192 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3193 {
3194         struct nfs4_exception exception = { };
3195         int err;
3196         do {
3197                 err = nfs4_handle_exception(NFS_SERVER(inode),
3198                                 __nfs4_proc_set_acl(inode, buf, buflen),
3199                                 &exception);
3200         } while (exception.retry);
3201         return err;
3202 }
3203
3204 static int
3205 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3206 {
3207         struct nfs_client *clp = server->nfs_client;
3208
3209         if (!clp || task->tk_status >= 0)
3210                 return 0;
3211         switch(task->tk_status) {
3212                 case -NFS4ERR_ADMIN_REVOKED:
3213                 case -NFS4ERR_BAD_STATEID:
3214                 case -NFS4ERR_OPENMODE:
3215                         if (state == NULL)
3216                                 break;
3217                         nfs4_state_mark_reclaim_nograce(clp, state);
3218                 case -NFS4ERR_STALE_CLIENTID:
3219                 case -NFS4ERR_STALE_STATEID:
3220                 case -NFS4ERR_EXPIRED:
3221                         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3222                         nfs4_schedule_state_recovery(clp);
3223                         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3224                                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3225                         task->tk_status = 0;
3226                         return -EAGAIN;
3227                 case -NFS4ERR_DELAY:
3228                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3229                 case -NFS4ERR_GRACE:
3230                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3231                         task->tk_status = 0;
3232                         return -EAGAIN;
3233                 case -NFS4ERR_OLD_STATEID:
3234                         task->tk_status = 0;
3235                         return -EAGAIN;
3236         }
3237         task->tk_status = nfs4_map_errors(task->tk_status);
3238         return 0;
3239 }
3240
3241 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program, unsigned short port, struct rpc_cred *cred)
3242 {
3243         nfs4_verifier sc_verifier;
3244         struct nfs4_setclientid setclientid = {
3245                 .sc_verifier = &sc_verifier,
3246                 .sc_prog = program,
3247         };
3248         struct rpc_message msg = {
3249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3250                 .rpc_argp = &setclientid,
3251                 .rpc_resp = clp,
3252                 .rpc_cred = cred,
3253         };
3254         __be32 *p;
3255         int loop = 0;
3256         int status;
3257
3258         p = (__be32*)sc_verifier.data;
3259         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3260         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3261
3262         for(;;) {
3263                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3264                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3265                                 clp->cl_ipaddr,
3266                                 rpc_peeraddr2str(clp->cl_rpcclient,
3267                                                         RPC_DISPLAY_ADDR),
3268                                 rpc_peeraddr2str(clp->cl_rpcclient,
3269                                                         RPC_DISPLAY_PROTO),
3270                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3271                                 clp->cl_id_uniquifier);
3272                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3273                                 sizeof(setclientid.sc_netid),
3274                                 rpc_peeraddr2str(clp->cl_rpcclient,
3275                                                         RPC_DISPLAY_NETID));
3276                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3277                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3278                                 clp->cl_ipaddr, port >> 8, port & 255);
3279
3280                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3281                 if (status != -NFS4ERR_CLID_INUSE)
3282                         break;
3283                 if (signalled())
3284                         break;
3285                 if (loop++ & 1)
3286                         ssleep(clp->cl_lease_time + 1);
3287                 else
3288                         if (++clp->cl_id_uniquifier == 0)
3289                                 break;
3290         }
3291         return status;
3292 }
3293
3294 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3295 {
3296         struct nfs_fsinfo fsinfo;
3297         struct rpc_message msg = {
3298                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3299                 .rpc_argp = clp,
3300                 .rpc_resp = &fsinfo,
3301                 .rpc_cred = cred,
