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/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/nfs_idmap.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #include "nfs4trace.h"
  70
  71 #define NFSDBG_FACILITY         NFSDBG_PROC
  72
  73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  75
  76 struct nfs4_opendata;
  77 static int _nfs4_proc_open(struct nfs4_opendata *data);
  78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state, struct nfs4_label *ilabel,
  87                             struct nfs4_label *olabel);
  88 #ifdef CONFIG_NFS_V4_1
  89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  90                 struct rpc_cred *);
  91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  92                 struct rpc_cred *);
  93 #endif
  94
  95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  96 static inline struct nfs4_label *
  97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  98         struct iattr *sattr, struct nfs4_label *label)
  99 {
 100         int err;
 101
 102         if (label == NULL)
 103                 return NULL;
 104
 105         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 106                 return NULL;
 107
 108         err = security_dentry_init_security(dentry, sattr->ia_mode,
 109                                 &dentry->d_name, (void **)&label->label, &label->len);
 110         if (err == 0)
 111                 return label;
 112
 113         return NULL;
 114 }
 115 static inline void
 116 nfs4_label_release_security(struct nfs4_label *label)
 117 {
 118         if (label)
 119                 security_release_secctx(label->label, label->len);
 120 }
 121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 122 {
 123         if (label)
 124                 return server->attr_bitmask;
 125
 126         return server->attr_bitmask_nl;
 127 }
 128 #else
 129 static inline struct nfs4_label *
 130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 131         struct iattr *sattr, struct nfs4_label *l)
 132 { return NULL; }
 133 static inline void
 134 nfs4_label_release_security(struct nfs4_label *label)
 135 { return; }
 136 static inline u32 *
 137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 { return server->attr_bitmask; }
 139 #endif
 140
 141 /* Prevent leaks of NFSv4 errors into userland */
 142 static int nfs4_map_errors(int err)
 143 {
 144         if (err >= -1000)
 145                 return err;
 146         switch (err) {
 147         case -NFS4ERR_RESOURCE:
 148         case -NFS4ERR_LAYOUTTRYLATER:
 149         case -NFS4ERR_RECALLCONFLICT:
 150                 return -EREMOTEIO;
 151         case -NFS4ERR_WRONGSEC:
 152         case -NFS4ERR_WRONG_CRED:
 153                 return -EPERM;
 154         case -NFS4ERR_BADOWNER:
 155         case -NFS4ERR_BADNAME:
 156                 return -EINVAL;
 157         case -NFS4ERR_SHARE_DENIED:
 158                 return -EACCES;
 159         case -NFS4ERR_MINOR_VERS_MISMATCH:
 160                 return -EPROTONOSUPPORT;
 161         case -NFS4ERR_ACCESS:
 162                 return -EACCES;
 163         case -NFS4ERR_FILE_OPEN:
 164                 return -EBUSY;
 165         default:
 166                 dprintk("%s could not handle NFSv4 error %d\n",
 167                                 __func__, -err);
 168                 break;
 169         }
 170         return -EIO;
 171 }
 172
 173 /*
 174  * This is our standard bitmap for GETATTR requests.
 175  */
 176 const u32 nfs4_fattr_bitmap[3] = {
 177         FATTR4_WORD0_TYPE
 178         | FATTR4_WORD0_CHANGE
 179         | FATTR4_WORD0_SIZE
 180         | FATTR4_WORD0_FSID
 181         | FATTR4_WORD0_FILEID,
 182         FATTR4_WORD1_MODE
 183         | FATTR4_WORD1_NUMLINKS
 184         | FATTR4_WORD1_OWNER
 185         | FATTR4_WORD1_OWNER_GROUP
 186         | FATTR4_WORD1_RAWDEV
 187         | FATTR4_WORD1_SPACE_USED
 188         | FATTR4_WORD1_TIME_ACCESS
 189         | FATTR4_WORD1_TIME_METADATA
 190         | FATTR4_WORD1_TIME_MODIFY,
 191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 192         FATTR4_WORD2_SECURITY_LABEL
 193 #endif
 194 };
 195
 196 static const u32 nfs4_pnfs_open_bitmap[3] = {
 197         FATTR4_WORD0_TYPE
 198         | FATTR4_WORD0_CHANGE
 199         | FATTR4_WORD0_SIZE
 200         | FATTR4_WORD0_FSID
 201         | FATTR4_WORD0_FILEID,
 202         FATTR4_WORD1_MODE
 203         | FATTR4_WORD1_NUMLINKS
 204         | FATTR4_WORD1_OWNER
 205         | FATTR4_WORD1_OWNER_GROUP
 206         | FATTR4_WORD1_RAWDEV
 207         | FATTR4_WORD1_SPACE_USED
 208         | FATTR4_WORD1_TIME_ACCESS
 209         | FATTR4_WORD1_TIME_METADATA
 210         | FATTR4_WORD1_TIME_MODIFY,
 211         FATTR4_WORD2_MDSTHRESHOLD
 212 };
 213
 214 static const u32 nfs4_open_noattr_bitmap[3] = {
 215         FATTR4_WORD0_TYPE
 216         | FATTR4_WORD0_CHANGE
 217         | FATTR4_WORD0_FILEID,
 218 };
 219
 220 const u32 nfs4_statfs_bitmap[3] = {
 221         FATTR4_WORD0_FILES_AVAIL
 222         | FATTR4_WORD0_FILES_FREE
 223         | FATTR4_WORD0_FILES_TOTAL,
 224         FATTR4_WORD1_SPACE_AVAIL
 225         | FATTR4_WORD1_SPACE_FREE
 226         | FATTR4_WORD1_SPACE_TOTAL
 227 };
 228
 229 const u32 nfs4_pathconf_bitmap[3] = {
 230         FATTR4_WORD0_MAXLINK
 231         | FATTR4_WORD0_MAXNAME,
 232         0
 233 };
 234
 235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 236                         | FATTR4_WORD0_MAXREAD
 237                         | FATTR4_WORD0_MAXWRITE
 238                         | FATTR4_WORD0_LEASE_TIME,
 239                         FATTR4_WORD1_TIME_DELTA
 240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 241                         FATTR4_WORD2_LAYOUT_BLKSIZE
 242 };
 243
 244 const u32 nfs4_fs_locations_bitmap[3] = {
 245         FATTR4_WORD0_TYPE
 246         | FATTR4_WORD0_CHANGE
 247         | FATTR4_WORD0_SIZE
 248         | FATTR4_WORD0_FSID
 249         | FATTR4_WORD0_FILEID
 250         | FATTR4_WORD0_FS_LOCATIONS,
 251         FATTR4_WORD1_MODE
 252         | FATTR4_WORD1_NUMLINKS
 253         | FATTR4_WORD1_OWNER
 254         | FATTR4_WORD1_OWNER_GROUP
 255         | FATTR4_WORD1_RAWDEV
 256         | FATTR4_WORD1_SPACE_USED
 257         | FATTR4_WORD1_TIME_ACCESS
 258         | FATTR4_WORD1_TIME_METADATA
 259         | FATTR4_WORD1_TIME_MODIFY
 260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 261 };
 262
 263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 264                 struct nfs4_readdir_arg *readdir)
 265 {
