fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/utsname.h>
  40 #include <linux/delay.h>
  41 #include <linux/errno.h>
  42 #include <linux/string.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/nfs.h>
  45 #include <linux/nfs4.h>
  46 #include <linux/nfs_fs.h>
  47 #include <linux/nfs_page.h>
  48 #include <linux/smp_lock.h>
  49 #include <linux/namei.h>
  50
  51 #include "nfs4_fs.h"
  52 #include "delegation.h"
  53
  54 #define NFSDBG_FACILITY         NFSDBG_PROC
  55
  56 #define NFS4_POLL_RETRY_MIN     (1*HZ)
  57 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  58
  59 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  60 static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *);
  61 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
  62 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
  63 extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus);
  64 extern struct rpc_procinfo nfs4_procedures[];
  65
  66 /* Prevent leaks of NFSv4 errors into userland */
  67 int nfs4_map_errors(int err)
  68 {
  69         if (err < -1000) {
  70                 dprintk("%s could not handle NFSv4 error %d\n",
  71                                 __FUNCTION__, -err);
  72                 return -EIO;
  73         }
  74         return err;
  75 }
  76
  77 /*
  78  * This is our standard bitmap for GETATTR requests.
  79  */
  80 const u32 nfs4_fattr_bitmap[2] = {
  81         FATTR4_WORD0_TYPE
  82         | FATTR4_WORD0_CHANGE
  83         | FATTR4_WORD0_SIZE
  84         | FATTR4_WORD0_FSID
  85         | FATTR4_WORD0_FILEID,
  86         FATTR4_WORD1_MODE
  87         | FATTR4_WORD1_NUMLINKS
  88         | FATTR4_WORD1_OWNER
  89         | FATTR4_WORD1_OWNER_GROUP
  90         | FATTR4_WORD1_RAWDEV
  91         | FATTR4_WORD1_SPACE_USED
  92         | FATTR4_WORD1_TIME_ACCESS
  93         | FATTR4_WORD1_TIME_METADATA
  94         | FATTR4_WORD1_TIME_MODIFY
  95 };
  96
  97 const u32 nfs4_statfs_bitmap[2] = {
  98         FATTR4_WORD0_FILES_AVAIL
  99         | FATTR4_WORD0_FILES_FREE
 100         | FATTR4_WORD0_FILES_TOTAL,
 101         FATTR4_WORD1_SPACE_AVAIL
 102         | FATTR4_WORD1_SPACE_FREE
 103         | FATTR4_WORD1_SPACE_TOTAL
 104 };
 105
 106 const u32 nfs4_pathconf_bitmap[2] = {
 107         FATTR4_WORD0_MAXLINK
 108         | FATTR4_WORD0_MAXNAME,
 109         0
 110 };
 111
 112 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 113                         | FATTR4_WORD0_MAXREAD
 114                         | FATTR4_WORD0_MAXWRITE
 115                         | FATTR4_WORD0_LEASE_TIME,
 116                         0
 117 };
 118
 119 static void nfs4_setup_readdir(u64 cookie, u32 *verifier, struct dentry *dentry,
 120                 struct nfs4_readdir_arg *readdir)
 121 {
 122         u32 *start, *p;
 123
 124         BUG_ON(readdir->count < 80);
 125         if (cookie > 2) {
 126                 readdir->cookie = (cookie > 2) ? cookie : 0;
 127                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 128                 return;
 129         }
 130
 131         readdir->cookie = 0;
 132         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 133         if (cookie == 2)
 134                 return;
 135
 136         /*
 137          * NFSv4 servers do not return entries for '.' and '..'
 138          * Therefore, we fake these entries here.  We let '.'
 139          * have cookie 0 and '..' have cookie 1.  Note that
 140          * when talking to the server, we always send cookie 0
 141          * instead of 1 or 2.
 142          */
 143         start = p = (u32 *)kmap_atomic(*readdir->pages, KM_USER0);
 144
 145         if (cookie == 0) {
 146                 *p++ = xdr_one;                                  /* next */
 147                 *p++ = xdr_zero;                   /* cookie, first word */
 148                 *p++ = xdr_one;                   /* cookie, second word */
 149                 *p++ = xdr_one;                             /* entry len */
 150                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 151                 p++;
 152                 *p++ = xdr_one;                         /* bitmap length */
 153                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 154                 *p++ = htonl(8);              /* attribute buffer length */
 155                 p = xdr_encode_hyper(p, dentry->d_inode->i_ino);
 156         }
 157
 158         *p++ = xdr_one;                                  /* next */
 159         *p++ = xdr_zero;                   /* cookie, first word */
 160         *p++ = xdr_two;                   /* cookie, second word */
 161         *p++ = xdr_two;                             /* entry len */
 162         memcpy(p, "..\0\0", 4);                         /* entry */
 163         p++;
 164         *p++ = xdr_one;                         /* bitmap length */
 165         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 166         *p++ = htonl(8);              /* attribute buffer length */
 167         p = xdr_encode_hyper(p, dentry->d_parent->d_inode->i_ino);
 168
 169         readdir->pgbase = (char *)p - (char *)start;
 170         readdir->count -= readdir->pgbase;
 171         kunmap_atomic(start, KM_USER0);
 172 }
 173
 174 static void
 175 renew_lease(struct nfs_server *server, unsigned long timestamp)
 176 {
 177         struct nfs4_client *clp = server->nfs4_state;
 178         spin_lock(&clp->cl_lock);
 179         if (time_before(clp->cl_last_renewal,timestamp))
 180                 clp->cl_last_renewal = timestamp;
 181         spin_unlock(&clp->cl_lock);
 182 }
 183
 184 static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinfo)
 185 {
 186         struct nfs_inode *nfsi = NFS_I(inode);
 187
 188         if (cinfo->before == nfsi->change_attr && cinfo->atomic)
 189                 nfsi->change_attr = cinfo->after;
 190 }
 191
 192 static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
 193 {
 194         struct inode *inode = state->inode;
 195
 196         open_flags &= (FMODE_READ|FMODE_WRITE);
 197         /* Protect against nfs4_find_state() */
 198         spin_lock(&inode->i_lock);
 199         state->state |= open_flags;
 200         /* NB! List reordering - see the reclaim code for why.  */
 201         if ((open_flags & FMODE_WRITE) && 0 == state->nwriters++)
 202                 list_move(&state->open_states, &state->owner->so_states);
 203         if (open_flags & FMODE_READ)
 204                 state->nreaders++;
 205         memcpy(&state->stateid, stateid, sizeof(state->stateid));
 206         spin_unlock(&inode->i_lock);
 207 }
 208
 209 /*
 210  * OPEN_RECLAIM:
 211  *      reclaim state on the server after a reboot.
 212  *      Assumes caller is holding the sp->so_sem
 213  */
 214 static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
 215 {
 216         struct inode *inode = state->inode;
 217         struct nfs_server *server = NFS_SERVER(inode);
 218         struct nfs_delegation *delegation = NFS_I(inode)->delegation;
 219         struct nfs_openargs o_arg = {
 220                 .fh = NFS_FH(inode),
 221                 .seqid = sp->so_seqid,
 222                 .id = sp->so_id,
 223                 .open_flags = state->state,
 224                 .clientid = server->nfs4_state->cl_clientid,
 225                 .claim = NFS4_OPEN_CLAIM_PREVIOUS,
 226                 .bitmask = server->attr_bitmask,
 227         };
 228         struct nfs_openres o_res = {
 229                 .server = server,       /* Grrr */
 230         };
 231         struct rpc_message msg = {
 232                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
 233                 .rpc_argp       = &o_arg,
 234                 .rpc_resp       = &o_res,
 235                 .rpc_cred       = sp->so_cred,
 236         };
 237         int status;
 238
 239         if (delegation != NULL) {
 240                 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
 241                         memcpy(&state->stateid, &delegation->stateid,
 242                                         sizeof(state->stateid));
 243                         set_bit(NFS_DELEGATED_STATE, &state->flags);
 244                         return 0;
 245                 }
 246                 o_arg.u.delegation_type = delegation->type;
 247         }
 248         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
 249         nfs4_increment_seqid(status, sp);
 250         if (status == 0) {
 251                 memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
 252                 if (o_res.delegation_type != 0) {
 253                         nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
 254                         /* Did the server issue an immediate delegation recall? */
 255                         if (o_res.do_recall)
 256                                 nfs_async_inode_return_delegation(inode, &o_res.stateid);
 257                 }
 258         }
 259         clear_bit(NFS_DELEGATED_STATE, &state->flags);
 260         /* Ensure we update the inode attributes */
 261         NFS_CACHEINV(inode);
 262         return status;
 263 }
 264
 265 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
 266 {
 267         struct nfs_server *server = NFS_SERVER(state->inode);
 268         struct nfs4_exception exception = { };
 269         int err;
 270         do {
 271                 err = _nfs4_open_reclaim(sp, state);
 272                 switch (err) {
 273                         case 0:
 274                         case -NFS4ERR_STALE_CLIENTID:
 275                         case -NFS4ERR_STALE_STATEID:
 276                         case -NFS4ERR_EXPIRED:
 277                                 return err;
 278                 }
 279                 err = nfs4_handle_exception(server, err, &exception);
 280         } while (exception.retry);
 281         return err;
 282 }
 283
 284 static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
 285 {
 286         struct nfs4_state_owner  *sp  = state->owner;
 287         struct inode *inode = dentry->d_inode;
 288         struct nfs_server *server = NFS_SERVER(inode);
 289         struct dentry *parent = dget_parent(dentry);
 290         struct nfs_openargs arg = {
 291                 .fh = NFS_FH(parent->d_inode),
 292                 .clientid = server->nfs4_state->cl_clientid,
 293                 .name = &dentry->d_name,
 294                 .id = sp->so_id,
 295                 .server = server,
 296                 .bitmask = server->attr_bitmask,
 297                 .claim = NFS4_OPEN_CLAIM_DELEGATE_CUR,
 298         };
 299         struct nfs_openres res = {
 300                 .server = server,
 301         };
 302         struct  rpc_message msg = {
 303                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
 304                 .rpc_argp       = &arg,
 305                 .rpc_resp       = &res,
 306                 .rpc_cred       = sp->so_cred,
 307         };
 308         int status = 0;
 309
 310         down(&sp->so_sema);
 311         if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
 312                 goto out;
 313         if (state->state == 0)
 314                 goto out;
 315         arg.seqid = sp->so_seqid;
 316         arg.open_flags = state->state;
 317         memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data));
 318         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
 319         nfs4_increment_seqid(status, sp);
 320         if (status >= 0) {
 321                 memcpy(state->stateid.data, res.stateid.data,
 322                                 sizeof(state->stateid.data));
 323                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
 324         }
 325 out:
 326         up(&sp->so_sema);
 327         dput(parent);
 328         return status;
 329 }
 330
 331 int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
 332 {
 333         struct nfs4_exception exception = { };
 334         struct nfs_server *server = NFS_SERVER(dentry->d_inode);
 335         int err;
 336         do {
 337                 err = _nfs4_open_delegation_recall(dentry, state);
 338                 switch (err) {
 339                         case 0:
 340                                 return err;
 341                         case -NFS4ERR_STALE_CLIENTID:
 342                         case -NFS4ERR_STALE_STATEID:
 343                         case -NFS4ERR_EXPIRED:
 344                                 /* Don't recall a delegation if it was lost */
 345                                 nfs4_schedule_state_recovery(server->nfs4_state);
 346                                 return err;
 347                 }
 348                 err = nfs4_handle_exception(server, err, &exception);
 349         } while (exception.retry);
 350         return err;
 351 }
 352
 353 static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid)
 354 {
 355         struct nfs_open_confirmargs arg = {
 356                 .fh             = fh,
 357                 .seqid          = sp->so_seqid,
 358                 .stateid        = *stateid,
 359         };
 360         struct nfs_open_confirmres res;
 361         struct  rpc_message msg = {
 362                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
 363                 .rpc_argp       = &arg,
 364                 .rpc_resp       = &res,
 365                 .rpc_cred       = sp->so_cred,
 366         };
 367         int status;
 368
 369         status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR);
 370         nfs4_increment_seqid(status, sp);
 371         if (status >= 0)
 372                 memcpy(stateid, &res.stateid, sizeof(*stateid));
 373         return status;
 374 }
 375
 376 static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner  *sp, struct nfs_openargs *o_arg, struct nfs_openres *o_res)
 377 {
 378         struct nfs_server *server = NFS_SERVER(dir);
 379         struct rpc_message msg = {
 380                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
 381                 .rpc_argp = o_arg,
 382                 .rpc_resp = o_res,
 383                 .rpc_cred = sp->so_cred,
 384         };
 385         int status;
 386
 387         /* Update sequence id. The caller must serialize! */
 388         o_arg->seqid = sp->so_seqid;
 389         o_arg->id = sp->so_id;
 390         o_arg->clientid = sp->so_client->cl_clientid;
 391
 392         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
 393         nfs4_increment_seqid(status, sp);
 394         if (status != 0)
 395                 goto out;
 396         update_changeattr(dir, &o_res->cinfo);
 397         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
 398                 status = _nfs4_proc_open_confirm(server->client, &o_res->fh,
 399                                 sp, &o_res->stateid);
 400                 if (status != 0)
 401                         goto out;
 402         }
 403         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
 404                 status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
 405 out:
 406         return status;
 407 }
 408
 409 static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openflags)
 410 {
 411         struct nfs_access_entry cache;
 412         int mask = 0;
 413         int status;
 414
 415         if (openflags & FMODE_READ)
 416                 mask |= MAY_READ;
 417         if (openflags & FMODE_WRITE)
 418                 mask |= MAY_WRITE;
 419         status = nfs_access_get_cached(inode, cred, &cache);
 420         if (status == 0)
 421                 goto out;
 422
 423         /* Be clever: ask server to check for all possible rights */
 424         cache.mask = MAY_EXEC | MAY_WRITE | MAY_READ;
 425         cache.cred = cred;
 426         cache.jiffies = jiffies;
 427         status = _nfs4_proc_access(inode, &cache);
 428         if (status != 0)
 429                 return status;
 430         nfs_access_add_cache(inode, &cache);
 431 out:
 432         if ((cache.mask & mask) == mask)
 433                 return 0;
 434         return -EACCES;
 435 }
 436
 437 /*
 438  * OPEN_EXPIRED:
 439  *      reclaim state on the server after a network partition.