3302         };
3303         unsigned long now;
3304         int status;
3305
3306         now = jiffies;
3307         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3308         if (status == 0) {
3309                 spin_lock(&clp->cl_lock);
3310                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3311                 clp->cl_last_renewal = now;
3312                 spin_unlock(&clp->cl_lock);
3313         }
3314         return status;
3315 }
3316
3317 int nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3318 {
3319         long timeout = 0;
3320         int err;
3321         do {
3322                 err = _nfs4_proc_setclientid_confirm(clp, cred);
3323                 switch (err) {
3324                         case 0:
3325                                 return err;
3326                         case -NFS4ERR_RESOURCE:
3327                                 /* The IBM lawyers misread another document! */
3328                         case -NFS4ERR_DELAY:
3329                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3330                 }
3331         } while (err == 0);
3332         return err;
3333 }
3334
3335 struct nfs4_delegreturndata {
3336         struct nfs4_delegreturnargs args;
3337         struct nfs4_delegreturnres res;
3338         struct nfs_fh fh;
3339         nfs4_stateid stateid;
3340         unsigned long timestamp;
3341         struct nfs_fattr fattr;
3342         int rpc_status;
3343 };
3344
3345 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3346 {
3347         struct nfs4_delegreturndata *data = calldata;
3348         data->rpc_status = task->tk_status;
3349         if (data->rpc_status == 0)
3350                 renew_lease(data->res.server, data->timestamp);
3351 }
3352
3353 static void nfs4_delegreturn_release(void *calldata)
3354 {
3355         kfree(calldata);
3356 }
3357
3358 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3359         .rpc_call_done = nfs4_delegreturn_done,
3360         .rpc_release = nfs4_delegreturn_release,
3361 };
3362
3363 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3364 {
3365         struct nfs4_delegreturndata *data;
3366         struct nfs_server *server = NFS_SERVER(inode);
3367         struct rpc_task *task;
3368         struct rpc_message msg = {
3369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3370                 .rpc_cred = cred,
3371         };
3372         struct rpc_task_setup task_setup_data = {
3373                 .rpc_client = server->client,
3374                 .rpc_message = &msg,
3375                 .callback_ops = &nfs4_delegreturn_ops,
3376                 .flags = RPC_TASK_ASYNC,
3377         };
3378         int status = 0;
3379
3380         data = kmalloc(sizeof(*data), GFP_KERNEL);
3381         if (data == NULL)
3382                 return -ENOMEM;
3383         data->args.fhandle = &data->fh;
3384         data->args.stateid = &data->stateid;
3385         data->args.bitmask = server->attr_bitmask;
3386         nfs_copy_fh(&data->fh, NFS_FH(inode));
3387         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3388         data->res.fattr = &data->fattr;
3389         data->res.server = server;
3390         data->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3391         nfs_fattr_init(data->res.fattr);
3392         data->timestamp = jiffies;
3393         data->rpc_status = 0;
3394
3395         task_setup_data.callback_data = data;
3396         msg.rpc_argp = &data->args,
3397         msg.rpc_resp = &data->res,
3398         task = rpc_run_task(&task_setup_data);
3399         if (IS_ERR(task))
3400                 return PTR_ERR(task);
3401         if (!issync)
3402                 goto out;
3403         status = nfs4_wait_for_completion_rpc_task(task);
3404         if (status != 0)
3405                 goto out;
3406         status = data->rpc_status;
3407         if (status != 0)
3408                 goto out;
3409         nfs_refresh_inode(inode, &data->fattr);
3410 out:
3411         rpc_put_task(task);
3412         return status;
3413 }
3414
3415 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3416 {
3417         struct nfs_server *server = NFS_SERVER(inode);
3418         struct nfs4_exception exception = { };
3419         int err;
3420         do {
3421                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3422                 switch (err) {
3423                         case -NFS4ERR_STALE_STATEID:
3424                         case -NFS4ERR_EXPIRED:
3425                         case 0:
3426                                 return 0;
3427                 }
3428                 err = nfs4_handle_exception(server, err, &exception);
3429         } while (exception.retry);
3430         return err;
3431 }
3432
3433 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3434 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3435
3436 /* 
3437  * sleep, with exponential backoff, and retry the LOCK operation. 