 266         __be32 *start, *p;
 267
 268         if (cookie > 2) {
 269                 readdir->cookie = cookie;
 270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 271                 return;
 272         }
 273
 274         readdir->cookie = 0;
 275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 276         if (cookie == 2)
 277                 return;
 278
 279         /*
 280          * NFSv4 servers do not return entries for '.' and '..'
 281          * Therefore, we fake these entries here.  We let '.'
 282          * have cookie 0 and '..' have cookie 1.  Note that
 283          * when talking to the server, we always send cookie 0
 284          * instead of 1 or 2.
 285          */
 286         start = p = kmap_atomic(*readdir->pages);
 287
 288         if (cookie == 0) {
 289                 *p++ = xdr_one;                                  /* next */
 290                 *p++ = xdr_zero;                   /* cookie, first word */
 291                 *p++ = xdr_one;                   /* cookie, second word */
 292                 *p++ = xdr_one;                             /* entry len */
 293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 294                 p++;
 295                 *p++ = xdr_one;                         /* bitmap length */
 296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 297                 *p++ = htonl(8);              /* attribute buffer length */
 298                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 299         }
 300
 301         *p++ = xdr_one;                                  /* next */
 302         *p++ = xdr_zero;                   /* cookie, first word */
 303         *p++ = xdr_two;                   /* cookie, second word */
 304         *p++ = xdr_two;                             /* entry len */
 305         memcpy(p, "..\0\0", 4);                         /* entry */
 306         p++;
 307         *p++ = xdr_one;                         /* bitmap length */
 308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309         *p++ = htonl(8);              /* attribute buffer length */
 310         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 311
 312         readdir->pgbase = (char *)p - (char *)start;
 313         readdir->count -= readdir->pgbase;
 314         kunmap_atomic(start);
 315 }
 316
 317 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 318 {
 319         int res = 0;
 320
 321         might_sleep();
 322
 323         if (*timeout <= 0)
 324                 *timeout = NFS4_POLL_RETRY_MIN;
 325         if (*timeout > NFS4_POLL_RETRY_MAX)
 326                 *timeout = NFS4_POLL_RETRY_MAX;
 327         freezable_schedule_timeout_killable_unsafe(*timeout);
 328         if (fatal_signal_pending(current))
 329                 res = -ERESTARTSYS;
 330         *timeout <<= 1;
 331         return res;
 332 }
 333
 334 /* This is the error handling routine for processes that are allowed
 335  * to sleep.
 336  */
 337 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 338 {
 339         struct nfs_client *clp = server->nfs_client;
 340         struct nfs4_state *state = exception->state;
 341         struct inode *inode = exception->inode;
 342         int ret = errorcode;
 343
 344         exception->retry = 0;
 345         switch(errorcode) {
 346                 case 0:
 347                         return 0;
 348                 case -NFS4ERR_OPENMODE:
 349                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 350                                 nfs4_inode_return_delegation(inode);
 351                                 exception->retry = 1;
 352                                 return 0;
 353                         }
 354                         if (state == NULL)
 355                                 break;
 356                         ret = nfs4_schedule_stateid_recovery(server, state);
 357                         if (ret < 0)
 358                                 break;
 359                         goto wait_on_recovery;
 360                 case -NFS4ERR_DELEG_REVOKED:
 361                 case -NFS4ERR_ADMIN_REVOKED:
 362                 case -NFS4ERR_BAD_STATEID:
 363                         if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
 364                                 nfs_remove_bad_delegation(inode);
 365                                 exception->retry = 1;
 366                                 break;
 367                         }
 368                         if (state == NULL)
 369                                 break;
 370                         ret = nfs4_schedule_stateid_recovery(server, state);
 371                         if (ret < 0)
 372                                 break;
 373                         goto wait_on_recovery;
 374                 case -NFS4ERR_EXPIRED:
 375                         if (state != NULL) {
 376                                 ret = nfs4_schedule_stateid_recovery(server, state);
 377                                 if (ret < 0)
 378                                         break;
 379                         }
 380                 case -NFS4ERR_STALE_STATEID:
 381                 case -NFS4ERR_STALE_CLIENTID:
 382                         nfs4_schedule_lease_recovery(clp);
 383                         goto wait_on_recovery;
 384                 case -NFS4ERR_MOVED:
 385                         ret = nfs4_schedule_migration_recovery(server);
 386                         if (ret < 0)
 387                                 break;
 388                         goto wait_on_recovery;
 389                 case -NFS4ERR_LEASE_MOVED:
 390                         nfs4_schedule_lease_moved_recovery(clp);
 391                         goto wait_on_recovery;
 392 #if defined(CONFIG_NFS_V4_1)
 393                 case -NFS4ERR_BADSESSION:
 394                 case -NFS4ERR_BADSLOT:
 395                 case -NFS4ERR_BAD_HIGH_SLOT:
 396                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 397                 case -NFS4ERR_DEADSESSION:
 398                 case -NFS4ERR_SEQ_FALSE_RETRY:
 399                 case -NFS4ERR_SEQ_MISORDERED:
 400                         dprintk("%s ERROR: %d Reset session\n", __func__,
 401                                 errorcode);
 402                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 403                         goto wait_on_recovery;
 404 #endif /* defined(CONFIG_NFS_V4_1) */
 405                 case -NFS4ERR_FILE_OPEN:
 406                         if (exception->timeout > HZ) {