 440  *      Assumes caller holds the appropriate lock
 441  */
 442 static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
 443 {
 444         struct dentry *parent = dget_parent(dentry);
 445         struct inode *dir = parent->d_inode;
 446         struct inode *inode = state->inode;
 447         struct nfs_server *server = NFS_SERVER(dir);
 448         struct nfs_delegation *delegation = NFS_I(inode)->delegation;
 449         struct nfs_fattr        f_attr = {
 450                 .valid = 0,
 451         };
 452         struct nfs_openargs o_arg = {
 453                 .fh = NFS_FH(dir),
 454                 .open_flags = state->state,
 455                 .name = &dentry->d_name,
 456                 .bitmask = server->attr_bitmask,
 457                 .claim = NFS4_OPEN_CLAIM_NULL,
 458         };
 459         struct nfs_openres o_res = {
 460                 .f_attr = &f_attr,
 461                 .server = server,
 462         };
 463         int status = 0;
 464
 465         if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
 466                 status = _nfs4_do_access(inode, sp->so_cred, state->state);
 467                 if (status < 0)
 468                         goto out;
 469                 memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid));
 470                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 471                 goto out;
 472         }
 473         status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
 474         if (status != 0)
 475                 goto out_nodeleg;
 476         /* Check if files differ */
 477         if ((f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT))
 478                 goto out_stale;
 479         /* Has the file handle changed? */
 480         if (nfs_compare_fh(&o_res.fh, NFS_FH(inode)) != 0) {
 481                 /* Verify if the change attributes are the same */
 482                 if (f_attr.change_attr != NFS_I(inode)->change_attr)
 483                         goto out_stale;
 484                 if (nfs_size_to_loff_t(f_attr.size) != inode->i_size)
 485                         goto out_stale;
 486                 /* Lets just pretend that this is the same file */
 487                 nfs_copy_fh(NFS_FH(inode), &o_res.fh);
 488                 NFS_I(inode)->fileid = f_attr.fileid;
 489         }
 490         memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
 491         if (o_res.delegation_type != 0) {
 492                 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM))
 493                         nfs_inode_set_delegation(inode, sp->so_cred, &o_res);
 494                 else
 495                         nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
 496         }
 497 out_nodeleg:
 498         clear_bit(NFS_DELEGATED_STATE, &state->flags);
 499 out:
 500         dput(parent);
 501         return status;
 502 out_stale:
 503         status = -ESTALE;
 504         /* Invalidate the state owner so we don't ever use it again */
 505         nfs4_drop_state_owner(sp);
 506         d_drop(dentry);
 507         /* Should we be trying to close that stateid? */
 508         goto out_nodeleg;
 509 }
 510
 511 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
 512 {
 513         struct nfs_inode *nfsi = NFS_I(state->inode);
 514         struct nfs_open_context *ctx;
 515         int status;
 516
 517         spin_lock(&state->inode->i_lock);
 518         list_for_each_entry(ctx, &nfsi->open_files, list) {
 519                 if (ctx->state != state)
 520                         continue;
 521                 get_nfs_open_context(ctx);
 522                 spin_unlock(&state->inode->i_lock);
 523                 status = _nfs4_open_expired(sp, state, ctx->dentry);
 524                 put_nfs_open_context(ctx);
 525                 return status;
 526         }
 527         spin_unlock(&state->inode->i_lock);
 528         return -ENOENT;
 529 }
 530
 531 /*
 532  * Returns an nfs4_state + an extra reference to the inode
 533  */
 534 static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred, struct nfs4_state **res)
 535 {
 536         struct nfs_delegation *delegation;
 537         struct nfs_server *server = NFS_SERVER(inode);
 538         struct nfs4_client *clp = server->nfs4_state;
 539         struct nfs_inode *nfsi = NFS_I(inode);
 540         struct nfs4_state_owner *sp = NULL;
 541         struct nfs4_state *state = NULL;
 542         int open_flags = flags & (FMODE_READ|FMODE_WRITE);
 543         int err;
 544
 545         /* Protect against reboot recovery - NOTE ORDER! */
 546         down_read(&clp->cl_sem);
 547         /* Protect against delegation recall */
 548         down_read(&nfsi->rwsem);
 549         delegation = NFS_I(inode)->delegation;
 550         err = -ENOENT;
 551         if (delegation == NULL || (delegation->type & open_flags) != open_flags)
 552                 goto out_err;
 553         err = -ENOMEM;
 554         if (!(sp = nfs4_get_state_owner(server, cred))) {
 555                 dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
 556                 goto out_err;
 557         }
 558         down(&sp->so_sema);
 559         state = nfs4_get_open_state(inode, sp);
 560         if (state == NULL)
 561                 goto out_err;
 562
 563         err = -ENOENT;
 564         if ((state->state & open_flags) == open_flags) {
 565                 spin_lock(&inode->i_lock);
 566                 if (open_flags & FMODE_READ)
 567                         state->nreaders++;
 568                 if (open_flags & FMODE_WRITE)
 569                         state->nwriters++;
 570                 spin_unlock(&inode->i_lock);
 571                 goto out_ok;
 572         } else if (state->state != 0)
 573                 goto out_err;
 574
 575         lock_kernel();
 576         err = _nfs4_do_access(inode, cred, open_flags);
 577         unlock_kernel();
 578         if (err != 0)
 579                 goto out_err;
 580         set_bit(NFS_DELEGATED_STATE, &state->flags);
 581         update_open_stateid(state, &delegation->stateid, open_flags);
 582 out_ok:
 583         up(&sp->so_sema);
 584         nfs4_put_state_owner(sp);
 585         up_read(&nfsi->rwsem);
 586         up_read(&clp->cl_sem);
 587         igrab(inode);
 588         *res = state;
 589         return 0;
 590 out_err:
 591         if (sp != NULL) {
 592                 if (state != NULL)
 593                         nfs4_put_open_state(state);
 594                 up(&sp->so_sema);
 595                 nfs4_put_state_owner(sp);
 596         }
 597         up_read(&nfsi->rwsem);
 598         up_read(&clp->cl_sem);
 599         return err;
 600 }
 601
 602 static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred)
 603 {
 604         struct nfs4_exception exception = { };
 605         struct nfs4_state *res;
 606         int err;
 607
 608         do {
 609                 err = _nfs4_open_delegated(inode, flags, cred, &res);
 610                 if (err == 0)
 611                         break;
 612                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(inode),
 613                                         err, &exception));
 614         } while (exception.retry);
 615         return res;
 616 }
 617
 618 /*
 619  * Returns an nfs4_state + an referenced inode
 620  */
 621 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
 622 {
 623         struct nfs4_state_owner  *sp;
 624         struct nfs4_state     *state = NULL;
 625         struct nfs_server       *server = NFS_SERVER(dir);
 626         struct nfs4_client *clp = server->nfs4_state;
 627         struct inode *inode = NULL;
 628         int                     status;
 629         struct nfs_fattr        f_attr = {
 630                 .valid          = 0,
 631         };
 632         struct nfs_openargs o_arg = {
 633                 .fh             = NFS_FH(dir),
 634                 .open_flags     = flags,
 635                 .name           = &dentry->d_name,
 636                 .server         = server,
 637                 .bitmask = server->attr_bitmask,
 638                 .claim = NFS4_OPEN_CLAIM_NULL,
 639         };
 640         struct nfs_openres o_res = {
 641                 .f_attr         = &f_attr,
 642                 .server         = server,
 643         };
 644
 645         /* Protect against reboot recovery conflicts */
 646         down_read(&clp->cl_sem);
 647         status = -ENOMEM;
 648         if (!(sp = nfs4_get_state_owner(server, cred))) {
 649                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
 650                 goto out_err;
 651         }
 652         if (flags & O_EXCL) {
 653                 u32 *p = (u32 *) o_arg.u.verifier.data;
 654                 p[0] = jiffies;
 655                 p[1] = current->pid;
 656         } else
 657                 o_arg.u.attrs = sattr;
 658         /* Serialization for the sequence id */
 659         down(&sp->so_sema);
 660
 661         status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
 662         if (status != 0)
 663                 goto out_err;
 664
 665         status = -ENOMEM;
 666         inode = nfs_fhget(dir->i_sb, &o_res.fh, &f_attr);
 667         if (!inode)
 668                 goto out_err;
 669         state = nfs4_get_open_state(inode, sp);
 670         if (!state)
 671                 goto out_err;
 672         update_open_stateid(state, &o_res.stateid, flags);
 673         if (o_res.delegation_type != 0)
 674                 nfs_inode_set_delegation(inode, cred, &o_res);
 675         up(&sp->so_sema);
 676         nfs4_put_state_owner(sp);
 677         up_read(&clp->cl_sem);
 678         *res = state;
 679         return 0;
 680 out_err:
 681         if (sp != NULL) {
 682                 if (state != NULL)
 683                         nfs4_put_open_state(state);
 684                 up(&sp->so_sema);
 685                 nfs4_put_state_owner(sp);
 686         }
 687         /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */
 688         up_read(&clp->cl_sem);
 689         if (inode != NULL)
 690                 iput(inode);
 691         *res = NULL;
 692         return status;
 693 }
 694
 695
 696 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred)
 697 {
 698         struct nfs4_exception exception = { };
 699         struct nfs4_state *res;
 700         int status;
 701
 702         do {
 703                 status = _nfs4_do_open(dir, dentry, flags, sattr, cred, &res);
 704                 if (status == 0)
 705                         break;
 706                 /* NOTE: BAD_SEQID means the server and client disagree about the
 707                  * book-keeping w.r.t. state-changing operations
 708                  * (OPEN/CLOSE/LOCK/LOCKU...)
