fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/utsname.h>
  40 #include <linux/delay.h>
  41 #include <linux/errno.h>
  42 #include <linux/string.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/nfs.h>
  45 #include <linux/nfs4.h>
  46 #include <linux/nfs_fs.h>
  47 #include <linux/nfs_page.h>
  48 #include <linux/smp_lock.h>
  49 #include <linux/namei.h>
  50
  51 #include "nfs4_fs.h"
  52 #include "delegation.h"
  53
  54 #define NFSDBG_FACILITY         NFSDBG_PROC
  55
  56 #define NFS4_POLL_RETRY_MIN     (1*HZ)
  57 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  58
  59 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  60 static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *);
  61 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
  62 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
  63 extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus);
  64 extern struct rpc_procinfo nfs4_procedures[];
  65
  66 /* Prevent leaks of NFSv4 errors into userland */
  67 int nfs4_map_errors(int err)
  68 {
  69         if (err < -1000) {
  70                 dprintk("%s could not handle NFSv4 error %d\n",
  71                                 __FUNCTION__, -err);
  72                 return -EIO;
  73         }
  74         return err;
  75 }
  76
  77 /*
  78  * This is our standard bitmap for GETATTR requests.
  79  */
  80 const u32 nfs4_fattr_bitmap[2] = {
  81         FATTR4_WORD0_TYPE
  82         | FATTR4_WORD0_CHANGE
  83         | FATTR4_WORD0_SIZE
  84         | FATTR4_WORD0_FSID
  85         | FATTR4_WORD0_FILEID,
  86         FATTR4_WORD1_MODE
  87         | FATTR4_WORD1_NUMLINKS
  88         | FATTR4_WORD1_OWNER
  89         | FATTR4_WORD1_OWNER_GROUP
  90         | FATTR4_WORD1_RAWDEV
  91         | FATTR4_WORD1_SPACE_USED
  92         | FATTR4_WORD1_TIME_ACCESS
  93         | FATTR4_WORD1_TIME_METADATA
  94         | FATTR4_WORD1_TIME_MODIFY
  95 };
  96
  97 const u32 nfs4_statfs_bitmap[2] = {
  98         FATTR4_WORD0_FILES_AVAIL
  99         | FATTR4_WORD0_FILES_FREE
 100         | FATTR4_WORD0_FILES_TOTAL,
 101         FATTR4_WORD1_SPACE_AVAIL
 102         | FATTR4_WORD1_SPACE_FREE
 103         | FATTR4_WORD1_SPACE_TOTAL
 104 };
 105
 106 const u32 nfs4_pathconf_bitmap[2] = {
 107         FATTR4_WORD0_MAXLINK
 108         | FATTR4_WORD0_MAXNAME,
 109         0
 110 };
 111
 112 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 113                         | FATTR4_WORD0_MAXREAD
 114                         | FATTR4_WORD0_MAXWRITE
 115                         | FATTR4_WORD0_LEASE_TIME,
 116                         0
 117 };
 118
 119 static void nfs4_setup_readdir(u64 cookie, u32 *verifier, struct dentry *dentry,
 120                 struct nfs4_readdir_arg *readdir)
 121 {
 122         u32 *start, *p;
 123
 124         BUG_ON(readdir->count < 80);
 125         if (cookie > 2) {
 126                 readdir->cookie = cookie;
 127                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 128                 return;
 129         }
 130
 131         readdir->cookie = 0;
 132         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 133         if (cookie == 2)
 134                 return;
 135
 136         /*
 137          * NFSv4 servers do not return entries for '.' and '..'
 138          * Therefore, we fake these entries here.  We let '.'
 139          * have cookie 0 and '..' have cookie 1.  Note that
 140          * when talking to the server, we always send cookie 0
 141          * instead of 1 or 2.
 142          */
 143         start = p = (u32 *)kmap_atomic(*readdir->pages, KM_USER0);
 144
 145         if (cookie == 0) {
 146                 *p++ = xdr_one;                                  /* next */
 147                 *p++ = xdr_zero;                   /* cookie, first word */
 148                 *p++ = xdr_one;                   /* cookie, second word */
 149                 *p++ = xdr_one;                             /* entry len */
 150                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 151                 p++;
 152                 *p++ = xdr_one;                         /* bitmap length */
 153                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 154                 *p++ = htonl(8);              /* attribute buffer length */
 155                 p = xdr_encode_hyper(p, dentry->d_inode->i_ino);
 156         }
 157
 158         *p++ = xdr_one;                                  /* next */
 159         *p++ = xdr_zero;                   /* cookie, first word */
 160         *p++ = xdr_two;                   /* cookie, second word */
 161         *p++ = xdr_two;                             /* entry len */
 162         memcpy(p, "..\0\0", 4);                         /* entry */
 163         p++;
 164         *p++ = xdr_one;                         /* bitmap length */
 165         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 166         *p++ = htonl(8);              /* attribute buffer length */
 167         p = xdr_encode_hyper(p, dentry->d_parent->d_inode->i_ino);
 168
 169         readdir->pgbase = (char *)p - (char *)start;
 170         readdir->count -= readdir->pgbase;
 171         kunmap_atomic(start, KM_USER0);
 172 }
 173
 174 static void
 175 renew_lease(struct nfs_server *server, unsigned long timestamp)
 176 {
 177         struct nfs4_client *clp = server->nfs4_state;
 178         spin_lock(&clp->cl_lock);
 179         if (time_before(clp->cl_last_renewal,timestamp))
 180                 clp->cl_last_renewal = timestamp;
 181         spin_unlock(&clp->cl_lock);
 182 }
 183
 184 static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinfo)
 185 {
 186         struct nfs_inode *nfsi = NFS_I(inode);
 187
 188         if (cinfo->before == nfsi->change_attr && cinfo->atomic)
 189                 nfsi->change_attr = cinfo->after;
 190 }
 191
 192 static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
 193 {
 194         struct inode *inode = state->inode;
 195
 196         open_flags &= (FMODE_READ|FMODE_WRITE);
 197         /* Protect against nfs4_find_state() */
 198         spin_lock(&inode->i_lock);
 199         state->state |= open_flags;
 200         /* NB! List reordering - see the reclaim code for why.  */
 201         if ((open_flags & FMODE_WRITE) && 0 == state->nwriters++)
 202                 list_move(&state->open_states, &state->owner->so_states);
 203         if (open_flags & FMODE_READ)
 204                 state->nreaders++;
 205         memcpy(&state->stateid, stateid, sizeof(state->stateid));
 206         spin_unlock(&inode->i_lock);
 207 }
 208
 209 /*
 210  * OPEN_RECLAIM:
 211  *      reclaim state on the server after a reboot.
 212  *      Assumes caller is holding the sp->so_sem
 213  */
 214 static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
 215 {
 216         struct inode *inode = state->inode;
 217         struct nfs_server *server = NFS_SERVER(inode);
 218         struct nfs_delegation *delegation = NFS_I(inode)->delegation;
 219         struct nfs_openargs o_arg = {
 220                 .fh = NFS_FH(inode),
 221                 .seqid = sp->so_seqid,
 222                 .id = sp->so_id,
 223                 .open_flags = state->state,
 224                 .clientid = server->nfs4_state->cl_clientid,
 225                 .claim = NFS4_OPEN_CLAIM_PREVIOUS,
 226                 .bitmask = server->attr_bitmask,
 227         };
 228         struct nfs_openres o_res = {
 229                 .server = server,       /* Grrr */
 230         };
 231         struct rpc_message msg = {
 232                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
 233                 .rpc_argp       = &o_arg,
 234                 .rpc_resp       = &o_res,
 235                 .rpc_cred       = sp->so_cred,
 236         };
 237         int status;
 238
 239         if (delegation != NULL) {
 240                 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
 241                         memcpy(&state->stateid, &delegation->stateid,
 242                                         sizeof(state->stateid));
 243                         set_bit(NFS_DELEGATED_STATE, &state->flags);
 244                         return 0;
 245                 }
 246                 o_arg.u.delegation_type = delegation->type;
 247         }
 248         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
 249         nfs4_increment_seqid(status, sp);
 250         if (status == 0) {
 251                 memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
 252                 if (o_res.delegation_type != 0) {
 253                         nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
 254                         /* Did the server issue an immediate delegation recall? */
 255                         if (o_res.do_recall)
 256                                 nfs_async_inode_return_delegation(inode, &o_res.stateid);
 257                 }
 258         }
 259         clear_bit(NFS_DELEGATED_STATE, &state->flags);
 260         /* Ensure we update the inode attributes */
 261         NFS_CACHEINV(inode);
 262         return status;
 263 }
 264
 265 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
 266 {
 267         struct nfs_server *server = NFS_SERVER(state->inode);
 268         struct nfs4_exception exception = { };
 269         int err;
 270         do {
 271                 err = _nfs4_open_reclaim(sp, state);
 272                 if (err != -NFS4ERR_DELAY)
 273                         break;
 274                 nfs4_handle_exception(server, err, &exception);
 275         } while (exception.retry);
 276         return err;
 277 }
 278
 279 static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
 280 {
 281         struct nfs4_state_owner  *sp  = state->owner;
 282         struct inode *inode = dentry->d_inode;
 283         struct nfs_server *server = NFS_SERVER(inode);
 284         struct dentry *parent = dget_parent(dentry);
 285         struct nfs_openargs arg = {
 286                 .fh = NFS_FH(parent->d_inode),
 287                 .clientid = server->nfs4_state->cl_clientid,
 288                 .name = &dentry->d_name,
 289                 .id = sp->so_id,
 290                 .server = server,
 291                 .bitmask = server->attr_bitmask,
 292                 .claim = NFS4_OPEN_CLAIM_DELEGATE_CUR,
 293         };
 294         struct nfs_openres res = {
 295                 .server = server,
 296         };
 297         struct  rpc_message msg = {
 298                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
 299                 .rpc_argp       = &arg,
 300                 .rpc_resp       = &res,
 301                 .rpc_cred       = sp->so_cred,
 302         };
 303         int status = 0;
 304
 305         down(&sp->so_sema);
 306         if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
 307                 goto out;
 308         if (state->state == 0)
 309                 goto out;
 310         arg.seqid = sp->so_seqid;
 311         arg.open_flags = state->state;
 312         memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data));
 313         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
 314         nfs4_increment_seqid(status, sp);
 315         if (status >= 0) {
 316                 memcpy(state->stateid.data, res.stateid.data,
 317                                 sizeof(state->stateid.data));
 318                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
 319         }
 320 out:
 321         up(&sp->so_sema);
 322         dput(parent);
 323         return status;
 324 }
 325
 326 int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
 327 {
 328         struct nfs4_exception exception = { };
 329         struct nfs_server *server = NFS_SERVER(dentry->d_inode);
 330         int err;
 331         do {
 332                 err = _nfs4_open_delegation_recall(dentry, state);
 333                 switch (err) {
 334                         case 0:
 335                                 return err;
 336                         case -NFS4ERR_STALE_CLIENTID:
 337                         case -NFS4ERR_STALE_STATEID:
 338                         case -NFS4ERR_EXPIRED:
 339                                 /* Don't recall a delegation if it was lost */
 340                                 nfs4_schedule_state_recovery(server->nfs4_state);
 341                                 return err;
 342                 }
 343                 err = nfs4_handle_exception(server, err, &exception);
 344         } while (exception.retry);
 345         return err;
 346 }
 347
 348 static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid)
 349 {
 350         struct nfs_open_confirmargs arg = {
 351                 .fh             = fh,
 352                 .seqid          = sp->so_seqid,
 353                 .stateid        = *stateid,
 354         };
 355         struct nfs_open_confirmres res;
 356         struct  rpc_message msg = {
 357                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
 358                 .rpc_argp       = &arg,
 359                 .rpc_resp       = &res,
 360                 .rpc_cred       = sp->so_cred,
 361         };
 362         int status;
 363
 364         status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR);
 365         nfs4_increment_seqid(status, sp);
 366         if (status >= 0)
 367                 memcpy(stateid, &res.stateid, sizeof(*stateid));
 368         return status;
 369 }
 370
 371 static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner  *sp, struct nfs_openargs *o_arg, struct nfs_openres *o_res)
 372 {
 373         struct nfs_server *server = NFS_SERVER(dir);
 374         struct rpc_message msg = {
 375                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
 376                 .rpc_argp = o_arg,
 377                 .rpc_resp = o_res,
 378                 .rpc_cred = sp->so_cred,
 379         };
 380         int status;
 381
 382         /* Update sequence id. The caller must serialize! */
 383         o_arg->seqid = sp->so_seqid;
 384         o_arg->id = sp->so_id;
 385         o_arg->clientid = sp->so_client->cl_clientid;
 386
 387         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
 388         nfs4_increment_seqid(status, sp);
 389         if (status != 0)
 390                 goto out;
 391         update_changeattr(dir, &o_res->cinfo);
 392         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
 393                 status = _nfs4_proc_open_confirm(server->client, &o_res->fh,
 394                                 sp, &o_res->stateid);
 395                 if (status != 0)
 396                         goto out;
 397         }
 398         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
 399                 status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
 400 out:
 401         return status;
 402 }
 403
 404 static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openflags)
 405 {
 406         struct nfs_access_entry cache;
 407         int mask = 0;
 408         int status;
 409
 410         if (openflags & FMODE_READ)
 411                 mask |= MAY_READ;
 412         if (openflags & FMODE_WRITE)
 413                 mask |= MAY_WRITE;
 414         status = nfs_access_get_cached(inode, cred, &cache);
 415         if (status == 0)
 416                 goto out;
 417
 418         /* Be clever: ask server to check for all possible rights */
 419         cache.mask = MAY_EXEC | MAY_WRITE | MAY_READ;
 420         cache.cred = cred;
 421         cache.jiffies = jiffies;
 422         status = _nfs4_proc_access(inode, &cache);
 423         if (status != 0)
 424                 return status;
 425         nfs_access_add_cache(inode, &cache);
 426 out:
 427         if ((cache.mask & mask) == mask)
 428                 return 0;
 429         return -EACCES;
 430 }
 431
 432 /*
 433  * OPEN_EXPIRED:
 434  *      reclaim state on the server after a network partition.