 407                                 /* We have retried a decent amount, time to
 408                                  * fail
 409                                  */
 410                                 ret = -EBUSY;
 411                                 break;
 412                         }
 413                 case -NFS4ERR_GRACE:
 414                 case -NFS4ERR_DELAY:
 415                         ret = nfs4_delay(server->client, &exception->timeout);
 416                         if (ret != 0)
 417                                 break;
 418                 case -NFS4ERR_RETRY_UNCACHED_REP:
 419                 case -NFS4ERR_OLD_STATEID:
 420                         exception->retry = 1;
 421                         break;
 422                 case -NFS4ERR_BADOWNER:
 423                         /* The following works around a Linux server bug! */
 424                 case -NFS4ERR_BADNAME:
 425                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 426                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 427                                 exception->retry = 1;
 428                                 printk(KERN_WARNING "NFS: v4 server %s "
 429                                                 "does not accept raw "
 430                                                 "uid/gids. "
 431                                                 "Reenabling the idmapper.\n",
 432                                                 server->nfs_client->cl_hostname);
 433                         }
 434         }
 435         /* We failed to handle the error */
 436         return nfs4_map_errors(ret);
 437 wait_on_recovery:
 438         ret = nfs4_wait_clnt_recover(clp);
 439         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 440                 return -EIO;
 441         if (ret == 0)
 442                 exception->retry = 1;
 443         return ret;
 444 }
 445
 446 /*
 447  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 448  * or 'false' otherwise.
 449  */
 450 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 451 {
 452         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 453
 454         if (flavor == RPC_AUTH_GSS_KRB5I ||
 455             flavor == RPC_AUTH_GSS_KRB5P)
 456                 return true;
 457
 458         return false;
 459 }
 460
 461 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 462 {
 463         spin_lock(&clp->cl_lock);
 464         if (time_before(clp->cl_last_renewal,timestamp))
 465                 clp->cl_last_renewal = timestamp;
 466         spin_unlock(&clp->cl_lock);
 467 }
 468
 469 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 470 {
 471         do_renew_lease(server->nfs_client, timestamp);
 472 }
 473
 474 struct nfs4_call_sync_data {
 475         const struct nfs_server *seq_server;
 476         struct nfs4_sequence_args *seq_args;
 477         struct nfs4_sequence_res *seq_res;
 478 };
 479
 480 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
 481                                struct nfs4_sequence_res *res, int cache_reply)
 482 {
 483         args->sa_slot = NULL;
 484         args->sa_cache_this = cache_reply;
 485         args->sa_privileged = 0;
 486
 487         res->sr_slot = NULL;
 488 }
 489
 490 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 491 {
 492         args->sa_privileged = 1;
 493 }
 494
 495 static int nfs40_setup_sequence(const struct nfs_server *server,
 496                                 struct nfs4_sequence_args *args,
 497                                 struct nfs4_sequence_res *res,
 498                                 struct rpc_task *task)
 499 {
 500         struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
 501         struct nfs4_slot *slot;
 502
 503         /* slot already allocated? */
 504         if (res->sr_slot != NULL)
 505                 goto out_start;
 506
 507         spin_lock(&tbl->slot_tbl_lock);
 508         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 509                 goto out_sleep;
 510
 511         slot = nfs4_alloc_slot(tbl);
 512         if (IS_ERR(slot)) {
 513                 if (slot == ERR_PTR(-ENOMEM))
 514                         task->tk_timeout = HZ >> 2;
 515                 goto out_sleep;
 516         }
 517         spin_unlock(&tbl->slot_tbl_lock);
 518
 519         args->sa_slot = slot;
 520         res->sr_slot = slot;
 521
 522 out_start:
 523         rpc_call_start(task);
 524         return 0;
 525
 526 out_sleep:
 527         if (args->sa_privileged)
 528                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 529                                 NULL, RPC_PRIORITY_PRIVILEGED);
 530         else
 531                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 532         spin_unlock(&tbl->slot_tbl_lock);
 533         return -EAGAIN;
 534 }
 535
 536 static int nfs40_sequence_done(struct rpc_task *task,
 537                                struct nfs4_sequence_res *res)
 538 {
 539         struct nfs4_slot *slot = res->sr_slot;
 540         struct nfs4_slot_table *tbl;
 541
 542         if (slot == NULL)
 543                 goto out;
 544
 545         tbl = slot->table;
 546         spin_lock(&tbl->slot_tbl_lock);
 547         if (!nfs41_wake_and_assign_slot(tbl, slot))
 548                 nfs4_free_slot(tbl, slot);
 549         spin_unlock(&tbl->slot_tbl_lock);
 550
 551         res->sr_slot = NULL;
 552 out:
 553         return 1;
 554 }
 555
 556 #if defined(CONFIG_NFS_V4_1)
 557
 558 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 559 {
 560         struct nfs4_session *session;
 561         struct nfs4_slot_table *tbl;
 562         struct nfs4_slot *slot = res->sr_slot;
 563         bool send_new_highest_used_slotid = false;
 564
 565         tbl = slot->table;
 566         session = tbl->session;
 567
 568         spin_lock(&tbl->slot_tbl_lock);
 569         /* Be nice to the server: try to ensure that the last transmitted
 570          * value for highest_user_slotid <= target_highest_slotid
 571          */
 572         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 573                 send_new_highest_used_slotid = true;
 574
 575         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 576                 send_new_highest_used_slotid = false;
 577                 goto out_unlock;
 578         }
 579         nfs4_free_slot(tbl, slot);
 580
 581         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 582                 send_new_highest_used_slotid = false;
 583 out_unlock:
 584         spin_unlock(&tbl->slot_tbl_lock);