 709                  * It is actually a sign of a bug on the client or on the server.
 710                  *
 711                  * If we receive a BAD_SEQID error in the particular case of
 712                  * doing an OPEN, we assume that nfs4_increment_seqid() will
 713                  * have unhashed the old state_owner for us, and that we can
 714                  * therefore safely retry using a new one. We should still warn
 715                  * the user though...
 716                  */
 717                 if (status == -NFS4ERR_BAD_SEQID) {
 718                         printk(KERN_WARNING "NFS: v4 server returned a bad sequence-id error!\n");
 719                         exception.retry = 1;
 720                         continue;
 721                 }
 722                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
 723                                         status, &exception));
 724         } while (exception.retry);
 725         return res;
 726 }
 727
 728 static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
 729                 struct nfs_fh *fhandle, struct iattr *sattr,
 730                 struct nfs4_state *state)
 731 {
 732         struct nfs_setattrargs  arg = {
 733                 .fh             = fhandle,
 734                 .iap            = sattr,
 735                 .server         = server,
 736                 .bitmask = server->attr_bitmask,
 737         };
 738         struct nfs_setattrres  res = {
 739                 .fattr          = fattr,
 740                 .server         = server,
 741         };
 742         struct rpc_message msg = {
 743                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
 744                 .rpc_argp       = &arg,
 745                 .rpc_resp       = &res,
 746         };
 747
 748         fattr->valid = 0;
 749
 750         if (state != NULL)
 751                 msg.rpc_cred = state->owner->so_cred;
 752         if (sattr->ia_valid & ATTR_SIZE)
 753                 nfs4_copy_stateid(&arg.stateid, state, NULL);
 754         else
 755                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
 756
 757         return rpc_call_sync(server->client, &msg, 0);
 758 }
 759
 760 static int nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
 761                 struct nfs_fh *fhandle, struct iattr *sattr,
 762                 struct nfs4_state *state)
 763 {
 764         struct nfs4_exception exception = { };
 765         int err;
 766         do {
 767                 err = nfs4_handle_exception(server,
 768                                 _nfs4_do_setattr(server, fattr, fhandle, sattr,
 769                                         state),
 770                                 &exception);
 771         } while (exception.retry);
 772         return err;
 773 }
 774
 775 struct nfs4_closedata {
 776         struct inode *inode;
 777         struct nfs4_state *state;
 778         struct nfs_closeargs arg;
 779         struct nfs_closeres res;
 780 };
 781
 782 static void nfs4_close_done(struct rpc_task *task)
 783 {
 784         struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
 785         struct nfs4_state *state = calldata->state;
 786         struct nfs4_state_owner *sp = state->owner;
 787         struct nfs_server *server = NFS_SERVER(calldata->inode);
 788
 789         /* hmm. we are done with the inode, and in the process of freeing
 790          * the state_owner. we keep this around to process errors
 791          */
 792         nfs4_increment_seqid(task->tk_status, sp);
 793         switch (task->tk_status) {
 794                 case 0:
 795                         memcpy(&state->stateid, &calldata->res.stateid,
 796                                         sizeof(state->stateid));
 797                         break;
 798                 case -NFS4ERR_STALE_STATEID:
 799                 case -NFS4ERR_EXPIRED:
 800                         state->state = calldata->arg.open_flags;
 801                         nfs4_schedule_state_recovery(server->nfs4_state);
 802                         break;
 803                 default:
 804                         if (nfs4_async_handle_error(task, server) == -EAGAIN) {
 805                                 rpc_restart_call(task);
 806                                 return;
 807                         }
 808         }
 809         state->state = calldata->arg.open_flags;
 810         nfs4_put_open_state(state);
 811         up(&sp->so_sema);
 812         nfs4_put_state_owner(sp);
 813         up_read(&server->nfs4_state->cl_sem);
 814         kfree(calldata);
 815 }
 816
 817 static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata)
 818 {
 819         struct rpc_message msg = {
 820                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
 821                 .rpc_argp = &calldata->arg,
 822                 .rpc_resp = &calldata->res,
 823                 .rpc_cred = calldata->state->owner->so_cred,
 824         };
 825         if (calldata->arg.open_flags != 0)
 826                 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
 827         return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata);
 828 }
 829
 830 /*
 831  * It is possible for data to be read/written from a mem-mapped file
 832  * after the sys_close call (which hits the vfs layer as a flush).
 833  * This means that we can't safely call nfsv4 close on a file until
 834  * the inode is cleared. This in turn means that we are not good
 835  * NFSv4 citizens - we do not indicate to the server to update the file's
 836  * share state even when we are done with one of the three share
 837  * stateid's in the inode.
 838  *
 839  * NOTE: Caller must be holding the sp->so_owner semaphore!
 840  */
 841 int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode)
 842 {
 843         struct nfs4_closedata *calldata;
 844         int status;
 845
 846         /* Tell caller we're done */
 847         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
 848                 state->state = mode;
 849                 return 0;
 850         }
 851         calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL);
 852         if (calldata == NULL)
 853                 return -ENOMEM;
 854         calldata->inode = inode;
 855         calldata->state = state;
 856         calldata->arg.fh = NFS_FH(inode);
 857         /* Serialization for the sequence id */
 858         calldata->arg.seqid = state->owner->so_seqid;
 859         calldata->arg.open_flags = mode;
 860         memcpy(&calldata->arg.stateid, &state->stateid,
 861                         sizeof(calldata->arg.stateid));
 862         status = nfs4_close_call(NFS_SERVER(inode)->client, calldata);
 863         /*
 864          * Return -EINPROGRESS on success in order to indicate to the
 865          * caller that an asynchronous RPC call has been launched, and
 866          * that it will release the semaphores on completion.
 867          */
 868         return (status == 0) ? -EINPROGRESS : status;
 869 }
 870
 871 struct inode *
 872 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
 873 {
 874         struct iattr attr;
 875         struct rpc_cred *cred;
 876         struct nfs4_state *state;
 877
 878         if (nd->flags & LOOKUP_CREATE) {
 879                 attr.ia_mode = nd->intent.open.create_mode;
 880                 attr.ia_valid = ATTR_MODE;
 881                 if (!IS_POSIXACL(dir))
 882                         attr.ia_mode &= ~current->fs->umask;
 883         } else {
 884                 attr.ia_valid = 0;
 885                 BUG_ON(nd->intent.open.flags & O_CREAT);
 886         }
 887
 888         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
 889         if (IS_ERR(cred))
 890                 return (struct inode *)cred;
 891         state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred);
 892         put_rpccred(cred);
 893         if (IS_ERR(state))
 894                 return (struct inode *)state;
 895         return state->inode;
 896 }
 897
 898 int
 899 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags)
 900 {
 901         struct rpc_cred *cred;
 902         struct nfs4_state *state;
 903         struct inode *inode;
 904
 905         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
 906         if (IS_ERR(cred))
 907                 return PTR_ERR(cred);
 908         state = nfs4_open_delegated(dentry->d_inode, openflags, cred);
 909         if (IS_ERR(state))
 910                 state = nfs4_do_open(dir, dentry, openflags, NULL, cred);
 911         put_rpccred(cred);
 912         if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0)
 913                 return 1;
 914         if (IS_ERR(state))
 915                 return 0;
 916         inode = state->inode;
 917         if (inode == dentry->d_inode) {
 918                 iput(inode);
 919                 return 1;
 920         }
 921         d_drop(dentry);
 922         nfs4_close_state(state, openflags);
 923         iput(inode);
 924         return 0;
 925 }
 926
 927
 928 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
 929 {
 930         struct nfs4_server_caps_res res = {};
 931         struct rpc_message msg = {
 932                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
 933                 .rpc_argp = fhandle,
 934                 .rpc_resp = &res,
 935         };
 936         int status;
 937
 938         status = rpc_call_sync(server->client, &msg, 0);
 939         if (status == 0) {
 940                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
 941                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
 942                         server->caps |= NFS_CAP_ACLS;
 943                 if (res.has_links != 0)
 944                         server->caps |= NFS_CAP_HARDLINKS;
 945                 if (res.has_symlinks != 0)
 946                         server->caps |= NFS_CAP_SYMLINKS;
 947                 server->acl_bitmask = res.acl_bitmask;
 948         }
 949         return status;
 950 }
 951
 952 static int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
 953 {
 954         struct nfs4_exception exception = { };
 955         int err;
 956         do {
 957                 err = nfs4_handle_exception(server,
 958                                 _nfs4_server_capabilities(server, fhandle),
 959                                 &exception);
 960         } while (exception.retry);
 961         return err;
 962 }
 963
 964 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
 965                 struct nfs_fsinfo *info)
 966 {
 967         struct nfs_fattr *      fattr = info->fattr;
 968         struct nfs4_lookup_root_arg args = {
 969                 .bitmask = nfs4_fattr_bitmap,
 970         };
 971         struct nfs4_lookup_res res = {
 972                 .server = server,
 973                 .fattr = fattr,
 974                 .fh = fhandle,
 975         };
 976         struct rpc_message msg = {
 977                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
 978                 .rpc_argp = &args,
 979                 .rpc_resp = &res,
 980         };
 981         fattr->valid = 0;
 982         return rpc_call_sync(server->client, &msg, 0);
 983 }
 984
 985 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
 986                 struct nfs_fsinfo *info)
 987 {
 988         struct nfs4_exception exception = { };
 989         int err;
 990         do {
 991                 err = nfs4_handle_exception(server,
 992                                 _nfs4_lookup_root(server, fhandle, info),
 993                                 &exception);
 994         } while (exception.retry);
 995         return err;
 996 }
 997
 998 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
 999                 struct nfs_fsinfo *info)
1000 {
1001         struct nfs_fattr *      fattr = info->fattr;
1002         unsigned char *         p;
1003         struct qstr             q;
1004         struct nfs4_lookup_arg args = {
1005                 .dir_fh = fhandle,
1006                 .name = &q,
1007                 .bitmask = nfs4_fattr_bitmap,
1008         };
1009         struct nfs4_lookup_res res = {
1010                 .server = server,
1011                 .fattr = fattr,
1012                 .fh = fhandle,
1013         };
1014         struct rpc_message msg = {
1015                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1016                 .rpc_argp = &args,
1017                 .rpc_resp = &res,
1018         };
1019         int status;
1020
1021         /*
1022          * Now we do a separate LOOKUP for each component of the mount path.
1023          * The LOOKUPs are done separately so that we can conveniently
1024          * catch an ERR_WRONGSEC if it occurs along the way...
1025          */
1026         status = nfs4_lookup_root(server, fhandle, info);
1027         if (status)
1028                 goto out;
1029
1030         p = server->mnt_path;
1031         for (;;) {
1032                 struct nfs4_exception exception = { };
1033
1034                 while (*p == '/')
1035                         p++;
1036                 if (!*p)
1037                         break;
1038                 q.name = p;
1039                 while (*p && (*p != '/'))
1040                         p++;
1041                 q.len = p - q.name;
1042
1043                 do {
1044                         fattr->valid = 0;
1045                         status = nfs4_handle_exception(server,
1046                                         rpc_call_sync(server->client, &msg, 0),
1047                                         &exception);
1048                 } while (exception.retry);
1049                 if (status == 0)
1050                         continue;
1051                 if (status == -ENOENT) {
1052                         printk(KERN_NOTICE "NFS: mount path %s does not exist!\n", server->mnt_path);
1053                         printk(KERN_NOTICE "NFS: suggestion: try mounting '/' instead.\n");
1054                 }
1055                 break;
1056         }
1057         if (status == 0)
1058                 status = nfs4_server_capabilities(server, fhandle);
1059         if (status == 0)
1060                 status = nfs4_do_fsinfo(server, fhandle, info);
1061 out:
1062         return status;
1063 }
1064
1065 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1066 {
1067         struct nfs4_getattr_arg args = {
1068                 .fh = fhandle,
1069                 .bitmask = server->attr_bitmask,
1070         };
1071         struct nfs4_getattr_res res = {
1072                 .fattr = fattr,
1073                 .server = server,
1074         };
1075         struct rpc_message msg = {
1076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
1077                 .rpc_argp = &args,
1078                 .rpc_resp = &res,
1079         };
1080
1081         fattr->valid = 0;
1082         return rpc_call_sync(server->client, &msg, 0);
1083 }
1084
1085 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1086 {
1087         struct nfs4_exception exception = { };
1088         int err;
1089         do {
1090                 err = nfs4_handle_exception(server,
1091                                 _nfs4_proc_getattr(server, fhandle, fattr),
1092                                 &exception);
1093         } while (exception.retry);
1094         return err;
1095 }
1096
1097 /*
1098  * The file is not closed if it is opened due to the a request to change
1099  * the size of the file. The open call will not be needed once the
1100  * VFS layer lookup-intents are implemented.