 435  *      Assumes caller holds the appropriate lock
 436  */
 437 static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
 438 {
 439         struct dentry *parent = dget_parent(dentry);
 440         struct inode *dir = parent->d_inode;
 441         struct inode *inode = state->inode;
 442         struct nfs_server *server = NFS_SERVER(dir);
 443         struct nfs_delegation *delegation = NFS_I(inode)->delegation;
 444         struct nfs_fattr        f_attr = {
 445                 .valid = 0,
 446         };
 447         struct nfs_openargs o_arg = {
 448                 .fh = NFS_FH(dir),
 449                 .open_flags = state->state,
 450                 .name = &dentry->d_name,
 451                 .bitmask = server->attr_bitmask,
 452                 .claim = NFS4_OPEN_CLAIM_NULL,
 453         };
 454         struct nfs_openres o_res = {
 455                 .f_attr = &f_attr,
 456                 .server = server,
 457         };
 458         int status = 0;
 459
 460         if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
 461                 status = _nfs4_do_access(inode, sp->so_cred, state->state);
 462                 if (status < 0)
 463                         goto out;
 464                 memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid));
 465                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 466                 goto out;
 467         }
 468         status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
 469         if (status != 0)
 470                 goto out_nodeleg;
 471         /* Check if files differ */
 472         if ((f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT))
 473                 goto out_stale;
 474         /* Has the file handle changed? */
 475         if (nfs_compare_fh(&o_res.fh, NFS_FH(inode)) != 0) {
 476                 /* Verify if the change attributes are the same */
 477                 if (f_attr.change_attr != NFS_I(inode)->change_attr)
 478                         goto out_stale;
 479                 if (nfs_size_to_loff_t(f_attr.size) != inode->i_size)
 480                         goto out_stale;
 481                 /* Lets just pretend that this is the same file */
 482                 nfs_copy_fh(NFS_FH(inode), &o_res.fh);
 483                 NFS_I(inode)->fileid = f_attr.fileid;
 484         }
 485         memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
 486         if (o_res.delegation_type != 0) {
 487                 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM))
 488                         nfs_inode_set_delegation(inode, sp->so_cred, &o_res);
 489                 else
 490                         nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
 491         }
 492 out_nodeleg:
 493         clear_bit(NFS_DELEGATED_STATE, &state->flags);
 494 out:
 495         dput(parent);
 496         return status;
 497 out_stale:
 498         status = -ESTALE;
 499         /* Invalidate the state owner so we don't ever use it again */
 500         nfs4_drop_state_owner(sp);
 501         d_drop(dentry);
 502         /* Should we be trying to close that stateid? */
 503         goto out_nodeleg;
 504 }
 505
 506 static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
 507 {
 508         struct nfs_server *server = NFS_SERVER(dentry->d_inode);
 509         struct nfs4_exception exception = { };
 510         int err;
 511
 512         do {
 513                 err = _nfs4_open_expired(sp, state, dentry);
 514                 if (err == -NFS4ERR_DELAY)
 515                         nfs4_handle_exception(server, err, &exception);
 516         } while (exception.retry);
 517         return err;
 518 }
 519
 520 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
 521 {
 522         struct nfs_inode *nfsi = NFS_I(state->inode);
 523         struct nfs_open_context *ctx;
 524         int status;
 525
 526         spin_lock(&state->inode->i_lock);
 527         list_for_each_entry(ctx, &nfsi->open_files, list) {
 528                 if (ctx->state != state)
 529                         continue;
 530                 get_nfs_open_context(ctx);
 531                 spin_unlock(&state->inode->i_lock);
 532                 status = nfs4_do_open_expired(sp, state, ctx->dentry);
 533                 put_nfs_open_context(ctx);
 534                 return status;
 535         }
 536         spin_unlock(&state->inode->i_lock);
 537         return -ENOENT;
 538 }
 539
 540 /*
 541  * Returns an nfs4_state + an extra reference to the inode
 542  */
 543 static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred, struct nfs4_state **res)
 544 {
 545         struct nfs_delegation *delegation;
 546         struct nfs_server *server = NFS_SERVER(inode);
 547         struct nfs4_client *clp = server->nfs4_state;
 548         struct nfs_inode *nfsi = NFS_I(inode);
 549         struct nfs4_state_owner *sp = NULL;
 550         struct nfs4_state *state = NULL;
 551         int open_flags = flags & (FMODE_READ|FMODE_WRITE);
 552         int err;
 553
 554         /* Protect against reboot recovery - NOTE ORDER! */
 555         down_read(&clp->cl_sem);
 556         /* Protect against delegation recall */
 557         down_read(&nfsi->rwsem);
 558         delegation = NFS_I(inode)->delegation;
 559         err = -ENOENT;
 560         if (delegation == NULL || (delegation->type & open_flags) != open_flags)
 561                 goto out_err;
 562         err = -ENOMEM;
 563         if (!(sp = nfs4_get_state_owner(server, cred))) {
 564                 dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
 565                 goto out_err;
 566         }
 567         down(&sp->so_sema);
 568         state = nfs4_get_open_state(inode, sp);
 569         if (state == NULL)
 570                 goto out_err;
 571
 572         err = -ENOENT;
 573         if ((state->state & open_flags) == open_flags) {
 574                 spin_lock(&inode->i_lock);
 575                 if (open_flags & FMODE_READ)
 576                         state->nreaders++;
 577                 if (open_flags & FMODE_WRITE)
 578                         state->nwriters++;
 579                 spin_unlock(&inode->i_lock);
 580                 goto out_ok;
 581         } else if (state->state != 0)
 582                 goto out_err;
 583
 584         lock_kernel();
 585         err = _nfs4_do_access(inode, cred, open_flags);
 586         unlock_kernel();
 587         if (err != 0)
 588                 goto out_err;
 589         set_bit(NFS_DELEGATED_STATE, &state->flags);
 590         update_open_stateid(state, &delegation->stateid, open_flags);
 591 out_ok:
 592         up(&sp->so_sema);
 593         nfs4_put_state_owner(sp);
 594         up_read(&nfsi->rwsem);
 595         up_read(&clp->cl_sem);
 596         igrab(inode);
 597         *res = state;
 598         return 0;
 599 out_err:
 600         if (sp != NULL) {
 601                 if (state != NULL)
 602                         nfs4_put_open_state(state);
 603                 up(&sp->so_sema);
 604                 nfs4_put_state_owner(sp);
 605         }
 606         up_read(&nfsi->rwsem);
 607         up_read(&clp->cl_sem);
 608         return err;
 609 }
 610
 611 static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred)
 612 {
 613         struct nfs4_exception exception = { };
 614         struct nfs4_state *res;
 615         int err;
 616
 617         do {
 618                 err = _nfs4_open_delegated(inode, flags, cred, &res);
 619                 if (err == 0)
 620                         break;
 621                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(inode),
 622                                         err, &exception));
 623         } while (exception.retry);
 624         return res;
 625 }
 626
 627 /*
 628  * Returns an nfs4_state + an referenced inode
 629  */
 630 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
 631 {
 632         struct nfs4_state_owner  *sp;
 633         struct nfs4_state     *state = NULL;
 634         struct nfs_server       *server = NFS_SERVER(dir);
 635         struct nfs4_client *clp = server->nfs4_state;
 636         struct inode *inode = NULL;
 637         int                     status;
 638         struct nfs_fattr        f_attr = {
 639                 .valid          = 0,
 640         };
 641         struct nfs_openargs o_arg = {
 642                 .fh             = NFS_FH(dir),
 643                 .open_flags     = flags,
 644                 .name           = &dentry->d_name,
 645                 .server         = server,
 646                 .bitmask = server->attr_bitmask,
 647                 .claim = NFS4_OPEN_CLAIM_NULL,
 648         };
 649         struct nfs_openres o_res = {
 650                 .f_attr         = &f_attr,
 651                 .server         = server,
 652         };
 653
 654         /* Protect against reboot recovery conflicts */
 655         down_read(&clp->cl_sem);
 656         status = -ENOMEM;
 657         if (!(sp = nfs4_get_state_owner(server, cred))) {
 658                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
 659                 goto out_err;
 660         }
 661         if (flags & O_EXCL) {
 662                 u32 *p = (u32 *) o_arg.u.verifier.data;
 663                 p[0] = jiffies;
 664                 p[1] = current->pid;
 665         } else
 666                 o_arg.u.attrs = sattr;
 667         /* Serialization for the sequence id */
 668         down(&sp->so_sema);
 669
 670         status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
 671         if (status != 0)
 672                 goto out_err;
 673
 674         status = -ENOMEM;
 675         inode = nfs_fhget(dir->i_sb, &o_res.fh, &f_attr);
 676         if (!inode)
 677                 goto out_err;
 678         state = nfs4_get_open_state(inode, sp);
 679         if (!state)
 680                 goto out_err;
 681         update_open_stateid(state, &o_res.stateid, flags);
 682         if (o_res.delegation_type != 0)
 683                 nfs_inode_set_delegation(inode, cred, &o_res);
 684         up(&sp->so_sema);
 685         nfs4_put_state_owner(sp);
 686         up_read(&clp->cl_sem);
 687         *res = state;
 688         return 0;
 689 out_err:
 690         if (sp != NULL) {
 691                 if (state != NULL)
 692                         nfs4_put_open_state(state);
 693                 up(&sp->so_sema);
 694                 nfs4_put_state_owner(sp);
 695         }
 696         /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */
 697         up_read(&clp->cl_sem);
 698         if (inode != NULL)
 699                 iput(inode);
 700         *res = NULL;
 701         return status;
 702 }
 703
 704
 705 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred)
 706 {
 707         struct nfs4_exception exception = { };
 708         struct nfs4_state *res;
 709         int status;
 710
 711         do {
 712                 status = _nfs4_do_open(dir, dentry, flags, sattr, cred, &res);
 713                 if (status == 0)
 714                         break;
 715                 /* NOTE: BAD_SEQID means the server and client disagree about the
 716                  * book-keeping w.r.t. state-changing operations
 717                  * (OPEN/CLOSE/LOCK/LOCKU...)
 718                  * It is actually a sign of a bug on the client or on the server.
 719                  *
 720                  * If we receive a BAD_SEQID error in the particular case of
 721                  * doing an OPEN, we assume that nfs4_increment_seqid() will
 722                  * have unhashed the old state_owner for us, and that we can
 723                  * therefore safely retry using a new one. We should still warn
 724                  * the user though...
