[PATCH] NFSv4: Ensure that propagate NFSv4 state errors to the reclaim code
[linux-2.6-block.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/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
51 #include "nfs4_fs.h"
52 #include "delegation.h"
53
54 #define NFSDBG_FACILITY         NFSDBG_PROC
55
56 #define NFS4_POLL_RETRY_MIN     (1*HZ)
57 #define NFS4_POLL_RETRY_MAX     (15*HZ)
58
59 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
60 static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *);
61 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
62 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
63 extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus);
64 extern struct rpc_procinfo nfs4_procedures[];
65
66 /* Prevent leaks of NFSv4 errors into userland */
67 int nfs4_map_errors(int err)
68 {
69         if (err < -1000) {
70                 dprintk("%s could not handle NFSv4 error %d\n",
71                                 __FUNCTION__, -err);
72                 return -EIO;
73         }
74         return err;
75 }
76
77 /*
78  * This is our standard bitmap for GETATTR requests.
79  */
80 const u32 nfs4_fattr_bitmap[2] = {
81         FATTR4_WORD0_TYPE
82         | FATTR4_WORD0_CHANGE
83         | FATTR4_WORD0_SIZE
84         | FATTR4_WORD0_FSID
85         | FATTR4_WORD0_FILEID,
86         FATTR4_WORD1_MODE
87         | FATTR4_WORD1_NUMLINKS
88         | FATTR4_WORD1_OWNER
89         | FATTR4_WORD1_OWNER_GROUP
90         | FATTR4_WORD1_RAWDEV
91         | FATTR4_WORD1_SPACE_USED
92         | FATTR4_WORD1_TIME_ACCESS
93         | FATTR4_WORD1_TIME_METADATA
94         | FATTR4_WORD1_TIME_MODIFY
95 };
96
97 const u32 nfs4_statfs_bitmap[2] = {
98         FATTR4_WORD0_FILES_AVAIL
99         | FATTR4_WORD0_FILES_FREE
100         | FATTR4_WORD0_FILES_TOTAL,
101         FATTR4_WORD1_SPACE_AVAIL
102         | FATTR4_WORD1_SPACE_FREE
103         | FATTR4_WORD1_SPACE_TOTAL
104 };
105
106 const u32 nfs4_pathconf_bitmap[2] = {
107         FATTR4_WORD0_MAXLINK
108         | FATTR4_WORD0_MAXNAME,
109         0
110 };
111
112 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
113                         | FATTR4_WORD0_MAXREAD
114                         | FATTR4_WORD0_MAXWRITE
115                         | FATTR4_WORD0_LEASE_TIME,
116                         0
117 };
118
119 static void nfs4_setup_readdir(u64 cookie, u32 *verifier, struct dentry *dentry,
120                 struct nfs4_readdir_arg *readdir)
121 {
122         u32 *start, *p;
123
124         BUG_ON(readdir->count < 80);
125         if (cookie > 2) {
126                 readdir->cookie = cookie;
127                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
128                 return;
129         }
130
131         readdir->cookie = 0;
132         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
133         if (cookie == 2)
134                 return;
135         
136         /*
137          * NFSv4 servers do not return entries for '.' and '..'
138          * Therefore, we fake these entries here.  We let '.'
139          * have cookie 0 and '..' have cookie 1.  Note that
140          * when talking to the server, we always send cookie 0
141          * instead of 1 or 2.
142          */
143         start = p = (u32 *)kmap_atomic(*readdir->pages, KM_USER0);
144         
145         if (cookie == 0) {
146                 *p++ = xdr_one;                                  /* next */
147                 *p++ = xdr_zero;                   /* cookie, first word */
148                 *p++ = xdr_one;                   /* cookie, second word */
149                 *p++ = xdr_one;                             /* entry len */
150                 memcpy(p, ".\0\0\0", 4);                        /* entry */
151                 p++;
152                 *p++ = xdr_one;                         /* bitmap length */
153                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
154                 *p++ = htonl(8);              /* attribute buffer length */
155                 p = xdr_encode_hyper(p, dentry->d_inode->i_ino);
156         }
157         
158         *p++ = xdr_one;                                  /* next */
159         *p++ = xdr_zero;                   /* cookie, first word */
160         *p++ = xdr_two;                   /* cookie, second word */
161         *p++ = xdr_two;                             /* entry len */
162         memcpy(p, "..\0\0", 4);                         /* entry */
163         p++;
164         *p++ = xdr_one;                         /* bitmap length */
165         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
166         *p++ = htonl(8);              /* attribute buffer length */
167         p = xdr_encode_hyper(p, dentry->d_parent->d_inode->i_ino);
168
169         readdir->pgbase = (char *)p - (char *)start;
170         readdir->count -= readdir->pgbase;
171         kunmap_atomic(start, KM_USER0);
172 }
173
174 static void
175 renew_lease(struct nfs_server *server, unsigned long timestamp)
176 {
177         struct nfs4_client *clp = server->nfs4_state;
178         spin_lock(&clp->cl_lock);
179         if (time_before(clp->cl_last_renewal,timestamp))
180                 clp->cl_last_renewal = timestamp;
181         spin_unlock(&clp->cl_lock);
182 }
183
184 static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinfo)
185 {
186         struct nfs_inode *nfsi = NFS_I(inode);
187
188         if (cinfo->before == nfsi->change_attr && cinfo->atomic)
189                 nfsi->change_attr = cinfo->after;
190 }
191
192 static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
193 {
194         struct inode *inode = state->inode;
195
196         open_flags &= (FMODE_READ|FMODE_WRITE);
197         /* Protect against nfs4_find_state() */
198         spin_lock(&inode->i_lock);
199         state->state |= open_flags;
200         /* NB! List reordering - see the reclaim code for why.  */
201         if ((open_flags & FMODE_WRITE) && 0 == state->nwriters++)
202                 list_move(&state->open_states, &state->owner->so_states);
203         if (open_flags & FMODE_READ)
204                 state->nreaders++;
205         memcpy(&state->stateid, stateid, sizeof(state->stateid));
206         spin_unlock(&inode->i_lock);
207 }
208
209 /*
210  * OPEN_RECLAIM:
211  *      reclaim state on the server after a reboot.
212  *      Assumes caller is holding the sp->so_sem
213  */
214 static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
215 {
216         struct inode *inode = state->inode;
217         struct nfs_server *server = NFS_SERVER(inode);
218         struct nfs_delegation *delegation = NFS_I(inode)->delegation;
219         struct nfs_openargs o_arg = {
220                 .fh = NFS_FH(inode),
221                 .seqid = sp->so_seqid,
222                 .id = sp->so_id,
223                 .open_flags = state->state,
224                 .clientid = server->nfs4_state->cl_clientid,
225                 .claim = NFS4_OPEN_CLAIM_PREVIOUS,
226                 .bitmask = server->attr_bitmask,
227         };
228         struct nfs_openres o_res = {
229                 .server = server,       /* Grrr */
230         };
231         struct rpc_message msg = {
232                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
233                 .rpc_argp       = &o_arg,
234                 .rpc_resp       = &o_res,
235                 .rpc_cred       = sp->so_cred,
236         };
237         int status;
238
239         if (delegation != NULL) {
240                 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
241                         memcpy(&state->stateid, &delegation->stateid,
242                                         sizeof(state->stateid));
243                         set_bit(NFS_DELEGATED_STATE, &state->flags);
244                         return 0;
245                 }
246                 o_arg.u.delegation_type = delegation->type;
247         }
248         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
249         nfs4_increment_seqid(status, sp);
250         if (status == 0) {
251                 memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
252                 if (o_res.delegation_type != 0) {
253                         nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
254                         /* Did the server issue an immediate delegation recall? */
255                         if (o_res.do_recall)
256                                 nfs_async_inode_return_delegation(inode, &o_res.stateid);
257                 }
258         }
259         clear_bit(NFS_DELEGATED_STATE, &state->flags);
260         /* Ensure we update the inode attributes */
261         NFS_CACHEINV(inode);
262         return status;
263 }
264
265 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
266 {
267         struct nfs_server *server = NFS_SERVER(state->inode);
268         struct nfs4_exception exception = { };
269         int err;
270         do {
271                 err = _nfs4_open_reclaim(sp, state);
272                 if (err != -NFS4ERR_DELAY)
273                         break;
274                 nfs4_handle_exception(server, err, &exception);
275         } while (exception.retry);
276         return err;
277 }
278
279 static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
280 {
281         struct nfs4_state_owner  *sp  = state->owner;
282         struct inode *inode = dentry->d_inode;
283         struct nfs_server *server = NFS_SERVER(inode);
284         struct dentry *parent = dget_parent(dentry);
285         struct nfs_openargs arg = {
286                 .fh = NFS_FH(parent->d_inode),
287                 .clientid = server->nfs4_state->cl_clientid,
288                 .name = &dentry->d_name,
289                 .id = sp->so_id,
290                 .server = server,
291                 .bitmask = server->attr_bitmask,
292                 .claim = NFS4_OPEN_CLAIM_DELEGATE_CUR,
293         };
294         struct nfs_openres res = {
295                 .server = server,
296         };
297         struct  rpc_message msg = {
298                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
299                 .rpc_argp       = &arg,
300                 .rpc_resp       = &res,
301                 .rpc_cred       = sp->so_cred,
302         };
303         int status = 0;
304
305         down(&sp->so_sema);
306         if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
307                 goto out;
308         if (state->state == 0)
309                 goto out;
310         arg.seqid = sp->so_seqid;
311         arg.open_flags = state->state;
312         memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data));
313         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
314         nfs4_increment_seqid(status, sp);
315         if (status >= 0) {
316                 memcpy(state->stateid.data, res.stateid.data,
317                                 sizeof(state->stateid.data));
318                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
319         }
320 out:
321         up(&sp->so_sema);
322         dput(parent);
323         return status;
324 }
325
326 int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
327 {
328         struct nfs4_exception exception = { };
329         struct nfs_server *server = NFS_SERVER(dentry->d_inode);
330         int err;
331         do {
332                 err = _nfs4_open_delegation_recall(dentry, state);
333                 switch (err) {
334                         case 0:
335                                 return err;
336                         case -NFS4ERR_STALE_CLIENTID:
337                         case -NFS4ERR_STALE_STATEID:
338                         case -NFS4ERR_EXPIRED:
339                                 /* Don't recall a delegation if it was lost */
340                                 nfs4_schedule_state_recovery(server->nfs4_state);
341                                 return err;
342                 }
343                 err = nfs4_handle_exception(server, err, &exception);
344         } while (exception.retry);
345         return err;
346 }
347
348 static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid)
349 {
350         struct nfs_open_confirmargs arg = {
351                 .fh             = fh,
352                 .seqid          = sp->so_seqid,
353                 .stateid        = *stateid,
354         };
355         struct nfs_open_confirmres res;
356         struct  rpc_message msg = {
357                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
358                 .rpc_argp       = &arg,
359                 .rpc_resp       = &res,
360                 .rpc_cred       = sp->so_cred,
361         };
362         int status;
363
364         status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR);
365         nfs4_increment_seqid(status, sp);
366         if (status >= 0)
367                 memcpy(stateid, &res.stateid, sizeof(*stateid));
368         return status;
369 }
370
371 static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner  *sp, struct nfs_openargs *o_arg, struct nfs_openres *o_res)
372 {
373         struct nfs_server *server = NFS_SERVER(dir);
374         struct rpc_message msg = {
375                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
376                 .rpc_argp = o_arg,
377                 .rpc_resp = o_res,
378                 .rpc_cred = sp->so_cred,
379         };
380         int status;
381
382         /* Update sequence id. The caller must serialize! */
383         o_arg->seqid = sp->so_seqid;
384         o_arg->id = sp->so_id;
385         o_arg->clientid = sp->so_client->cl_clientid;
386
387         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
388         nfs4_increment_seqid(status, sp);
389         if (status != 0)
390                 goto out;
391         update_changeattr(dir, &o_res->cinfo);
392         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
393                 status = _nfs4_proc_open_confirm(server->client, &o_res->fh,
394                                 sp, &o_res->stateid);
395                 if (status != 0)
396                         goto out;
397         }
398         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
399                 status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
400 out:
401         return status;
402 }
403
404 static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openflags)
405 {
406         struct nfs_access_entry cache;
407         int mask = 0;
408         int status;
409
410         if (openflags & FMODE_READ)
411                 mask |= MAY_READ;
412         if (openflags & FMODE_WRITE)
413                 mask |= MAY_WRITE;
414         status = nfs_access_get_cached(inode, cred, &cache);
415         if (status == 0)
416                 goto out;
417
418         /* Be clever: ask server to check for all possible rights */
419         cache.mask = MAY_EXEC | MAY_WRITE | MAY_READ;
420         cache.cred = cred;
421         cache.jiffies = jiffies;
422         status = _nfs4_proc_access(inode, &cache);
423         if (status != 0)
424                 return status;
425         nfs_access_add_cache(inode, &cache);
426 out:
427         if ((cache.mask & mask) == mask)
428                 return 0;
429         return -EACCES;
430 }
431
432 /*
433  * OPEN_EXPIRED:
434  *      reclaim state on the server after a network partition.