1101  *
1102  * Close is called when the inode is destroyed.
1103  * If we haven't opened the file for O_WRONLY, we
1104  * need to in the size_change case to obtain a stateid.
1105  *
1106  * Got race?
1107  * Because OPEN is always done by name in nfsv4, it is
1108  * possible that we opened a different file by the same
1109  * name.  We can recognize this race condition, but we
1110  * can't do anything about it besides returning an error.
1111  *
1112  * This will be fixed with VFS changes (lookup-intent).
1113  */
1114 static int
1115 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
1116                   struct iattr *sattr)
1117 {
1118         struct inode *          inode = dentry->d_inode;
1119         int                     size_change = sattr->ia_valid & ATTR_SIZE;
1120         struct nfs4_state       *state = NULL;
1121         int need_iput = 0;
1122         int status;
1123
1124         fattr->valid = 0;
1125
1126         if (size_change) {
1127                 struct rpc_cred *cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
1128                 if (IS_ERR(cred))
1129                         return PTR_ERR(cred);
1130                 state = nfs4_find_state(inode, cred, FMODE_WRITE);
1131                 if (state == NULL) {
1132                         state = nfs4_open_delegated(dentry->d_inode,
1133                                         FMODE_WRITE, cred);
1134                         if (IS_ERR(state))
1135                                 state = nfs4_do_open(dentry->d_parent->d_inode,
1136                                                 dentry, FMODE_WRITE,
1137                                                 NULL, cred);
1138                         need_iput = 1;
1139                 }
1140                 put_rpccred(cred);
1141                 if (IS_ERR(state))
1142                         return PTR_ERR(state);
1143
1144                 if (state->inode != inode) {
1145                         printk(KERN_WARNING "nfs: raced in setattr (%p != %p), returning -EIO\n", inode, state->inode);
1146                         status = -EIO;
1147                         goto out;
1148                 }
1149         }
1150         status = nfs4_do_setattr(NFS_SERVER(inode), fattr,
1151                         NFS_FH(inode), sattr, state);
1152 out:
1153         if (state) {
1154                 inode = state->inode;
1155                 nfs4_close_state(state, FMODE_WRITE);
1156                 if (need_iput)
1157                         iput(inode);
1158         }
1159         return status;
1160 }
1161
1162 static int _nfs4_proc_lookup(struct inode *dir, struct qstr *name,
1163                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1164 {
1165         int                    status;
1166         struct nfs_server *server = NFS_SERVER(dir);
1167         struct nfs4_lookup_arg args = {
1168                 .bitmask = server->attr_bitmask,
1169                 .dir_fh = NFS_FH(dir),
1170                 .name = name,
1171         };
1172         struct nfs4_lookup_res res = {
1173                 .server = server,
1174                 .fattr = fattr,
1175                 .fh = fhandle,
1176         };
1177         struct rpc_message msg = {
1178                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1179                 .rpc_argp = &args,
1180                 .rpc_resp = &res,
1181         };
1182
1183         fattr->valid = 0;
1184
1185         dprintk("NFS call  lookup %s\n", name->name);
1186         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1187         dprintk("NFS reply lookup: %d\n", status);
1188         return status;
1189 }
1190
1191 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1192 {
1193         struct nfs4_exception exception = { };
1194         int err;
1195         do {
1196                 err = nfs4_handle_exception(NFS_SERVER(dir),
1197                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
1198                                 &exception);
1199         } while (exception.retry);
1200         return err;
1201 }
1202
1203 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
1204 {
1205         struct nfs4_accessargs args = {
1206                 .fh = NFS_FH(inode),
1207         };
1208         struct nfs4_accessres res = { 0 };
1209         struct rpc_message msg = {
1210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
1211                 .rpc_argp = &args,
1212                 .rpc_resp = &res,
1213                 .rpc_cred = entry->cred,
1214         };
1215         int mode = entry->mask;
1216         int status;
1217
1218         /*
1219          * Determine which access bits we want to ask for...
1220          */
1221         if (mode & MAY_READ)
1222                 args.access |= NFS4_ACCESS_READ;
1223         if (S_ISDIR(inode->i_mode)) {
1224                 if (mode & MAY_WRITE)
1225                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
1226                 if (mode & MAY_EXEC)
1227                         args.access |= NFS4_ACCESS_LOOKUP;
1228         } else {
1229                 if (mode & MAY_WRITE)
1230                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
1231                 if (mode & MAY_EXEC)
1232                         args.access |= NFS4_ACCESS_EXECUTE;
1233         }
1234         status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1235         if (!status) {
1236                 entry->mask = 0;
1237                 if (res.access & NFS4_ACCESS_READ)
1238                         entry->mask |= MAY_READ;
1239                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
1240                         entry->mask |= MAY_WRITE;
1241                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
1242                         entry->mask |= MAY_EXEC;
1243         }
1244         return status;
1245 }
1246
1247 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
1248 {
1249         struct nfs4_exception exception = { };
1250         int err;
1251         do {
1252                 err = nfs4_handle_exception(NFS_SERVER(inode),
1253                                 _nfs4_proc_access(inode, entry),
1254                                 &exception);
1255         } while (exception.retry);
1256         return err;
1257 }
1258
1259 /*
1260  * TODO: For the time being, we don't try to get any attributes
1261  * along with any of the zero-copy operations READ, READDIR,
1262  * READLINK, WRITE.
1263  *
1264  * In the case of the first three, we want to put the GETATTR
1265  * after the read-type operation -- this is because it is hard
1266  * to predict the length of a GETATTR response in v4, and thus
1267  * align the READ data correctly.  This means that the GETATTR
1268  * may end up partially falling into the page cache, and we should
1269  * shift it into the 'tail' of the xdr_buf before processing.
1270  * To do this efficiently, we need to know the total length
1271  * of data received, which doesn't seem to be available outside
1272  * of the RPC layer.
1273  *
1274  * In the case of WRITE, we also want to put the GETATTR after
1275  * the operation -- in this case because we want to make sure
1276  * we get the post-operation mtime and size.  This means that
1277  * we can't use xdr_encode_pages() as written: we need a variant
1278  * of it which would leave room in the 'tail' iovec.
1279  *
1280  * Both of these changes to the XDR layer would in fact be quite
1281  * minor, but I decided to leave them for a subsequent patch.
1282  */
1283 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
1284                 unsigned int pgbase, unsigned int pglen)
1285 {
1286         struct nfs4_readlink args = {
1287                 .fh       = NFS_FH(inode),
1288                 .pgbase   = pgbase,
1289                 .pglen    = pglen,
1290                 .pages    = &page,
1291         };
1292         struct rpc_message msg = {
1293                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
1294                 .rpc_argp = &args,
1295                 .rpc_resp = NULL,
1296         };
1297
1298         return rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1299 }
1300
1301 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
1302                 unsigned int pgbase, unsigned int pglen)
1303 {
1304         struct nfs4_exception exception = { };
1305         int err;
1306         do {
1307                 err = nfs4_handle_exception(NFS_SERVER(inode),
1308                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
1309                                 &exception);
1310         } while (exception.retry);
1311         return err;
1312 }
1313
1314 static int _nfs4_proc_read(struct nfs_read_data *rdata)
1315 {
1316         int flags = rdata->flags;
1317         struct inode *inode = rdata->inode;
1318         struct nfs_fattr *fattr = rdata->res.fattr;
1319         struct nfs_server *server = NFS_SERVER(inode);
1320         struct rpc_message msg = {
1321                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_READ],
1322                 .rpc_argp       = &rdata->args,
1323                 .rpc_resp       = &rdata->res,
1324                 .rpc_cred       = rdata->cred,
1325         };
1326         unsigned long timestamp = jiffies;
1327         int status;
1328
1329         dprintk("NFS call  read %d @ %Ld\n", rdata->args.count,
1330                         (long long) rdata->args.offset);
1331
1332         fattr->valid = 0;
1333         status = rpc_call_sync(server->client, &msg, flags);
1334         if (!status)
1335                 renew_lease(server, timestamp);
1336         dprintk("NFS reply read: %d\n", status);
1337         return status;
1338 }
1339
1340 static int nfs4_proc_read(struct nfs_read_data *rdata)
1341 {
1342         struct nfs4_exception exception = { };
1343         int err;
1344         do {
1345                 err = nfs4_handle_exception(NFS_SERVER(rdata->inode),
1346                                 _nfs4_proc_read(rdata),
1347                                 &exception);
1348         } while (exception.retry);
1349         return err;
1350 }
1351
1352 static int _nfs4_proc_write(struct nfs_write_data *wdata)
1353 {
1354         int rpcflags = wdata->flags;
1355         struct inode *inode = wdata->inode;
1356         struct nfs_fattr *fattr = wdata->res.fattr;
1357         struct nfs_server *server = NFS_SERVER(inode);
1358         struct rpc_message msg = {
1359                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
1360                 .rpc_argp       = &wdata->args,
1361                 .rpc_resp       = &wdata->res,
1362                 .rpc_cred       = wdata->cred,
1363         };
1364         int status;
1365
1366         dprintk("NFS call  write %d @ %Ld\n", wdata->args.count,
1367                         (long long) wdata->args.offset);
1368
1369         fattr->valid = 0;
1370         status = rpc_call_sync(server->client, &msg, rpcflags);
1371         dprintk("NFS reply write: %d\n", status);
1372         return status;
1373 }
1374
1375 static int nfs4_proc_write(struct nfs_write_data *wdata)
1376 {
1377         struct nfs4_exception exception = { };
1378         int err;
1379         do {
1380                 err = nfs4_handle_exception(NFS_SERVER(wdata->inode),
1381                                 _nfs4_proc_write(wdata),
1382                                 &exception);
1383         } while (exception.retry);
1384         return err;
1385 }
1386
1387 static int _nfs4_proc_commit(struct nfs_write_data *cdata)
1388 {
1389         struct inode *inode = cdata->inode;
1390         struct nfs_fattr *fattr = cdata->res.fattr;
1391         struct nfs_server *server = NFS_SERVER(inode);
1392         struct rpc_message msg = {
1393                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
1394                 .rpc_argp       = &cdata->args,
1395                 .rpc_resp       = &cdata->res,
1396                 .rpc_cred       = cdata->cred,
1397         };
1398         int status;
1399
1400         dprintk("NFS call  commit %d @ %Ld\n", cdata->args.count,
1401                         (long long) cdata->args.offset);
1402
1403         fattr->valid = 0;
1404         status = rpc_call_sync(server->client, &msg, 0);
1405         dprintk("NFS reply commit: %d\n", status);
1406         return status;
1407 }
1408
1409 static int nfs4_proc_commit(struct nfs_write_data *cdata)
1410 {
1411         struct nfs4_exception exception = { };
1412         int err;
1413         do {
1414                 err = nfs4_handle_exception(NFS_SERVER(cdata->inode),
1415                                 _nfs4_proc_commit(cdata),
1416                                 &exception);
1417         } while (exception.retry);
1418         return err;
1419 }
1420
1421 /*
1422  * Got race?
1423  * We will need to arrange for the VFS layer to provide an atomic open.
1424  * Until then, this create/open method is prone to inefficiency and race
1425  * conditions due to the lookup, create, and open VFS calls from sys_open()
1426  * placed on the wire.