 725                  */
 726                 if (status == -NFS4ERR_BAD_SEQID) {
 727                         printk(KERN_WARNING "NFS: v4 server returned a bad sequence-id error!\n");
 728                         exception.retry = 1;
 729                         continue;
 730                 }
 731                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
 732                                         status, &exception));
 733         } while (exception.retry);
 734         return res;
 735 }
 736
 737 static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
 738                 struct nfs_fh *fhandle, struct iattr *sattr,
 739                 struct nfs4_state *state)
 740 {
 741         struct nfs_setattrargs  arg = {
 742                 .fh             = fhandle,
 743                 .iap            = sattr,
 744                 .server         = server,
 745                 .bitmask = server->attr_bitmask,
 746         };
 747         struct nfs_setattrres  res = {
 748                 .fattr          = fattr,
 749                 .server         = server,
 750         };
 751         struct rpc_message msg = {
 752                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
 753                 .rpc_argp       = &arg,
 754                 .rpc_resp       = &res,
 755         };
 756         int status;
 757
 758         fattr->valid = 0;
 759
 760         if (state != NULL) {
 761                 msg.rpc_cred = state->owner->so_cred;
 762                 nfs4_copy_stateid(&arg.stateid, state, current->files);
 763         } else
 764                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
 765
 766         status = rpc_call_sync(server->client, &msg, 0);
 767         return status;
 768 }
 769
 770 static int nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
 771                 struct nfs_fh *fhandle, struct iattr *sattr,
 772                 struct nfs4_state *state)
 773 {
 774         struct nfs4_exception exception = { };
 775         int err;
 776         do {
 777                 err = nfs4_handle_exception(server,
 778                                 _nfs4_do_setattr(server, fattr, fhandle, sattr,
 779                                         state),
 780                                 &exception);
 781         } while (exception.retry);
 782         return err;
 783 }
 784
 785 struct nfs4_closedata {
 786         struct inode *inode;
 787         struct nfs4_state *state;
 788         struct nfs_closeargs arg;
 789         struct nfs_closeres res;
 790 };
 791
 792 static void nfs4_close_done(struct rpc_task *task)
 793 {
 794         struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
 795         struct nfs4_state *state = calldata->state;
 796         struct nfs4_state_owner *sp = state->owner;
 797         struct nfs_server *server = NFS_SERVER(calldata->inode);
 798
 799         /* hmm. we are done with the inode, and in the process of freeing
 800          * the state_owner. we keep this around to process errors
 801          */
 802         nfs4_increment_seqid(task->tk_status, sp);
 803         switch (task->tk_status) {
 804                 case 0:
 805                         memcpy(&state->stateid, &calldata->res.stateid,
 806                                         sizeof(state->stateid));
 807                         break;
 808                 case -NFS4ERR_STALE_STATEID:
 809                 case -NFS4ERR_EXPIRED:
 810                         state->state = calldata->arg.open_flags;
 811                         nfs4_schedule_state_recovery(server->nfs4_state);
 812                         break;
 813                 default:
 814                         if (nfs4_async_handle_error(task, server) == -EAGAIN) {
 815                                 rpc_restart_call(task);
 816                                 return;
 817                         }
 818         }
 819         state->state = calldata->arg.open_flags;
 820         nfs4_put_open_state(state);
 821         up(&sp->so_sema);
 822         nfs4_put_state_owner(sp);
 823         up_read(&server->nfs4_state->cl_sem);
 824         kfree(calldata);
 825 }
 826
 827 static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata)
 828 {
 829         struct rpc_message msg = {
 830                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
 831                 .rpc_argp = &calldata->arg,
 832                 .rpc_resp = &calldata->res,
 833                 .rpc_cred = calldata->state->owner->so_cred,
 834         };
 835         if (calldata->arg.open_flags != 0)
 836                 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
 837         return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata);
 838 }
 839
 840 /*
 841  * It is possible for data to be read/written from a mem-mapped file
 842  * after the sys_close call (which hits the vfs layer as a flush).
 843  * This means that we can't safely call nfsv4 close on a file until
 844  * the inode is cleared. This in turn means that we are not good
 845  * NFSv4 citizens - we do not indicate to the server to update the file's
 846  * share state even when we are done with one of the three share
 847  * stateid's in the inode.
 848  *
 849  * NOTE: Caller must be holding the sp->so_owner semaphore!
 850  */
 851 int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode)
 852 {
 853         struct nfs4_closedata *calldata;
 854         int status;
 855
 856         /* Tell caller we're done */
 857         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
 858                 state->state = mode;
 859                 return 0;
 860         }
 861         calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL);
 862         if (calldata == NULL)
 863                 return -ENOMEM;
 864         calldata->inode = inode;
 865         calldata->state = state;
 866         calldata->arg.fh = NFS_FH(inode);
 867         /* Serialization for the sequence id */
 868         calldata->arg.seqid = state->owner->so_seqid;
 869         calldata->arg.open_flags = mode;
 870         memcpy(&calldata->arg.stateid, &state->stateid,
 871                         sizeof(calldata->arg.stateid));
 872         status = nfs4_close_call(NFS_SERVER(inode)->client, calldata);
 873         /*
 874          * Return -EINPROGRESS on success in order to indicate to the
 875          * caller that an asynchronous RPC call has been launched, and
 876          * that it will release the semaphores on completion.
 877          */
 878         return (status == 0) ? -EINPROGRESS : status;
 879 }
 880
 881 struct inode *
 882 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
 883 {
 884         struct iattr attr;
 885         struct rpc_cred *cred;
 886         struct nfs4_state *state;
 887
 888         if (nd->flags & LOOKUP_CREATE) {
 889                 attr.ia_mode = nd->intent.open.create_mode;
 890                 attr.ia_valid = ATTR_MODE;
 891                 if (!IS_POSIXACL(dir))
 892                         attr.ia_mode &= ~current->fs->umask;
 893         } else {
 894                 attr.ia_valid = 0;
 895                 BUG_ON(nd->intent.open.flags & O_CREAT);
 896         }
 897
 898         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
 899         if (IS_ERR(cred))
 900                 return (struct inode *)cred;
 901         state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred);
 902         put_rpccred(cred);
 903         if (IS_ERR(state))
 904                 return (struct inode *)state;
 905         return state->inode;
 906 }
 907
 908 int
 909 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags)
 910 {
 911         struct rpc_cred *cred;
 912         struct nfs4_state *state;
 913         struct inode *inode;
 914
 915         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
 916         if (IS_ERR(cred))
 917                 return PTR_ERR(cred);
 918         state = nfs4_open_delegated(dentry->d_inode, openflags, cred);
 919         if (IS_ERR(state))
 920                 state = nfs4_do_open(dir, dentry, openflags, NULL, cred);
 921         put_rpccred(cred);
 922         if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0)
 923                 return 1;
 924         if (IS_ERR(state))
 925                 return 0;
 926         inode = state->inode;
 927         if (inode == dentry->d_inode) {
 928                 iput(inode);
 929                 return 1;
 930         }
 931         d_drop(dentry);
 932         nfs4_close_state(state, openflags);
 933         iput(inode);
 934         return 0;
 935 }
 936
 937
 938 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
 939 {
 940         struct nfs4_server_caps_res res = {};
 941         struct rpc_message msg = {
 942                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
 943                 .rpc_argp = fhandle,
 944                 .rpc_resp = &res,
 945         };
 946         int status;
 947
 948         status = rpc_call_sync(server->client, &msg, 0);
 949         if (status == 0) {
 950                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
 951                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
 952                         server->caps |= NFS_CAP_ACLS;
 953                 if (res.has_links != 0)
 954                         server->caps |= NFS_CAP_HARDLINKS;
 955                 if (res.has_symlinks != 0)
 956                         server->caps |= NFS_CAP_SYMLINKS;
 957                 server->acl_bitmask = res.acl_bitmask;
 958         }
 959         return status;
 960 }
 961
 962 static int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
 963 {
 964         struct nfs4_exception exception = { };
 965         int err;
 966         do {
 967                 err = nfs4_handle_exception(server,
 968                                 _nfs4_server_capabilities(server, fhandle),
 969                                 &exception);
 970         } while (exception.retry);
 971         return err;
 972 }
 973
 974 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
 975                 struct nfs_fsinfo *info)
 976 {
 977         struct nfs_fattr *      fattr = info->fattr;
 978         struct nfs4_lookup_root_arg args = {
 979                 .bitmask = nfs4_fattr_bitmap,
 980         };
 981         struct nfs4_lookup_res res = {
 982                 .server = server,
 983                 .fattr = fattr,
 984                 .fh = fhandle,
 985         };
 986         struct rpc_message msg = {
 987                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
 988                 .rpc_argp = &args,
 989                 .rpc_resp = &res,
 990         };
 991         fattr->valid = 0;
 992         return rpc_call_sync(server->client, &msg, 0);
 993 }
 994
 995 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
 996                 struct nfs_fsinfo *info)
 997 {
 998         struct nfs4_exception exception = { };
 999         int err;
1000         do {
1001                 err = nfs4_handle_exception(server,
1002                                 _nfs4_lookup_root(server, fhandle, info),
1003                                 &exception);
1004         } while (exception.retry);
1005         return err;
1006 }
1007
1008 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
1009                 struct nfs_fsinfo *info)
1010 {
1011         struct nfs_fattr *      fattr = info->fattr;
1012         unsigned char *         p;
1013         struct qstr             q;
1014         struct nfs4_lookup_arg args = {
1015                 .dir_fh = fhandle,
1016                 .name = &q,
1017                 .bitmask = nfs4_fattr_bitmap,
1018         };
1019         struct nfs4_lookup_res res = {
1020                 .server = server,
1021                 .fattr = fattr,
1022                 .fh = fhandle,
1023         };
1024         struct rpc_message msg = {
1025                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1026                 .rpc_argp = &args,
1027                 .rpc_resp = &res,
1028         };
1029         int status;
1030
1031         /*
1032          * Now we do a separate LOOKUP for each component of the mount path.
1033          * The LOOKUPs are done separately so that we can conveniently
1034          * catch an ERR_WRONGSEC if it occurs along the way...
1035          */
1036         status = nfs4_lookup_root(server, fhandle, info);
1037         if (status)
1038                 goto out;
1039
1040         p = server->mnt_path;
1041         for (;;) {
1042                 struct nfs4_exception exception = { };
1043
1044                 while (*p == '/')
1045                         p++;
1046                 if (!*p)
1047                         break;
1048                 q.name = p;
1049                 while (*p && (*p != '/'))
1050                         p++;
1051                 q.len = p - q.name;
1052
1053                 do {
1054                         fattr->valid = 0;
1055                         status = nfs4_handle_exception(server,
1056                                         rpc_call_sync(server->client, &msg, 0),
1057                                         &exception);
1058                 } while (exception.retry);
1059                 if (status == 0)
1060                         continue;
1061                 if (status == -ENOENT) {
1062                         printk(KERN_NOTICE "NFS: mount path %s does not exist!\n", server->mnt_path);
1063                         printk(KERN_NOTICE "NFS: suggestion: try mounting '/' instead.\n");
1064                 }
1065                 break;
1066         }
1067         if (status == 0)
1068                 status = nfs4_server_capabilities(server, fhandle);
1069         if (status == 0)
1070                 status = nfs4_do_fsinfo(server, fhandle, info);
1071 out:
1072         return status;
1073 }
1074
1075 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1076 {
1077         struct nfs4_getattr_arg args = {
1078                 .fh = fhandle,
1079                 .bitmask = server->attr_bitmask,
1080         };
1081         struct nfs4_getattr_res res = {
1082                 .fattr = fattr,
1083                 .server = server,
1084         };
1085         struct rpc_message msg = {
1086                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
1087                 .rpc_argp = &args,
1088                 .rpc_resp = &res,
1089         };
1090
1091         fattr->valid = 0;
1092         return rpc_call_sync(server->client, &msg, 0);
1093 }
1094
1095 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1096 {
1097         struct nfs4_exception exception = { };
1098         int err;
1099         do {
1100                 err = nfs4_handle_exception(server,
1101                                 _nfs4_proc_getattr(server, fhandle, fattr),
1102                                 &exception);
1103         } while (exception.retry);
1104         return err;
1105 }
1106
1107 /*
1108  * The file is not closed if it is opened due to the a request to change
1109  * the size of the file. The open call will not be needed once the
1110  * VFS layer lookup-intents are implemented.
1111  *
1112  * Close is called when the inode is destroyed.
1113  * If we haven't opened the file for O_WRONLY, we
1114  * need to in the size_change case to obtain a stateid.
1115  *
1116  * Got race?
1117  * Because OPEN is always done by name in nfsv4, it is
1118  * possible that we opened a different file by the same
1119  * name.  We can recognize this race condition, but we
1120  * can't do anything about it besides returning an error.