3438  */
3439 static unsigned long
3440 nfs4_set_lock_task_retry(unsigned long timeout)
3441 {
3442         schedule_timeout_killable(timeout);
3443         timeout <<= 1;
3444         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3445                 return NFS4_LOCK_MAXTIMEOUT;
3446         return timeout;
3447 }
3448
3449 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3450 {
3451         struct inode *inode = state->inode;
3452         struct nfs_server *server = NFS_SERVER(inode);
3453         struct nfs_client *clp = server->nfs_client;
3454         struct nfs_lockt_args arg = {
3455                 .fh = NFS_FH(inode),
3456                 .fl = request,
3457         };
3458         struct nfs_lockt_res res = {
3459                 .denied = request,
3460         };
3461         struct rpc_message msg = {
3462                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3463                 .rpc_argp       = &arg,
3464                 .rpc_resp       = &res,
3465                 .rpc_cred       = state->owner->so_cred,
3466         };
3467         struct nfs4_lock_state *lsp;
3468         int status;
3469
3470         arg.lock_owner.clientid = clp->cl_clientid;
3471         status = nfs4_set_lock_state(state, request);
3472         if (status != 0)
3473                 goto out;
3474         lsp = request->fl_u.nfs4_fl.owner;
3475         arg.lock_owner.id = lsp->ls_id.id;
3476         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
3477         switch (status) {
3478                 case 0:
3479                         request->fl_type = F_UNLCK;
3480                         break;
3481                 case -NFS4ERR_DENIED:
3482                         status = 0;
3483         }
3484         request->fl_ops->fl_release_private(request);
3485 out:
3486         return status;
3487 }
3488
3489 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3490 {
3491         struct nfs4_exception exception = { };
3492         int err;
3493
3494         do {
3495                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3496                                 _nfs4_proc_getlk(state, cmd, request),
3497                                 &exception);
3498         } while (exception.retry);
3499         return err;
3500 }
3501
3502 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3503 {
3504         int res = 0;
3505         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3506                 case FL_POSIX:
3507                         res = posix_lock_file_wait(file, fl);
3508                         break;
3509                 case FL_FLOCK:
3510                         res = flock_lock_file_wait(file, fl);
3511                         break;
3512                 default:
3513                         BUG();
3514         }
3515         return res;
3516 }
3517
3518 struct nfs4_unlockdata {
3519         struct nfs_locku_args arg;
3520         struct nfs_locku_res res;
3521         struct nfs4_lock_state *lsp;
3522         struct nfs_open_context *ctx;
3523         struct file_lock fl;
3524         const struct nfs_server *server;
3525         unsigned long timestamp;
3526 };
3527
3528 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3529                 struct nfs_open_context *ctx,
3530                 struct nfs4_lock_state *lsp,
3531                 struct nfs_seqid *seqid)
3532 {
3533         struct nfs4_unlockdata *p;
3534         struct inode *inode = lsp->ls_state->inode;
3535
3536         p = kmalloc(sizeof(*p), GFP_KERNEL);
3537         if (p == NULL)
3538                 return NULL;
3539         p->arg.fh = NFS_FH(inode);
3540         p->arg.fl = &p->fl;
3541         p->arg.seqid = seqid;
3542         p->res.seqid = seqid;
3543         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3544         p->arg.stateid = &lsp->ls_stateid;
3545         p->lsp = lsp;
3546         atomic_inc(&lsp->ls_count);
3547         /* Ensure we don't close file until we're done freeing locks! */
3548         p->ctx = get_nfs_open_context(ctx);
3549         memcpy(&p->fl, fl, sizeof(p->fl));
3550         p->server = NFS_SERVER(inode);
3551         return p;
3552 }
3553
3554 static void nfs4_locku_release_calldata(void *data)
3555 {
3556         struct nfs4_unlockdata *calldata = data;
3557         nfs_free_seqid(calldata->arg.seqid);
3558         nfs4_put_lock_state(calldata->lsp);
3559         put_nfs_open_context(calldata->ctx);
3560         kfree(calldata);
3561 }
3562
3563 static void nfs4_locku_done(struct rpc_task *task, void *data)
3564 {
3565         struct nfs4_unlockdata *calldata = data;
3566
3567         if (RPC_ASSASSINATED(task))
3568                 return;
3569         switch (task->tk_status) {
3570                 case 0:
3571                         memcpy(calldata->lsp->ls_stateid.data,