435  *      Assumes caller holds the appropriate lock
436  */
437 static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
438 {
439         struct dentry *parent = dget_parent(dentry);
440         struct inode *dir = parent->d_inode;
441         struct inode *inode = state->inode;
442         struct nfs_server *server = NFS_SERVER(dir);
443         struct nfs_delegation *delegation = NFS_I(inode)->delegation;
444         struct nfs_fattr        f_attr = {
445                 .valid = 0,
446         };
447         struct nfs_openargs o_arg = {
448                 .fh = NFS_FH(dir),
449                 .open_flags = state->state,
450                 .name = &dentry->d_name,
451                 .bitmask = server->attr_bitmask,
452                 .claim = NFS4_OPEN_CLAIM_NULL,
453         };
454         struct nfs_openres o_res = {
455                 .f_attr = &f_attr,
456                 .server = server,
457         };
458         int status = 0;
459
460         if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
461                 status = _nfs4_do_access(inode, sp->so_cred, state->state);
462                 if (status < 0)
463                         goto out;
464                 memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid));
465                 set_bit(NFS_DELEGATED_STATE, &state->flags);
466                 goto out;
467         }
468         status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
469         if (status != 0)
470                 goto out_nodeleg;
471         /* Check if files differ */
472         if ((f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT))
473                 goto out_stale;
474         /* Has the file handle changed? */
475         if (nfs_compare_fh(&o_res.fh, NFS_FH(inode)) != 0) {
476                 /* Verify if the change attributes are the same */
477                 if (f_attr.change_attr != NFS_I(inode)->change_attr)
478                         goto out_stale;
479                 if (nfs_size_to_loff_t(f_attr.size) != inode->i_size)
480                         goto out_stale;
481                 /* Lets just pretend that this is the same file */
482                 nfs_copy_fh(NFS_FH(inode), &o_res.fh);
483                 NFS_I(inode)->fileid = f_attr.fileid;
484         }
485         memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
486         if (o_res.delegation_type != 0) {
487                 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM))
488                         nfs_inode_set_delegation(inode, sp->so_cred, &o_res);
489                 else
490                         nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
491         }
492 out_nodeleg:
493         clear_bit(NFS_DELEGATED_STATE, &state->flags);
494 out:
495         dput(parent);
496         return status;
497 out_stale:
498         status = -ESTALE;
499         /* Invalidate the state owner so we don't ever use it again */
500         nfs4_drop_state_owner(sp);
501         d_drop(dentry);
502         /* Should we be trying to close that stateid? */
503         goto out_nodeleg;
504 }
505
506 static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
507 {
508         struct nfs_server *server = NFS_SERVER(dentry->d_inode);
509         struct nfs4_exception exception = { };
510         int err;
511
512         do {
513                 err = _nfs4_open_expired(sp, state, dentry);
514                 if (err == -NFS4ERR_DELAY)
515                         nfs4_handle_exception(server, err, &exception);
516         } while (exception.retry);
517         return err;
518 }
519
520 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
521 {
522         struct nfs_inode *nfsi = NFS_I(state->inode);
523         struct nfs_open_context *ctx;
524         int status;
525
526         spin_lock(&state->inode->i_lock);
527         list_for_each_entry(ctx, &nfsi->open_files, list) {
528                 if (ctx->state != state)
529                         continue;
530                 get_nfs_open_context(ctx);
531                 spin_unlock(&state->inode->i_lock);
532                 status = nfs4_do_open_expired(sp, state, ctx->dentry);
533                 put_nfs_open_context(ctx);
534                 return status;
535         }
536         spin_unlock(&state->inode->i_lock);
537         return -ENOENT;
538 }
539
540 /*
541  * Returns an nfs4_state + an extra reference to the inode
542  */
543 static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred, struct nfs4_state **res)
544 {
545         struct nfs_delegation *delegation;
546         struct nfs_server *server = NFS_SERVER(inode);
547         struct nfs4_client *clp = server->nfs4_state;
548         struct nfs_inode *nfsi = NFS_I(inode);
549         struct nfs4_state_owner *sp = NULL;
550         struct nfs4_state *state = NULL;
551         int open_flags = flags & (FMODE_READ|FMODE_WRITE);
552         int err;
553
554         /* Protect against reboot recovery - NOTE ORDER! */
555         down_read(&clp->cl_sem);
556         /* Protect against delegation recall */
557         down_read(&nfsi->rwsem);
558         delegation = NFS_I(inode)->delegation;
559         err = -ENOENT;
560         if (delegation == NULL || (delegation->type & open_flags) != open_flags)
561                 goto out_err;
562         err = -ENOMEM;
563         if (!(sp = nfs4_get_state_owner(server, cred))) {
564                 dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
565                 goto out_err;
566         }
567         down(&sp->so_sema);
568         state = nfs4_get_open_state(inode, sp);
569         if (state == NULL)
570                 goto out_err;
571
572         err = -ENOENT;
573         if ((state->state & open_flags) == open_flags) {
574                 spin_lock(&inode->i_lock);
575                 if (open_flags & FMODE_READ)
576                         state->nreaders++;
577                 if (open_flags & FMODE_WRITE)
578                         state->nwriters++;
579                 spin_unlock(&inode->i_lock);
580                 goto out_ok;
581         } else if (state->state != 0)
582                 goto out_err;
583
584         lock_kernel();
585         err = _nfs4_do_access(inode, cred, open_flags);
586         unlock_kernel();
587         if (err != 0)
588                 goto out_err;
589         set_bit(NFS_DELEGATED_STATE, &state->flags);
590         update_open_stateid(state, &delegation->stateid, open_flags);
591 out_ok:
592         up(&sp->so_sema);
593         nfs4_put_state_owner(sp);
594         up_read(&nfsi->rwsem);
595         up_read(&clp->cl_sem);
596         igrab(inode);
597         *res = state;
598         return 0; 
599 out_err:
600         if (sp != NULL) {
601                 if (state != NULL)
602                         nfs4_put_open_state(state);
603                 up(&sp->so_sema);
604                 nfs4_put_state_owner(sp);
605         }
606         up_read(&nfsi->rwsem);
607         up_read(&clp->cl_sem);
608         return err;
609 }
610
611 static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred)
612 {
613         struct nfs4_exception exception = { };
614         struct nfs4_state *res;
615         int err;
616
617         do {
618                 err = _nfs4_open_delegated(inode, flags, cred, &res);
619                 if (err == 0)
620                         break;
621                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(inode),
622                                         err, &exception));
623         } while (exception.retry);
624         return res;
625 }
626
627 /*
628  * Returns an nfs4_state + an referenced inode
629  */
630 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
631 {
632         struct nfs4_state_owner  *sp;
633         struct nfs4_state     *state = NULL;
634         struct nfs_server       *server = NFS_SERVER(dir);
635         struct nfs4_client *clp = server->nfs4_state;
636         struct inode *inode = NULL;
637         int                     status;
638         struct nfs_fattr        f_attr = {
639                 .valid          = 0,
640         };
641         struct nfs_openargs o_arg = {
642                 .fh             = NFS_FH(dir),
643                 .open_flags     = flags,
644                 .name           = &dentry->d_name,
645                 .server         = server,
646                 .bitmask = server->attr_bitmask,
647                 .claim = NFS4_OPEN_CLAIM_NULL,
648         };
649         struct nfs_openres o_res = {
650                 .f_attr         = &f_attr,
651                 .server         = server,
652         };
653
654         /* Protect against reboot recovery conflicts */
655         down_read(&clp->cl_sem);
656         status = -ENOMEM;
657         if (!(sp = nfs4_get_state_owner(server, cred))) {
658                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
659                 goto out_err;
660         }
661         if (flags & O_EXCL) {
662                 u32 *p = (u32 *) o_arg.u.verifier.data;
663                 p[0] = jiffies;
664                 p[1] = current->pid;
665         } else
666                 o_arg.u.attrs = sattr;
667         /* Serialization for the sequence id */
668         down(&sp->so_sema);
669
670         status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
671         if (status != 0)
672                 goto out_err;
673
674         status = -ENOMEM;
675         inode = nfs_fhget(dir->i_sb, &o_res.fh, &f_attr);
676         if (!inode)
677                 goto out_err;
678         state = nfs4_get_open_state(inode, sp);
679         if (!state)
680                 goto out_err;
681         update_open_stateid(state, &o_res.stateid, flags);
682         if (o_res.delegation_type != 0)
683                 nfs_inode_set_delegation(inode, cred, &o_res);
684         up(&sp->so_sema);
685         nfs4_put_state_owner(sp);
686         up_read(&clp->cl_sem);
687         *res = state;
688         return 0;
689 out_err:
690         if (sp != NULL) {
691                 if (state != NULL)
692                         nfs4_put_open_state(state);
693                 up(&sp->so_sema);
694                 nfs4_put_state_owner(sp);
695         }
696         /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */
697         up_read(&clp->cl_sem);
698         if (inode != NULL)
699                 iput(inode);
700         *res = NULL;
701         return status;
702 }
703
704
705 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred)
706 {
707         struct nfs4_exception exception = { };
708         struct nfs4_state *res;
709         int status;
710
711         do {
712                 status = _nfs4_do_open(dir, dentry, flags, sattr, cred, &res);
713                 if (status == 0)
714                         break;
715                 /* NOTE: BAD_SEQID means the server and client disagree about the
716                  * book-keeping w.r.t. state-changing operations
717                  * (OPEN/CLOSE/LOCK/LOCKU...)
718                  * It is actually a sign of a bug on the client or on the server.
719                  *
720                  * If we receive a BAD_SEQID error in the particular case of
721                  * doing an OPEN, we assume that nfs4_increment_seqid() will
722                  * have unhashed the old state_owner for us, and that we can
723                  * therefore safely retry using a new one. We should still warn
724                  * the user though...