1121  *
1122  * This will be fixed with VFS changes (lookup-intent).
1123  */
1124 static int
1125 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
1126                   struct iattr *sattr)
1127 {
1128         struct rpc_cred *cred;
1129         struct inode *inode = dentry->d_inode;
1130         struct nfs4_state *state;
1131         int status;
1132
1133         fattr->valid = 0;
1134
1135         cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
1136         if (IS_ERR(cred))
1137                 return PTR_ERR(cred);
1138         /* Search for an existing WRITE delegation first */
1139         state = nfs4_open_delegated(inode, FMODE_WRITE, cred);
1140         if (!IS_ERR(state)) {
1141                 /* NB: nfs4_open_delegated() bumps the inode->i_count */
1142                 iput(inode);
1143         } else {
1144                 /* Search for an existing open(O_WRITE) stateid */
1145                 state = nfs4_find_state(inode, cred, FMODE_WRITE);
1146         }
1147
1148         status = nfs4_do_setattr(NFS_SERVER(inode), fattr,
1149                         NFS_FH(inode), sattr, state);
1150         if (status == 0)
1151                 nfs_setattr_update_inode(inode, sattr);
1152         if (state != NULL)
1153                 nfs4_close_state(state, FMODE_WRITE);
1154         put_rpccred(cred);
1155         return status;
1156 }
1157
1158 static int _nfs4_proc_lookup(struct inode *dir, struct qstr *name,
1159                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1160 {
1161         int                    status;
1162         struct nfs_server *server = NFS_SERVER(dir);
1163         struct nfs4_lookup_arg args = {
1164                 .bitmask = server->attr_bitmask,
1165                 .dir_fh = NFS_FH(dir),
1166                 .name = name,
1167         };
1168         struct nfs4_lookup_res res = {
1169                 .server = server,
1170                 .fattr = fattr,
1171                 .fh = fhandle,
1172         };
1173         struct rpc_message msg = {
1174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1175                 .rpc_argp = &args,
1176                 .rpc_resp = &res,
1177         };
1178
1179         fattr->valid = 0;
1180
1181         dprintk("NFS call  lookup %s\n", name->name);
1182         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1183         dprintk("NFS reply lookup: %d\n", status);
1184         return status;
1185 }
1186
1187 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1188 {
1189         struct nfs4_exception exception = { };
1190         int err;
1191         do {
1192                 err = nfs4_handle_exception(NFS_SERVER(dir),
1193                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
1194                                 &exception);
1195         } while (exception.retry);
1196         return err;
1197 }
1198
1199 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
1200 {
1201         struct nfs4_accessargs args = {
1202                 .fh = NFS_FH(inode),
1203         };
1204         struct nfs4_accessres res = { 0 };
1205         struct rpc_message msg = {
1206                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
1207                 .rpc_argp = &args,
1208                 .rpc_resp = &res,
1209                 .rpc_cred = entry->cred,
1210         };
1211         int mode = entry->mask;
1212         int status;
1213
1214         /*
1215          * Determine which access bits we want to ask for...
1216          */
1217         if (mode & MAY_READ)
1218                 args.access |= NFS4_ACCESS_READ;
1219         if (S_ISDIR(inode->i_mode)) {
1220                 if (mode & MAY_WRITE)
1221                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
1222                 if (mode & MAY_EXEC)
1223                         args.access |= NFS4_ACCESS_LOOKUP;
1224         } else {
1225                 if (mode & MAY_WRITE)
1226                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
1227                 if (mode & MAY_EXEC)
1228                         args.access |= NFS4_ACCESS_EXECUTE;
1229         }
1230         status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1231         if (!status) {
1232                 entry->mask = 0;
1233                 if (res.access & NFS4_ACCESS_READ)
1234                         entry->mask |= MAY_READ;
1235                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
1236                         entry->mask |= MAY_WRITE;
1237                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
1238                         entry->mask |= MAY_EXEC;
1239         }
1240         return status;
1241 }
1242
1243 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
1244 {
1245         struct nfs4_exception exception = { };
1246         int err;
1247         do {
1248                 err = nfs4_handle_exception(NFS_SERVER(inode),
1249                                 _nfs4_proc_access(inode, entry),
1250                                 &exception);
1251         } while (exception.retry);
1252         return err;
1253 }
1254
1255 /*
1256  * TODO: For the time being, we don't try to get any attributes
1257  * along with any of the zero-copy operations READ, READDIR,
1258  * READLINK, WRITE.
1259  *
1260  * In the case of the first three, we want to put the GETATTR
1261  * after the read-type operation -- this is because it is hard
1262  * to predict the length of a GETATTR response in v4, and thus
1263  * align the READ data correctly.  This means that the GETATTR
1264  * may end up partially falling into the page cache, and we should
1265  * shift it into the 'tail' of the xdr_buf before processing.
1266  * To do this efficiently, we need to know the total length
1267  * of data received, which doesn't seem to be available outside
1268  * of the RPC layer.
1269  *
1270  * In the case of WRITE, we also want to put the GETATTR after
1271  * the operation -- in this case because we want to make sure
1272  * we get the post-operation mtime and size.  This means that
1273  * we can't use xdr_encode_pages() as written: we need a variant
1274  * of it which would leave room in the 'tail' iovec.
1275  *
1276  * Both of these changes to the XDR layer would in fact be quite
1277  * minor, but I decided to leave them for a subsequent patch.
1278  */
1279 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
1280                 unsigned int pgbase, unsigned int pglen)
1281 {
1282         struct nfs4_readlink args = {
1283                 .fh       = NFS_FH(inode),
1284                 .pgbase   = pgbase,
1285                 .pglen    = pglen,
1286                 .pages    = &page,
1287         };
1288         struct rpc_message msg = {
1289                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
1290                 .rpc_argp = &args,
1291                 .rpc_resp = NULL,
1292         };
1293
1294         return rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1295 }
1296
1297 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
1298                 unsigned int pgbase, unsigned int pglen)
1299 {
1300         struct nfs4_exception exception = { };
1301         int err;
1302         do {
1303                 err = nfs4_handle_exception(NFS_SERVER(inode),
1304                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
1305                                 &exception);
1306         } while (exception.retry);
1307         return err;
1308 }
1309
1310 static int _nfs4_proc_read(struct nfs_read_data *rdata)
1311 {
1312         int flags = rdata->flags;
1313         struct inode *inode = rdata->inode;
1314         struct nfs_fattr *fattr = rdata->res.fattr;
1315         struct nfs_server *server = NFS_SERVER(inode);
1316         struct rpc_message msg = {
1317                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_READ],
1318                 .rpc_argp       = &rdata->args,
1319                 .rpc_resp       = &rdata->res,
1320                 .rpc_cred       = rdata->cred,
1321         };
1322         unsigned long timestamp = jiffies;
1323         int status;
1324
1325         dprintk("NFS call  read %d @ %Ld\n", rdata->args.count,
1326                         (long long) rdata->args.offset);
1327
1328         fattr->valid = 0;
1329         status = rpc_call_sync(server->client, &msg, flags);
1330         if (!status)
1331                 renew_lease(server, timestamp);
1332         dprintk("NFS reply read: %d\n", status);
1333         return status;
1334 }
1335
1336 static int nfs4_proc_read(struct nfs_read_data *rdata)
1337 {
1338         struct nfs4_exception exception = { };
1339         int err;
1340         do {
1341                 err = nfs4_handle_exception(NFS_SERVER(rdata->inode),
1342                                 _nfs4_proc_read(rdata),
1343                                 &exception);
1344         } while (exception.retry);
1345         return err;
1346 }
1347
1348 static int _nfs4_proc_write(struct nfs_write_data *wdata)
1349 {
1350         int rpcflags = wdata->flags;
1351         struct inode *inode = wdata->inode;
1352         struct nfs_fattr *fattr = wdata->res.fattr;
1353         struct nfs_server *server = NFS_SERVER(inode);
1354         struct rpc_message msg = {
1355                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
1356                 .rpc_argp       = &wdata->args,
1357                 .rpc_resp       = &wdata->res,
1358                 .rpc_cred       = wdata->cred,
1359         };
1360         int status;
1361
1362         dprintk("NFS call  write %d @ %Ld\n", wdata->args.count,
1363                         (long long) wdata->args.offset);
1364
1365         fattr->valid = 0;
1366         status = rpc_call_sync(server->client, &msg, rpcflags);
1367         dprintk("NFS reply write: %d\n", status);
1368         return status;
1369 }
1370
1371 static int nfs4_proc_write(struct nfs_write_data *wdata)
1372 {
1373         struct nfs4_exception exception = { };
1374         int err;
1375         do {
1376                 err = nfs4_handle_exception(NFS_SERVER(wdata->inode),
1377                                 _nfs4_proc_write(wdata),
1378                                 &exception);
1379         } while (exception.retry);
1380         return err;
1381 }
1382
1383 static int _nfs4_proc_commit(struct nfs_write_data *cdata)
1384 {
1385         struct inode *inode = cdata->inode;
1386         struct nfs_fattr *fattr = cdata->res.fattr;
1387         struct nfs_server *server = NFS_SERVER(inode);
1388         struct rpc_message msg = {
1389                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
1390                 .rpc_argp       = &cdata->args,
1391                 .rpc_resp       = &cdata->res,
1392                 .rpc_cred       = cdata->cred,
1393         };
1394         int status;
1395
1396         dprintk("NFS call  commit %d @ %Ld\n", cdata->args.count,
1397                         (long long) cdata->args.offset);
1398
1399         fattr->valid = 0;
1400         status = rpc_call_sync(server->client, &msg, 0);
1401         dprintk("NFS reply commit: %d\n", status);
1402         return status;
1403 }
1404
1405 static int nfs4_proc_commit(struct nfs_write_data *cdata)
1406 {
1407         struct nfs4_exception exception = { };
1408         int err;
1409         do {
1410                 err = nfs4_handle_exception(NFS_SERVER(cdata->inode),
1411                                 _nfs4_proc_commit(cdata),
1412                                 &exception);
1413         } while (exception.retry);
1414         return err;
1415 }
1416
1417 /*
1418  * Got race?
1419  * We will need to arrange for the VFS layer to provide an atomic open.
1420  * Until then, this create/open method is prone to inefficiency and race
1421  * conditions due to the lookup, create, and open VFS calls from sys_open()
1422  * placed on the wire.
1423  *
1424  * Given the above sorry state of affairs, I'm simply sending an OPEN.
1425  * The file will be opened again in the subsequent VFS open call
1426  * (nfs4_proc_file_open).