 585         res->sr_slot = NULL;
 586         if (send_new_highest_used_slotid)
 587                 nfs41_server_notify_highest_slotid_update(session->clp);
 588 }
 589
 590 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 591 {
 592         struct nfs4_session *session;
 593         struct nfs4_slot *slot = res->sr_slot;
 594         struct nfs_client *clp;
 595         bool interrupted = false;
 596         int ret = 1;
 597
 598         if (slot == NULL)
 599                 goto out_noaction;
 600         /* don't increment the sequence number if the task wasn't sent */
 601         if (!RPC_WAS_SENT(task))
 602                 goto out;
 603
 604         session = slot->table->session;
 605
 606         if (slot->interrupted) {
 607                 slot->interrupted = 0;
 608                 interrupted = true;
 609         }
 610
 611         trace_nfs4_sequence_done(session, res);
 612         /* Check the SEQUENCE operation status */
 613         switch (res->sr_status) {
 614         case 0:
 615                 /* Update the slot's sequence and clientid lease timer */
 616                 ++slot->seq_nr;
 617                 clp = session->clp;
 618                 do_renew_lease(clp, res->sr_timestamp);
 619                 /* Check sequence flags */
 620                 if (res->sr_status_flags != 0)
 621                         nfs4_schedule_lease_recovery(clp);
 622                 nfs41_update_target_slotid(slot->table, slot, res);
 623                 break;
 624         case 1:
 625                 /*
 626                  * sr_status remains 1 if an RPC level error occurred.
 627                  * The server may or may not have processed the sequence
 628                  * operation..
 629                  * Mark the slot as having hosted an interrupted RPC call.
 630                  */
 631                 slot->interrupted = 1;
 632                 goto out;
 633         case -NFS4ERR_DELAY:
 634                 /* The server detected a resend of the RPC call and
 635                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 636                  * of RFC5661.
 637                  */
 638                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 639                         __func__,
 640                         slot->slot_nr,
 641                         slot->seq_nr);
 642                 goto out_retry;
 643         case -NFS4ERR_BADSLOT:
 644                 /*
 645                  * The slot id we used was probably retired. Try again
 646                  * using a different slot id.
 647                  */
 648                 goto retry_nowait;
 649         case -NFS4ERR_SEQ_MISORDERED:
 650                 /*
 651                  * Was the last operation on this sequence interrupted?
 652                  * If so, retry after bumping the sequence number.
 653                  */
 654                 if (interrupted) {
 655                         ++slot->seq_nr;
 656                         goto retry_nowait;
 657                 }
 658                 /*
 659                  * Could this slot have been previously retired?
 660                  * If so, then the server may be expecting seq_nr = 1!
 661                  */
 662                 if (slot->seq_nr != 1) {
 663                         slot->seq_nr = 1;
 664                         goto retry_nowait;
 665                 }
 666                 break;
 667         case -NFS4ERR_SEQ_FALSE_RETRY:
 668                 ++slot->seq_nr;
 669                 goto retry_nowait;
 670         default:
 671                 /* Just update the slot sequence no. */
 672                 ++slot->seq_nr;
 673         }
 674 out:
 675         /* The session may be reset by one of the error handlers. */
 676         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 677         nfs41_sequence_free_slot(res);
 678 out_noaction:
 679         return ret;
 680 retry_nowait:
 681         if (rpc_restart_call_prepare(task)) {
 682                 task->tk_status = 0;
 683                 ret = 0;
 684         }
 685         goto out;
 686 out_retry:
 687         if (!rpc_restart_call(task))
 688                 goto out;
 689         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 690         return 0;
 691 }
 692 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 693
 694 static int nfs4_sequence_done(struct rpc_task *task,
 695                                struct nfs4_sequence_res *res)
 696 {
 697         if (res->sr_slot == NULL)
 698                 return 1;
 699         if (!res->sr_slot->table->session)
 700                 return nfs40_sequence_done(task, res);
 701         return nfs41_sequence_done(task, res);
 702 }
 703
 704 int nfs41_setup_sequence(struct nfs4_session *session,
 705                                 struct nfs4_sequence_args *args,
 706                                 struct nfs4_sequence_res *res,
 707                                 struct rpc_task *task)
 708 {
 709         struct nfs4_slot *slot;
 710         struct nfs4_slot_table *tbl;
 711
 712         dprintk("--> %s\n", __func__);
 713         /* slot already allocated? */
 714         if (res->sr_slot != NULL)
 715                 goto out_success;
 716
 717         tbl = &session->fc_slot_table;
 718
 719         task->tk_timeout = 0;
 720
 721         spin_lock(&tbl->slot_tbl_lock);
 722         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 723             !args->sa_privileged) {
 724                 /* The state manager will wait until the slot table is empty */
 725                 dprintk("%s session is draining\n", __func__);
 726                 goto out_sleep;
 727         }
 728
 729         slot = nfs4_alloc_slot(tbl);
 730         if (IS_ERR(slot)) {
 731                 /* If out of memory, try again in 1/4 second */
 732                 if (slot == ERR_PTR(-ENOMEM))
 733                         task->tk_timeout = HZ >> 2;
 734                 dprintk("<-- %s: no free slots\n", __func__);
 735                 goto out_sleep;
 736         }
 737         spin_unlock(&tbl->slot_tbl_lock);
 738
 739         args->sa_slot = slot;
 740
 741         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 742                         slot->slot_nr, slot->seq_nr);
 743
 744         res->sr_slot = slot;
 745         res->sr_timestamp = jiffies;
 746         res->sr_status_flags = 0;
 747         /*
 748          * sr_status is only set in decode_sequence, and so will remain
 749          * set to 1 if an rpc level failure occurs.
