fs/nfs/nfs4proc.c

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