3572                                         calldata->res.stateid.data,
3573                                         sizeof(calldata->lsp->ls_stateid.data));
3574                         renew_lease(calldata->server, calldata->timestamp);
3575                         break;
3576                 case -NFS4ERR_BAD_STATEID:
3577                 case -NFS4ERR_OLD_STATEID:
3578                 case -NFS4ERR_STALE_STATEID:
3579                 case -NFS4ERR_EXPIRED:
3580                         break;
3581                 default:
3582                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
3583                                 rpc_restart_call(task);
3584         }
3585 }
3586
3587 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
3588 {
3589         struct nfs4_unlockdata *calldata = data;
3590
3591         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3592                 return;
3593         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
3594                 /* Note: exit _without_ running nfs4_locku_done */
3595                 task->tk_action = NULL;
3596                 return;
3597         }
3598         calldata->timestamp = jiffies;
3599         rpc_call_start(task);
3600 }
3601
3602 static const struct rpc_call_ops nfs4_locku_ops = {
3603         .rpc_call_prepare = nfs4_locku_prepare,
3604         .rpc_call_done = nfs4_locku_done,
3605         .rpc_release = nfs4_locku_release_calldata,
3606 };
3607
3608 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
3609                 struct nfs_open_context *ctx,
3610                 struct nfs4_lock_state *lsp,
3611                 struct nfs_seqid *seqid)
3612 {
3613         struct nfs4_unlockdata *data;
3614         struct rpc_message msg = {
3615                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
3616                 .rpc_cred = ctx->cred,
3617         };
3618         struct rpc_task_setup task_setup_data = {
3619                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
3620                 .rpc_message = &msg,
3621                 .callback_ops = &nfs4_locku_ops,
3622                 .workqueue = nfsiod_workqueue,
3623                 .flags = RPC_TASK_ASYNC,
3624         };
3625
3626         /* Ensure this is an unlock - when canceling a lock, the
3627          * canceled lock is passed in, and it won't be an unlock.
3628          */
3629         fl->fl_type = F_UNLCK;
3630
3631         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
3632         if (data == NULL) {
3633                 nfs_free_seqid(seqid);
3634                 return ERR_PTR(-ENOMEM);
3635         }
3636
3637         msg.rpc_argp = &data->arg,
3638         msg.rpc_resp = &data->res,
3639         task_setup_data.callback_data = data;
3640         return rpc_run_task(&task_setup_data);
3641 }
3642
3643 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
3644 {
3645         struct nfs_inode *nfsi = NFS_I(state->inode);
3646         struct nfs_seqid *seqid;
3647         struct nfs4_lock_state *lsp;
3648         struct rpc_task *task;
3649         int status = 0;
3650         unsigned char fl_flags = request->fl_flags;
3651
3652         status = nfs4_set_lock_state(state, request);
3653         /* Unlock _before_ we do the RPC call */
3654         request->fl_flags |= FL_EXISTS;
3655         down_read(&nfsi->rwsem);
3656         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
3657                 up_read(&nfsi->rwsem);
3658                 goto out;
3659         }
3660         up_read(&nfsi->rwsem);
3661         if (status != 0)
3662                 goto out;
3663         /* Is this a delegated lock? */
3664         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
3665                 goto out;
3666         lsp = request->fl_u.nfs4_fl.owner;
3667         seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3668         status = -ENOMEM;
3669         if (seqid == NULL)
3670                 goto out;
3671         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
3672         status = PTR_ERR(task);
3673         if (IS_ERR(task))
3674                 goto out;
3675         status = nfs4_wait_for_completion_rpc_task(task);
3676         rpc_put_task(task);
3677 out:
3678         request->fl_flags = fl_flags;
3679         return status;
3680 }
3681
3682 struct nfs4_lockdata {
3683         struct nfs_lock_args arg;
3684         struct nfs_lock_res res;
3685         struct nfs4_lock_state *lsp;
3686         struct nfs_open_context *ctx;
3687         struct file_lock fl;
3688         unsigned long timestamp;
3689         int rpc_status;
3690         int cancelled;
3691         struct nfs_server *server;
3692 };
3693
3694 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