725                  */
726                 if (status == -NFS4ERR_BAD_SEQID) {
727                         printk(KERN_WARNING "NFS: v4 server returned a bad sequence-id error!\n");
728                         exception.retry = 1;
729                         continue;
730                 }
731                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
732                                         status, &exception));
733         } while (exception.retry);
734         return res;
735 }
736
737 static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
738                 struct nfs_fh *fhandle, struct iattr *sattr,
739                 struct nfs4_state *state)
740 {
741         struct nfs_setattrargs  arg = {
742                 .fh             = fhandle,
743                 .iap            = sattr,
744                 .server         = server,
745                 .bitmask = server->attr_bitmask,
746         };
747         struct nfs_setattrres  res = {
748                 .fattr          = fattr,
749                 .server         = server,
750         };
751         struct rpc_message msg = {
752                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
753                 .rpc_argp       = &arg,
754                 .rpc_resp       = &res,
755         };
756
757         fattr->valid = 0;
758
759         if (state != NULL)
760                 msg.rpc_cred = state->owner->so_cred;
761         if (sattr->ia_valid & ATTR_SIZE)
762                 nfs4_copy_stateid(&arg.stateid, state, NULL);
763         else
764                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
765
766         return rpc_call_sync(server->client, &msg, 0);
767 }
768
769 static int nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
770                 struct nfs_fh *fhandle, struct iattr *sattr,
771                 struct nfs4_state *state)
772 {
773         struct nfs4_exception exception = { };
774         int err;
775         do {
776                 err = nfs4_handle_exception(server,
777                                 _nfs4_do_setattr(server, fattr, fhandle, sattr,
778                                         state),
779                                 &exception);
780         } while (exception.retry);
781         return err;
782 }
783
784 struct nfs4_closedata {
785         struct inode *inode;
786         struct nfs4_state *state;
787         struct nfs_closeargs arg;
788         struct nfs_closeres res;
789 };
790
791 static void nfs4_close_done(struct rpc_task *task)
792 {
793         struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
794         struct nfs4_state *state = calldata->state;
795         struct nfs4_state_owner *sp = state->owner;
796         struct nfs_server *server = NFS_SERVER(calldata->inode);
797
798         /* hmm. we are done with the inode, and in the process of freeing
799          * the state_owner. we keep this around to process errors
800          */
801         nfs4_increment_seqid(task->tk_status, sp);
802         switch (task->tk_status) {
803                 case 0:
804                         memcpy(&state->stateid, &calldata->res.stateid,
805                                         sizeof(state->stateid));
806                         break;
807                 case -NFS4ERR_STALE_STATEID:
808                 case -NFS4ERR_EXPIRED:
809                         state->state = calldata->arg.open_flags;
810                         nfs4_schedule_state_recovery(server->nfs4_state);
811                         break;
812                 default:
813                         if (nfs4_async_handle_error(task, server) == -EAGAIN) {
814                                 rpc_restart_call(task);
815                                 return;
816                         }
817         }
818         state->state = calldata->arg.open_flags;
819         nfs4_put_open_state(state);
820         up(&sp->so_sema);
821         nfs4_put_state_owner(sp);
822         up_read(&server->nfs4_state->cl_sem);
823         kfree(calldata);
824 }
825
826 static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata)
827 {
828         struct rpc_message msg = {
829                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
830                 .rpc_argp = &calldata->arg,
831                 .rpc_resp = &calldata->res,
832                 .rpc_cred = calldata->state->owner->so_cred,
833         };
834         if (calldata->arg.open_flags != 0)
835                 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
836         return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata);
837 }
838
839 /* 
840  * It is possible for data to be read/written from a mem-mapped file 
841  * after the sys_close call (which hits the vfs layer as a flush).
842  * This means that we can't safely call nfsv4 close on a file until 
843  * the inode is cleared. This in turn means that we are not good
844  * NFSv4 citizens - we do not indicate to the server to update the file's 
845  * share state even when we are done with one of the three share 
846  * stateid's in the inode.
847  *
848  * NOTE: Caller must be holding the sp->so_owner semaphore!
849  */
850 int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) 
851 {
852         struct nfs4_closedata *calldata;
853         int status;
854
855         /* Tell caller we're done */
856         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
857                 state->state = mode;
858                 return 0;
859         }
860         calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL);
861         if (calldata == NULL)
862                 return -ENOMEM;
863         calldata->inode = inode;
864         calldata->state = state;
865         calldata->arg.fh = NFS_FH(inode);
866         /* Serialization for the sequence id */
867         calldata->arg.seqid = state->owner->so_seqid;
868         calldata->arg.open_flags = mode;
869         memcpy(&calldata->arg.stateid, &state->stateid,
870                         sizeof(calldata->arg.stateid));
871         status = nfs4_close_call(NFS_SERVER(inode)->client, calldata);
872         /*
873          * Return -EINPROGRESS on success in order to indicate to the
874          * caller that an asynchronous RPC call has been launched, and
875          * that it will release the semaphores on completion.
876          */
877         return (status == 0) ? -EINPROGRESS : status;
878 }
879
880 struct inode *
881 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
882 {
883         struct iattr attr;
884         struct rpc_cred *cred;
885         struct nfs4_state *state;
886
887         if (nd->flags & LOOKUP_CREATE) {
888                 attr.ia_mode = nd->intent.open.create_mode;
889                 attr.ia_valid = ATTR_MODE;
890                 if (!IS_POSIXACL(dir))
891                         attr.ia_mode &= ~current->fs->umask;
892         } else {
893                 attr.ia_valid = 0;
894                 BUG_ON(nd->intent.open.flags & O_CREAT);
895         }
896
897         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
898         if (IS_ERR(cred))
899                 return (struct inode *)cred;
900         state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred);
901         put_rpccred(cred);
902         if (IS_ERR(state))
903                 return (struct inode *)state;
904         return state->inode;
905 }
906
907 int
908 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags)
909 {
910         struct rpc_cred *cred;
911         struct nfs4_state *state;
912         struct inode *inode;
913
914         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
915         if (IS_ERR(cred))
916                 return PTR_ERR(cred);
917         state = nfs4_open_delegated(dentry->d_inode, openflags, cred);
918         if (IS_ERR(state))
919                 state = nfs4_do_open(dir, dentry, openflags, NULL, cred);
920         put_rpccred(cred);
921         if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0)
922                 return 1;
923         if (IS_ERR(state))
924                 return 0;
925         inode = state->inode;
926         if (inode == dentry->d_inode) {
927                 iput(inode);
928                 return 1;
929         }
930         d_drop(dentry);
931         nfs4_close_state(state, openflags);
932         iput(inode);
933         return 0;
934 }
935
936
937 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
938 {
939         struct nfs4_server_caps_res res = {};
940         struct rpc_message msg = {
941                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
942                 .rpc_argp = fhandle,
943                 .rpc_resp = &res,
944         };
945         int status;
946
947         status = rpc_call_sync(server->client, &msg, 0);
948         if (status == 0) {
949                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
950                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
951                         server->caps |= NFS_CAP_ACLS;
952                 if (res.has_links != 0)
953                         server->caps |= NFS_CAP_HARDLINKS;
954                 if (res.has_symlinks != 0)
955                         server->caps |= NFS_CAP_SYMLINKS;
956                 server->acl_bitmask = res.acl_bitmask;
957         }
958         return status;
959 }
960
961 static int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
962 {
963         struct nfs4_exception exception = { };
964         int err;
965         do {
966                 err = nfs4_handle_exception(server,
967                                 _nfs4_server_capabilities(server, fhandle),
968                                 &exception);
969         } while (exception.retry);
970         return err;
971 }
972
973 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
974                 struct nfs_fsinfo *info)
975 {
976         struct nfs_fattr *      fattr = info->fattr;
977         struct nfs4_lookup_root_arg args = {
978                 .bitmask = nfs4_fattr_bitmap,
979         };
980         struct nfs4_lookup_res res = {
981                 .server = server,
982                 .fattr = fattr,
983                 .fh = fhandle,
984         };
985         struct rpc_message msg = {
986                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
987                 .rpc_argp = &args,
988                 .rpc_resp = &res,
989         };
990         fattr->valid = 0;
991         return rpc_call_sync(server->client, &msg, 0);
992 }
993
994 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
995                 struct nfs_fsinfo *info)
996 {
997         struct nfs4_exception exception = { };
998         int err;
999         do {
1000                 err = nfs4_handle_exception(server,
1001                                 _nfs4_lookup_root(server, fhandle, info),
1002                                 &exception);
1003         } while (exception.retry);
1004         return err;
1005 }
1006
1007 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
1008                 struct nfs_fsinfo *info)
1009 {
1010         struct nfs_fattr *      fattr = info->fattr;
1011         unsigned char *         p;
1012         struct qstr             q;
1013         struct nfs4_lookup_arg args = {
1014                 .dir_fh = fhandle,
1015                 .name = &q,
1016                 .bitmask = nfs4_fattr_bitmap,
1017         };
1018         struct nfs4_lookup_res res = {
1019                 .server = server,
1020                 .fattr = fattr,
1021                 .fh = fhandle,
1022         };
1023         struct rpc_message msg = {
1024                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1025                 .rpc_argp = &args,
1026                 .rpc_resp = &res,
1027         };
1028         int status;
1029
1030         /*
1031          * Now we do a separate LOOKUP for each component of the mount path.
1032          * The LOOKUPs are done separately so that we can conveniently
1033          * catch an ERR_WRONGSEC if it occurs along the way...
1034          */
1035         status = nfs4_lookup_root(server, fhandle, info);
1036         if (status)
1037                 goto out;
1038
1039         p = server->mnt_path;
1040         for (;;) {
1041                 struct nfs4_exception exception = { };
1042
1043                 while (*p == '/')
1044                         p++;
1045                 if (!*p)
1046                         break;
1047                 q.name = p;
1048                 while (*p && (*p != '/'))
1049                         p++;
1050                 q.len = p - q.name;
1051
1052                 do {
1053                         fattr->valid = 0;
1054                         status = nfs4_handle_exception(server,
1055                                         rpc_call_sync(server->client, &msg, 0),
1056                                         &exception);
1057                 } while (exception.retry);
1058                 if (status == 0)
1059                         continue;
1060                 if (status == -ENOENT) {
1061                         printk(KERN_NOTICE "NFS: mount path %s does not exist!\n", server->mnt_path);
1062                         printk(KERN_NOTICE "NFS: suggestion: try mounting '/' instead.\n");
1063                 }
1064                 break;
1065         }
1066         if (status == 0)
1067                 status = nfs4_server_capabilities(server, fhandle);
1068         if (status == 0)
1069                 status = nfs4_do_fsinfo(server, fhandle, info);
1070 out:
1071         return status;
1072 }
1073
1074 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1075 {
1076         struct nfs4_getattr_arg args = {
1077                 .fh = fhandle,
1078                 .bitmask = server->attr_bitmask,
1079         };
1080         struct nfs4_getattr_res res = {
1081                 .fattr = fattr,
1082                 .server = server,
1083         };
1084         struct rpc_message msg = {
1085                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
1086                 .rpc_argp = &args,
1087                 .rpc_resp = &res,
1088         };
1089         
1090         fattr->valid = 0;
1091         return rpc_call_sync(server->client, &msg, 0);
1092 }
1093
1094 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1095 {
1096         struct nfs4_exception exception = { };
1097         int err;
1098         do {
1099                 err = nfs4_handle_exception(server,
1100                                 _nfs4_proc_getattr(server, fhandle, fattr),
1101                                 &exception);
1102         } while (exception.retry);
1103         return err;
1104 }
1105
1106 /* 
1107  * The file is not closed if it is opened due to the a request to change
1108  * the size of the file. The open call will not be needed once the
1109  * VFS layer lookup-intents are implemented.
1110  *
1111  * Close is called when the inode is destroyed.
1112  * If we haven't opened the file for O_WRONLY, we
1113  * need to in the size_change case to obtain a stateid.
1114  *
1115  * Got race?
1116  * Because OPEN is always done by name in nfsv4, it is
1117  * possible that we opened a different file by the same
1118  * name.  We can recognize this race condition, but we
1119  * can't do anything about it besides returning an error.
1120  *
1121  * This will be fixed with VFS changes (lookup-intent).