1427  *
1428  * Given the above sorry state of affairs, I'm simply sending an OPEN.
1429  * The file will be opened again in the subsequent VFS open call
1430  * (nfs4_proc_file_open).
1431  *
1432  * The open for read will just hang around to be used by any process that
1433  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
1434  */
1435
1436 static int
1437 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1438                  int flags)
1439 {
1440         struct nfs4_state *state;
1441         struct rpc_cred *cred;
1442         int status = 0;
1443
1444         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
1445         if (IS_ERR(cred)) {
1446                 status = PTR_ERR(cred);
1447                 goto out;
1448         }
1449         state = nfs4_do_open(dir, dentry, flags, sattr, cred);
1450         put_rpccred(cred);
1451         if (IS_ERR(state)) {
1452                 status = PTR_ERR(state);
1453                 goto out;
1454         }
1455         d_instantiate(dentry, state->inode);
1456         if (flags & O_EXCL) {
1457                 struct nfs_fattr fattr;
1458                 status = nfs4_do_setattr(NFS_SERVER(dir), &fattr,
1459                                      NFS_FH(state->inode), sattr, state);
1460                 if (status == 0)
1461                         goto out;
1462         } else if (flags != 0)
1463                 goto out;
1464         nfs4_close_state(state, flags);
1465 out:
1466         return status;
1467 }
1468
1469 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
1470 {
1471         struct nfs4_remove_arg args = {
1472                 .fh = NFS_FH(dir),
1473                 .name = name,
1474         };
1475         struct nfs4_change_info res;
1476         struct rpc_message msg = {
1477                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
1478                 .rpc_argp       = &args,
1479                 .rpc_resp       = &res,
1480         };
1481         int                     status;
1482
1483         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1484         if (status == 0)
1485                 update_changeattr(dir, &res);
1486         return status;
1487 }
1488
1489 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
1490 {
1491         struct nfs4_exception exception = { };
1492         int err;
1493         do {
1494                 err = nfs4_handle_exception(NFS_SERVER(dir),
1495                                 _nfs4_proc_remove(dir, name),
1496                                 &exception);
1497         } while (exception.retry);
1498         return err;
1499 }
1500
1501 struct unlink_desc {
1502         struct nfs4_remove_arg  args;
1503         struct nfs4_change_info res;
1504 };
1505
1506 static int nfs4_proc_unlink_setup(struct rpc_message *msg, struct dentry *dir,
1507                 struct qstr *name)
1508 {
1509         struct unlink_desc *up;
1510
1511         up = (struct unlink_desc *) kmalloc(sizeof(*up), GFP_KERNEL);
1512         if (!up)
1513                 return -ENOMEM;
1514
1515         up->args.fh = NFS_FH(dir->d_inode);
1516         up->args.name = name;
1517
1518         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
1519         msg->rpc_argp = &up->args;
1520         msg->rpc_resp = &up->res;
1521         return 0;
1522 }
1523
1524 static int nfs4_proc_unlink_done(struct dentry *dir, struct rpc_task *task)
1525 {
1526         struct rpc_message *msg = &task->tk_msg;
1527         struct unlink_desc *up;
1528
1529         if (msg->rpc_resp != NULL) {
1530                 up = container_of(msg->rpc_resp, struct unlink_desc, res);
1531                 update_changeattr(dir->d_inode, &up->res);
1532                 kfree(up);
1533                 msg->rpc_resp = NULL;
1534                 msg->rpc_argp = NULL;
1535         }
1536         return 0;
1537 }
1538
1539 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
1540                 struct inode *new_dir, struct qstr *new_name)
1541 {
1542         struct nfs4_rename_arg arg = {
1543                 .old_dir = NFS_FH(old_dir),
1544                 .new_dir = NFS_FH(new_dir),
1545                 .old_name = old_name,
1546                 .new_name = new_name,
1547         };
1548         struct nfs4_rename_res res = { };
1549         struct rpc_message msg = {
1550                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
1551                 .rpc_argp = &arg,
1552                 .rpc_resp = &res,
1553         };
1554         int                     status;
1555
1556         status = rpc_call_sync(NFS_CLIENT(old_dir), &msg, 0);
1557
1558         if (!status) {
1559                 update_changeattr(old_dir, &res.old_cinfo);
1560                 update_changeattr(new_dir, &res.new_cinfo);
1561         }
1562         return status;
1563 }
1564
1565 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
1566                 struct inode *new_dir, struct qstr *new_name)
1567 {
1568         struct nfs4_exception exception = { };
1569         int err;
1570         do {
1571                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
1572                                 _nfs4_proc_rename(old_dir, old_name,
1573                                         new_dir, new_name),
1574                                 &exception);
1575         } while (exception.retry);
1576         return err;
1577 }
1578
1579 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
1580 {
1581         struct nfs4_link_arg arg = {
1582                 .fh     = NFS_FH(inode),
1583                 .dir_fh = NFS_FH(dir),
1584                 .name   = name,
1585         };
1586         struct nfs4_change_info cinfo = { };
1587         struct rpc_message msg = {
1588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
1589                 .rpc_argp = &arg,
1590                 .rpc_resp = &cinfo,
1591         };
1592         int                     status;
1593
1594         status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1595         if (!status)
1596                 update_changeattr(dir, &cinfo);
1597
1598         return status;
1599 }
1600
1601 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
1602 {
1603         struct nfs4_exception exception = { };
1604         int err;
1605         do {
1606                 err = nfs4_handle_exception(NFS_SERVER(inode),
1607                                 _nfs4_proc_link(inode, dir, name),
1608                                 &exception);
1609         } while (exception.retry);
1610         return err;
1611 }
1612
1613 static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name,
1614                 struct qstr *path, struct iattr *sattr, struct nfs_fh *fhandle,
1615                 struct nfs_fattr *fattr)
1616 {
1617         struct nfs_server *server = NFS_SERVER(dir);
1618         struct nfs4_create_arg arg = {
1619                 .dir_fh = NFS_FH(dir),
1620                 .server = server,
1621                 .name = name,
1622                 .attrs = sattr,
1623                 .ftype = NF4LNK,
1624                 .bitmask = server->attr_bitmask,
1625         };
1626         struct nfs4_create_res res = {
1627                 .server = server,
1628                 .fh = fhandle,
1629                 .fattr = fattr,
1630         };
1631         struct rpc_message msg = {
1632                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK],
1633                 .rpc_argp = &arg,
1634                 .rpc_resp = &res,
1635         };
1636         int                     status;
1637
1638         if (path->len > NFS4_MAXPATHLEN)
1639                 return -ENAMETOOLONG;
1640         arg.u.symlink = path;
1641         fattr->valid = 0;
1642
1643         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1644         if (!status)
1645                 update_changeattr(dir, &res.dir_cinfo);
1646         return status;
1647 }
1648
1649 static int nfs4_proc_symlink(struct inode *dir, struct qstr *name,
1650                 struct qstr *path, struct iattr *sattr, struct nfs_fh *fhandle,
1651                 struct nfs_fattr *fattr)
1652 {
1653         struct nfs4_exception exception = { };
1654         int err;
1655         do {
1656                 err = nfs4_handle_exception(NFS_SERVER(dir),
1657                                 _nfs4_proc_symlink(dir, name, path, sattr,
1658                                         fhandle, fattr),
1659                                 &exception);
1660         } while (exception.retry);
1661         return err;
1662 }
1663
1664 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
1665                 struct iattr *sattr)
1666 {
1667         struct nfs_server *server = NFS_SERVER(dir);
1668         struct nfs_fh fhandle;
1669         struct nfs_fattr fattr;
1670         struct nfs4_create_arg arg = {
1671                 .dir_fh = NFS_FH(dir),
1672                 .server = server,
1673                 .name = &dentry->d_name,
1674                 .attrs = sattr,
1675                 .ftype = NF4DIR,
1676                 .bitmask = server->attr_bitmask,
1677         };
1678         struct nfs4_create_res res = {
1679                 .server = server,
1680                 .fh = &fhandle,
1681                 .fattr = &fattr,
1682         };
1683         struct rpc_message msg = {
1684                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE],
1685                 .rpc_argp = &arg,
1686                 .rpc_resp = &res,
1687         };
1688         int                     status;
1689
1690         fattr.valid = 0;
1691
1692         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1693         if (!status) {
1694                 update_changeattr(dir, &res.dir_cinfo);
1695                 status = nfs_instantiate(dentry, &fhandle, &fattr);
1696         }
1697         return status;
1698 }
1699
1700 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
1701                 struct iattr *sattr)
1702 {
1703         struct nfs4_exception exception = { };
1704         int err;
1705         do {
1706                 err = nfs4_handle_exception(NFS_SERVER(dir),
1707                                 _nfs4_proc_mkdir(dir, dentry, sattr),
1708                                 &exception);
1709         } while (exception.retry);
1710         return err;
1711 }
1712
1713 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
1714                   u64 cookie, struct page *page, unsigned int count, int plus)
1715 {
1716         struct inode            *dir = dentry->d_inode;
1717         struct nfs4_readdir_arg args = {
1718                 .fh = NFS_FH(dir),
1719                 .pages = &page,
1720                 .pgbase = 0,
1721                 .count = count,
1722                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
1723         };
1724         struct nfs4_readdir_res res;
1725         struct rpc_message msg = {
1726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
1727                 .rpc_argp = &args,
1728                 .rpc_resp = &res,
1729                 .rpc_cred = cred,
1730         };
1731         int                     status;
1732
1733         lock_kernel();
1734         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
1735         res.pgbase = args.pgbase;
1736         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1737         if (status == 0)
1738                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
1739         unlock_kernel();
1740         return status;
1741 }
1742
1743 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
1744                   u64 cookie, struct page *page, unsigned int count, int plus)
1745 {
1746         struct nfs4_exception exception = { };
1747         int err;
1748         do {
1749                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
1750                                 _nfs4_proc_readdir(dentry, cred, cookie,
1751                                         page, count, plus),
1752                                 &exception);
1753         } while (exception.retry);
1754         return err;
1755 }
1756
1757 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
1758                 struct iattr *sattr, dev_t rdev)
1759 {
1760         struct nfs_server *server = NFS_SERVER(dir);
1761         struct nfs_fh fh;
1762         struct nfs_fattr fattr;
1763         struct nfs4_create_arg arg = {
1764                 .dir_fh = NFS_FH(dir),
1765                 .server = server,
1766                 .name = &dentry->d_name,
1767                 .attrs = sattr,
1768                 .bitmask = server->attr_bitmask,
1769         };
1770         struct nfs4_create_res res = {
1771                 .server = server,
1772                 .fh = &fh,
1773                 .fattr = &fattr,
1774         };
1775         struct rpc_message msg = {
1776                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE],
1777                 .rpc_argp = &arg,
1778                 .rpc_resp = &res,
1779         };
1780         int                     status;
1781         int                     mode = sattr->ia_mode;
1782
1783         fattr.valid = 0;
1784
1785         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
1786         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
1787         if (S_ISFIFO(mode))
1788                 arg.ftype = NF4FIFO;
1789         else if (S_ISBLK(mode)) {
1790                 arg.ftype = NF4BLK;
1791                 arg.u.device.specdata1 = MAJOR(rdev);
1792                 arg.u.device.specdata2 = MINOR(rdev);
1793         }
1794         else if (S_ISCHR(mode)) {
1795                 arg.ftype = NF4CHR;