3695                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp)
3696 {
3697         struct nfs4_lockdata *p;
3698         struct inode *inode = lsp->ls_state->inode;
3699         struct nfs_server *server = NFS_SERVER(inode);
3700
3701         p = kzalloc(sizeof(*p), GFP_KERNEL);
3702         if (p == NULL)
3703                 return NULL;
3704
3705         p->arg.fh = NFS_FH(inode);
3706         p->arg.fl = &p->fl;
3707         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid);
3708         if (p->arg.open_seqid == NULL)
3709                 goto out_free;
3710         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3711         if (p->arg.lock_seqid == NULL)
3712                 goto out_free_seqid;
3713         p->arg.lock_stateid = &lsp->ls_stateid;
3714         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
3715         p->arg.lock_owner.id = lsp->ls_id.id;
3716         p->res.lock_seqid = p->arg.lock_seqid;
3717         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3718         p->lsp = lsp;
3719         p->server = server;
3720         atomic_inc(&lsp->ls_count);
3721         p->ctx = get_nfs_open_context(ctx);
3722         memcpy(&p->fl, fl, sizeof(p->fl));
3723         return p;
3724 out_free_seqid:
3725         nfs_free_seqid(p->arg.open_seqid);
3726 out_free:
3727         kfree(p);
3728         return NULL;
3729 }
3730
3731 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
3732 {
3733         struct nfs4_lockdata *data = calldata;
3734         struct nfs4_state *state = data->lsp->ls_state;
3735
3736         dprintk("%s: begin!\n", __func__);
3737         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
3738                 return;
3739         /* Do we need to do an open_to_lock_owner? */
3740         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
3741                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
3742                         return;
3743                 data->arg.open_stateid = &state->stateid;
3744                 data->arg.new_lock_owner = 1;
3745                 data->res.open_seqid = data->arg.open_seqid;
3746         } else
3747                 data->arg.new_lock_owner = 0;
3748         data->timestamp = jiffies;
3749         if (nfs4_setup_sequence(data->server->nfs_client, &data->arg.seq_args,
3750                                 &data->res.seq_res, 1, task))
3751                 return;
3752         rpc_call_start(task);
3753         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
3754 }
3755
3756 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
3757 {
3758         struct nfs4_lockdata *data = calldata;
3759
3760         dprintk("%s: begin!\n", __func__);
3761
3762         nfs4_sequence_done_free_slot(data->server, &data->res.seq_res,
3763                                      task->tk_status);
3764
3765         data->rpc_status = task->tk_status;
3766         if (RPC_ASSASSINATED(task))
3767                 goto out;
3768         if (data->arg.new_lock_owner != 0) {
3769                 if (data->rpc_status == 0)
3770                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
3771                 else
3772                         goto out;
3773         }
3774         if (data->rpc_status == 0) {
3775                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
3776                                         sizeof(data->lsp->ls_stateid.data));
3777                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
3778                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
3779         }
3780 out:
3781         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
3782 }
3783
3784 static void nfs4_lock_release(void *calldata)
3785 {
3786         struct nfs4_lockdata *data = calldata;
3787
3788         dprintk("%s: begin!\n", __func__);
3789         nfs_free_seqid(data->arg.open_seqid);
3790         if (data->cancelled != 0) {
3791                 struct rpc_task *task;
3792                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
3793                                 data->arg.lock_seqid);
3794                 if (!IS_ERR(task))
3795                         rpc_put_task(task);
3796                 dprintk("%s: cancelling lock!\n", __func__);
3797         } else
3798                 nfs_free_seqid(data->arg.lock_seqid);
3799         nfs4_put_lock_state(data->lsp);
3800         put_nfs_open_context(data->ctx);
3801         kfree(data);
3802         dprintk("%s: done!\n", __func__);
3803 }
3804
3805 static const struct rpc_call_ops nfs4_lock_ops = {
3806         .rpc_call_prepare = nfs4_lock_prepare,
3807         .rpc_call_done = nfs4_lock_done,
3808         .rpc_release = nfs4_lock_release,
3809 };
3810
3811 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int reclaim)
3812 {
3813         struct nfs4_lockdata *data;