1122  */
1123 static int
1124 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
1125                   struct iattr *sattr)
1126 {
1127         struct inode *          inode = dentry->d_inode;
1128         int                     size_change = sattr->ia_valid & ATTR_SIZE;
1129         struct nfs4_state       *state = NULL;
1130         int need_iput = 0;
1131         int status;
1132
1133         fattr->valid = 0;
1134         
1135         if (size_change) {
1136                 struct rpc_cred *cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
1137                 if (IS_ERR(cred))
1138                         return PTR_ERR(cred);
1139                 state = nfs4_find_state(inode, cred, FMODE_WRITE);
1140                 if (state == NULL) {
1141                         state = nfs4_open_delegated(dentry->d_inode,
1142                                         FMODE_WRITE, cred);
1143                         if (IS_ERR(state))
1144                                 state = nfs4_do_open(dentry->d_parent->d_inode,
1145                                                 dentry, FMODE_WRITE,
1146                                                 NULL, cred);
1147                         need_iput = 1;
1148                 }
1149                 put_rpccred(cred);
1150                 if (IS_ERR(state))
1151                         return PTR_ERR(state);
1152
1153                 if (state->inode != inode) {
1154                         printk(KERN_WARNING "nfs: raced in setattr (%p != %p), returning -EIO\n", inode, state->inode);
1155                         status = -EIO;
1156                         goto out;
1157                 }
1158         }
1159         status = nfs4_do_setattr(NFS_SERVER(inode), fattr,
1160                         NFS_FH(inode), sattr, state);
1161 out:
1162         if (state) {
1163                 inode = state->inode;
1164                 nfs4_close_state(state, FMODE_WRITE);
1165                 if (need_iput)
1166                         iput(inode);
1167         }
1168         return status;
1169 }
1170
1171 static int _nfs4_proc_lookup(struct inode *dir, struct qstr *name,
1172                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1173 {
1174         int                    status;
1175         struct nfs_server *server = NFS_SERVER(dir);
1176         struct nfs4_lookup_arg args = {
1177                 .bitmask = server->attr_bitmask,
1178                 .dir_fh = NFS_FH(dir),
1179                 .name = name,
1180         };
1181         struct nfs4_lookup_res res = {
1182                 .server = server,
1183                 .fattr = fattr,
1184                 .fh = fhandle,
1185         };
1186         struct rpc_message msg = {
1187                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1188                 .rpc_argp = &args,
1189                 .rpc_resp = &res,
1190         };
1191         
1192         fattr->valid = 0;
1193         
1194         dprintk("NFS call  lookup %s\n", name->name);
1195         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1196         dprintk("NFS reply lookup: %d\n", status);
1197         return status;
1198 }
1199
1200 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1201 {
1202         struct nfs4_exception exception = { };
1203         int err;
1204         do {
1205                 err = nfs4_handle_exception(NFS_SERVER(dir),
1206                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
1207                                 &exception);
1208         } while (exception.retry);
1209         return err;
1210 }
1211
1212 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
1213 {
1214         struct nfs4_accessargs args = {
1215                 .fh = NFS_FH(inode),
1216         };
1217         struct nfs4_accessres res = { 0 };
1218         struct rpc_message msg = {
1219                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
1220                 .rpc_argp = &args,
1221                 .rpc_resp = &res,
1222                 .rpc_cred = entry->cred,
1223         };
1224         int mode = entry->mask;
1225         int status;
1226
1227         /*
1228          * Determine which access bits we want to ask for...
1229          */
1230         if (mode & MAY_READ)
1231                 args.access |= NFS4_ACCESS_READ;
1232         if (S_ISDIR(inode->i_mode)) {
1233                 if (mode & MAY_WRITE)
1234                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
1235                 if (mode & MAY_EXEC)
1236                         args.access |= NFS4_ACCESS_LOOKUP;
1237         } else {
1238                 if (mode & MAY_WRITE)
1239                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
1240                 if (mode & MAY_EXEC)
1241                         args.access |= NFS4_ACCESS_EXECUTE;
1242         }
1243         status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1244         if (!status) {
1245                 entry->mask = 0;
1246                 if (res.access & NFS4_ACCESS_READ)
1247                         entry->mask |= MAY_READ;
1248                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
1249                         entry->mask |= MAY_WRITE;
1250                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
1251                         entry->mask |= MAY_EXEC;
1252         }
1253         return status;
1254 }
1255
1256 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
1257 {
1258         struct nfs4_exception exception = { };
1259         int err;
1260         do {
1261                 err = nfs4_handle_exception(NFS_SERVER(inode),
1262                                 _nfs4_proc_access(inode, entry),
1263                                 &exception);
1264         } while (exception.retry);
1265         return err;
1266 }
1267
1268 /*
1269  * TODO: For the time being, we don't try to get any attributes
1270  * along with any of the zero-copy operations READ, READDIR,
1271  * READLINK, WRITE.
1272  *
1273  * In the case of the first three, we want to put the GETATTR
1274  * after the read-type operation -- this is because it is hard
1275  * to predict the length of a GETATTR response in v4, and thus
1276  * align the READ data correctly.  This means that the GETATTR
1277  * may end up partially falling into the page cache, and we should
1278  * shift it into the 'tail' of the xdr_buf before processing.
1279  * To do this efficiently, we need to know the total length
1280  * of data received, which doesn't seem to be available outside
1281  * of the RPC layer.
1282  *
1283  * In the case of WRITE, we also want to put the GETATTR after
1284  * the operation -- in this case because we want to make sure
1285  * we get the post-operation mtime and size.  This means that
1286  * we can't use xdr_encode_pages() as written: we need a variant
1287  * of it which would leave room in the 'tail' iovec.
1288  *
1289  * Both of these changes to the XDR layer would in fact be quite
1290  * minor, but I decided to leave them for a subsequent patch.
1291  */
1292 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
1293                 unsigned int pgbase, unsigned int pglen)
1294 {
1295         struct nfs4_readlink args = {
1296                 .fh       = NFS_FH(inode),
1297                 .pgbase   = pgbase,
1298                 .pglen    = pglen,
1299                 .pages    = &page,
1300         };
1301         struct rpc_message msg = {
1302                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
1303                 .rpc_argp = &args,
1304                 .rpc_resp = NULL,
1305         };
1306
1307         return rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1308 }
1309
1310 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
1311                 unsigned int pgbase, unsigned int pglen)
1312 {
1313         struct nfs4_exception exception = { };
1314         int err;
1315         do {
1316                 err = nfs4_handle_exception(NFS_SERVER(inode),
1317                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
1318                                 &exception);
1319         } while (exception.retry);
1320         return err;
1321 }
1322
1323 static int _nfs4_proc_read(struct nfs_read_data *rdata)
1324 {
1325         int flags = rdata->flags;
1326         struct inode *inode = rdata->inode;
1327         struct nfs_fattr *fattr = rdata->res.fattr;
1328         struct nfs_server *server = NFS_SERVER(inode);
1329         struct rpc_message msg = {
1330                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_READ],
1331                 .rpc_argp       = &rdata->args,
1332                 .rpc_resp       = &rdata->res,
1333                 .rpc_cred       = rdata->cred,
1334         };
1335         unsigned long timestamp = jiffies;
1336         int status;
1337
1338         dprintk("NFS call  read %d @ %Ld\n", rdata->args.count,
1339                         (long long) rdata->args.offset);
1340
1341         fattr->valid = 0;
1342         status = rpc_call_sync(server->client, &msg, flags);
1343         if (!status)
1344                 renew_lease(server, timestamp);
1345         dprintk("NFS reply read: %d\n", status);
1346         return status;
1347 }
1348
1349 static int nfs4_proc_read(struct nfs_read_data *rdata)
1350 {
1351         struct nfs4_exception exception = { };
1352         int err;
1353         do {
1354                 err = nfs4_handle_exception(NFS_SERVER(rdata->inode),
1355                                 _nfs4_proc_read(rdata),
1356                                 &exception);
1357         } while (exception.retry);
1358         return err;
1359 }
1360
1361 static int _nfs4_proc_write(struct nfs_write_data *wdata)
1362 {
1363         int rpcflags = wdata->flags;
1364         struct inode *inode = wdata->inode;
1365         struct nfs_fattr *fattr = wdata->res.fattr;
1366         struct nfs_server *server = NFS_SERVER(inode);
1367         struct rpc_message msg = {
1368                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
1369                 .rpc_argp       = &wdata->args,
1370                 .rpc_resp       = &wdata->res,
1371                 .rpc_cred       = wdata->cred,
1372         };
1373         int status;
1374
1375         dprintk("NFS call  write %d @ %Ld\n", wdata->args.count,
1376                         (long long) wdata->args.offset);
1377
1378         fattr->valid = 0;
1379         status = rpc_call_sync(server->client, &msg, rpcflags);
1380         dprintk("NFS reply write: %d\n", status);
1381         return status;
1382 }
1383
1384 static int nfs4_proc_write(struct nfs_write_data *wdata)
1385 {
1386         struct nfs4_exception exception = { };
1387         int err;
1388         do {
1389                 err = nfs4_handle_exception(NFS_SERVER(wdata->inode),
1390                                 _nfs4_proc_write(wdata),
1391                                 &exception);
1392         } while (exception.retry);
1393         return err;
1394 }
1395
1396 static int _nfs4_proc_commit(struct nfs_write_data *cdata)
1397 {
1398         struct inode *inode = cdata->inode;
1399         struct nfs_fattr *fattr = cdata->res.fattr;
1400         struct nfs_server *server = NFS_SERVER(inode);
1401         struct rpc_message msg = {
1402                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
1403                 .rpc_argp       = &cdata->args,
1404                 .rpc_resp       = &cdata->res,
1405                 .rpc_cred       = cdata->cred,
1406         };
1407         int status;
1408
1409         dprintk("NFS call  commit %d @ %Ld\n", cdata->args.count,
1410                         (long long) cdata->args.offset);
1411
1412         fattr->valid = 0;
1413         status = rpc_call_sync(server->client, &msg, 0);
1414         dprintk("NFS reply commit: %d\n", status);
1415         return status;
1416 }
1417
1418 static int nfs4_proc_commit(struct nfs_write_data *cdata)
1419 {
1420         struct nfs4_exception exception = { };
1421         int err;
1422         do {
1423                 err = nfs4_handle_exception(NFS_SERVER(cdata->inode),
1424                                 _nfs4_proc_commit(cdata),
1425                                 &exception);
1426         } while (exception.retry);
1427         return err;
1428 }
1429
1430 /*
1431  * Got race?
1432  * We will need to arrange for the VFS layer to provide an atomic open.
1433  * Until then, this create/open method is prone to inefficiency and race
1434  * conditions due to the lookup, create, and open VFS calls from sys_open()
1435  * placed on the wire.
1436  *
1437  * Given the above sorry state of affairs, I'm simply sending an OPEN.
1438  * The file will be opened again in the subsequent VFS open call
1439  * (nfs4_proc_file_open).