 750          */
 751         res->sr_status = 1;
 752         trace_nfs4_setup_sequence(session, args);
 753 out_success:
 754         rpc_call_start(task);
 755         return 0;
 756 out_sleep:
 757         /* Privileged tasks are queued with top priority */
 758         if (args->sa_privileged)
 759                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 760                                 NULL, RPC_PRIORITY_PRIVILEGED);
 761         else
 762                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 763         spin_unlock(&tbl->slot_tbl_lock);
 764         return -EAGAIN;
 765 }
 766 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 767
 768 static int nfs4_setup_sequence(const struct nfs_server *server,
 769                                struct nfs4_sequence_args *args,
 770                                struct nfs4_sequence_res *res,
 771                                struct rpc_task *task)
 772 {
 773         struct nfs4_session *session = nfs4_get_session(server);
 774         int ret = 0;
 775
 776         if (!session)
 777                 return nfs40_setup_sequence(server, args, res, task);
 778
 779         dprintk("--> %s clp %p session %p sr_slot %u\n",
 780                 __func__, session->clp, session, res->sr_slot ?
 781                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 782
 783         ret = nfs41_setup_sequence(session, args, res, task);
 784
 785         dprintk("<-- %s status=%d\n", __func__, ret);
 786         return ret;
 787 }
 788
 789 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 790 {
 791         struct nfs4_call_sync_data *data = calldata;
 792         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 793
 794         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 795
 796         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 797 }
 798
 799 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 800 {
 801         struct nfs4_call_sync_data *data = calldata;
 802
 803         nfs41_sequence_done(task, data->seq_res);
 804 }
 805
 806 static const struct rpc_call_ops nfs41_call_sync_ops = {
 807         .rpc_call_prepare = nfs41_call_sync_prepare,
 808         .rpc_call_done = nfs41_call_sync_done,
 809 };
 810
 811 #else   /* !CONFIG_NFS_V4_1 */
 812
 813 static int nfs4_setup_sequence(const struct nfs_server *server,
 814                                struct nfs4_sequence_args *args,
 815                                struct nfs4_sequence_res *res,
 816                                struct rpc_task *task)
 817 {
 818         return nfs40_setup_sequence(server, args, res, task);
 819 }
 820
 821 static int nfs4_sequence_done(struct rpc_task *task,
 822                                struct nfs4_sequence_res *res)
 823 {
 824         return nfs40_sequence_done(task, res);
 825 }
 826
 827 #endif  /* !CONFIG_NFS_V4_1 */
 828
 829 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 830 {
 831         struct nfs4_call_sync_data *data = calldata;
 832         nfs4_setup_sequence(data->seq_server,
 833                                 data->seq_args, data->seq_res, task);
 834 }
 835
 836 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 837 {
 838         struct nfs4_call_sync_data *data = calldata;
 839         nfs4_sequence_done(task, data->seq_res);
 840 }
 841
 842 static const struct rpc_call_ops nfs40_call_sync_ops = {
 843         .rpc_call_prepare = nfs40_call_sync_prepare,
 844         .rpc_call_done = nfs40_call_sync_done,
 845 };
 846
 847 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 848                                    struct nfs_server *server,
 849                                    struct rpc_message *msg,
 850                                    struct nfs4_sequence_args *args,
 851                                    struct nfs4_sequence_res *res)
 852 {
 853         int ret;
 854         struct rpc_task *task;
 855         struct nfs_client *clp = server->nfs_client;
 856         struct nfs4_call_sync_data data = {
 857                 .seq_server = server,
 858                 .seq_args = args,
 859                 .seq_res = res,
 860         };
 861         struct rpc_task_setup task_setup = {
 862                 .rpc_client = clnt,
 863                 .rpc_message = msg,
 864                 .callback_ops = clp->cl_mvops->call_sync_ops,
 865                 .callback_data = &data
 866         };
 867
 868         task = rpc_run_task(&task_setup);
 869         if (IS_ERR(task))
 870                 ret = PTR_ERR(task);
 871         else {
 872                 ret = task->tk_status;
 873                 rpc_put_task(task);
 874         }
 875         return ret;
 876 }
 877
 878 static
 879 int nfs4_call_sync(struct rpc_clnt *clnt,
 880                    struct nfs_server *server,
 881                    struct rpc_message *msg,
 882                    struct nfs4_sequence_args *args,
 883                    struct nfs4_sequence_res *res,
 884                    int cache_reply)
 885 {
 886         nfs4_init_sequence(args, res, cache_reply);
 887         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 888 }
 889
 890 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 891 {
 892         struct nfs_inode *nfsi = NFS_I(dir);
 893
 894         spin_lock(&dir->i_lock);
 895         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 896         if (!cinfo->atomic || cinfo->before != dir->i_version)
 897                 nfs_force_lookup_revalidate(dir);
 898         dir->i_version = cinfo->after;
 899         nfs_fscache_invalidate(dir);
 900         spin_unlock(&dir->i_lock);
 901 }
 902
 903 struct nfs4_opendata {
 904         struct kref kref;
 905         struct nfs_openargs o_arg;
 906         struct nfs_openres o_res;
 907         struct nfs_open_confirmargs c_arg;
 908         struct nfs_open_confirmres c_res;
 909         struct nfs4_string owner_name;
 910         struct nfs4_string group_name;
 911         struct nfs_fattr f_attr;
 912         struct nfs4_label *f_label;
 913         struct dentry *dir;
 914         struct dentry *dentry;
 915         struct nfs4_state_owner *owner;
 916         struct nfs4_state *state;
 917         struct iattr attrs;
 918         unsigned long timestamp;
 919         unsigned int rpc_done : 1;
 920         unsigned int file_created : 1;
 921         unsigned int is_recover : 1;
 922         int rpc_status;
 923         int cancelled;
 924 };
 925
 926 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 927                 int err, struct nfs4_exception *exception)
 928 {
 929         if (err != -EINVAL)
 930                 return false;
 931         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 932                 return false;
 933         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 934         exception->retry = 1;
 935         return true;
 936 }
 937
 938 static enum open_claim_type4
 939 nfs4_map_atomic_open_claim(struct nfs_server *server,
 940                 enum open_claim_type4 claim)