3814         struct rpc_task *task;
3815         struct rpc_message msg = {
3816                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
3817                 .rpc_cred = state->owner->so_cred,
3818         };
3819         struct rpc_task_setup task_setup_data = {
3820                 .rpc_client = NFS_CLIENT(state->inode),
3821                 .rpc_message = &msg,
3822                 .callback_ops = &nfs4_lock_ops,
3823                 .workqueue = nfsiod_workqueue,
3824                 .flags = RPC_TASK_ASYNC,
3825         };
3826         int ret;
3827
3828         dprintk("%s: begin!\n", __func__);
3829         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
3830                         fl->fl_u.nfs4_fl.owner);
3831         if (data == NULL)
3832                 return -ENOMEM;
3833         if (IS_SETLKW(cmd))
3834                 data->arg.block = 1;
3835         if (reclaim != 0)
3836                 data->arg.reclaim = 1;
3837         msg.rpc_argp = &data->arg,
3838         msg.rpc_resp = &data->res,
3839         task_setup_data.callback_data = data;
3840         task = rpc_run_task(&task_setup_data);
3841         if (IS_ERR(task))
3842                 return PTR_ERR(task);
3843         ret = nfs4_wait_for_completion_rpc_task(task);
3844         if (ret == 0) {
3845                 ret = data->rpc_status;
3846                 if (ret == -NFS4ERR_DENIED)
3847                         ret = -EAGAIN;
3848         } else
3849                 data->cancelled = 1;
3850         rpc_put_task(task);
3851         dprintk("%s: done, ret = %d!\n", __func__, ret);
3852         return ret;
3853 }
3854
3855 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
3856 {
3857         struct nfs_server *server = NFS_SERVER(state->inode);
3858         struct nfs4_exception exception = { };
3859         int err;
3860
3861         do {
3862                 /* Cache the lock if possible... */
3863                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
3864                         return 0;
3865                 err = _nfs4_do_setlk(state, F_SETLK, request, 1);
3866                 if (err != -NFS4ERR_DELAY)
3867                         break;
3868                 nfs4_handle_exception(server, err, &exception);
3869         } while (exception.retry);
3870         return err;
3871 }
3872
3873 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
3874 {
3875         struct nfs_server *server = NFS_SERVER(state->inode);
3876         struct nfs4_exception exception = { };
3877         int err;
3878
3879         err = nfs4_set_lock_state(state, request);
3880         if (err != 0)
3881                 return err;
3882         do {
3883                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
3884                         return 0;
3885                 err = _nfs4_do_setlk(state, F_SETLK, request, 0);
3886                 if (err != -NFS4ERR_DELAY)
3887                         break;
3888                 nfs4_handle_exception(server, err, &exception);
3889         } while (exception.retry);
3890         return err;
3891 }
3892
3893 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3894 {
3895         struct nfs_inode *nfsi = NFS_I(state->inode);
3896         unsigned char fl_flags = request->fl_flags;
3897         int status;
3898
3899         /* Is this a delegated open? */
3900         status = nfs4_set_lock_state(state, request);
3901         if (status != 0)
3902                 goto out;
3903         request->fl_flags |= FL_ACCESS;
3904         status = do_vfs_lock(request->fl_file, request);
3905         if (status < 0)
3906                 goto out;
3907         down_read(&nfsi->rwsem);
3908         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
3909                 /* Yes: cache locks! */
3910                 /* ...but avoid races with delegation recall... */
3911                 request->fl_flags = fl_flags & ~FL_SLEEP;
3912                 status = do_vfs_lock(request->fl_file, request);
3913                 goto out_unlock;
3914         }
3915         status = _nfs4_do_setlk(state, cmd, request, 0);
3916         if (status != 0)
3917                 goto out_unlock;
3918         /* Note: we always want to sleep here! */
3919         request->fl_flags = fl_flags | FL_SLEEP;
3920         if (do_vfs_lock(request->fl_file, request) < 0)
3921                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
3922 out_unlock:
3923         up_read(&nfsi->rwsem);
3924 out:
3925         request->fl_flags = fl_flags;
3926         return status;
3927 }
3928
3929 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3930 {
3931         struct nfs4_exception exception = { };
3932         int err;
3933
3934         do {