1440  *
1441  * The open for read will just hang around to be used by any process that
1442  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
1443  */
1444
1445 static int
1446 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1447                  int flags)
1448 {
1449         struct nfs4_state *state;
1450         struct rpc_cred *cred;
1451         int status = 0;
1452
1453         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
1454         if (IS_ERR(cred)) {
1455                 status = PTR_ERR(cred);
1456                 goto out;
1457         }
1458         state = nfs4_do_open(dir, dentry, flags, sattr, cred);
1459         put_rpccred(cred);
1460         if (IS_ERR(state)) {
1461                 status = PTR_ERR(state);
1462                 goto out;
1463         }
1464         d_instantiate(dentry, state->inode);
1465         if (flags & O_EXCL) {
1466                 struct nfs_fattr fattr;
1467                 status = nfs4_do_setattr(NFS_SERVER(dir), &fattr,
1468                                      NFS_FH(state->inode), sattr, state);
1469                 if (status == 0)
1470                         goto out;
1471         } else if (flags != 0)
1472                 goto out;
1473         nfs4_close_state(state, flags);
1474 out:
1475         return status;
1476 }
1477
1478 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
1479 {
1480         struct nfs4_remove_arg args = {
1481                 .fh = NFS_FH(dir),
1482                 .name = name,
1483         };
1484         struct nfs4_change_info res;
1485         struct rpc_message msg = {
1486                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
1487                 .rpc_argp       = &args,
1488                 .rpc_resp       = &res,
1489         };
1490         int                     status;
1491
1492         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1493         if (status == 0)
1494                 update_changeattr(dir, &res);
1495         return status;
1496 }
1497
1498 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
1499 {
1500         struct nfs4_exception exception = { };
1501         int err;
1502         do {
1503                 err = nfs4_handle_exception(NFS_SERVER(dir),
1504                                 _nfs4_proc_remove(dir, name),
1505                                 &exception);
1506         } while (exception.retry);
1507         return err;
1508 }
1509
1510 struct unlink_desc {
1511         struct nfs4_remove_arg  args;
1512         struct nfs4_change_info res;
1513 };
1514
1515 static int nfs4_proc_unlink_setup(struct rpc_message *msg, struct dentry *dir,
1516                 struct qstr *name)
1517 {
1518         struct unlink_desc *up;
1519
1520         up = (struct unlink_desc *) kmalloc(sizeof(*up), GFP_KERNEL);
1521         if (!up)
1522                 return -ENOMEM;
1523         
1524         up->args.fh = NFS_FH(dir->d_inode);
1525         up->args.name = name;
1526         
1527         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
1528         msg->rpc_argp = &up->args;
1529         msg->rpc_resp = &up->res;
1530         return 0;
1531 }
1532
1533 static int nfs4_proc_unlink_done(struct dentry *dir, struct rpc_task *task)
1534 {
1535         struct rpc_message *msg = &task->tk_msg;
1536         struct unlink_desc *up;
1537         
1538         if (msg->rpc_resp != NULL) {
1539                 up = container_of(msg->rpc_resp, struct unlink_desc, res);
1540                 update_changeattr(dir->d_inode, &up->res);
1541                 kfree(up);
1542                 msg->rpc_resp = NULL;
1543                 msg->rpc_argp = NULL;
1544         }
1545         return 0;
1546 }
1547
1548 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
1549                 struct inode *new_dir, struct qstr *new_name)
1550 {
1551         struct nfs4_rename_arg arg = {
1552                 .old_dir = NFS_FH(old_dir),
1553                 .new_dir = NFS_FH(new_dir),
1554                 .old_name = old_name,
1555                 .new_name = new_name,
1556         };
1557         struct nfs4_rename_res res = { };
1558         struct rpc_message msg = {
1559                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
1560                 .rpc_argp = &arg,
1561                 .rpc_resp = &res,
1562         };
1563         int                     status;
1564         
1565         status = rpc_call_sync(NFS_CLIENT(old_dir), &msg, 0);
1566
1567         if (!status) {
1568                 update_changeattr(old_dir, &res.old_cinfo);
1569                 update_changeattr(new_dir, &res.new_cinfo);
1570         }
1571         return status;
1572 }
1573
1574 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
1575                 struct inode *new_dir, struct qstr *new_name)
1576 {
1577         struct nfs4_exception exception = { };
1578         int err;
1579         do {
1580                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
1581                                 _nfs4_proc_rename(old_dir, old_name,
1582                                         new_dir, new_name),
1583                                 &exception);
1584         } while (exception.retry);
1585         return err;
1586 }
1587
1588 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
1589 {
1590         struct nfs4_link_arg arg = {
1591                 .fh     = NFS_FH(inode),
1592                 .dir_fh = NFS_FH(dir),
1593                 .name   = name,
1594         };
1595         struct nfs4_change_info cinfo = { };
1596         struct rpc_message msg = {
1597                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
1598                 .rpc_argp = &arg,
1599                 .rpc_resp = &cinfo,
1600         };
1601         int                     status;
1602
1603         status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1604         if (!status)
1605                 update_changeattr(dir, &cinfo);
1606
1607         return status;
1608 }
1609
1610 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
1611 {
1612         struct nfs4_exception exception = { };
1613         int err;
1614         do {
1615                 err = nfs4_handle_exception(NFS_SERVER(inode),
1616                                 _nfs4_proc_link(inode, dir, name),
1617                                 &exception);
1618         } while (exception.retry);
1619         return err;
1620 }
1621
1622 static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name,
1623                 struct qstr *path, struct iattr *sattr, struct nfs_fh *fhandle,
1624                 struct nfs_fattr *fattr)
1625 {
1626         struct nfs_server *server = NFS_SERVER(dir);
1627         struct nfs4_create_arg arg = {
1628                 .dir_fh = NFS_FH(dir),
1629                 .server = server,
1630                 .name = name,
1631                 .attrs = sattr,
1632                 .ftype = NF4LNK,
1633                 .bitmask = server->attr_bitmask,
1634         };
1635         struct nfs4_create_res res = {
1636                 .server = server,
1637                 .fh = fhandle,
1638                 .fattr = fattr,
1639         };
1640         struct rpc_message msg = {
1641                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK],
1642                 .rpc_argp = &arg,
1643                 .rpc_resp = &res,
1644         };
1645         int                     status;
1646
1647         if (path->len > NFS4_MAXPATHLEN)
1648                 return -ENAMETOOLONG;
1649         arg.u.symlink = path;
1650         fattr->valid = 0;
1651         
1652         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1653         if (!status)
1654                 update_changeattr(dir, &res.dir_cinfo);
1655         return status;
1656 }
1657
1658 static int nfs4_proc_symlink(struct inode *dir, struct qstr *name,
1659                 struct qstr *path, struct iattr *sattr, struct nfs_fh *fhandle,
1660                 struct nfs_fattr *fattr)
1661 {
1662         struct nfs4_exception exception = { };
1663         int err;
1664         do {
1665                 err = nfs4_handle_exception(NFS_SERVER(dir),
1666                                 _nfs4_proc_symlink(dir, name, path, sattr,
1667                                         fhandle, fattr),
1668                                 &exception);
1669         } while (exception.retry);
1670         return err;
1671 }
1672
1673 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
1674                 struct iattr *sattr)
1675 {
1676         struct nfs_server *server = NFS_SERVER(dir);
1677         struct nfs_fh fhandle;
1678         struct nfs_fattr fattr;
1679         struct nfs4_create_arg arg = {
1680                 .dir_fh = NFS_FH(dir),
1681                 .server = server,
1682                 .name = &dentry->d_name,
1683                 .attrs = sattr,
1684                 .ftype = NF4DIR,
1685                 .bitmask = server->attr_bitmask,
1686         };
1687         struct nfs4_create_res res = {
1688                 .server = server,
1689                 .fh = &fhandle,
1690                 .fattr = &fattr,
1691         };
1692         struct rpc_message msg = {
1693                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE],
1694                 .rpc_argp = &arg,
1695                 .rpc_resp = &res,
1696         };
1697         int                     status;
1698
1699         fattr.valid = 0;
1700         
1701         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1702         if (!status) {
1703                 update_changeattr(dir, &res.dir_cinfo);
1704                 status = nfs_instantiate(dentry, &fhandle, &fattr);
1705         }
1706         return status;
1707 }
1708
1709 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
1710                 struct iattr *sattr)
1711 {
1712         struct nfs4_exception exception = { };
1713         int err;
1714         do {
1715                 err = nfs4_handle_exception(NFS_SERVER(dir),
1716                                 _nfs4_proc_mkdir(dir, dentry, sattr),
1717                                 &exception);
1718         } while (exception.retry);
1719         return err;
1720 }
1721
1722 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
1723                   u64 cookie, struct page *page, unsigned int count, int plus)
1724 {
1725         struct inode            *dir = dentry->d_inode;
1726         struct nfs4_readdir_arg args = {
1727                 .fh = NFS_FH(dir),
1728                 .pages = &page,
1729                 .pgbase = 0,
1730                 .count = count,
1731                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
1732         };
1733         struct nfs4_readdir_res res;
1734         struct rpc_message msg = {
1735                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
1736                 .rpc_argp = &args,
1737                 .rpc_resp = &res,
1738                 .rpc_cred = cred,
1739         };
1740         int                     status;
1741
1742         lock_kernel();
1743         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
1744         res.pgbase = args.pgbase;
1745         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1746         if (status == 0)
1747                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
1748         unlock_kernel();
1749         return status;
1750 }
1751
1752 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
1753                   u64 cookie, struct page *page, unsigned int count, int plus)
1754 {
1755         struct nfs4_exception exception = { };
1756         int err;
1757         do {
1758                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
1759                                 _nfs4_proc_readdir(dentry, cred, cookie,
1760                                         page, count, plus),
1761                                 &exception);
1762         } while (exception.retry);
1763         return err;
1764 }
1765
1766 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
1767                 struct iattr *sattr, dev_t rdev)
1768 {
1769         struct nfs_server *server = NFS_SERVER(dir);
1770         struct nfs_fh fh;
1771         struct nfs_fattr fattr;
1772         struct nfs4_create_arg arg = {
1773                 .dir_fh = NFS_FH(dir),
1774                 .server = server,
1775                 .name = &dentry->d_name,
1776                 .attrs = sattr,
1777                 .bitmask = server->attr_bitmask,
1778         };
1779         struct nfs4_create_res res = {
1780                 .server = server,
1781                 .fh = &fh,
1782                 .fattr = &fattr,
1783         };
1784         struct rpc_message msg = {
1785                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE],
1786                 .rpc_argp = &arg,
1787                 .rpc_resp = &res,
1788         };
1789         int                     status;
1790         int                     mode = sattr->ia_mode;
1791
1792         fattr.valid = 0;
1793
1794         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
1795         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
1796         if (S_ISFIFO(mode))
1797                 arg.ftype = NF4FIFO;
1798         else if (S_ISBLK(mode)) {
1799                 arg.ftype = NF4BLK;
1800                 arg.u.device.specdata1 = MAJOR(rdev);
1801                 arg.u.device.specdata2 = MINOR(rdev);
1802         }
1803         else if (S_ISCHR(mode)) {
1804                 arg.ftype = NF4CHR;
1805                 arg.u.device.specdata1 = MAJOR(rdev);
1806                 arg.u.device.specdata2 = MINOR(rdev);
1807         }
1808         else
1809                 arg.ftype = NF4SOCK;
1810         
1811         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1812         if (status == 0) {
1813                 update_changeattr(dir, &res.dir_cinfo);
1814                 status = nfs_instantiate(dentry, &fh, &fattr);
1815         }
1816         return status;
1817 }
1818
1819 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
1820                 struct iattr *sattr, dev_t rdev)
1821 {
1822         struct nfs4_exception exception = { };
1823         int err;
1824         do {
1825                 err = nfs4_handle_exception(NFS_SERVER(dir),
1826                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
1827                                 &exception);
1828         } while (exception.retry);
1829         return err;
1830 }
1831
1832 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
1833                  struct nfs_fsstat *fsstat)
1834 {
1835         struct nfs4_statfs_arg args = {
1836                 .fh = fhandle,
1837                 .bitmask = server->attr_bitmask,
1838         };
1839         struct rpc_message msg = {
1840                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
1841                 .rpc_argp = &args,
1842                 .rpc_resp = fsstat,
1843         };
1844
1845         fsstat->fattr->valid = 0;
1846         return rpc_call_sync(server->client, &msg, 0);
1847 }
1848
1849 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
1850 {
1851         struct nfs4_exception exception = { };
1852         int err;
1853         do {
1854                 err = nfs4_handle_exception(server,
1855                                 _nfs4_proc_statfs(server, fhandle, fsstat),
1856                                 &exception);
1857         } while (exception.retry);
1858         return err;
1859 }
1860
1861 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
1862                 struct nfs_fsinfo *fsinfo)
1863 {
1864         struct nfs4_fsinfo_arg args = {
1865                 .fh = fhandle,
1866                 .bitmask = server->attr_bitmask,
1867         };
1868         struct rpc_message msg = {
1869                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
1870                 .rpc_argp = &args,
1871                 .rpc_resp = fsinfo,
1872         };
1873
1874         return rpc_call_sync(server->client, &msg, 0);
1875 }
1876
1877 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
1878 {
1879         struct nfs4_exception exception = { };
1880         int err;
1881
1882         do {
1883                 err = nfs4_handle_exception(server,
1884                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
1885                                 &exception);
1886         } while (exception.retry);
1887         return err;
1888 }
1889
1890 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
1891 {
1892         fsinfo->fattr->valid = 0;
1893         return nfs4_do_fsinfo(server, fhandle, fsinfo);
1894 }
1895
1896 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
1897                 struct nfs_pathconf *pathconf)
1898 {
1899         struct nfs4_pathconf_arg args = {
1900                 .fh = fhandle,
1901                 .bitmask = server->attr_bitmask,
1902         };
1903         struct rpc_message msg = {
1904                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
1905                 .rpc_argp = &args,
1906                 .rpc_resp = pathconf,
1907         };
1908
1909         /* None of the pathconf attributes are mandatory to implement */
1910         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
1911                 memset(pathconf, 0, sizeof(*pathconf));
1912                 return 0;
1913         }
1914
1915         pathconf->fattr->valid = 0;
1916         return rpc_call_sync(server->client, &msg, 0);
1917 }
1918
1919 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
1920                 struct nfs_pathconf *pathconf)
1921 {
1922         struct nfs4_exception exception = { };
1923         int err;
1924
1925         do {
1926                 err = nfs4_handle_exception(server,
1927                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
1928                                 &exception);
1929         } while (exception.retry);
1930         return err;
1931 }
1932
1933 static void
1934 nfs4_read_done(struct rpc_task *task)
1935 {
1936         struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata;
1937         struct inode *inode = data->inode;
1938
1939         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
1940                 rpc_restart_call(task);
1941                 return;
1942         }
1943         if (task->tk_status > 0)
1944                 renew_lease(NFS_SERVER(inode), data->timestamp);
1945         /* Call back common NFS readpage processing */
1946         nfs_readpage_result(task);
1947 }
1948
1949 static void
1950 nfs4_proc_read_setup(struct nfs_read_data *data)
1951 {
1952         struct rpc_task *task = &data->task;
1953         struct rpc_message msg = {
1954                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ],
1955                 .rpc_argp = &data->args,
1956                 .rpc_resp = &data->res,
1957                 .rpc_cred = data->cred,
1958         };
1959         struct inode *inode = data->inode;
1960         int flags;
1961
1962         data->timestamp   = jiffies;
1963
1964         /* N.B. Do we need to test? Never called for swapfile inode */
1965         flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
1966
1967         /* Finalize the task. */
1968         rpc_init_task(task, NFS_CLIENT(inode), nfs4_read_done, flags);
1969         rpc_call_setup(task, &msg, 0);
1970 }
1971
1972 static void
1973 nfs4_write_done(struct rpc_task *task)
1974 {
1975         struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
1976         struct inode *inode = data->inode;
1977         
1978         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
1979                 rpc_restart_call(task);
1980                 return;
1981         }
1982         if (task->tk_status >= 0)
1983                 renew_lease(NFS_SERVER(inode), data->timestamp);
1984         /* Call back common NFS writeback processing */
1985         nfs_writeback_done(task);
1986 }
1987
1988 static void
1989 nfs4_proc_write_setup(struct nfs_write_data *data, int how)
1990 {
1991         struct rpc_task *task = &data->task;
1992         struct rpc_message msg = {
1993                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
1994                 .rpc_argp = &data->args,
1995                 .rpc_resp = &data->res,
1996                 .rpc_cred = data->cred,
1997         };
1998         struct inode *inode = data->inode;
1999         int stable;
2000         int flags;
2001         
2002         if (how & FLUSH_STABLE) {
2003                 if (!NFS_I(inode)->ncommit)
2004                         stable = NFS_FILE_SYNC;
2005                 else
2006                         stable = NFS_DATA_SYNC;
2007         } else
2008                 stable = NFS_UNSTABLE;
2009         data->args.stable = stable;
2010
2011         data->timestamp   = jiffies;
2012
2013         /* Set the initial flags for the task.  */
2014         flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
2015
2016         /* Finalize the task. */
2017         rpc_init_task(task, NFS_CLIENT(inode), nfs4_write_done, flags);
2018         rpc_call_setup(task, &msg, 0);
2019 }
2020
2021 static void
2022 nfs4_commit_done(struct rpc_task *task)
2023 {
2024         struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
2025         struct inode *inode = data->inode;
2026         
2027         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
2028                 rpc_restart_call(task);
2029                 return;
2030         }
2031         /* Call back common NFS writeback processing */
2032         nfs_commit_done(task);
2033 }
2034
2035 static void
2036 nfs4_proc_commit_setup(struct nfs_write_data *data, int how)
2037 {
2038         struct rpc_task *task = &data->task;
2039         struct rpc_message msg = {
2040                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
2041                 .rpc_argp = &data->args,
2042                 .rpc_resp = &data->res,
2043                 .rpc_cred = data->cred,
2044         };      
2045         struct inode *inode = data->inode;
2046         int flags;
2047         
2048         /* Set the initial flags for the task.  */
2049         flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
2050
2051         /* Finalize the task. */
2052         rpc_init_task(task, NFS_CLIENT(inode), nfs4_commit_done, flags);
2053         rpc_call_setup(task, &msg, 0);  
2054 }
2055
2056 /*
2057  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
2058  * standalone procedure for queueing an asynchronous RENEW.