1796                 arg.u.device.specdata1 = MAJOR(rdev);
1797                 arg.u.device.specdata2 = MINOR(rdev);
1798         }
1799         else
1800                 arg.ftype = NF4SOCK;
1801
1802         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1803         if (status == 0) {
1804                 update_changeattr(dir, &res.dir_cinfo);
1805                 status = nfs_instantiate(dentry, &fh, &fattr);
1806         }
1807         return status;
1808 }
1809
1810 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
1811                 struct iattr *sattr, dev_t rdev)
1812 {
1813         struct nfs4_exception exception = { };
1814         int err;
1815         do {
1816                 err = nfs4_handle_exception(NFS_SERVER(dir),
1817                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
1818                                 &exception);
1819         } while (exception.retry);
1820         return err;
1821 }
1822
1823 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
1824                  struct nfs_fsstat *fsstat)
1825 {
1826         struct nfs4_statfs_arg args = {
1827                 .fh = fhandle,
1828                 .bitmask = server->attr_bitmask,
1829         };
1830         struct rpc_message msg = {
1831                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
1832                 .rpc_argp = &args,
1833                 .rpc_resp = fsstat,
1834         };
1835
1836         fsstat->fattr->valid = 0;
1837         return rpc_call_sync(server->client, &msg, 0);
1838 }
1839
1840 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
1841 {
1842         struct nfs4_exception exception = { };
1843         int err;
1844         do {
1845                 err = nfs4_handle_exception(server,
1846                                 _nfs4_proc_statfs(server, fhandle, fsstat),
1847                                 &exception);
1848         } while (exception.retry);
1849         return err;
1850 }
1851
1852 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
1853                 struct nfs_fsinfo *fsinfo)
1854 {
1855         struct nfs4_fsinfo_arg args = {
1856                 .fh = fhandle,
1857                 .bitmask = server->attr_bitmask,
1858         };
1859         struct rpc_message msg = {
1860                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
1861                 .rpc_argp = &args,
1862                 .rpc_resp = fsinfo,
1863         };
1864
1865         return rpc_call_sync(server->client, &msg, 0);
1866 }
1867
1868 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
1869 {
1870         struct nfs4_exception exception = { };
1871         int err;
1872
1873         do {
1874                 err = nfs4_handle_exception(server,
1875                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
1876                                 &exception);
1877         } while (exception.retry);
1878         return err;
1879 }
1880
1881 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
1882 {
1883         fsinfo->fattr->valid = 0;
1884         return nfs4_do_fsinfo(server, fhandle, fsinfo);
1885 }
1886
1887 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
1888                 struct nfs_pathconf *pathconf)
1889 {
1890         struct nfs4_pathconf_arg args = {
1891                 .fh = fhandle,
1892                 .bitmask = server->attr_bitmask,
1893         };
1894         struct rpc_message msg = {
1895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
1896                 .rpc_argp = &args,
1897                 .rpc_resp = pathconf,
1898         };
1899
1900         /* None of the pathconf attributes are mandatory to implement */
1901         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
1902                 memset(pathconf, 0, sizeof(*pathconf));
1903                 return 0;
1904         }
1905
1906         pathconf->fattr->valid = 0;
1907         return rpc_call_sync(server->client, &msg, 0);
1908 }
1909
1910 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
1911                 struct nfs_pathconf *pathconf)
1912 {
1913         struct nfs4_exception exception = { };
1914         int err;
1915
1916         do {
1917                 err = nfs4_handle_exception(server,
1918                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
1919                                 &exception);
1920         } while (exception.retry);
1921         return err;
1922 }
1923
1924 static void
1925 nfs4_read_done(struct rpc_task *task)
1926 {
1927         struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata;
1928         struct inode *inode = data->inode;
1929
1930         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
1931                 rpc_restart_call(task);
1932                 return;
1933         }
1934         if (task->tk_status > 0)
1935                 renew_lease(NFS_SERVER(inode), data->timestamp);
1936         /* Call back common NFS readpage processing */
1937         nfs_readpage_result(task);
1938 }
1939
1940 static void
1941 nfs4_proc_read_setup(struct nfs_read_data *data)
1942 {
1943         struct rpc_task *task = &data->task;
1944         struct rpc_message msg = {
1945                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ],
1946                 .rpc_argp = &data->args,
1947                 .rpc_resp = &data->res,
1948                 .rpc_cred = data->cred,
1949         };
1950         struct inode *inode = data->inode;
1951         int flags;
1952
1953         data->timestamp   = jiffies;
1954
1955         /* N.B. Do we need to test? Never called for swapfile inode */
1956         flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
1957
1958         /* Finalize the task. */
1959         rpc_init_task(task, NFS_CLIENT(inode), nfs4_read_done, flags);
1960         rpc_call_setup(task, &msg, 0);
1961 }
1962
1963 static void
1964 nfs4_write_done(struct rpc_task *task)
1965 {
1966         struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
1967         struct inode *inode = data->inode;
1968
1969         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
1970                 rpc_restart_call(task);
1971                 return;
1972         }
1973         if (task->tk_status >= 0)
1974                 renew_lease(NFS_SERVER(inode), data->timestamp);
1975         /* Call back common NFS writeback processing */
1976         nfs_writeback_done(task);
1977 }
1978
1979 static void
1980 nfs4_proc_write_setup(struct nfs_write_data *data, int how)
1981 {
1982         struct rpc_task *task = &data->task;
1983         struct rpc_message msg = {
1984                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
1985                 .rpc_argp = &data->args,
1986                 .rpc_resp = &data->res,
1987                 .rpc_cred = data->cred,
1988         };
1989         struct inode *inode = data->inode;
1990         int stable;
1991         int flags;
1992
1993         if (how & FLUSH_STABLE) {
1994                 if (!NFS_I(inode)->ncommit)
1995                         stable = NFS_FILE_SYNC;
1996                 else
1997                         stable = NFS_DATA_SYNC;
1998         } else
1999                 stable = NFS_UNSTABLE;
2000         data->args.stable = stable;
2001
2002         data->timestamp   = jiffies;
2003
2004         /* Set the initial flags for the task.  */
2005         flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
2006
2007         /* Finalize the task. */
2008         rpc_init_task(task, NFS_CLIENT(inode), nfs4_write_done, flags);
2009         rpc_call_setup(task, &msg, 0);
2010 }
2011
2012 static void
2013 nfs4_commit_done(struct rpc_task *task)
2014 {
2015         struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
2016         struct inode *inode = data->inode;
2017
2018         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
2019                 rpc_restart_call(task);
2020                 return;
2021         }
2022         /* Call back common NFS writeback processing */
2023         nfs_commit_done(task);
2024 }
2025
2026 static void
2027 nfs4_proc_commit_setup(struct nfs_write_data *data, int how)
2028 {
2029         struct rpc_task *task = &data->task;
2030         struct rpc_message msg = {
2031                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
2032                 .rpc_argp = &data->args,
2033                 .rpc_resp = &data->res,
2034                 .rpc_cred = data->cred,
2035         };
2036         struct inode *inode = data->inode;
2037         int flags;
2038
2039         /* Set the initial flags for the task.  */
2040         flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
2041
2042         /* Finalize the task. */
2043         rpc_init_task(task, NFS_CLIENT(inode), nfs4_commit_done, flags);
2044         rpc_call_setup(task, &msg, 0);
2045 }
2046
2047 /*
2048  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
2049  * standalone procedure for queueing an asynchronous RENEW.
2050  */
2051 static void
2052 renew_done(struct rpc_task *task)
2053 {
2054         struct nfs4_client *clp = (struct nfs4_client *)task->tk_msg.rpc_argp;
2055         unsigned long timestamp = (unsigned long)task->tk_calldata;
2056
2057         if (task->tk_status < 0) {
2058                 switch (task->tk_status) {
2059                         case -NFS4ERR_STALE_CLIENTID:
2060                         case -NFS4ERR_EXPIRED:
2061                         case -NFS4ERR_CB_PATH_DOWN:
2062                                 nfs4_schedule_state_recovery(clp);
2063                 }
2064                 return;
2065         }
2066         spin_lock(&clp->cl_lock);
2067         if (time_before(clp->cl_last_renewal,timestamp))
2068                 clp->cl_last_renewal = timestamp;
2069         spin_unlock(&clp->cl_lock);
2070 }
2071
2072 int
2073 nfs4_proc_async_renew(struct nfs4_client *clp)
2074 {
2075         struct rpc_message msg = {
2076                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2077                 .rpc_argp       = clp,
2078                 .rpc_cred       = clp->cl_cred,
2079         };
2080
2081         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
2082                         renew_done, (void *)jiffies);
2083 }
2084
2085 int
2086 nfs4_proc_renew(struct nfs4_client *clp)
2087 {
2088         struct rpc_message msg = {
2089                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2090                 .rpc_argp       = clp,
2091                 .rpc_cred       = clp->cl_cred,
2092         };
2093         unsigned long now = jiffies;
2094         int status;
2095
2096         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2097         if (status < 0)
2098                 return status;
2099         spin_lock(&clp->cl_lock);
2100         if (time_before(clp->cl_last_renewal,now))
2101                 clp->cl_last_renewal = now;
2102         spin_unlock(&clp->cl_lock);
2103         return 0;
2104 }
2105
2106 /*
2107  * We will need to arrange for the VFS layer to provide an atomic open.
2108  * Until then, this open method is prone to inefficiency and race conditions
2109  * due to the lookup, potential create, and open VFS calls from sys_open()
2110  * placed on the wire.
2111  */
2112 static int
2113 nfs4_proc_file_open(struct inode *inode, struct file *filp)
2114 {
2115         struct dentry *dentry = filp->f_dentry;
2116         struct nfs_open_context *ctx;
2117         struct nfs4_state *state = NULL;
2118         struct rpc_cred *cred;
2119         int status = -ENOMEM;
2120
2121         dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n",
2122                                (int)dentry->d_parent->d_name.len,
2123                                dentry->d_parent->d_name.name,
2124                                (int)dentry->d_name.len, dentry->d_name.name);
2125
2126
2127         /* Find our open stateid */
2128         cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
2129         if (IS_ERR(cred))
2130                 return PTR_ERR(cred);
2131         ctx = alloc_nfs_open_context(dentry, cred);
2132         put_rpccred(cred);
2133         if (unlikely(ctx == NULL))
2134                 return -ENOMEM;
2135         status = -EIO; /* ERACE actually */
2136         state = nfs4_find_state(inode, cred, filp->f_mode);
2137         if (unlikely(state == NULL))
2138                 goto no_state;
2139         ctx->state = state;
2140         nfs4_close_state(state, filp->f_mode);
2141         ctx->mode = filp->f_mode;
2142         nfs_file_set_open_context(filp, ctx);
2143         put_nfs_open_context(ctx);
2144         if (filp->f_mode & FMODE_WRITE)
2145                 nfs_begin_data_update(inode);
2146         return 0;
2147 no_state:
2148         printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__);
2149         put_nfs_open_context(ctx);
2150         return status;
2151 }
2152
2153 /*
2154  * Release our state
2155  */
2156 static int
2157 nfs4_proc_file_release(struct inode *inode, struct file *filp)
2158 {
2159         if (filp->f_mode & FMODE_WRITE)
2160                 nfs_end_data_update(inode);
2161         nfs_file_clear_open_context(filp);
2162         return 0;
2163 }
2164
2165 static inline int nfs4_server_supports_acls(struct nfs_server *server)
2166 {
2167         return (server->caps & NFS_CAP_ACLS)
2168                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2169                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
2170 }
2171
2172 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
2173  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
2174  * the stack.