 941 {
 942         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 943                 return claim;
 944         switch (claim) {
 945         default:
 946                 return claim;
 947         case NFS4_OPEN_CLAIM_FH:
 948                 return NFS4_OPEN_CLAIM_NULL;
 949         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 950                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 951         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 952                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 953         }
 954 }
 955
 956 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 957 {
 958         p->o_res.f_attr = &p->f_attr;
 959         p->o_res.f_label = p->f_label;
 960         p->o_res.seqid = p->o_arg.seqid;
 961         p->c_res.seqid = p->c_arg.seqid;
 962         p->o_res.server = p->o_arg.server;
 963         p->o_res.access_request = p->o_arg.access;
 964         nfs_fattr_init(&p->f_attr);
 965         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 966 }
 967
 968 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 969                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 970                 const struct iattr *attrs,
 971                 struct nfs4_label *label,
 972                 enum open_claim_type4 claim,
 973                 gfp_t gfp_mask)
 974 {
 975         struct dentry *parent = dget_parent(dentry);
 976         struct inode *dir = parent->d_inode;
 977         struct nfs_server *server = NFS_SERVER(dir);
 978         struct nfs4_opendata *p;
 979
 980         p = kzalloc(sizeof(*p), gfp_mask);
 981         if (p == NULL)
 982                 goto err;
 983
 984         p->f_label = nfs4_label_alloc(server, gfp_mask);
 985         if (IS_ERR(p->f_label))
 986                 goto err_free_p;
 987
 988         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 989         if (p->o_arg.seqid == NULL)
 990                 goto err_free_label;
 991         nfs_sb_active(dentry->d_sb);
 992         p->dentry = dget(dentry);
 993         p->dir = parent;
 994         p->owner = sp;
 995         atomic_inc(&sp->so_count);
 996         p->o_arg.open_flags = flags;
 997         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 998         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 999          * will return permission denied for all bits until close */
1000         if (!(flags & O_EXCL)) {
1001                 /* ask server to check for all possible rights as results
1002                  * are cached */
1003                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1004                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1005         }
1006         p->o_arg.clientid = server->nfs_client->cl_clientid;
1007         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1008         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1009         p->o_arg.name = &dentry->d_name;
1010         p->o_arg.server = server;
1011         p->o_arg.bitmask = nfs4_bitmask(server, label);
1012         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1013         p->o_arg.label = label;
1014         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1015         switch (p->o_arg.claim) {
1016         case NFS4_OPEN_CLAIM_NULL:
1017         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1018         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1019                 p->o_arg.fh = NFS_FH(dir);
1020                 break;
1021         case NFS4_OPEN_CLAIM_PREVIOUS:
1022         case NFS4_OPEN_CLAIM_FH:
1023         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1024         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1025                 p->o_arg.fh = NFS_FH(dentry->d_inode);
1026         }
1027         if (attrs != NULL && attrs->ia_valid != 0) {
1028                 __u32 verf[2];
1029
1030                 p->o_arg.u.attrs = &p->attrs;
1031                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1032
1033                 verf[0] = jiffies;
1034                 verf[1] = current->pid;
1035                 memcpy(p->o_arg.u.verifier.data, verf,
1036                                 sizeof(p->o_arg.u.verifier.data));
1037         }
1038         p->c_arg.fh = &p->o_res.fh;
1039         p->c_arg.stateid = &p->o_res.stateid;
1040         p->c_arg.seqid = p->o_arg.seqid;
1041         nfs4_init_opendata_res(p);
1042         kref_init(&p->kref);
1043         return p;
1044
1045 err_free_label:
1046         nfs4_label_free(p->f_label);
1047 err_free_p:
1048         kfree(p);
1049 err:
1050         dput(parent);
1051         return NULL;
1052 }
1053
1054 static void nfs4_opendata_free(struct kref *kref)
1055 {
1056         struct nfs4_opendata *p = container_of(kref,
1057                         struct nfs4_opendata, kref);
1058         struct super_block *sb = p->dentry->d_sb;
1059
1060         nfs_free_seqid(p->o_arg.seqid);
1061         if (p->state != NULL)
1062                 nfs4_put_open_state(p->state);
1063         nfs4_put_state_owner(p->owner);
1064
1065         nfs4_label_free(p->f_label);
1066
1067         dput(p->dir);
1068         dput(p->dentry);
1069         nfs_sb_deactive(sb);
1070         nfs_fattr_free_names(&p->f_attr);
1071         kfree(p->f_attr.mdsthreshold);
1072         kfree(p);
1073 }
1074
1075 static void nfs4_opendata_put(struct nfs4_opendata *p)
1076 {
1077         if (p != NULL)
1078                 kref_put(&p->kref, nfs4_opendata_free);
1079 }
1080
1081 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1082 {
1083         int ret;
1084
1085         ret = rpc_wait_for_completion_task(task);
1086         return ret;
1087 }
1088
1089 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1090 {
1091         int ret = 0;
1092
1093         if (open_mode & (O_EXCL|O_TRUNC))
1094                 goto out;
1095         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1096                 case FMODE_READ:
1097                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1098                                 && state->n_rdonly != 0;
1099                         break;
1100                 case FMODE_WRITE:
1101                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1102                                 && state->n_wronly != 0;
1103                         break;
1104                 case FMODE_READ|FMODE_WRITE:
1105                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1106                                 && state->n_rdwr != 0;
1107         }
1108 out:
1109         return ret;
1110 }
1111
1112 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1113 {
1114         if (delegation == NULL)
1115                 return 0;
1116         if ((delegation->type & fmode) != fmode)
1117                 return 0;
1118         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1119                 return 0;
1120         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1121                 return 0;