1427  *
1428  * The open for read will just hang around to be used by any process that
1429  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
1430  */
1431
1432 static int
1433 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1434                  int flags)
1435 {
1436         struct nfs4_state *state;
1437         struct rpc_cred *cred;
1438         int status = 0;
1439
1440         cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
1441         if (IS_ERR(cred)) {
1442                 status = PTR_ERR(cred);
1443                 goto out;
1444         }
1445         state = nfs4_do_open(dir, dentry, flags, sattr, cred);
1446         put_rpccred(cred);
1447         if (IS_ERR(state)) {
1448                 status = PTR_ERR(state);
1449                 goto out;
1450         }
1451         d_instantiate(dentry, state->inode);
1452         if (flags & O_EXCL) {
1453                 struct nfs_fattr fattr;
1454                 status = nfs4_do_setattr(NFS_SERVER(dir), &fattr,
1455                                      NFS_FH(state->inode), sattr, state);
1456                 if (status == 0) {
1457                         nfs_setattr_update_inode(state->inode, sattr);
1458                         goto out;
1459                 }
1460         } else if (flags != 0)
1461                 goto out;
1462         nfs4_close_state(state, flags);
1463 out:
1464         return status;
1465 }
1466
1467 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
1468 {
1469         struct nfs4_remove_arg args = {
1470                 .fh = NFS_FH(dir),
1471                 .name = name,
1472         };
1473         struct nfs4_change_info res;
1474         struct rpc_message msg = {
1475                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
1476                 .rpc_argp       = &args,
1477                 .rpc_resp       = &res,
1478         };
1479         int                     status;
1480
1481         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1482         if (status == 0)
1483                 update_changeattr(dir, &res);
1484         return status;
1485 }
1486
1487 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
1488 {
1489         struct nfs4_exception exception = { };
1490         int err;
1491         do {
1492                 err = nfs4_handle_exception(NFS_SERVER(dir),
1493                                 _nfs4_proc_remove(dir, name),
1494                                 &exception);
1495         } while (exception.retry);
1496         return err;
1497 }
1498
1499 struct unlink_desc {
1500         struct nfs4_remove_arg  args;
1501         struct nfs4_change_info res;
1502 };
1503
1504 static int nfs4_proc_unlink_setup(struct rpc_message *msg, struct dentry *dir,
1505                 struct qstr *name)
1506 {
1507         struct unlink_desc *up;
1508
1509         up = (struct unlink_desc *) kmalloc(sizeof(*up), GFP_KERNEL);
1510         if (!up)
1511                 return -ENOMEM;
1512
1513         up->args.fh = NFS_FH(dir->d_inode);
1514         up->args.name = name;
1515
1516         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
1517         msg->rpc_argp = &up->args;
1518         msg->rpc_resp = &up->res;
1519         return 0;
1520 }
1521
1522 static int nfs4_proc_unlink_done(struct dentry *dir, struct rpc_task *task)
1523 {
1524         struct rpc_message *msg = &task->tk_msg;
1525         struct unlink_desc *up;
1526
1527         if (msg->rpc_resp != NULL) {
1528                 up = container_of(msg->rpc_resp, struct unlink_desc, res);
1529                 update_changeattr(dir->d_inode, &up->res);
1530                 kfree(up);
1531                 msg->rpc_resp = NULL;
1532                 msg->rpc_argp = NULL;
1533         }
1534         return 0;
1535 }
1536
1537 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
1538                 struct inode *new_dir, struct qstr *new_name)
1539 {
1540         struct nfs4_rename_arg arg = {
1541                 .old_dir = NFS_FH(old_dir),
1542                 .new_dir = NFS_FH(new_dir),
1543                 .old_name = old_name,
1544                 .new_name = new_name,
1545         };
1546         struct nfs4_rename_res res = { };
1547         struct rpc_message msg = {
1548                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
1549                 .rpc_argp = &arg,
1550                 .rpc_resp = &res,
1551         };
1552         int                     status;
1553
1554         status = rpc_call_sync(NFS_CLIENT(old_dir), &msg, 0);
1555
1556         if (!status) {
1557                 update_changeattr(old_dir, &res.old_cinfo);
1558                 update_changeattr(new_dir, &res.new_cinfo);
1559         }
1560         return status;
1561 }
1562
1563 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
1564                 struct inode *new_dir, struct qstr *new_name)
1565 {
1566         struct nfs4_exception exception = { };
1567         int err;
1568         do {
1569                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
1570                                 _nfs4_proc_rename(old_dir, old_name,
1571                                         new_dir, new_name),
1572                                 &exception);
1573         } while (exception.retry);
1574         return err;
1575 }
1576
1577 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
1578 {
1579         struct nfs4_link_arg arg = {
1580                 .fh     = NFS_FH(inode),
1581                 .dir_fh = NFS_FH(dir),
1582                 .name   = name,
1583         };
1584         struct nfs4_change_info cinfo = { };
1585         struct rpc_message msg = {
1586                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
1587                 .rpc_argp = &arg,
1588                 .rpc_resp = &cinfo,
1589         };
1590         int                     status;
1591
1592         status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
1593         if (!status)
1594                 update_changeattr(dir, &cinfo);
1595
1596         return status;
1597 }
1598
1599 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
1600 {
1601         struct nfs4_exception exception = { };
1602         int err;
1603         do {
1604                 err = nfs4_handle_exception(NFS_SERVER(inode),
1605                                 _nfs4_proc_link(inode, dir, name),
1606                                 &exception);
1607         } while (exception.retry);
1608         return err;
1609 }
1610
1611 static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name,
1612                 struct qstr *path, struct iattr *sattr, struct nfs_fh *fhandle,
1613                 struct nfs_fattr *fattr)
1614 {
1615         struct nfs_server *server = NFS_SERVER(dir);
1616         struct nfs4_create_arg arg = {
1617                 .dir_fh = NFS_FH(dir),
1618                 .server = server,
1619                 .name = name,
1620                 .attrs = sattr,
1621                 .ftype = NF4LNK,
1622                 .bitmask = server->attr_bitmask,
1623         };
1624         struct nfs4_create_res res = {
1625                 .server = server,
1626                 .fh = fhandle,
1627                 .fattr = fattr,
1628         };
1629         struct rpc_message msg = {
1630                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK],
1631                 .rpc_argp = &arg,
1632                 .rpc_resp = &res,
1633         };
1634         int                     status;
1635
1636         if (path->len > NFS4_MAXPATHLEN)
1637                 return -ENAMETOOLONG;
1638         arg.u.symlink = path;
1639         fattr->valid = 0;
1640
1641         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1642         if (!status)
1643                 update_changeattr(dir, &res.dir_cinfo);
1644         return status;
1645 }
1646
1647 static int nfs4_proc_symlink(struct inode *dir, struct qstr *name,
1648                 struct qstr *path, struct iattr *sattr, struct nfs_fh *fhandle,
1649                 struct nfs_fattr *fattr)
1650 {
1651         struct nfs4_exception exception = { };
1652         int err;
1653         do {
1654                 err = nfs4_handle_exception(NFS_SERVER(dir),
1655                                 _nfs4_proc_symlink(dir, name, path, sattr,
1656                                         fhandle, fattr),
1657                                 &exception);
1658         } while (exception.retry);
1659         return err;
1660 }
1661
1662 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
1663                 struct iattr *sattr)
1664 {
1665         struct nfs_server *server = NFS_SERVER(dir);
1666         struct nfs_fh fhandle;
1667         struct nfs_fattr fattr;
1668         struct nfs4_create_arg arg = {
1669                 .dir_fh = NFS_FH(dir),
1670                 .server = server,
1671                 .name = &dentry->d_name,
1672                 .attrs = sattr,
1673                 .ftype = NF4DIR,
1674                 .bitmask = server->attr_bitmask,
1675         };
1676         struct nfs4_create_res res = {
1677                 .server = server,
1678                 .fh = &fhandle,
1679                 .fattr = &fattr,
1680         };
1681         struct rpc_message msg = {
1682                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE],
1683                 .rpc_argp = &arg,
1684                 .rpc_resp = &res,
1685         };
1686         int                     status;
1687
1688         fattr.valid = 0;
1689
1690         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1691         if (!status) {
1692                 update_changeattr(dir, &res.dir_cinfo);
1693                 status = nfs_instantiate(dentry, &fhandle, &fattr);
1694         }
1695         return status;
1696 }
1697
1698 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
1699                 struct iattr *sattr)
1700 {
1701         struct nfs4_exception exception = { };
1702         int err;
1703         do {
1704                 err = nfs4_handle_exception(NFS_SERVER(dir),
1705                                 _nfs4_proc_mkdir(dir, dentry, sattr),
1706                                 &exception);
1707         } while (exception.retry);
1708         return err;
1709 }
1710
1711 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
1712                   u64 cookie, struct page *page, unsigned int count, int plus)
1713 {
1714         struct inode            *dir = dentry->d_inode;
1715         struct nfs4_readdir_arg args = {
1716                 .fh = NFS_FH(dir),
1717                 .pages = &page,
1718                 .pgbase = 0,
1719                 .count = count,
1720                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
1721         };
1722         struct nfs4_readdir_res res;
1723         struct rpc_message msg = {
1724                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
1725                 .rpc_argp = &args,
1726                 .rpc_resp = &res,
1727                 .rpc_cred = cred,
1728         };
1729         int                     status;
1730
1731         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __FUNCTION__,
1732                         dentry->d_parent->d_name.name,
1733                         dentry->d_name.name,
1734                         (unsigned long long)cookie);
1735         lock_kernel();
1736         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
1737         res.pgbase = args.pgbase;
1738         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1739         if (status == 0)
1740                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
1741         unlock_kernel();
1742         dprintk("%s: returns %d\n", __FUNCTION__, status);
1743         return status;
1744 }
1745
1746 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
1747                   u64 cookie, struct page *page, unsigned int count, int plus)
1748 {
1749         struct nfs4_exception exception = { };
1750         int err;
1751         do {
1752                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
1753                                 _nfs4_proc_readdir(dentry, cred, cookie,
1754                                         page, count, plus),
1755                                 &exception);
1756         } while (exception.retry);
1757         return err;
1758 }
1759
1760 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
1761                 struct iattr *sattr, dev_t rdev)
1762 {
1763         struct nfs_server *server = NFS_SERVER(dir);
1764         struct nfs_fh fh;
1765         struct nfs_fattr fattr;
1766         struct nfs4_create_arg arg = {
1767                 .dir_fh = NFS_FH(dir),
1768                 .server = server,
1769                 .name = &dentry->d_name,
1770                 .attrs = sattr,
1771                 .bitmask = server->attr_bitmask,
1772         };
1773         struct nfs4_create_res res = {
1774                 .server = server,
1775                 .fh = &fh,
1776                 .fattr = &fattr,
1777         };
1778         struct rpc_message msg = {
1779                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE],
1780                 .rpc_argp = &arg,
1781                 .rpc_resp = &res,
1782         };
1783         int                     status;
1784         int                     mode = sattr->ia_mode;
1785
1786         fattr.valid = 0;
1787
1788         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
1789         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
1790         if (S_ISFIFO(mode))
1791                 arg.ftype = NF4FIFO;
1792         else if (S_ISBLK(mode)) {
1793                 arg.ftype = NF4BLK;
1794                 arg.u.device.specdata1 = MAJOR(rdev);
1795                 arg.u.device.specdata2 = MINOR(rdev);
1796         }
1797         else if (S_ISCHR(mode)) {
1798                 arg.ftype = NF4CHR;
1799                 arg.u.device.specdata1 = MAJOR(rdev);
1800                 arg.u.device.specdata2 = MINOR(rdev);
1801         }
1802         else
1803                 arg.ftype = NF4SOCK;
1804
1805         status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