2175  */
2176 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
2177
2178 static void buf_to_pages(const void *buf, size_t buflen,
2179                 struct page **pages, unsigned int *pgbase)
2180 {
2181         const void *p = buf;
2182
2183         *pgbase = offset_in_page(buf);
2184         p -= *pgbase;
2185         while (p < buf + buflen) {
2186                 *(pages++) = virt_to_page(p);
2187                 p += PAGE_CACHE_SIZE;
2188         }
2189 }
2190
2191 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
2192 {
2193         struct nfs_server *server = NFS_SERVER(inode);
2194         struct page *pages[NFS4ACL_MAXPAGES];
2195         struct nfs_getaclargs args = {
2196                 .fh = NFS_FH(inode),
2197                 .acl_pages = pages,
2198                 .acl_len = buflen,
2199         };
2200         size_t resp_len = buflen;
2201         struct rpc_message msg = {
2202                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
2203                 .rpc_argp = &args,
2204                 .rpc_resp = &resp_len,
2205         };
2206         int ret;
2207
2208         if (!nfs4_server_supports_acls(server))
2209                 return -EOPNOTSUPP;
2210         buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
2211         ret = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
2212         if (buflen && resp_len > buflen)
2213                 return -ERANGE;
2214         if (ret == 0)
2215                 ret = resp_len;
2216         return ret;
2217 }
2218
2219 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
2220 {
2221         struct nfs_server *server = NFS_SERVER(inode);
2222         struct page *pages[NFS4ACL_MAXPAGES];
2223         struct nfs_setaclargs arg = {
2224                 .fh             = NFS_FH(inode),
2225                 .acl_pages      = pages,
2226                 .acl_len        = buflen,
2227         };
2228         struct rpc_message msg = {
2229                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
2230                 .rpc_argp       = &arg,
2231                 .rpc_resp       = NULL,
2232         };
2233         int ret;
2234
2235         if (!nfs4_server_supports_acls(server))
2236                 return -EOPNOTSUPP;
2237         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
2238         ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0);
2239         return ret;
2240 }
2241
2242 static int
2243 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server)
2244 {
2245         struct nfs4_client *clp = server->nfs4_state;
2246
2247         if (!clp || task->tk_status >= 0)
2248                 return 0;
2249         switch(task->tk_status) {
2250                 case -NFS4ERR_STALE_CLIENTID:
2251                 case -NFS4ERR_STALE_STATEID:
2252                 case -NFS4ERR_EXPIRED:
2253                         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL, NULL);
2254                         nfs4_schedule_state_recovery(clp);
2255                         if (test_bit(NFS4CLNT_OK, &clp->cl_state))
2256                                 rpc_wake_up_task(task);
2257                         task->tk_status = 0;
2258                         return -EAGAIN;
2259                 case -NFS4ERR_GRACE:
2260                 case -NFS4ERR_DELAY:
2261                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
2262                         task->tk_status = 0;
2263                         return -EAGAIN;
2264                 case -NFS4ERR_OLD_STATEID:
2265                         task->tk_status = 0;
2266                         return -EAGAIN;
2267         }
2268         task->tk_status = nfs4_map_errors(task->tk_status);
2269         return 0;
2270 }
2271
2272 static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs4_client *clp)
2273 {
2274         DEFINE_WAIT(wait);
2275         sigset_t oldset;
2276         int interruptible, res = 0;
2277
2278         might_sleep();
2279
2280         rpc_clnt_sigmask(clnt, &oldset);
2281         interruptible = TASK_UNINTERRUPTIBLE;
2282         if (clnt->cl_intr)
2283                 interruptible = TASK_INTERRUPTIBLE;
2284         prepare_to_wait(&clp->cl_waitq, &wait, interruptible);
2285         nfs4_schedule_state_recovery(clp);
2286         if (clnt->cl_intr && signalled())
2287                 res = -ERESTARTSYS;
2288         else if (!test_bit(NFS4CLNT_OK, &clp->cl_state))
2289                 schedule();
2290         finish_wait(&clp->cl_waitq, &wait);
2291         rpc_clnt_sigunmask(clnt, &oldset);
2292         return res;
2293 }
2294
2295 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
2296 {
2297         sigset_t oldset;
2298         int res = 0;
2299
2300         might_sleep();
2301
2302         if (*timeout <= 0)
2303                 *timeout = NFS4_POLL_RETRY_MIN;
2304         if (*timeout > NFS4_POLL_RETRY_MAX)
2305                 *timeout = NFS4_POLL_RETRY_MAX;
2306         rpc_clnt_sigmask(clnt, &oldset);
2307         if (clnt->cl_intr) {
2308                 set_current_state(TASK_INTERRUPTIBLE);
2309                 schedule_timeout(*timeout);
2310                 if (signalled())
2311                         res = -ERESTARTSYS;
2312         } else {
2313                 set_current_state(TASK_UNINTERRUPTIBLE);
2314                 schedule_timeout(*timeout);
2315         }
2316         rpc_clnt_sigunmask(clnt, &oldset);
2317         *timeout <<= 1;
2318         return res;
2319 }
2320
2321 /* This is the error handling routine for processes that are allowed
2322  * to sleep.
2323  */
2324 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
2325 {
2326         struct nfs4_client *clp = server->nfs4_state;
2327         int ret = errorcode;
2328
2329         exception->retry = 0;
2330         switch(errorcode) {
2331                 case 0:
2332                         return 0;
2333                 case -NFS4ERR_STALE_CLIENTID:
2334                 case -NFS4ERR_STALE_STATEID:
2335                 case -NFS4ERR_EXPIRED:
2336                         ret = nfs4_wait_clnt_recover(server->client, clp);
2337                         if (ret == 0)
2338                                 exception->retry = 1;
2339                         break;
2340                 case -NFS4ERR_GRACE:
2341                 case -NFS4ERR_DELAY:
2342                         ret = nfs4_delay(server->client, &exception->timeout);
2343                         if (ret == 0)
2344                                 exception->retry = 1;
2345                         break;
2346                 case -NFS4ERR_OLD_STATEID:
2347                         if (ret == 0)
2348                                 exception->retry = 1;
2349         }
2350         /* We failed to handle the error */
2351         return nfs4_map_errors(ret);
2352 }
2353
2354 int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short port)
2355 {
2356         nfs4_verifier sc_verifier;
2357         struct nfs4_setclientid setclientid = {
2358                 .sc_verifier = &sc_verifier,
2359                 .sc_prog = program,
2360         };
2361         struct rpc_message msg = {
2362                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
2363                 .rpc_argp = &setclientid,
2364                 .rpc_resp = clp,
2365                 .rpc_cred = clp->cl_cred,
2366         };
2367         u32 *p;
2368         int loop = 0;
2369         int status;
2370
2371         p = (u32*)sc_verifier.data;
2372         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
2373         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
2374
2375         for(;;) {
2376                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
2377                                 sizeof(setclientid.sc_name), "%s/%u.%u.%u.%u %s %u",
2378                                 clp->cl_ipaddr, NIPQUAD(clp->cl_addr.s_addr),
2379                                 clp->cl_cred->cr_ops->cr_name,
2380                                 clp->cl_id_uniquifier);
2381                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
2382                                 sizeof(setclientid.sc_netid), "tcp");
2383                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
2384                                 sizeof(setclientid.sc_uaddr), "%s.%d.%d",
2385                                 clp->cl_ipaddr, port >> 8, port & 255);
2386
2387                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2388                 if (status != -NFS4ERR_CLID_INUSE)
2389                         break;
2390                 if (signalled())
2391                         break;
2392                 if (loop++ & 1)
2393                         ssleep(clp->cl_lease_time + 1);
2394                 else
2395                         if (++clp->cl_id_uniquifier == 0)
2396                                 break;
2397         }
2398         return status;
2399 }
2400
2401 int
2402 nfs4_proc_setclientid_confirm(struct nfs4_client *clp)
2403 {
2404         struct nfs_fsinfo fsinfo;
2405         struct rpc_message msg = {
2406                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
2407                 .rpc_argp = clp,
2408                 .rpc_resp = &fsinfo,
2409                 .rpc_cred = clp->cl_cred,
2410         };
2411         unsigned long now;
2412         int status;
2413
2414         now = jiffies;
2415         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2416         if (status == 0) {
2417                 spin_lock(&clp->cl_lock);
2418                 clp->cl_lease_time = fsinfo.lease_time * HZ;
2419                 clp->cl_last_renewal = now;
2420                 spin_unlock(&clp->cl_lock);
2421         }
2422         return status;
2423 }
2424
2425 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid)
2426 {
2427         struct nfs4_delegreturnargs args = {
2428                 .fhandle = NFS_FH(inode),
2429                 .stateid = stateid,
2430         };
2431         struct rpc_message msg = {
2432                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
2433                 .rpc_argp = &args,
2434                 .rpc_cred = cred,
2435         };
2436
2437         return rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
2438 }
2439
2440 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid)
2441 {
2442         struct nfs_server *server = NFS_SERVER(inode);
2443         struct nfs4_exception exception = { };
2444         int err;
2445         do {
2446                 err = _nfs4_proc_delegreturn(inode, cred, stateid);
2447                 switch (err) {
2448                         case -NFS4ERR_STALE_STATEID:
2449                         case -NFS4ERR_EXPIRED:
2450                                 nfs4_schedule_state_recovery(server->nfs4_state);
2451                         case 0:
2452                                 return 0;
2453                 }
2454                 err = nfs4_handle_exception(server, err, &exception);
2455         } while (exception.retry);
2456         return err;
2457 }
2458
2459 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
2460 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
2461
2462 /*
2463  * sleep, with exponential backoff, and retry the LOCK operation.