1122         nfs_mark_delegation_referenced(delegation);
1123         return 1;
1124 }
1125
1126 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1127 {
1128         switch (fmode) {
1129                 case FMODE_WRITE:
1130                         state->n_wronly++;
1131                         break;
1132                 case FMODE_READ:
1133                         state->n_rdonly++;
1134                         break;
1135                 case FMODE_READ|FMODE_WRITE:
1136                         state->n_rdwr++;
1137         }
1138         nfs4_state_set_mode_locked(state, state->state | fmode);
1139 }
1140
1141 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1142 {
1143         struct nfs_client *clp = state->owner->so_server->nfs_client;
1144         bool need_recover = false;
1145
1146         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1147                 need_recover = true;
1148         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1149                 need_recover = true;
1150         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1151                 need_recover = true;
1152         if (need_recover)
1153                 nfs4_state_mark_reclaim_nograce(clp, state);
1154 }
1155
1156 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1157                 nfs4_stateid *stateid)
1158 {
1159         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1160                 return true;
1161         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1162                 nfs_test_and_clear_all_open_stateid(state);
1163                 return true;
1164         }
1165         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1166                 return true;
1167         return false;
1168 }
1169
1170 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1171                 nfs4_stateid *stateid, fmode_t fmode)
1172 {
1173         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1174         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1175         case FMODE_WRITE:
1176                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1177                 break;
1178         case FMODE_READ:
1179                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1180                 break;
1181         case 0:
1182                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1183                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1184                 clear_bit(NFS_OPEN_STATE, &state->flags);
1185         }
1186         if (stateid == NULL)
1187                 return;
1188         if (!nfs_need_update_open_stateid(state, stateid))
1189                 return;
1190         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1191                 nfs4_stateid_copy(&state->stateid, stateid);
1192         nfs4_stateid_copy(&state->open_stateid, stateid);
1193 }
1194
1195 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1196 {
1197         write_seqlock(&state->seqlock);
1198         nfs_clear_open_stateid_locked(state, stateid, fmode);
1199         write_sequnlock(&state->seqlock);
1200         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1201                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1202 }
1203
1204 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1205 {
1206         switch (fmode) {
1207                 case FMODE_READ:
1208                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1209                         break;
1210                 case FMODE_WRITE:
1211                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1212                         break;
1213                 case FMODE_READ|FMODE_WRITE:
1214                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1215         }
1216         if (!nfs_need_update_open_stateid(state, stateid))
1217                 return;
1218         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1219                 nfs4_stateid_copy(&state->stateid, stateid);
1220         nfs4_stateid_copy(&state->open_stateid, stateid);
1221 }
1222
1223 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1224 {
1225         /*
1226          * Protect the call to nfs4_state_set_mode_locked and
1227          * serialise the stateid update
1228          */
1229         write_seqlock(&state->seqlock);
1230         if (deleg_stateid != NULL) {
1231                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1232                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1233         }
1234         if (open_stateid != NULL)
1235                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1236         write_sequnlock(&state->seqlock);
1237         spin_lock(&state->owner->so_lock);
1238         update_open_stateflags(state, fmode);
1239         spin_unlock(&state->owner->so_lock);
1240 }
1241
1242 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1243 {
1244         struct nfs_inode *nfsi = NFS_I(state->inode);
1245         struct nfs_delegation *deleg_cur;
1246         int ret = 0;
1247
1248         fmode &= (FMODE_READ|FMODE_WRITE);
1249
1250         rcu_read_lock();
1251         deleg_cur = rcu_dereference(nfsi->delegation);
1252         if (deleg_cur == NULL)
1253                 goto no_delegation;
1254
1255         spin_lock(&deleg_cur->lock);
1256         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1257            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1258             (deleg_cur->type & fmode) != fmode)
1259                 goto no_delegation_unlock;
1260
1261         if (delegation == NULL)
1262                 delegation = &deleg_cur->stateid;
1263         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1264                 goto no_delegation_unlock;
1265
1266         nfs_mark_delegation_referenced(deleg_cur);
1267         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1268         ret = 1;
1269 no_delegation_unlock:
1270         spin_unlock(&deleg_cur->lock);
1271 no_delegation:
1272         rcu_read_unlock();
1273
1274         if (!ret && open_stateid != NULL) {
1275                 __update_open_stateid(state, open_stateid, NULL, fmode);
1276                 ret = 1;
1277         }
1278         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1279                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1280
1281         return ret;
1282 }
1283
1284
1285 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1286 {
1287         struct nfs_delegation *delegation;
1288
1289         rcu_read_lock();
1290         delegation = rcu_dereference(NFS_I(inode)->delegation);
1291         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1292                 rcu_read_unlock();
1293                 return;
1294         }
1295         rcu_read_unlock();
1296         nfs4_inode_return_delegation(inode);
1297 }
1298
1299 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1300 {
1301         struct nfs4_state *state = opendata->state;