1806         if (status == 0) {
1807                 update_changeattr(dir, &res.dir_cinfo);
1808                 status = nfs_instantiate(dentry, &fh, &fattr);
1809         }
1810         return status;
1811 }
1812
1813 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
1814                 struct iattr *sattr, dev_t rdev)
1815 {
1816         struct nfs4_exception exception = { };
1817         int err;
1818         do {
1819                 err = nfs4_handle_exception(NFS_SERVER(dir),
1820                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
1821                                 &exception);
1822         } while (exception.retry);
1823         return err;
1824 }
1825
1826 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
1827                  struct nfs_fsstat *fsstat)
1828 {
1829         struct nfs4_statfs_arg args = {
1830                 .fh = fhandle,
1831                 .bitmask = server->attr_bitmask,
1832         };
1833         struct rpc_message msg = {
1834                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
1835                 .rpc_argp = &args,
1836                 .rpc_resp = fsstat,
1837         };
1838
1839         fsstat->fattr->valid = 0;
1840         return rpc_call_sync(server->client, &msg, 0);
1841 }
1842
1843 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
1844 {
1845         struct nfs4_exception exception = { };
1846         int err;
1847         do {
1848                 err = nfs4_handle_exception(server,
1849                                 _nfs4_proc_statfs(server, fhandle, fsstat),
1850                                 &exception);
1851         } while (exception.retry);
1852         return err;
1853 }
1854
1855 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
1856                 struct nfs_fsinfo *fsinfo)
1857 {
1858         struct nfs4_fsinfo_arg args = {
1859                 .fh = fhandle,
1860                 .bitmask = server->attr_bitmask,
1861         };
1862         struct rpc_message msg = {
1863                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
1864                 .rpc_argp = &args,
1865                 .rpc_resp = fsinfo,
1866         };
1867
1868         return rpc_call_sync(server->client, &msg, 0);
1869 }
1870
1871 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
1872 {
1873         struct nfs4_exception exception = { };
1874         int err;
1875
1876         do {
1877                 err = nfs4_handle_exception(server,
1878                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
1879                                 &exception);
1880         } while (exception.retry);
1881         return err;
1882 }
1883
1884 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
1885 {
1886         fsinfo->fattr->valid = 0;
1887         return nfs4_do_fsinfo(server, fhandle, fsinfo);
1888 }
1889
1890 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
1891                 struct nfs_pathconf *pathconf)
1892 {
1893         struct nfs4_pathconf_arg args = {
1894                 .fh = fhandle,
1895                 .bitmask = server->attr_bitmask,
1896         };
1897         struct rpc_message msg = {
1898                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
1899                 .rpc_argp = &args,
1900                 .rpc_resp = pathconf,
1901         };
1902
1903         /* None of the pathconf attributes are mandatory to implement */
1904         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
1905                 memset(pathconf, 0, sizeof(*pathconf));
1906                 return 0;
1907         }
1908
1909         pathconf->fattr->valid = 0;
1910         return rpc_call_sync(server->client, &msg, 0);
1911 }
1912
1913 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
1914                 struct nfs_pathconf *pathconf)
1915 {
1916         struct nfs4_exception exception = { };
1917         int err;
1918
1919         do {
1920                 err = nfs4_handle_exception(server,
1921                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
1922                                 &exception);
1923         } while (exception.retry);
1924         return err;
1925 }
1926
1927 static void
1928 nfs4_read_done(struct rpc_task *task)
1929 {
1930         struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata;
1931         struct inode *inode = data->inode;
1932
1933         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
1934                 rpc_restart_call(task);
1935                 return;
1936         }
1937         if (task->tk_status > 0)
1938                 renew_lease(NFS_SERVER(inode), data->timestamp);
1939         /* Call back common NFS readpage processing */
1940         nfs_readpage_result(task);
1941 }
1942
1943 static void
1944 nfs4_proc_read_setup(struct nfs_read_data *data)
1945 {
1946         struct rpc_task *task = &data->task;
1947         struct rpc_message msg = {
1948                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ],
1949                 .rpc_argp = &data->args,
1950                 .rpc_resp = &data->res,
1951                 .rpc_cred = data->cred,
1952         };
1953         struct inode *inode = data->inode;
1954         int flags;
1955
1956         data->timestamp   = jiffies;
1957
1958         /* N.B. Do we need to test? Never called for swapfile inode */
1959         flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
1960
1961         /* Finalize the task. */
1962         rpc_init_task(task, NFS_CLIENT(inode), nfs4_read_done, flags);
1963         rpc_call_setup(task, &msg, 0);
1964 }
1965
1966 static void
1967 nfs4_write_done(struct rpc_task *task)
1968 {
1969         struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
1970         struct inode *inode = data->inode;
1971
1972         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
1973                 rpc_restart_call(task);
1974                 return;
1975         }
1976         if (task->tk_status >= 0)
1977                 renew_lease(NFS_SERVER(inode), data->timestamp);
1978         /* Call back common NFS writeback processing */
1979         nfs_writeback_done(task);
1980 }
1981
1982 static void
1983 nfs4_proc_write_setup(struct nfs_write_data *data, int how)
1984 {
1985         struct rpc_task *task = &data->task;
1986         struct rpc_message msg = {
1987                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
1988                 .rpc_argp = &data->args,
1989                 .rpc_resp = &data->res,
1990                 .rpc_cred = data->cred,
1991         };
1992         struct inode *inode = data->inode;
1993         int stable;
1994         int flags;
1995
1996         if (how & FLUSH_STABLE) {
1997                 if (!NFS_I(inode)->ncommit)
1998                         stable = NFS_FILE_SYNC;
1999                 else
2000                         stable = NFS_DATA_SYNC;
2001         } else
2002                 stable = NFS_UNSTABLE;
2003         data->args.stable = stable;
2004
2005         data->timestamp   = jiffies;
2006
2007         /* Set the initial flags for the task.  */
2008         flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
2009
2010         /* Finalize the task. */
2011         rpc_init_task(task, NFS_CLIENT(inode), nfs4_write_done, flags);
2012         rpc_call_setup(task, &msg, 0);
2013 }
2014
2015 static void
2016 nfs4_commit_done(struct rpc_task *task)
2017 {
2018         struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
2019         struct inode *inode = data->inode;
2020
2021         if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
2022                 rpc_restart_call(task);
2023                 return;
2024         }
2025         /* Call back common NFS writeback processing */
2026         nfs_commit_done(task);
2027 }
2028
2029 static void
2030 nfs4_proc_commit_setup(struct nfs_write_data *data, int how)
2031 {
2032         struct rpc_task *task = &data->task;
2033         struct rpc_message msg = {
2034                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
2035                 .rpc_argp = &data->args,
2036                 .rpc_resp = &data->res,
2037                 .rpc_cred = data->cred,
2038         };
2039         struct inode *inode = data->inode;
2040         int flags;
2041
2042         /* Set the initial flags for the task.  */
2043         flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
2044
2045         /* Finalize the task. */
2046         rpc_init_task(task, NFS_CLIENT(inode), nfs4_commit_done, flags);
2047         rpc_call_setup(task, &msg, 0);
2048 }
2049
2050 /*
2051  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
2052  * standalone procedure for queueing an asynchronous RENEW.
2053  */
2054 static void
2055 renew_done(struct rpc_task *task)
2056 {
2057         struct nfs4_client *clp = (struct nfs4_client *)task->tk_msg.rpc_argp;
2058         unsigned long timestamp = (unsigned long)task->tk_calldata;
2059
2060         if (task->tk_status < 0) {
2061                 switch (task->tk_status) {
2062                         case -NFS4ERR_STALE_CLIENTID:
2063                         case -NFS4ERR_EXPIRED:
2064                         case -NFS4ERR_CB_PATH_DOWN:
2065                                 nfs4_schedule_state_recovery(clp);
2066                 }
2067                 return;
2068         }
2069         spin_lock(&clp->cl_lock);
2070         if (time_before(clp->cl_last_renewal,timestamp))
2071                 clp->cl_last_renewal = timestamp;
2072         spin_unlock(&clp->cl_lock);
2073 }
2074
2075 int
2076 nfs4_proc_async_renew(struct nfs4_client *clp)
2077 {
2078         struct rpc_message msg = {
2079                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2080                 .rpc_argp       = clp,
2081                 .rpc_cred       = clp->cl_cred,
2082         };
2083
2084         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
2085                         renew_done, (void *)jiffies);
2086 }
2087
2088 int
2089 nfs4_proc_renew(struct nfs4_client *clp)
2090 {
2091         struct rpc_message msg = {
2092                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2093                 .rpc_argp       = clp,
2094                 .rpc_cred       = clp->cl_cred,
2095         };
2096         unsigned long now = jiffies;
2097         int status;
2098
2099         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2100         if (status < 0)
2101                 return status;
2102         spin_lock(&clp->cl_lock);
2103         if (time_before(clp->cl_last_renewal,now))
2104                 clp->cl_last_renewal = now;
2105         spin_unlock(&clp->cl_lock);
2106         return 0;
2107 }
2108
2109 /*
2110  * We will need to arrange for the VFS layer to provide an atomic open.
2111  * Until then, this open method is prone to inefficiency and race conditions
2112  * due to the lookup, potential create, and open VFS calls from sys_open()
2113  * placed on the wire.
2114  */
2115 static int
2116 nfs4_proc_file_open(struct inode *inode, struct file *filp)
2117 {
2118         struct dentry *dentry = filp->f_dentry;
2119         struct nfs_open_context *ctx;
2120         struct nfs4_state *state = NULL;
2121         struct rpc_cred *cred;
2122         int status = -ENOMEM;
2123
2124         dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n",
2125                                (int)dentry->d_parent->d_name.len,
2126                                dentry->d_parent->d_name.name,
2127                                (int)dentry->d_name.len, dentry->d_name.name);
2128
2129
2130         /* Find our open stateid */
2131         cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
2132         if (IS_ERR(cred))
2133                 return PTR_ERR(cred);
2134         ctx = alloc_nfs_open_context(dentry, cred);
2135         put_rpccred(cred);
2136         if (unlikely(ctx == NULL))
2137                 return -ENOMEM;
2138         status = -EIO; /* ERACE actually */
2139         state = nfs4_find_state(inode, cred, filp->f_mode);
2140         if (unlikely(state == NULL))
2141                 goto no_state;
2142         ctx->state = state;
2143         nfs4_close_state(state, filp->f_mode);
2144         ctx->mode = filp->f_mode;
2145         nfs_file_set_open_context(filp, ctx);
2146         put_nfs_open_context(ctx);
2147         if (filp->f_mode & FMODE_WRITE)
2148                 nfs_begin_data_update(inode);
2149         return 0;
2150 no_state:
2151         printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__);
2152         put_nfs_open_context(ctx);
2153         return status;
2154 }
2155
2156 /*
2157  * Release our state
2158  */
2159 static int
2160 nfs4_proc_file_release(struct inode *inode, struct file *filp)
2161 {
2162         if (filp->f_mode & FMODE_WRITE)
2163                 nfs_end_data_update(inode);
2164         nfs_file_clear_open_context(filp);
2165         return 0;
2166 }
2167
2168 static inline int nfs4_server_supports_acls(struct nfs_server *server)
2169 {
2170         return (server->caps & NFS_CAP_ACLS)
2171                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2172                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
2173 }
2174
2175 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
2176  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
2177  * the stack.
2178  */
2179 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
2180
2181 static void buf_to_pages(const void *buf, size_t buflen,
2182                 struct page **pages, unsigned int *pgbase)
2183 {
2184         const void *p = buf;
2185
2186         *pgbase = offset_in_page(buf);
2187         p -= *pgbase;
2188         while (p < buf + buflen) {
2189                 *(pages++) = virt_to_page(p);
2190                 p += PAGE_CACHE_SIZE;
2191         }
2192 }
2193
2194 struct nfs4_cached_acl {
2195         int cached;
2196         size_t len;
2197         char data[0];
2198 };
2199
2200 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
2201 {
2202         struct nfs_inode *nfsi = NFS_I(inode);
2203
2204         spin_lock(&inode->i_lock);
2205         kfree(nfsi->nfs4_acl);