2059  */
2060 static void
2061 renew_done(struct rpc_task *task)
2062 {
2063         struct nfs4_client *clp = (struct nfs4_client *)task->tk_msg.rpc_argp;
2064         unsigned long timestamp = (unsigned long)task->tk_calldata;
2065
2066         if (task->tk_status < 0) {
2067                 switch (task->tk_status) {
2068                         case -NFS4ERR_STALE_CLIENTID:
2069                         case -NFS4ERR_EXPIRED:
2070                         case -NFS4ERR_CB_PATH_DOWN:
2071                                 nfs4_schedule_state_recovery(clp);
2072                 }
2073                 return;
2074         }
2075         spin_lock(&clp->cl_lock);
2076         if (time_before(clp->cl_last_renewal,timestamp))
2077                 clp->cl_last_renewal = timestamp;
2078         spin_unlock(&clp->cl_lock);
2079 }
2080
2081 int
2082 nfs4_proc_async_renew(struct nfs4_client *clp)
2083 {
2084         struct rpc_message msg = {
2085                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2086                 .rpc_argp       = clp,
2087                 .rpc_cred       = clp->cl_cred,
2088         };
2089
2090         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
2091                         renew_done, (void *)jiffies);
2092 }
2093
2094 int
2095 nfs4_proc_renew(struct nfs4_client *clp)
2096 {
2097         struct rpc_message msg = {
2098                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2099                 .rpc_argp       = clp,
2100                 .rpc_cred       = clp->cl_cred,
2101         };
2102         unsigned long now = jiffies;
2103         int status;
2104
2105         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2106         if (status < 0)
2107                 return status;
2108         spin_lock(&clp->cl_lock);
2109         if (time_before(clp->cl_last_renewal,now))
2110                 clp->cl_last_renewal = now;
2111         spin_unlock(&clp->cl_lock);
2112         return 0;
2113 }
2114
2115 /*
2116  * We will need to arrange for the VFS layer to provide an atomic open.
2117  * Until then, this open method is prone to inefficiency and race conditions
2118  * due to the lookup, potential create, and open VFS calls from sys_open()
2119  * placed on the wire.
2120  */
2121 static int
2122 nfs4_proc_file_open(struct inode *inode, struct file *filp)
2123 {
2124         struct dentry *dentry = filp->f_dentry;
2125         struct nfs_open_context *ctx;
2126         struct nfs4_state *state = NULL;
2127         struct rpc_cred *cred;
2128         int status = -ENOMEM;
2129
2130         dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n",
2131                                (int)dentry->d_parent->d_name.len,
2132                                dentry->d_parent->d_name.name,
2133                                (int)dentry->d_name.len, dentry->d_name.name);
2134
2135
2136         /* Find our open stateid */
2137         cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
2138         if (IS_ERR(cred))
2139                 return PTR_ERR(cred);
2140         ctx = alloc_nfs_open_context(dentry, cred);
2141         put_rpccred(cred);
2142         if (unlikely(ctx == NULL))
2143                 return -ENOMEM;
2144         status = -EIO; /* ERACE actually */
2145         state = nfs4_find_state(inode, cred, filp->f_mode);
2146         if (unlikely(state == NULL))
2147                 goto no_state;
2148         ctx->state = state;
2149         nfs4_close_state(state, filp->f_mode);
2150         ctx->mode = filp->f_mode;
2151         nfs_file_set_open_context(filp, ctx);
2152         put_nfs_open_context(ctx);
2153         if (filp->f_mode & FMODE_WRITE)
2154                 nfs_begin_data_update(inode);
2155         return 0;
2156 no_state:
2157         printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__);
2158         put_nfs_open_context(ctx);
2159         return status;
2160 }
2161
2162 /*
2163  * Release our state
2164  */
2165 static int
2166 nfs4_proc_file_release(struct inode *inode, struct file *filp)
2167 {
2168         if (filp->f_mode & FMODE_WRITE)
2169                 nfs_end_data_update(inode);
2170         nfs_file_clear_open_context(filp);
2171         return 0;
2172 }
2173
2174 static inline int nfs4_server_supports_acls(struct nfs_server *server)
2175 {
2176         return (server->caps & NFS_CAP_ACLS)
2177                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2178                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
2179 }
2180
2181 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
2182  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
2183  * the stack.
2184  */
2185 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
2186
2187 static void buf_to_pages(const void *buf, size_t buflen,
2188                 struct page **pages, unsigned int *pgbase)
2189 {
2190         const void *p = buf;
2191
2192         *pgbase = offset_in_page(buf);
2193         p -= *pgbase;
2194         while (p < buf + buflen) {
2195                 *(pages++) = virt_to_page(p);
2196                 p += PAGE_CACHE_SIZE;
2197         }
2198 }
2199
2200 struct nfs4_cached_acl {
2201         int cached;
2202         size_t len;
2203         char data[0];
2204 };
2205
2206 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
2207 {
2208         struct nfs_inode *nfsi = NFS_I(inode);
2209
2210         spin_lock(&inode->i_lock);
2211         kfree(nfsi->nfs4_acl);
2212         nfsi->nfs4_acl = acl;
2213         spin_unlock(&inode->i_lock);
2214 }
2215
2216 static void nfs4_zap_acl_attr(struct inode *inode)
2217 {
2218         nfs4_set_cached_acl(inode, NULL);
2219 }
2220
2221 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
2222 {
2223         struct nfs_inode *nfsi = NFS_I(inode);
2224         struct nfs4_cached_acl *acl;
2225         int ret = -ENOENT;
2226
2227         spin_lock(&inode->i_lock);
2228         acl = nfsi->nfs4_acl;
2229         if (acl == NULL)
2230                 goto out;
2231         if (buf == NULL) /* user is just asking for length */
2232                 goto out_len;
2233         if (acl->cached == 0)
2234                 goto out;
2235         ret = -ERANGE; /* see getxattr(2) man page */
2236         if (acl->len > buflen)
2237                 goto out;
2238         memcpy(buf, acl->data, acl->len);
2239 out_len:
2240         ret = acl->len;
2241 out:
2242         spin_unlock(&inode->i_lock);
2243         return ret;
2244 }
2245
2246 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
2247 {
2248         struct nfs4_cached_acl *acl;
2249
2250         if (buf && acl_len <= PAGE_SIZE) {
2251                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
2252                 if (acl == NULL)
2253                         goto out;
2254                 acl->cached = 1;
2255                 memcpy(acl->data, buf, acl_len);
2256         } else {
2257                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
2258                 if (acl == NULL)
2259                         goto out;
2260                 acl->cached = 0;
2261         }
2262         acl->len = acl_len;
2263 out:
2264         nfs4_set_cached_acl(inode, acl);
2265 }
2266
2267 static inline ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
2268 {
2269         struct page *pages[NFS4ACL_MAXPAGES];
2270         struct nfs_getaclargs args = {
2271                 .fh = NFS_FH(inode),
2272                 .acl_pages = pages,
2273                 .acl_len = buflen,
2274         };
2275         size_t resp_len = buflen;
2276         void *resp_buf;
2277         struct rpc_message msg = {
2278                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
2279                 .rpc_argp = &args,
2280                 .rpc_resp = &resp_len,
2281         };
2282         struct page *localpage = NULL;
2283         int ret;
2284
2285         if (buflen < PAGE_SIZE) {
2286                 /* As long as we're doing a round trip to the server anyway,
2287                  * let's be prepared for a page of acl data. */
2288                 localpage = alloc_page(GFP_KERNEL);
2289                 resp_buf = page_address(localpage);
2290                 if (localpage == NULL)
2291                         return -ENOMEM;
2292                 args.acl_pages[0] = localpage;
2293                 args.acl_pgbase = 0;
2294                 args.acl_len = PAGE_SIZE;
2295         } else {
2296                 resp_buf = buf;
2297                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
2298         }
2299         ret = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
2300         if (ret)
2301                 goto out_free;
2302         if (resp_len > args.acl_len)
2303                 nfs4_write_cached_acl(inode, NULL, resp_len);
2304         else
2305                 nfs4_write_cached_acl(inode, resp_buf, resp_len);
2306         if (buf) {
2307                 ret = -ERANGE;
2308                 if (resp_len > buflen)
2309                         goto out_free;
2310                 if (localpage)
2311                         memcpy(buf, resp_buf, resp_len);
2312         }
2313         ret = resp_len;
2314 out_free:
2315         if (localpage)
2316                 __free_page(localpage);
2317         return ret;
2318 }
2319
2320 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
2321 {
2322         struct nfs_server *server = NFS_SERVER(inode);
2323         int ret;
2324
2325         if (!nfs4_server_supports_acls(server))
2326                 return -EOPNOTSUPP;
2327         ret = nfs_revalidate_inode(server, inode);
2328         if (ret < 0)
2329                 return ret;
2330         ret = nfs4_read_cached_acl(inode, buf, buflen);
2331         if (ret != -ENOENT)
2332                 return ret;
2333         return nfs4_get_acl_uncached(inode, buf, buflen);
2334 }
2335
2336 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
2337 {
2338         struct nfs_server *server = NFS_SERVER(inode);
2339         struct page *pages[NFS4ACL_MAXPAGES];
2340         struct nfs_setaclargs arg = {
2341                 .fh             = NFS_FH(inode),
2342                 .acl_pages      = pages,
2343                 .acl_len        = buflen,
2344         };
2345         struct rpc_message msg = {
2346                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
2347                 .rpc_argp       = &arg,
2348                 .rpc_resp       = NULL,
2349         };
2350         int ret;
2351
2352         if (!nfs4_server_supports_acls(server))
2353                 return -EOPNOTSUPP;
2354         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
2355         ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0);
2356         if (ret == 0)
2357                 nfs4_write_cached_acl(inode, buf, buflen);
2358         return ret;
2359 }
2360
2361 static int
2362 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server)
2363 {
2364         struct nfs4_client *clp = server->nfs4_state;
2365
2366         if (!clp || task->tk_status >= 0)
2367                 return 0;
2368         switch(task->tk_status) {
2369                 case -NFS4ERR_STALE_CLIENTID:
2370                 case -NFS4ERR_STALE_STATEID:
2371                 case -NFS4ERR_EXPIRED:
2372                         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL, NULL);
2373                         nfs4_schedule_state_recovery(clp);
2374                         if (test_bit(NFS4CLNT_OK, &clp->cl_state))
2375                                 rpc_wake_up_task(task);
2376                         task->tk_status = 0;
2377                         return -EAGAIN;
2378                 case -NFS4ERR_GRACE:
2379                 case -NFS4ERR_DELAY:
2380                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
2381                         task->tk_status = 0;
2382                         return -EAGAIN;
2383                 case -NFS4ERR_OLD_STATEID:
2384                         task->tk_status = 0;
2385                         return -EAGAIN;
2386         }
2387         task->tk_status = nfs4_map_errors(task->tk_status);
2388         return 0;
2389 }
2390
2391 static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs4_client *clp)
2392 {
2393         DEFINE_WAIT(wait);
2394         sigset_t oldset;
2395         int interruptible, res = 0;
2396
2397         might_sleep();
2398
2399         rpc_clnt_sigmask(clnt, &oldset);
2400         interruptible = TASK_UNINTERRUPTIBLE;
2401         if (clnt->cl_intr)
2402                 interruptible = TASK_INTERRUPTIBLE;
2403         prepare_to_wait(&clp->cl_waitq, &wait, interruptible);
2404         nfs4_schedule_state_recovery(clp);
2405         if (clnt->cl_intr && signalled())