3935                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3936                                 _nfs4_proc_setlk(state, cmd, request),
3937                                 &exception);
3938         } while (exception.retry);
3939         return err;
3940 }
3941
3942 static int
3943 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
3944 {
3945         struct nfs_open_context *ctx;
3946         struct nfs4_state *state;
3947         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
3948         int status;
3949
3950         /* verify open state */
3951         ctx = nfs_file_open_context(filp);
3952         state = ctx->state;
3953
3954         if (request->fl_start < 0 || request->fl_end < 0)
3955                 return -EINVAL;
3956
3957         if (IS_GETLK(cmd))
3958                 return nfs4_proc_getlk(state, F_GETLK, request);
3959
3960         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
3961                 return -EINVAL;
3962
3963         if (request->fl_type == F_UNLCK)
3964                 return nfs4_proc_unlck(state, cmd, request);
3965
3966         do {
3967                 status = nfs4_proc_setlk(state, cmd, request);
3968                 if ((status != -EAGAIN) || IS_SETLK(cmd))
3969                         break;
3970                 timeout = nfs4_set_lock_task_retry(timeout);
3971                 status = -ERESTARTSYS;
3972                 if (signalled())
3973                         break;
3974         } while(status < 0);
3975         return status;
3976 }
3977
3978 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
3979 {
3980         struct nfs_server *server = NFS_SERVER(state->inode);
3981         struct nfs4_exception exception = { };
3982         int err;
3983
3984         err = nfs4_set_lock_state(state, fl);
3985         if (err != 0)
3986                 goto out;
3987         do {
3988                 err = _nfs4_do_setlk(state, F_SETLK, fl, 0);
3989                 if (err != -NFS4ERR_DELAY)
3990                         break;
3991                 err = nfs4_handle_exception(server, err, &exception);
3992         } while (exception.retry);
3993 out:
3994         return err;
3995 }
3996
3997 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
3998
3999 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
4000                 size_t buflen, int flags)
4001 {
4002         struct inode *inode = dentry->d_inode;
4003
4004         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4005                 return -EOPNOTSUPP;
4006
4007         return nfs4_proc_set_acl(inode, buf, buflen);
4008 }
4009
4010 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
4011  * and that's what we'll do for e.g. user attributes that haven't been set.
4012  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
4013  * attributes in kernel-managed attribute namespaces. */
4014 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
4015                 size_t buflen)
4016 {
4017         struct inode *inode = dentry->d_inode;
4018
4019         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4020                 return -EOPNOTSUPP;
4021
4022         return nfs4_proc_get_acl(inode, buf, buflen);
4023 }
4024
4025 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
4026 {
4027         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
4028
4029         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4030                 return 0;
4031         if (buf && buflen < len)
4032                 return -ERANGE;
4033         if (buf)
4034                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
4035         return len;
4036 }
4037
4038 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4039 {
4040         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4041                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4042                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4043                 return;
4044
4045         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4046                 NFS_ATTR_FATTR_NLINK;
4047         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4048         fattr->nlink = 2;
4049 }
4050
4051 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4052                 struct nfs4_fs_locations *fs_locations, struct page *page)
4053 {
4054         struct nfs_server *server = NFS_SERVER(dir);
4055         u32 bitmask[2] = {
4056                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4057                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4058         };
4059         struct nfs4_fs_locations_arg args = {
4060                 .dir_fh = NFS_FH(dir),
4061                 .name = name,
4062                 .page = page,
4063                 .bitmask = bitmask,
4064         };
4065         struct nfs4_fs_locations_res res = {