2206         nfsi->nfs4_acl = acl;
2207         spin_unlock(&inode->i_lock);
2208 }
2209
2210 static void nfs4_zap_acl_attr(struct inode *inode)
2211 {
2212         nfs4_set_cached_acl(inode, NULL);
2213 }
2214
2215 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
2216 {
2217         struct nfs_inode *nfsi = NFS_I(inode);
2218         struct nfs4_cached_acl *acl;
2219         int ret = -ENOENT;
2220
2221         spin_lock(&inode->i_lock);
2222         acl = nfsi->nfs4_acl;
2223         if (acl == NULL)
2224                 goto out;
2225         if (buf == NULL) /* user is just asking for length */
2226                 goto out_len;
2227         if (acl->cached == 0)
2228                 goto out;
2229         ret = -ERANGE; /* see getxattr(2) man page */
2230         if (acl->len > buflen)
2231                 goto out;
2232         memcpy(buf, acl->data, acl->len);
2233 out_len:
2234         ret = acl->len;
2235 out:
2236         spin_unlock(&inode->i_lock);
2237         return ret;
2238 }
2239
2240 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
2241 {
2242         struct nfs4_cached_acl *acl;
2243
2244         if (buf && acl_len <= PAGE_SIZE) {
2245                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
2246                 if (acl == NULL)
2247                         goto out;
2248                 acl->cached = 1;
2249                 memcpy(acl->data, buf, acl_len);
2250         } else {
2251                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
2252                 if (acl == NULL)
2253                         goto out;
2254                 acl->cached = 0;
2255         }
2256         acl->len = acl_len;
2257 out:
2258         nfs4_set_cached_acl(inode, acl);
2259 }
2260
2261 static inline ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
2262 {
2263         struct page *pages[NFS4ACL_MAXPAGES];
2264         struct nfs_getaclargs args = {
2265                 .fh = NFS_FH(inode),
2266                 .acl_pages = pages,
2267                 .acl_len = buflen,
2268         };
2269         size_t resp_len = buflen;
2270         void *resp_buf;
2271         struct rpc_message msg = {
2272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
2273                 .rpc_argp = &args,
2274                 .rpc_resp = &resp_len,
2275         };
2276         struct page *localpage = NULL;
2277         int ret;
2278
2279         if (buflen < PAGE_SIZE) {
2280                 /* As long as we're doing a round trip to the server anyway,
2281                  * let's be prepared for a page of acl data. */
2282                 localpage = alloc_page(GFP_KERNEL);
2283                 resp_buf = page_address(localpage);
2284                 if (localpage == NULL)
2285                         return -ENOMEM;
2286                 args.acl_pages[0] = localpage;
2287                 args.acl_pgbase = 0;
2288                 args.acl_len = PAGE_SIZE;
2289         } else {
2290                 resp_buf = buf;
2291                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
2292         }
2293         ret = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
2294         if (ret)
2295                 goto out_free;
2296         if (resp_len > args.acl_len)
2297                 nfs4_write_cached_acl(inode, NULL, resp_len);
2298         else
2299                 nfs4_write_cached_acl(inode, resp_buf, resp_len);
2300         if (buf) {
2301                 ret = -ERANGE;
2302                 if (resp_len > buflen)
2303                         goto out_free;
2304                 if (localpage)
2305                         memcpy(buf, resp_buf, resp_len);
2306         }
2307         ret = resp_len;
2308 out_free:
2309         if (localpage)
2310                 __free_page(localpage);
2311         return ret;
2312 }
2313
2314 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
2315 {
2316         struct nfs_server *server = NFS_SERVER(inode);
2317         int ret;
2318
2319         if (!nfs4_server_supports_acls(server))
2320                 return -EOPNOTSUPP;
2321         ret = nfs_revalidate_inode(server, inode);
2322         if (ret < 0)
2323                 return ret;
2324         ret = nfs4_read_cached_acl(inode, buf, buflen);
2325         if (ret != -ENOENT)
2326                 return ret;
2327         return nfs4_get_acl_uncached(inode, buf, buflen);
2328 }
2329
2330 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
2331 {
2332         struct nfs_server *server = NFS_SERVER(inode);
2333         struct page *pages[NFS4ACL_MAXPAGES];
2334         struct nfs_setaclargs arg = {
2335                 .fh             = NFS_FH(inode),
2336                 .acl_pages      = pages,
2337                 .acl_len        = buflen,
2338         };
2339         struct rpc_message msg = {
2340                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
2341                 .rpc_argp       = &arg,
2342                 .rpc_resp       = NULL,
2343         };
2344         int ret;
2345
2346         if (!nfs4_server_supports_acls(server))
2347                 return -EOPNOTSUPP;
2348         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
2349         ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0);
2350         if (ret == 0)
2351                 nfs4_write_cached_acl(inode, buf, buflen);
2352         return ret;
2353 }
2354
2355 static int
2356 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server)
2357 {
2358         struct nfs4_client *clp = server->nfs4_state;
2359
2360         if (!clp || task->tk_status >= 0)
2361                 return 0;
2362         switch(task->tk_status) {
2363                 case -NFS4ERR_STALE_CLIENTID:
2364                 case -NFS4ERR_STALE_STATEID:
2365                 case -NFS4ERR_EXPIRED:
2366                         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL, NULL);
2367                         nfs4_schedule_state_recovery(clp);
2368                         if (test_bit(NFS4CLNT_OK, &clp->cl_state))
2369                                 rpc_wake_up_task(task);
2370                         task->tk_status = 0;
2371                         return -EAGAIN;
2372                 case -NFS4ERR_GRACE:
2373                 case -NFS4ERR_DELAY:
2374                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
2375                         task->tk_status = 0;
2376                         return -EAGAIN;
2377                 case -NFS4ERR_OLD_STATEID:
2378                         task->tk_status = 0;
2379                         return -EAGAIN;
2380         }
2381         task->tk_status = nfs4_map_errors(task->tk_status);
2382         return 0;
2383 }
2384
2385 static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs4_client *clp)
2386 {
2387         DEFINE_WAIT(wait);
2388         sigset_t oldset;
2389         int interruptible, res = 0;
2390
2391         might_sleep();
2392
2393         rpc_clnt_sigmask(clnt, &oldset);
2394         interruptible = TASK_UNINTERRUPTIBLE;
2395         if (clnt->cl_intr)
2396                 interruptible = TASK_INTERRUPTIBLE;
2397         prepare_to_wait(&clp->cl_waitq, &wait, interruptible);
2398         nfs4_schedule_state_recovery(clp);
2399         if (clnt->cl_intr && signalled())
2400                 res = -ERESTARTSYS;
2401         else if (!test_bit(NFS4CLNT_OK, &clp->cl_state))
2402                 schedule();
2403         finish_wait(&clp->cl_waitq, &wait);
2404         rpc_clnt_sigunmask(clnt, &oldset);
2405         return res;
2406 }
2407
2408 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
2409 {
2410         sigset_t oldset;
2411         int res = 0;
2412
2413         might_sleep();
2414
2415         if (*timeout <= 0)
2416                 *timeout = NFS4_POLL_RETRY_MIN;
2417         if (*timeout > NFS4_POLL_RETRY_MAX)
2418                 *timeout = NFS4_POLL_RETRY_MAX;
2419         rpc_clnt_sigmask(clnt, &oldset);
2420         if (clnt->cl_intr) {
2421                 set_current_state(TASK_INTERRUPTIBLE);
2422                 schedule_timeout(*timeout);
2423                 if (signalled())
2424                         res = -ERESTARTSYS;
2425         } else {
2426                 set_current_state(TASK_UNINTERRUPTIBLE);
2427                 schedule_timeout(*timeout);
2428         }
2429         rpc_clnt_sigunmask(clnt, &oldset);
2430         *timeout <<= 1;
2431         return res;
2432 }
2433
2434 /* This is the error handling routine for processes that are allowed
2435  * to sleep.
2436  */
2437 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
2438 {
2439         struct nfs4_client *clp = server->nfs4_state;
2440         int ret = errorcode;
2441
2442         exception->retry = 0;
2443         switch(errorcode) {
2444                 case 0:
2445                         return 0;
2446                 case -NFS4ERR_STALE_CLIENTID:
2447                 case -NFS4ERR_STALE_STATEID:
2448                 case -NFS4ERR_EXPIRED:
2449                         ret = nfs4_wait_clnt_recover(server->client, clp);
2450                         if (ret == 0)
2451                                 exception->retry = 1;
2452                         break;
2453                 case -NFS4ERR_GRACE:
2454                 case -NFS4ERR_DELAY:
2455                         ret = nfs4_delay(server->client, &exception->timeout);
2456                         if (ret == 0)
2457                                 exception->retry = 1;
2458                         break;
2459                 case -NFS4ERR_OLD_STATEID:
2460                         if (ret == 0)
2461                                 exception->retry = 1;
2462         }
2463         /* We failed to handle the error */
2464         return nfs4_map_errors(ret);
2465 }
2466
2467 int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short port)
2468 {
2469         nfs4_verifier sc_verifier;
2470         struct nfs4_setclientid setclientid = {
2471                 .sc_verifier = &sc_verifier,
2472                 .sc_prog = program,
2473         };
2474         struct rpc_message msg = {
2475                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
2476                 .rpc_argp = &setclientid,
2477                 .rpc_resp = clp,
2478                 .rpc_cred = clp->cl_cred,
2479         };
2480         u32 *p;
2481         int loop = 0;
2482         int status;
2483
2484         p = (u32*)sc_verifier.data;
2485         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
2486         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
2487
2488         for(;;) {
2489                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
2490                                 sizeof(setclientid.sc_name), "%s/%u.%u.%u.%u %s %u",
2491                                 clp->cl_ipaddr, NIPQUAD(clp->cl_addr.s_addr),
2492                                 clp->cl_cred->cr_ops->cr_name,
2493                                 clp->cl_id_uniquifier);
2494                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
2495                                 sizeof(setclientid.sc_netid), "tcp");
2496                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
2497                                 sizeof(setclientid.sc_uaddr), "%s.%d.%d",
2498                                 clp->cl_ipaddr, port >> 8, port & 255);
2499
2500                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2501                 if (status != -NFS4ERR_CLID_INUSE)
2502                         break;
2503                 if (signalled())
2504                         break;
2505                 if (loop++ & 1)
2506                         ssleep(clp->cl_lease_time + 1);
2507                 else
2508                         if (++clp->cl_id_uniquifier == 0)
2509                                 break;
2510         }
2511         return status;
2512 }
2513
2514 int
2515 nfs4_proc_setclientid_confirm(struct nfs4_client *clp)
2516 {
2517         struct nfs_fsinfo fsinfo;
2518         struct rpc_message msg = {
2519                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
2520                 .rpc_argp = clp,
2521                 .rpc_resp = &fsinfo,
2522                 .rpc_cred = clp->cl_cred,
2523         };
2524         unsigned long now;
2525         int status;
2526
2527         now = jiffies;
2528         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2529         if (status == 0) {
2530                 spin_lock(&clp->cl_lock);
2531                 clp->cl_lease_time = fsinfo.lease_time * HZ;
2532                 clp->cl_last_renewal = now;
2533                 spin_unlock(&clp->cl_lock);
2534         }
2535         return status;
2536 }
2537
2538 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid)
2539 {
2540         struct nfs4_delegreturnargs args = {
2541                 .fhandle = NFS_FH(inode),
2542                 .stateid = stateid,
2543         };
2544         struct rpc_message msg = {
2545                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
2546                 .rpc_argp = &args,
2547                 .rpc_cred = cred,
2548         };
2549
2550         return rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
2551 }
2552
2553 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid)
2554 {
2555         struct nfs_server *server = NFS_SERVER(inode);
2556         struct nfs4_exception exception = { };
2557         int err;
2558         do {
2559                 err = _nfs4_proc_delegreturn(inode, cred, stateid);
2560                 switch (err) {
2561                         case -NFS4ERR_STALE_STATEID:
2562                         case -NFS4ERR_EXPIRED:
2563                                 nfs4_schedule_state_recovery(server->nfs4_state);
2564                         case 0:
2565                                 return 0;
2566                 }
2567                 err = nfs4_handle_exception(server, err, &exception);
2568         } while (exception.retry);
2569         return err;
2570 }
2571
2572 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
2573 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
2574
2575 /*
2576  * sleep, with exponential backoff, and retry the LOCK operation.