2406                 res = -ERESTARTSYS;
2407         else if (!test_bit(NFS4CLNT_OK, &clp->cl_state))
2408                 schedule();
2409         finish_wait(&clp->cl_waitq, &wait);
2410         rpc_clnt_sigunmask(clnt, &oldset);
2411         return res;
2412 }
2413
2414 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
2415 {
2416         sigset_t oldset;
2417         int res = 0;
2418
2419         might_sleep();
2420
2421         if (*timeout <= 0)
2422                 *timeout = NFS4_POLL_RETRY_MIN;
2423         if (*timeout > NFS4_POLL_RETRY_MAX)
2424                 *timeout = NFS4_POLL_RETRY_MAX;
2425         rpc_clnt_sigmask(clnt, &oldset);
2426         if (clnt->cl_intr) {
2427                 set_current_state(TASK_INTERRUPTIBLE);
2428                 schedule_timeout(*timeout);
2429                 if (signalled())
2430                         res = -ERESTARTSYS;
2431         } else {
2432                 set_current_state(TASK_UNINTERRUPTIBLE);
2433                 schedule_timeout(*timeout);
2434         }
2435         rpc_clnt_sigunmask(clnt, &oldset);
2436         *timeout <<= 1;
2437         return res;
2438 }
2439
2440 /* This is the error handling routine for processes that are allowed
2441  * to sleep.
2442  */
2443 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
2444 {
2445         struct nfs4_client *clp = server->nfs4_state;
2446         int ret = errorcode;
2447
2448         exception->retry = 0;
2449         switch(errorcode) {
2450                 case 0:
2451                         return 0;
2452                 case -NFS4ERR_STALE_CLIENTID:
2453                 case -NFS4ERR_STALE_STATEID:
2454                 case -NFS4ERR_EXPIRED:
2455                         ret = nfs4_wait_clnt_recover(server->client, clp);
2456                         if (ret == 0)
2457                                 exception->retry = 1;
2458                         break;
2459                 case -NFS4ERR_GRACE:
2460                 case -NFS4ERR_DELAY:
2461                         ret = nfs4_delay(server->client, &exception->timeout);
2462                         if (ret == 0)
2463                                 exception->retry = 1;
2464                         break;
2465                 case -NFS4ERR_OLD_STATEID:
2466                         if (ret == 0)
2467                                 exception->retry = 1;
2468         }
2469         /* We failed to handle the error */
2470         return nfs4_map_errors(ret);
2471 }
2472
2473 int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short port)
2474 {
2475         nfs4_verifier sc_verifier;
2476         struct nfs4_setclientid setclientid = {
2477                 .sc_verifier = &sc_verifier,
2478                 .sc_prog = program,
2479         };
2480         struct rpc_message msg = {
2481                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
2482                 .rpc_argp = &setclientid,
2483                 .rpc_resp = clp,
2484                 .rpc_cred = clp->cl_cred,
2485         };
2486         u32 *p;
2487         int loop = 0;
2488         int status;
2489
2490         p = (u32*)sc_verifier.data;
2491         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
2492         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
2493
2494         for(;;) {
2495                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
2496                                 sizeof(setclientid.sc_name), "%s/%u.%u.%u.%u %s %u",
2497                                 clp->cl_ipaddr, NIPQUAD(clp->cl_addr.s_addr),
2498                                 clp->cl_cred->cr_ops->cr_name,
2499                                 clp->cl_id_uniquifier);
2500                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
2501                                 sizeof(setclientid.sc_netid), "tcp");
2502                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
2503                                 sizeof(setclientid.sc_uaddr), "%s.%d.%d",
2504                                 clp->cl_ipaddr, port >> 8, port & 255);
2505
2506                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2507                 if (status != -NFS4ERR_CLID_INUSE)
2508                         break;
2509                 if (signalled())
2510                         break;
2511                 if (loop++ & 1)
2512                         ssleep(clp->cl_lease_time + 1);
2513                 else
2514                         if (++clp->cl_id_uniquifier == 0)
2515                                 break;
2516         }
2517         return status;
2518 }
2519
2520 int
2521 nfs4_proc_setclientid_confirm(struct nfs4_client *clp)
2522 {
2523         struct nfs_fsinfo fsinfo;
2524         struct rpc_message msg = {
2525                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
2526                 .rpc_argp = clp,
2527                 .rpc_resp = &fsinfo,
2528                 .rpc_cred = clp->cl_cred,
2529         };
2530         unsigned long now;
2531         int status;
2532
2533         now = jiffies;
2534         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2535         if (status == 0) {
2536                 spin_lock(&clp->cl_lock);
2537                 clp->cl_lease_time = fsinfo.lease_time * HZ;
2538                 clp->cl_last_renewal = now;
2539                 spin_unlock(&clp->cl_lock);
2540         }
2541         return status;
2542 }
2543
2544 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid)
2545 {
2546         struct nfs4_delegreturnargs args = {
2547                 .fhandle = NFS_FH(inode),
2548                 .stateid = stateid,
2549         };
2550         struct rpc_message msg = {
2551                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
2552                 .rpc_argp = &args,
2553                 .rpc_cred = cred,
2554         };
2555
2556         return rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
2557 }
2558
2559 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid)
2560 {
2561         struct nfs_server *server = NFS_SERVER(inode);
2562         struct nfs4_exception exception = { };
2563         int err;
2564         do {
2565                 err = _nfs4_proc_delegreturn(inode, cred, stateid);
2566                 switch (err) {
2567                         case -NFS4ERR_STALE_STATEID:
2568                         case -NFS4ERR_EXPIRED:
2569                                 nfs4_schedule_state_recovery(server->nfs4_state);
2570                         case 0:
2571                                 return 0;
2572                 }
2573                 err = nfs4_handle_exception(server, err, &exception);
2574         } while (exception.retry);
2575         return err;
2576 }
2577
2578 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
2579 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
2580
2581 /* 
2582  * sleep, with exponential backoff, and retry the LOCK operation. 
2583  */
2584 static unsigned long
2585 nfs4_set_lock_task_retry(unsigned long timeout)
2586 {
2587         current->state = TASK_INTERRUPTIBLE;
2588         schedule_timeout(timeout);
2589         timeout <<= 1;
2590         if (timeout > NFS4_LOCK_MAXTIMEOUT)
2591                 return NFS4_LOCK_MAXTIMEOUT;
2592         return timeout;
2593 }
2594
2595 static inline int
2596 nfs4_lck_type(int cmd, struct file_lock *request)
2597 {
2598         /* set lock type */
2599         switch (request->fl_type) {
2600                 case F_RDLCK:
2601                         return IS_SETLKW(cmd) ? NFS4_READW_LT : NFS4_READ_LT;
2602                 case F_WRLCK:
2603                         return IS_SETLKW(cmd) ? NFS4_WRITEW_LT : NFS4_WRITE_LT;
2604                 case F_UNLCK:
2605                         return NFS4_WRITE_LT; 
2606         }
2607         BUG();
2608         return 0;
2609 }
2610
2611 static inline uint64_t
2612 nfs4_lck_length(struct file_lock *request)
2613 {
2614         if (request->fl_end == OFFSET_MAX)
2615                 return ~(uint64_t)0;
2616         return request->fl_end - request->fl_start + 1;
2617 }
2618
2619 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2620 {
2621         struct inode *inode = state->inode;
2622         struct nfs_server *server = NFS_SERVER(inode);
2623         struct nfs4_client *clp = server->nfs4_state;
2624         struct nfs_lockargs arg = {
2625                 .fh = NFS_FH(inode),
2626                 .type = nfs4_lck_type(cmd, request),
2627                 .offset = request->fl_start,
2628                 .length = nfs4_lck_length(request),
2629         };
2630         struct nfs_lockres res = {
2631                 .server = server,
2632         };
2633         struct rpc_message msg = {
2634                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
2635                 .rpc_argp       = &arg,
2636                 .rpc_resp       = &res,
2637                 .rpc_cred       = state->owner->so_cred,
2638         };
2639         struct nfs_lowner nlo;
2640         struct nfs4_lock_state *lsp;
2641         int status;
2642
2643         down_read(&clp->cl_sem);
2644         nlo.clientid = clp->cl_clientid;
2645         down(&state->lock_sema);
2646         lsp = nfs4_find_lock_state(state, request->fl_owner);
2647         if (lsp)
2648                 nlo.id = lsp->ls_id; 
2649         else {
2650                 spin_lock(&clp->cl_lock);
2651                 nlo.id = nfs4_alloc_lockowner_id(clp);
2652                 spin_unlock(&clp->cl_lock);
2653         }
2654         arg.u.lockt = &nlo;
2655         status = rpc_call_sync(server->client, &msg, 0);
2656         if (!status) {
2657                 request->fl_type = F_UNLCK;
2658         } else if (status == -NFS4ERR_DENIED) {
2659                 int64_t len, start, end;
2660                 start = res.u.denied.offset;
2661                 len = res.u.denied.length;
2662                 end = start + len - 1;
2663                 if (end < 0 || len == 0)
2664                         request->fl_end = OFFSET_MAX;
2665                 else
2666                         request->fl_end = (loff_t)end;
2667                 request->fl_start = (loff_t)start;
2668                 request->fl_type = F_WRLCK;
2669                 if (res.u.denied.type & 1)
2670                         request->fl_type = F_RDLCK;
2671                 request->fl_pid = 0;
2672                 status = 0;
2673         }
2674         if (lsp)
2675                 nfs4_put_lock_state(lsp);
2676         up(&state->lock_sema);
2677         up_read(&clp->cl_sem);
2678         return status;
2679 }
2680
2681 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2682 {
2683         struct nfs4_exception exception = { };
2684         int err;
2685
2686         do {
2687                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2688                                 _nfs4_proc_getlk(state, cmd, request),
2689                                 &exception);
2690         } while (exception.retry);
2691         return err;
2692 }
2693
2694 static int do_vfs_lock(struct file *file, struct file_lock *fl)
2695 {
2696         int res = 0;
2697         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
2698                 case FL_POSIX:
2699                         res = posix_lock_file_wait(file, fl);
2700                         break;
2701                 case FL_FLOCK:
2702                         res = flock_lock_file_wait(file, fl);
2703                         break;
2704                 default:
2705                         BUG();
2706         }
2707         if (res < 0)
2708                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n",
2709                                 __FUNCTION__);
2710         return res;
2711 }
2712
2713 static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2714 {
2715         struct inode *inode = state->inode;
2716         struct nfs_server *server = NFS_SERVER(inode);
2717         struct nfs4_client *clp = server->nfs4_state;
2718         struct nfs_lockargs arg = {
2719                 .fh = NFS_FH(inode),
2720                 .type = nfs4_lck_type(cmd, request),
2721                 .offset = request->fl_start,
2722                 .length = nfs4_lck_length(request),
2723         };
2724         struct nfs_lockres res = {
2725                 .server = server,
2726         };
2727         struct rpc_message msg = {
2728                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
2729                 .rpc_argp       = &arg,
2730                 .rpc_resp       = &res,
2731                 .rpc_cred       = state->owner->so_cred,
2732         };
2733         struct nfs4_lock_state *lsp;
2734         struct nfs_locku_opargs luargs;