4066                 .fs_locations = fs_locations,
4067         };
4068         struct rpc_message msg = {
4069                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4070                 .rpc_argp = &args,
4071                 .rpc_resp = &res,
4072         };
4073         int status;
4074
4075         dprintk("%s: start\n", __func__);
4076         nfs_fattr_init(&fs_locations->fattr);
4077         fs_locations->server = server;
4078         fs_locations->nlocations = 0;
4079         status = nfs4_call_sync(server, &msg, &args, &res, 0);
4080         nfs_fixup_referral_attributes(&fs_locations->fattr);
4081         dprintk("%s: returned status = %d\n", __func__, status);
4082         return status;
4083 }
4084
4085 #ifdef CONFIG_NFS_V4_1
4086 /* Destroy the slot table */
4087 static void nfs4_destroy_slot_table(struct nfs4_session *session)
4088 {
4089         if (session->fc_slot_table.slots == NULL)
4090                 return;
4091         kfree(session->fc_slot_table.slots);
4092         session->fc_slot_table.slots = NULL;
4093         return;
4094 }
4095
4096 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4097 {
4098         struct nfs4_session *session;
4099         struct nfs4_slot_table *tbl;
4100
4101         session = kzalloc(sizeof(struct nfs4_session), GFP_KERNEL);
4102         if (!session)
4103                 return NULL;
4104         tbl = &session->fc_slot_table;
4105         spin_lock_init(&tbl->slot_tbl_lock);
4106         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "Slot table");
4107         session->clp = clp;
4108         return session;
4109 }
4110
4111 void nfs4_destroy_session(struct nfs4_session *session)
4112 {
4113         nfs4_destroy_slot_table(session);
4114         kfree(session);
4115 }
4116
4117 #endif /* CONFIG_NFS_V4_1 */
4118
4119 struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops = {
4120         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
4121         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
4122         .recover_open   = nfs4_open_reclaim,
4123         .recover_lock   = nfs4_lock_reclaim,
4124 };
4125
4126 struct nfs4_state_recovery_ops nfs4_nograce_recovery_ops = {
4127         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
4128         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
4129         .recover_open   = nfs4_open_expired,
4130         .recover_lock   = nfs4_lock_expired,
4131 };
4132
4133 static const struct inode_operations nfs4_file_inode_operations = {
4134         .permission     = nfs_permission,
4135         .getattr        = nfs_getattr,
4136         .setattr        = nfs_setattr,
4137         .getxattr       = nfs4_getxattr,
4138         .setxattr       = nfs4_setxattr,
4139         .listxattr      = nfs4_listxattr,
4140 };
4141
4142 const struct nfs_rpc_ops nfs_v4_clientops = {
4143         .version        = 4,                    /* protocol version */
4144         .dentry_ops     = &nfs4_dentry_operations,
4145         .dir_inode_ops  = &nfs4_dir_inode_operations,
4146         .file_inode_ops = &nfs4_file_inode_operations,
4147         .getroot        = nfs4_proc_get_root,
4148         .getattr        = nfs4_proc_getattr,
4149         .setattr        = nfs4_proc_setattr,
4150         .lookupfh       = nfs4_proc_lookupfh,
4151         .lookup         = nfs4_proc_lookup,
4152         .access         = nfs4_proc_access,
4153         .readlink       = nfs4_proc_readlink,
4154         .create         = nfs4_proc_create,
4155         .remove         = nfs4_proc_remove,
4156         .unlink_setup   = nfs4_proc_unlink_setup,
4157         .unlink_done    = nfs4_proc_unlink_done,
4158         .rename         = nfs4_proc_rename,
4159         .link           = nfs4_proc_link,
4160         .symlink        = nfs4_proc_symlink,
4161         .mkdir          = nfs4_proc_mkdir,
4162         .rmdir          = nfs4_proc_remove,
4163         .readdir        = nfs4_proc_readdir,
4164         .mknod          = nfs4_proc_mknod,
4165         .statfs         = nfs4_proc_statfs,
4166         .fsinfo         = nfs4_proc_fsinfo,
4167         .pathconf       = nfs4_proc_pathconf,
4168         .set_capabilities = nfs4_server_capabilities,
4169         .decode_dirent  = nfs4_decode_dirent,
4170         .read_setup     = nfs4_proc_read_setup,
4171         .read_done      = nfs4_read_done,
4172         .write_setup    = nfs4_proc_write_setup,
4173         .write_done     = nfs4_write_done,
4174         .commit_setup   = nfs4_proc_commit_setup,
4175         .commit_done    = nfs4_commit_done,
4176         .lock           = nfs4_proc_lock,
4177         .clear_acl_cache = nfs4_zap_acl_attr,
4178         .close_context  = nfs4_close_context,
4179 };
4180
4181 /*
4182  * Local variables:
4183  *  c-basic-offset: 8
4184  * End:
4185  */