2464  */
2465 static unsigned long
2466 nfs4_set_lock_task_retry(unsigned long timeout)
2467 {
2468         current->state = TASK_INTERRUPTIBLE;
2469         schedule_timeout(timeout);
2470         timeout <<= 1;
2471         if (timeout > NFS4_LOCK_MAXTIMEOUT)
2472                 return NFS4_LOCK_MAXTIMEOUT;
2473         return timeout;
2474 }
2475
2476 static inline int
2477 nfs4_lck_type(int cmd, struct file_lock *request)
2478 {
2479         /* set lock type */
2480         switch (request->fl_type) {
2481                 case F_RDLCK:
2482                         return IS_SETLKW(cmd) ? NFS4_READW_LT : NFS4_READ_LT;
2483                 case F_WRLCK:
2484                         return IS_SETLKW(cmd) ? NFS4_WRITEW_LT : NFS4_WRITE_LT;
2485                 case F_UNLCK:
2486                         return NFS4_WRITE_LT;
2487         }
2488         BUG();
2489         return 0;
2490 }
2491
2492 static inline uint64_t
2493 nfs4_lck_length(struct file_lock *request)
2494 {
2495         if (request->fl_end == OFFSET_MAX)
2496                 return ~(uint64_t)0;
2497         return request->fl_end - request->fl_start + 1;
2498 }
2499
2500 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2501 {
2502         struct inode *inode = state->inode;
2503         struct nfs_server *server = NFS_SERVER(inode);
2504         struct nfs4_client *clp = server->nfs4_state;
2505         struct nfs_lockargs arg = {
2506                 .fh = NFS_FH(inode),
2507                 .type = nfs4_lck_type(cmd, request),
2508                 .offset = request->fl_start,
2509                 .length = nfs4_lck_length(request),
2510         };
2511         struct nfs_lockres res = {
2512                 .server = server,
2513         };
2514         struct rpc_message msg = {
2515                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
2516                 .rpc_argp       = &arg,
2517                 .rpc_resp       = &res,
2518                 .rpc_cred       = state->owner->so_cred,
2519         };
2520         struct nfs_lowner nlo;
2521         struct nfs4_lock_state *lsp;
2522         int status;
2523
2524         down_read(&clp->cl_sem);
2525         nlo.clientid = clp->cl_clientid;
2526         down(&state->lock_sema);
2527         lsp = nfs4_find_lock_state(state, request->fl_owner);
2528         if (lsp)
2529                 nlo.id = lsp->ls_id;
2530         else {
2531                 spin_lock(&clp->cl_lock);
2532                 nlo.id = nfs4_alloc_lockowner_id(clp);
2533                 spin_unlock(&clp->cl_lock);
2534         }
2535         arg.u.lockt = &nlo;
2536         status = rpc_call_sync(server->client, &msg, 0);
2537         if (!status) {
2538                 request->fl_type = F_UNLCK;
2539         } else if (status == -NFS4ERR_DENIED) {
2540                 int64_t len, start, end;
2541                 start = res.u.denied.offset;
2542                 len = res.u.denied.length;
2543                 end = start + len - 1;
2544                 if (end < 0 || len == 0)
2545                         request->fl_end = OFFSET_MAX;
2546                 else
2547                         request->fl_end = (loff_t)end;
2548                 request->fl_start = (loff_t)start;
2549                 request->fl_type = F_WRLCK;
2550                 if (res.u.denied.type & 1)
2551                         request->fl_type = F_RDLCK;
2552                 request->fl_pid = 0;
2553                 status = 0;
2554         }
2555         if (lsp)
2556                 nfs4_put_lock_state(lsp);
2557         up(&state->lock_sema);
2558         up_read(&clp->cl_sem);
2559         return status;
2560 }
2561
2562 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2563 {
2564         struct nfs4_exception exception = { };
2565         int err;
2566
2567         do {
2568                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2569                                 _nfs4_proc_getlk(state, cmd, request),
2570                                 &exception);
2571         } while (exception.retry);
2572         return err;
2573 }
2574
2575 static int do_vfs_lock(struct file *file, struct file_lock *fl)
2576 {
2577         int res = 0;
2578         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
2579                 case FL_POSIX:
2580                         res = posix_lock_file_wait(file, fl);
2581                         break;
2582                 case FL_FLOCK:
2583                         res = flock_lock_file_wait(file, fl);
2584                         break;
2585                 default:
2586                         BUG();
2587         }
2588         if (res < 0)
2589                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n",
2590                                 __FUNCTION__);
2591         return res;
2592 }
2593
2594 static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2595 {
2596         struct inode *inode = state->inode;
2597         struct nfs_server *server = NFS_SERVER(inode);
2598         struct nfs4_client *clp = server->nfs4_state;
2599         struct nfs_lockargs arg = {
2600                 .fh = NFS_FH(inode),
2601                 .type = nfs4_lck_type(cmd, request),
2602                 .offset = request->fl_start,
2603                 .length = nfs4_lck_length(request),
2604         };
2605         struct nfs_lockres res = {
2606                 .server = server,
2607         };
2608         struct rpc_message msg = {
2609                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
2610                 .rpc_argp       = &arg,
2611                 .rpc_resp       = &res,
2612                 .rpc_cred       = state->owner->so_cred,
2613         };
2614         struct nfs4_lock_state *lsp;
2615         struct nfs_locku_opargs luargs;
2616         int status = 0;
2617
2618         down_read(&clp->cl_sem);
2619         down(&state->lock_sema);
2620         lsp = nfs4_find_lock_state(state, request->fl_owner);
2621         if (!lsp)
2622                 goto out;
2623         /* We might have lost the locks! */
2624         if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) {
2625                 luargs.seqid = lsp->ls_seqid;
2626                 memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid));
2627                 arg.u.locku = &luargs;
2628                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2629                 nfs4_increment_lock_seqid(status, lsp);
2630         }
2631
2632         if (status == 0) {
2633                 memcpy(&lsp->ls_stateid,  &res.u.stateid,
2634                                 sizeof(lsp->ls_stateid));
2635                 nfs4_notify_unlck(state, request, lsp);
2636         }
2637         nfs4_put_lock_state(lsp);
2638 out:
2639         up(&state->lock_sema);
2640         if (status == 0)
2641                 do_vfs_lock(request->fl_file, request);
2642         up_read(&clp->cl_sem);
2643         return status;
2644 }
2645
2646 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2647 {
2648         struct nfs4_exception exception = { };
2649         int err;
2650
2651         do {
2652                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2653                                 _nfs4_proc_unlck(state, cmd, request),
2654                                 &exception);
2655         } while (exception.retry);
2656         return err;
2657 }
2658
2659 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim)
2660 {
2661         struct inode *inode = state->inode;
2662         struct nfs_server *server = NFS_SERVER(inode);
2663         struct nfs4_lock_state *lsp;
2664         struct nfs_lockargs arg = {
2665                 .fh = NFS_FH(inode),
2666                 .type = nfs4_lck_type(cmd, request),
2667                 .offset = request->fl_start,
2668                 .length = nfs4_lck_length(request),
2669         };
2670         struct nfs_lockres res = {
2671                 .server = server,
2672         };
2673         struct rpc_message msg = {
2674                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
2675                 .rpc_argp       = &arg,
2676                 .rpc_resp       = &res,
2677                 .rpc_cred       = state->owner->so_cred,
2678         };
2679         struct nfs_lock_opargs largs = {
2680                 .reclaim = reclaim,
2681                 .new_lock_owner = 0,
2682         };
2683         int status;
2684
2685         lsp = nfs4_get_lock_state(state, request->fl_owner);
2686         if (lsp == NULL)
2687                 return -ENOMEM;
2688         if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) {
2689                 struct nfs4_state_owner *owner = state->owner;
2690                 struct nfs_open_to_lock otl = {
2691                         .lock_owner = {
2692                                 .clientid = server->nfs4_state->cl_clientid,
2693                         },
2694                 };
2695
2696                 otl.lock_seqid = lsp->ls_seqid;
2697                 otl.lock_owner.id = lsp->ls_id;
2698                 memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid));
2699                 largs.u.open_lock = &otl;
2700                 largs.new_lock_owner = 1;
2701                 arg.u.lock = &largs;
2702                 down(&owner->so_sema);
2703                 otl.open_seqid = owner->so_seqid;
2704                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2705                 /* increment open_owner seqid on success, and
2706                 * seqid mutating errors */
2707                 nfs4_increment_seqid(status, owner);
2708                 up(&owner->so_sema);
2709         } else {
2710                 struct nfs_exist_lock el = {
2711                         .seqid = lsp->ls_seqid,
2712                 };
2713                 memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid));
2714                 largs.u.exist_lock = &el;
2715                 largs.new_lock_owner = 0;
2716                 arg.u.lock = &largs;
2717                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2718         }
2719         /* increment seqid on success, and * seqid mutating errors*/
2720         nfs4_increment_lock_seqid(status, lsp);
2721         /* save the returned stateid. */
2722         if (status == 0) {
2723                 memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid));
2724                 lsp->ls_flags |= NFS_LOCK_INITIALIZED;
2725                 if (!reclaim)
2726                         nfs4_notify_setlk(state, request, lsp);
2727         } else if (status == -NFS4ERR_DENIED)
2728                 status = -EAGAIN;
2729         nfs4_put_lock_state(lsp);
2730         return status;
2731 }
2732
2733 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
2734 {
2735         return _nfs4_do_setlk(state, F_SETLK, request, 1);
2736 }
2737
2738 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
2739 {
2740         return _nfs4_do_setlk(state, F_SETLK, request, 0);
2741 }
2742
2743 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2744 {
2745         struct nfs4_client *clp = state->owner->so_client;
2746         int status;
2747
2748         down_read(&clp->cl_sem);
2749         down(&state->lock_sema);
2750         status = _nfs4_do_setlk(state, cmd, request, 0);
2751         up(&state->lock_sema);
2752         if (status == 0) {
2753                 /* Note: we always want to sleep here! */
2754                 request->fl_flags |= FL_SLEEP;
2755                 if (do_vfs_lock(request->fl_file, request) < 0)
2756                         printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __FUNCTION__);
2757         }
2758         up_read(&clp->cl_sem);
2759         return status;
2760 }
2761
2762 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2763 {
2764         struct nfs4_exception exception = { };
2765         int err;
2766
2767         do {
2768                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2769                                 _nfs4_proc_setlk(state, cmd, request),
2770                                 &exception);
2771         } while (exception.retry);
2772         return err;
2773 }
2774
2775 static int
2776 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
2777 {
2778         struct nfs_open_context *ctx;
2779         struct nfs4_state *state;
2780         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
2781         int status;
2782
2783         /* verify open state */
2784         ctx = (struct nfs_open_context *)filp->private_data;
2785         state = ctx->state;
2786
2787         if (request->fl_start < 0 || request->fl_end < 0)
2788                 return -EINVAL;
2789
2790         if (IS_GETLK(cmd))
2791                 return nfs4_proc_getlk(state, F_GETLK, request);
2792
2793         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
2794                 return -EINVAL;
2795
2796         if (request->fl_type == F_UNLCK)
2797                 return nfs4_proc_unlck(state, cmd, request);
2798
2799         do {
2800                 status = nfs4_proc_setlk(state, cmd, request);
2801                 if ((status != -EAGAIN) || IS_SETLK(cmd))
2802                         break;
2803                 timeout = nfs4_set_lock_task_retry(timeout);
2804                 status = -ERESTARTSYS;
2805                 if (signalled())
2806                         break;
2807         } while(status < 0);
2808
2809         return status;
2810 }
2811
2812
2813 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
2814
2815 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
2816                 size_t buflen, int flags)
2817 {
2818         struct inode *inode = dentry->d_inode;
2819
2820         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
2821                 return -EOPNOTSUPP;
2822
2823         if (!S_ISREG(inode->i_mode) &&
2824             (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX))
2825                 return -EPERM;
2826
2827         return nfs4_proc_set_acl(inode, buf, buflen);
2828 }
2829
2830 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
2831  * and that's what we'll do for e.g. user attributes that haven't been set.
2832  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
2833  * attributes in kernel-managed attribute namespaces. */
2834 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
2835                 size_t buflen)
2836 {
2837         struct inode *inode = dentry->d_inode;
2838
2839         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
2840                 return -EOPNOTSUPP;
2841
2842         return nfs4_proc_get_acl(inode, buf, buflen);
2843 }
2844
2845 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
2846 {
2847         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
2848
2849         if (buf && buflen < len)
2850                 return -ERANGE;
2851         if (buf)
2852                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
2853         return len;
2854 }
2855
2856 struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops = {
2857         .recover_open   = nfs4_open_reclaim,
2858         .recover_lock   = nfs4_lock_reclaim,
2859 };
2860
2861 struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops = {
2862         .recover_open   = nfs4_open_expired,
2863         .recover_lock   = nfs4_lock_expired,
2864 };
2865
2866 static struct inode_operations nfs4_file_inode_operations = {
2867         .permission     = nfs_permission,
2868         .getattr        = nfs_getattr,
2869         .setattr        = nfs_setattr,
2870         .getxattr       = nfs4_getxattr,
2871         .setxattr       = nfs4_setxattr,
2872         .listxattr      = nfs4_listxattr,
2873 };
2874
2875 struct nfs_rpc_ops      nfs_v4_clientops = {
2876         .version        = 4,                    /* protocol version */
2877         .dentry_ops     = &nfs4_dentry_operations,
2878         .dir_inode_ops  = &nfs4_dir_inode_operations,
2879         .file_inode_ops = &nfs4_file_inode_operations,
2880         .getroot        = nfs4_proc_get_root,
2881         .getattr        = nfs4_proc_getattr,
2882         .setattr        = nfs4_proc_setattr,
2883         .lookup         = nfs4_proc_lookup,
2884         .access         = nfs4_proc_access,
2885         .readlink       = nfs4_proc_readlink,
2886         .read           = nfs4_proc_read,
2887         .write          = nfs4_proc_write,
2888         .commit         = nfs4_proc_commit,
2889         .create         = nfs4_proc_create,
2890         .remove         = nfs4_proc_remove,
2891         .unlink_setup   = nfs4_proc_unlink_setup,
2892         .unlink_done    = nfs4_proc_unlink_done,
2893         .rename         = nfs4_proc_rename,
2894         .link           = nfs4_proc_link,
2895         .symlink        = nfs4_proc_symlink,
2896         .mkdir          = nfs4_proc_mkdir,
2897         .rmdir          = nfs4_proc_remove,
2898         .readdir        = nfs4_proc_readdir,
2899         .mknod          = nfs4_proc_mknod,
2900         .statfs         = nfs4_proc_statfs,
2901         .fsinfo         = nfs4_proc_fsinfo,
2902         .pathconf       = nfs4_proc_pathconf,
2903         .decode_dirent  = nfs4_decode_dirent,
2904         .read_setup     = nfs4_proc_read_setup,
2905         .write_setup    = nfs4_proc_write_setup,
2906         .commit_setup   = nfs4_proc_commit_setup,
2907         .file_open      = nfs4_proc_file_open,
2908         .file_release   = nfs4_proc_file_release,
2909         .lock           = nfs4_proc_lock,
2910 };
2911
2912 /*
2913  * Local variables:
2914  *  c-basic-offset: 8
2915  * End:
2916  */