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