2735         int status = 0;
2736                         
2737         down_read(&clp->cl_sem);
2738         down(&state->lock_sema);
2739         lsp = nfs4_find_lock_state(state, request->fl_owner);
2740         if (!lsp)
2741                 goto out;
2742         /* We might have lost the locks! */
2743         if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) {
2744                 luargs.seqid = lsp->ls_seqid;
2745                 memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid));
2746                 arg.u.locku = &luargs;
2747                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2748                 nfs4_increment_lock_seqid(status, lsp);
2749         }
2750
2751         if (status == 0) {
2752                 memcpy(&lsp->ls_stateid,  &res.u.stateid, 
2753                                 sizeof(lsp->ls_stateid));
2754                 nfs4_notify_unlck(state, request, lsp);
2755         }
2756         nfs4_put_lock_state(lsp);
2757 out:
2758         up(&state->lock_sema);
2759         if (status == 0)
2760                 do_vfs_lock(request->fl_file, request);
2761         up_read(&clp->cl_sem);
2762         return status;
2763 }
2764
2765 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2766 {
2767         struct nfs4_exception exception = { };
2768         int err;
2769
2770         do {
2771                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2772                                 _nfs4_proc_unlck(state, cmd, request),
2773                                 &exception);
2774         } while (exception.retry);
2775         return err;
2776 }
2777
2778 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim)
2779 {
2780         struct inode *inode = state->inode;
2781         struct nfs_server *server = NFS_SERVER(inode);
2782         struct nfs4_lock_state *lsp;
2783         struct nfs_lockargs arg = {
2784                 .fh = NFS_FH(inode),
2785                 .type = nfs4_lck_type(cmd, request),
2786                 .offset = request->fl_start,
2787                 .length = nfs4_lck_length(request),
2788         };
2789         struct nfs_lockres res = {
2790                 .server = server,
2791         };
2792         struct rpc_message msg = {
2793                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
2794                 .rpc_argp       = &arg,
2795                 .rpc_resp       = &res,
2796                 .rpc_cred       = state->owner->so_cred,
2797         };
2798         struct nfs_lock_opargs largs = {
2799                 .reclaim = reclaim,
2800                 .new_lock_owner = 0,
2801         };
2802         int status;
2803
2804         lsp = nfs4_get_lock_state(state, request->fl_owner);
2805         if (lsp == NULL)
2806                 return -ENOMEM;
2807         if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) {
2808                 struct nfs4_state_owner *owner = state->owner;
2809                 struct nfs_open_to_lock otl = {
2810                         .lock_owner = {
2811                                 .clientid = server->nfs4_state->cl_clientid,
2812                         },
2813                 };
2814
2815                 otl.lock_seqid = lsp->ls_seqid;
2816                 otl.lock_owner.id = lsp->ls_id;
2817                 memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid));
2818                 largs.u.open_lock = &otl;
2819                 largs.new_lock_owner = 1;
2820                 arg.u.lock = &largs;
2821                 down(&owner->so_sema);
2822                 otl.open_seqid = owner->so_seqid;
2823                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2824                 /* increment open_owner seqid on success, and 
2825                 * seqid mutating errors */
2826                 nfs4_increment_seqid(status, owner);
2827                 up(&owner->so_sema);
2828         } else {
2829                 struct nfs_exist_lock el = {
2830                         .seqid = lsp->ls_seqid,
2831                 };
2832                 memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid));
2833                 largs.u.exist_lock = &el;
2834                 largs.new_lock_owner = 0;
2835                 arg.u.lock = &largs;
2836                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2837         }
2838         /* increment seqid on success, and * seqid mutating errors*/
2839         nfs4_increment_lock_seqid(status, lsp);
2840         /* save the returned stateid. */
2841         if (status == 0) {
2842                 memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid));
2843                 lsp->ls_flags |= NFS_LOCK_INITIALIZED;
2844                 if (!reclaim)
2845                         nfs4_notify_setlk(state, request, lsp);
2846         } else if (status == -NFS4ERR_DENIED)
2847                 status = -EAGAIN;
2848         nfs4_put_lock_state(lsp);
2849         return status;
2850 }
2851
2852 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
2853 {
2854         struct nfs_server *server = NFS_SERVER(state->inode);
2855         struct nfs4_exception exception = { };
2856         int err;
2857
2858         do {
2859                 err = _nfs4_do_setlk(state, F_SETLK, request, 1);
2860                 if (err != -NFS4ERR_DELAY)
2861                         break;
2862                 nfs4_handle_exception(server, err, &exception);
2863         } while (exception.retry);
2864         return err;
2865 }
2866
2867 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
2868 {
2869         struct nfs_server *server = NFS_SERVER(state->inode);
2870         struct nfs4_exception exception = { };
2871         int err;
2872
2873         do {
2874                 err = _nfs4_do_setlk(state, F_SETLK, request, 0);
2875                 if (err != -NFS4ERR_DELAY)
2876                         break;
2877                 nfs4_handle_exception(server, err, &exception);
2878         } while (exception.retry);
2879         return err;
2880 }
2881
2882 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2883 {
2884         struct nfs4_client *clp = state->owner->so_client;
2885         int status;
2886
2887         down_read(&clp->cl_sem);
2888         down(&state->lock_sema);
2889         status = _nfs4_do_setlk(state, cmd, request, 0);
2890         up(&state->lock_sema);
2891         if (status == 0) {
2892                 /* Note: we always want to sleep here! */
2893                 request->fl_flags |= FL_SLEEP;
2894                 if (do_vfs_lock(request->fl_file, request) < 0)
2895                         printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __FUNCTION__);
2896         }
2897         up_read(&clp->cl_sem);
2898         return status;
2899 }
2900
2901 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2902 {
2903         struct nfs4_exception exception = { };
2904         int err;
2905
2906         do {
2907                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2908                                 _nfs4_proc_setlk(state, cmd, request),
2909                                 &exception);
2910         } while (exception.retry);
2911         return err;
2912 }
2913
2914 static int
2915 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
2916 {
2917         struct nfs_open_context *ctx;
2918         struct nfs4_state *state;
2919         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
2920         int status;
2921
2922         /* verify open state */
2923         ctx = (struct nfs_open_context *)filp->private_data;
2924         state = ctx->state;
2925
2926         if (request->fl_start < 0 || request->fl_end < 0)
2927                 return -EINVAL;
2928
2929         if (IS_GETLK(cmd))
2930                 return nfs4_proc_getlk(state, F_GETLK, request);
2931
2932         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
2933                 return -EINVAL;
2934
2935         if (request->fl_type == F_UNLCK)
2936                 return nfs4_proc_unlck(state, cmd, request);
2937
2938         do {
2939                 status = nfs4_proc_setlk(state, cmd, request);
2940                 if ((status != -EAGAIN) || IS_SETLK(cmd))
2941                         break;
2942                 timeout = nfs4_set_lock_task_retry(timeout);
2943                 status = -ERESTARTSYS;
2944                 if (signalled())
2945                         break;
2946         } while(status < 0);
2947
2948         return status;
2949 }
2950
2951
2952 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
2953
2954 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
2955                 size_t buflen, int flags)
2956 {
2957         struct inode *inode = dentry->d_inode;
2958
2959         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
2960                 return -EOPNOTSUPP;
2961
2962         if (!S_ISREG(inode->i_mode) &&
2963             (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX))
2964                 return -EPERM;
2965
2966         return nfs4_proc_set_acl(inode, buf, buflen);
2967 }
2968
2969 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
2970  * and that's what we'll do for e.g. user attributes that haven't been set.
2971  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
2972  * attributes in kernel-managed attribute namespaces. */
2973 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
2974                 size_t buflen)
2975 {
2976         struct inode *inode = dentry->d_inode;
2977
2978         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
2979                 return -EOPNOTSUPP;
2980
2981         return nfs4_proc_get_acl(inode, buf, buflen);
2982 }
2983
2984 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
2985 {
2986         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
2987
2988         if (buf && buflen < len)
2989                 return -ERANGE;
2990         if (buf)
2991                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
2992         return len;
2993 }
2994
2995 struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops = {
2996         .recover_open   = nfs4_open_reclaim,
2997         .recover_lock   = nfs4_lock_reclaim,
2998 };
2999
3000 struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops = {
3001         .recover_open   = nfs4_open_expired,
3002         .recover_lock   = nfs4_lock_expired,
3003 };
3004
3005 static struct inode_operations nfs4_file_inode_operations = {
3006         .permission     = nfs_permission,
3007         .getattr        = nfs_getattr,
3008         .setattr        = nfs_setattr,
3009         .getxattr       = nfs4_getxattr,
3010         .setxattr       = nfs4_setxattr,
3011         .listxattr      = nfs4_listxattr,
3012 };
3013
3014 struct nfs_rpc_ops      nfs_v4_clientops = {
3015         .version        = 4,                    /* protocol version */
3016         .dentry_ops     = &nfs4_dentry_operations,
3017         .dir_inode_ops  = &nfs4_dir_inode_operations,
3018         .file_inode_ops = &nfs4_file_inode_operations,
3019         .getroot        = nfs4_proc_get_root,
3020         .getattr        = nfs4_proc_getattr,
3021         .setattr        = nfs4_proc_setattr,
3022         .lookup         = nfs4_proc_lookup,
3023         .access         = nfs4_proc_access,
3024         .readlink       = nfs4_proc_readlink,
3025         .read           = nfs4_proc_read,
3026         .write          = nfs4_proc_write,
3027         .commit         = nfs4_proc_commit,
3028         .create         = nfs4_proc_create,
3029         .remove         = nfs4_proc_remove,
3030         .unlink_setup   = nfs4_proc_unlink_setup,
3031         .unlink_done    = nfs4_proc_unlink_done,
3032         .rename         = nfs4_proc_rename,
3033         .link           = nfs4_proc_link,
3034         .symlink        = nfs4_proc_symlink,
3035         .mkdir          = nfs4_proc_mkdir,
3036         .rmdir          = nfs4_proc_remove,
3037         .readdir        = nfs4_proc_readdir,
3038         .mknod          = nfs4_proc_mknod,
3039         .statfs         = nfs4_proc_statfs,
3040         .fsinfo         = nfs4_proc_fsinfo,
3041         .pathconf       = nfs4_proc_pathconf,
3042         .decode_dirent  = nfs4_decode_dirent,
3043         .read_setup     = nfs4_proc_read_setup,
3044         .write_setup    = nfs4_proc_write_setup,
3045         .commit_setup   = nfs4_proc_commit_setup,
3046         .file_open      = nfs4_proc_file_open,
3047         .file_release   = nfs4_proc_file_release,
3048         .lock           = nfs4_proc_lock,
3049         .clear_acl_cache = nfs4_zap_acl_attr,
3050 };
3051
3052 /*
3053  * Local variables:
3054  *  c-basic-offset: 8
3055  * End:
3056  */