2577  */
2578 static unsigned long
2579 nfs4_set_lock_task_retry(unsigned long timeout)
2580 {
2581         current->state = TASK_INTERRUPTIBLE;
2582         schedule_timeout(timeout);
2583         timeout <<= 1;
2584         if (timeout > NFS4_LOCK_MAXTIMEOUT)
2585                 return NFS4_LOCK_MAXTIMEOUT;
2586         return timeout;
2587 }
2588
2589 static inline int
2590 nfs4_lck_type(int cmd, struct file_lock *request)
2591 {
2592         /* set lock type */
2593         switch (request->fl_type) {
2594                 case F_RDLCK:
2595                         return IS_SETLKW(cmd) ? NFS4_READW_LT : NFS4_READ_LT;
2596                 case F_WRLCK:
2597                         return IS_SETLKW(cmd) ? NFS4_WRITEW_LT : NFS4_WRITE_LT;
2598                 case F_UNLCK:
2599                         return NFS4_WRITE_LT;
2600         }
2601         BUG();
2602         return 0;
2603 }
2604
2605 static inline uint64_t
2606 nfs4_lck_length(struct file_lock *request)
2607 {
2608         if (request->fl_end == OFFSET_MAX)
2609                 return ~(uint64_t)0;
2610         return request->fl_end - request->fl_start + 1;
2611 }
2612
2613 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2614 {
2615         struct inode *inode = state->inode;
2616         struct nfs_server *server = NFS_SERVER(inode);
2617         struct nfs4_client *clp = server->nfs4_state;
2618         struct nfs_lockargs arg = {
2619                 .fh = NFS_FH(inode),
2620                 .type = nfs4_lck_type(cmd, request),
2621                 .offset = request->fl_start,
2622                 .length = nfs4_lck_length(request),
2623         };
2624         struct nfs_lockres res = {
2625                 .server = server,
2626         };
2627         struct rpc_message msg = {
2628                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
2629                 .rpc_argp       = &arg,
2630                 .rpc_resp       = &res,
2631                 .rpc_cred       = state->owner->so_cred,
2632         };
2633         struct nfs_lowner nlo;
2634         struct nfs4_lock_state *lsp;
2635         int status;
2636
2637         down_read(&clp->cl_sem);
2638         nlo.clientid = clp->cl_clientid;
2639         down(&state->lock_sema);
2640         status = nfs4_set_lock_state(state, request);
2641         if (status != 0)
2642                 goto out;
2643         lsp = request->fl_u.nfs4_fl.owner;
2644         nlo.id = lsp->ls_id;
2645         arg.u.lockt = &nlo;
2646         status = rpc_call_sync(server->client, &msg, 0);
2647         if (!status) {
2648                 request->fl_type = F_UNLCK;
2649         } else if (status == -NFS4ERR_DENIED) {
2650                 int64_t len, start, end;
2651                 start = res.u.denied.offset;
2652                 len = res.u.denied.length;
2653                 end = start + len - 1;
2654                 if (end < 0 || len == 0)
2655                         request->fl_end = OFFSET_MAX;
2656                 else
2657                         request->fl_end = (loff_t)end;
2658                 request->fl_start = (loff_t)start;
2659                 request->fl_type = F_WRLCK;
2660                 if (res.u.denied.type & 1)
2661                         request->fl_type = F_RDLCK;
2662                 request->fl_pid = 0;
2663                 status = 0;
2664         }
2665 out:
2666         up(&state->lock_sema);
2667         up_read(&clp->cl_sem);
2668         return status;
2669 }
2670
2671 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2672 {
2673         struct nfs4_exception exception = { };
2674         int err;
2675
2676         do {
2677                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2678                                 _nfs4_proc_getlk(state, cmd, request),
2679                                 &exception);
2680         } while (exception.retry);
2681         return err;
2682 }
2683
2684 static int do_vfs_lock(struct file *file, struct file_lock *fl)
2685 {
2686         int res = 0;
2687         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
2688                 case FL_POSIX:
2689                         res = posix_lock_file_wait(file, fl);
2690                         break;
2691                 case FL_FLOCK:
2692                         res = flock_lock_file_wait(file, fl);
2693                         break;
2694                 default:
2695                         BUG();
2696         }
2697         if (res < 0)
2698                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n",
2699                                 __FUNCTION__);
2700         return res;
2701 }
2702
2703 static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2704 {
2705         struct inode *inode = state->inode;
2706         struct nfs_server *server = NFS_SERVER(inode);
2707         struct nfs4_client *clp = server->nfs4_state;
2708         struct nfs_lockargs arg = {
2709                 .fh = NFS_FH(inode),
2710                 .type = nfs4_lck_type(cmd, request),
2711                 .offset = request->fl_start,
2712                 .length = nfs4_lck_length(request),
2713         };
2714         struct nfs_lockres res = {
2715                 .server = server,
2716         };
2717         struct rpc_message msg = {
2718                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
2719                 .rpc_argp       = &arg,
2720                 .rpc_resp       = &res,
2721                 .rpc_cred       = state->owner->so_cred,
2722         };
2723         struct nfs4_lock_state *lsp;
2724         struct nfs_locku_opargs luargs;
2725         int status;
2726
2727         down_read(&clp->cl_sem);
2728         down(&state->lock_sema);
2729         status = nfs4_set_lock_state(state, request);
2730         if (status != 0)
2731                 goto out;
2732         lsp = request->fl_u.nfs4_fl.owner;
2733         /* We might have lost the locks! */
2734         if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0)
2735                 goto out;
2736         luargs.seqid = lsp->ls_seqid;
2737         memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid));
2738         arg.u.locku = &luargs;
2739         status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2740         nfs4_increment_lock_seqid(status, lsp);
2741
2742         if (status == 0)
2743                 memcpy(&lsp->ls_stateid,  &res.u.stateid,
2744                                 sizeof(lsp->ls_stateid));
2745 out:
2746         up(&state->lock_sema);
2747         if (status == 0)
2748                 do_vfs_lock(request->fl_file, request);
2749         up_read(&clp->cl_sem);
2750         return status;
2751 }
2752
2753 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
2754 {
2755         struct nfs4_exception exception = { };
2756         int err;
2757
2758         do {
2759                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2760                                 _nfs4_proc_unlck(state, cmd, request),
2761                                 &exception);
2762         } while (exception.retry);
2763         return err;
2764 }
2765
2766 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim)
2767 {
2768         struct inode *inode = state->inode;
2769         struct nfs_server *server = NFS_SERVER(inode);
2770         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
2771         struct nfs_lockargs arg = {
2772                 .fh = NFS_FH(inode),
2773                 .type = nfs4_lck_type(cmd, request),
2774                 .offset = request->fl_start,
2775                 .length = nfs4_lck_length(request),
2776         };
2777         struct nfs_lockres res = {
2778                 .server = server,
2779         };
2780         struct rpc_message msg = {
2781                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
2782                 .rpc_argp       = &arg,
2783                 .rpc_resp       = &res,
2784                 .rpc_cred       = state->owner->so_cred,
2785         };
2786         struct nfs_lock_opargs largs = {
2787                 .reclaim = reclaim,
2788                 .new_lock_owner = 0,
2789         };
2790         int status;
2791
2792         if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) {
2793                 struct nfs4_state_owner *owner = state->owner;
2794                 struct nfs_open_to_lock otl = {
2795                         .lock_owner = {
2796                                 .clientid = server->nfs4_state->cl_clientid,
2797                         },
2798                 };
2799
2800                 otl.lock_seqid = lsp->ls_seqid;
2801                 otl.lock_owner.id = lsp->ls_id;
2802                 memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid));
2803                 largs.u.open_lock = &otl;
2804                 largs.new_lock_owner = 1;
2805                 arg.u.lock = &largs;
2806                 down(&owner->so_sema);
2807                 otl.open_seqid = owner->so_seqid;
2808                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2809                 /* increment open_owner seqid on success, and
2810                 * seqid mutating errors */
2811                 nfs4_increment_seqid(status, owner);
2812                 up(&owner->so_sema);
2813                 if (status == 0) {
2814                         lsp->ls_flags |= NFS_LOCK_INITIALIZED;
2815                         lsp->ls_seqid++;
2816                 }
2817         } else {
2818                 struct nfs_exist_lock el = {
2819                         .seqid = lsp->ls_seqid,
2820                 };
2821                 memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid));
2822                 largs.u.exist_lock = &el;
2823                 arg.u.lock = &largs;
2824                 status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
2825                 /* increment seqid on success, and * seqid mutating errors*/
2826                 nfs4_increment_lock_seqid(status, lsp);
2827         }
2828         /* save the returned stateid. */
2829         if (status == 0)
2830                 memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid));
2831         else if (status == -NFS4ERR_DENIED)
2832                 status = -EAGAIN;
2833         return status;
2834 }
2835
2836 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
2837 {
2838         struct nfs_server *server = NFS_SERVER(state->inode);
2839         struct nfs4_exception exception = { };
2840         int err;
2841
2842         do {
2843                 err = _nfs4_do_setlk(state, F_SETLK, request, 1);
2844                 if (err != -NFS4ERR_DELAY)
2845                         break;
2846                 nfs4_handle_exception(server, err, &exception);
2847         } while (exception.retry);
2848         return err;
2849 }
2850
2851 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
2852 {
2853         struct nfs_server *server = NFS_SERVER(state->inode);
2854         struct nfs4_exception exception = { };
2855         int err;
2856
2857         do {
2858                 err = _nfs4_do_setlk(state, F_SETLK, request, 0);
2859                 if (err != -NFS4ERR_DELAY)
2860                         break;
2861                 nfs4_handle_exception(server, err, &exception);
2862         } while (exception.retry);
2863         return err;
2864 }
2865
2866 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2867 {
2868         struct nfs4_client *clp = state->owner->so_client;
2869         int status;
2870
2871         down_read(&clp->cl_sem);
2872         down(&state->lock_sema);
2873         status = nfs4_set_lock_state(state, request);
2874         if (status == 0)
2875                 status = _nfs4_do_setlk(state, cmd, request, 0);
2876         up(&state->lock_sema);
2877         if (status == 0) {
2878                 /* Note: we always want to sleep here! */
2879                 request->fl_flags |= FL_SLEEP;
2880                 if (do_vfs_lock(request->fl_file, request) < 0)
2881                         printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __FUNCTION__);
2882         }
2883         up_read(&clp->cl_sem);
2884         return status;
2885 }
2886
2887 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
2888 {
2889         struct nfs4_exception exception = { };
2890         int err;
2891
2892         do {
2893                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
2894                                 _nfs4_proc_setlk(state, cmd, request),
2895                                 &exception);
2896         } while (exception.retry);
2897         return err;
2898 }
2899
2900 static int
2901 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
2902 {
2903         struct nfs_open_context *ctx;
2904         struct nfs4_state *state;
2905         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
2906         int status;
2907
2908         /* verify open state */
2909         ctx = (struct nfs_open_context *)filp->private_data;
2910         state = ctx->state;
2911
2912         if (request->fl_start < 0 || request->fl_end < 0)
2913                 return -EINVAL;
2914
2915         if (IS_GETLK(cmd))
2916                 return nfs4_proc_getlk(state, F_GETLK, request);
2917
2918         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
2919                 return -EINVAL;
2920
2921         if (request->fl_type == F_UNLCK)
2922                 return nfs4_proc_unlck(state, cmd, request);
2923
2924         do {
2925                 status = nfs4_proc_setlk(state, cmd, request);
2926                 if ((status != -EAGAIN) || IS_SETLK(cmd))
2927                         break;
2928                 timeout = nfs4_set_lock_task_retry(timeout);
2929                 status = -ERESTARTSYS;
2930                 if (signalled())
2931                         break;
2932         } while(status < 0);
2933         return status;
2934 }
2935
2936
2937 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
2938
2939 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
2940                 size_t buflen, int flags)
2941 {
2942         struct inode *inode = dentry->d_inode;
2943
2944         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
2945                 return -EOPNOTSUPP;
2946
2947         if (!S_ISREG(inode->i_mode) &&
2948             (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX))
2949                 return -EPERM;
2950
2951         return nfs4_proc_set_acl(inode, buf, buflen);
2952 }
2953
2954 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
2955  * and that's what we'll do for e.g. user attributes that haven't been set.
2956  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
2957  * attributes in kernel-managed attribute namespaces. */
2958 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
2959                 size_t buflen)
2960 {
2961         struct inode *inode = dentry->d_inode;
2962
2963         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
2964                 return -EOPNOTSUPP;
2965
2966         return nfs4_proc_get_acl(inode, buf, buflen);
2967 }
2968
2969 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
2970 {
2971         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
2972
2973         if (buf && buflen < len)
2974                 return -ERANGE;
2975         if (buf)
2976                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
2977         return len;
2978 }
2979
2980 struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops = {
2981         .recover_open   = nfs4_open_reclaim,
2982         .recover_lock   = nfs4_lock_reclaim,
2983 };
2984
2985 struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops = {
2986         .recover_open   = nfs4_open_expired,
2987         .recover_lock   = nfs4_lock_expired,
2988 };
2989
2990 static struct inode_operations nfs4_file_inode_operations = {
2991         .permission     = nfs_permission,
2992         .getattr        = nfs_getattr,
2993         .setattr        = nfs_setattr,
2994         .getxattr       = nfs4_getxattr,
2995         .setxattr       = nfs4_setxattr,
2996         .listxattr      = nfs4_listxattr,
2997 };
2998
2999 struct nfs_rpc_ops      nfs_v4_clientops = {
3000         .version        = 4,                    /* protocol version */
3001         .dentry_ops     = &nfs4_dentry_operations,
3002         .dir_inode_ops  = &nfs4_dir_inode_operations,
3003         .file_inode_ops = &nfs4_file_inode_operations,
3004         .getroot        = nfs4_proc_get_root,
3005         .getattr        = nfs4_proc_getattr,
3006         .setattr        = nfs4_proc_setattr,
3007         .lookup         = nfs4_proc_lookup,
3008         .access         = nfs4_proc_access,
3009         .readlink       = nfs4_proc_readlink,
3010         .read           = nfs4_proc_read,
3011         .write          = nfs4_proc_write,
3012         .commit         = nfs4_proc_commit,
3013         .create         = nfs4_proc_create,
3014         .remove         = nfs4_proc_remove,
3015         .unlink_setup   = nfs4_proc_unlink_setup,
3016         .unlink_done    = nfs4_proc_unlink_done,
3017         .rename         = nfs4_proc_rename,
3018         .link           = nfs4_proc_link,
3019         .symlink        = nfs4_proc_symlink,
3020         .mkdir          = nfs4_proc_mkdir,
3021         .rmdir          = nfs4_proc_remove,
3022         .readdir        = nfs4_proc_readdir,
3023         .mknod          = nfs4_proc_mknod,
3024         .statfs         = nfs4_proc_statfs,
3025         .fsinfo         = nfs4_proc_fsinfo,
3026         .pathconf       = nfs4_proc_pathconf,
3027         .decode_dirent  = nfs4_decode_dirent,
3028         .read_setup     = nfs4_proc_read_setup,
3029         .write_setup    = nfs4_proc_write_setup,
3030         .commit_setup   = nfs4_proc_commit_setup,
3031         .file_open      = nfs4_proc_file_open,
3032         .file_release   = nfs4_proc_file_release,
3033         .lock           = nfs4_proc_lock,
3034         .clear_acl_cache = nfs4_zap_acl_attr,
3035 };
3036
3037 /*
3038  * Local variables:
3039  *  c-basic-offset: 8
3040  * End:
3041  */