1302         struct nfs_inode *nfsi = NFS_I(state->inode);
1303         struct nfs_delegation *delegation;
1304         int open_mode = opendata->o_arg.open_flags;
1305         fmode_t fmode = opendata->o_arg.fmode;
1306         nfs4_stateid stateid;
1307         int ret = -EAGAIN;
1308
1309         for (;;) {
1310                 if (can_open_cached(state, fmode, open_mode)) {
1311                         spin_lock(&state->owner->so_lock);
1312                         if (can_open_cached(state, fmode, open_mode)) {
1313                                 update_open_stateflags(state, fmode);
1314                                 spin_unlock(&state->owner->so_lock);
1315                                 goto out_return_state;
1316                         }
1317                         spin_unlock(&state->owner->so_lock);
1318                 }
1319                 rcu_read_lock();
1320                 delegation = rcu_dereference(nfsi->delegation);
1321                 if (!can_open_delegated(delegation, fmode)) {
1322                         rcu_read_unlock();
1323                         break;
1324                 }
1325                 /* Save the delegation */
1326                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1327                 rcu_read_unlock();
1328                 nfs_release_seqid(opendata->o_arg.seqid);
1329                 if (!opendata->is_recover) {
1330                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1331                         if (ret != 0)
1332                                 goto out;
1333                 }
1334                 ret = -EAGAIN;
1335
1336                 /* Try to update the stateid using the delegation */
1337                 if (update_open_stateid(state, NULL, &stateid, fmode))
1338                         goto out_return_state;
1339         }
1340 out:
1341         return ERR_PTR(ret);
1342 out_return_state:
1343         atomic_inc(&state->count);
1344         return state;
1345 }
1346
1347 static void
1348 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1349 {
1350         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1351         struct nfs_delegation *delegation;
1352         int delegation_flags = 0;
1353
1354         rcu_read_lock();
1355         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1356         if (delegation)
1357                 delegation_flags = delegation->flags;
1358         rcu_read_unlock();
1359         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1360                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1361                                    "returning a delegation for "
1362                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1363                                    clp->cl_hostname);
1364         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1365                 nfs_inode_set_delegation(state->inode,
1366                                          data->owner->so_cred,
1367                                          &data->o_res);
1368         else
1369                 nfs_inode_reclaim_delegation(state->inode,
1370                                              data->owner->so_cred,
1371                                              &data->o_res);
1372 }
1373
1374 /*
1375  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1376  * and update the nfs4_state.
1377  */
1378 static struct nfs4_state *
1379 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1380 {
1381         struct inode *inode = data->state->inode;
1382         struct nfs4_state *state = data->state;
1383         int ret;
1384
1385         if (!data->rpc_done) {
1386                 if (data->rpc_status) {
1387                         ret = data->rpc_status;
1388                         goto err;
1389                 }
1390                 /* cached opens have already been processed */
1391                 goto update;
1392         }
1393
1394         ret = nfs_refresh_inode(inode, &data->f_attr);
1395         if (ret)
1396                 goto err;
1397
1398         if (data->o_res.delegation_type != 0)
1399                 nfs4_opendata_check_deleg(data, state);
1400 update:
1401         update_open_stateid(state, &data->o_res.stateid, NULL,
1402                             data->o_arg.fmode);
1403         atomic_inc(&state->count);
1404
1405         return state;
1406 err:
1407         return ERR_PTR(ret);
1408
1409 }
1410
1411 static struct nfs4_state *
1412 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1413 {
1414         struct inode *inode;
1415         struct nfs4_state *state = NULL;
1416         int ret;
1417
1418         if (!data->rpc_done) {
1419                 state = nfs4_try_open_cached(data);
1420                 goto out;
1421         }
1422
1423         ret = -EAGAIN;
1424         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1425                 goto err;
1426         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1427         ret = PTR_ERR(inode);
1428         if (IS_ERR(inode))
1429                 goto err;
1430         ret = -ENOMEM;
1431         state = nfs4_get_open_state(inode, data->owner);
1432         if (state == NULL)
1433                 goto err_put_inode;
1434         if (data->o_res.delegation_type != 0)
1435                 nfs4_opendata_check_deleg(data, state);
1436         update_open_stateid(state, &data->o_res.stateid, NULL,
1437                         data->o_arg.fmode);
1438         iput(inode);
1439 out:
1440         nfs_release_seqid(data->o_arg.seqid);
1441         return state;
1442 err_put_inode:
1443         iput(inode);
1444 err:
1445         return ERR_PTR(ret);
1446 }
1447
1448 static struct nfs4_state *
1449 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1450 {
1451         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1452                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1453         return _nfs4_opendata_to_nfs4_state(data);
1454 }
1455
1456 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1457 {
1458         struct nfs_inode *nfsi = NFS_I(state->inode);
1459         struct nfs_open_context *ctx;
1460
1461         spin_lock(&state->inode->i_lock);
1462         list_for_each_entry(ctx, &nfsi->open_files, list) {
1463                 if (ctx->state != state)
1464                         continue;
1465                 get_nfs_open_context(ctx);
1466                 spin_unlock(&state->inode->i_lock);
1467                 return ctx;
1468         }
1469         spin_unlock(&state->inode->i_lock);
1470         return ERR_PTR(-ENOENT);
1471 }
1472
1473 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1474                 struct nfs4_state *state, enum open_claim_type4 claim)
1475 {
1476         struct nfs4_opendata *opendata;
1477
1478         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1479                         NULL, NULL, claim, GFP_NOFS);
1480         if (opendata == NULL)
1481                 return ERR_PTR(-ENOMEM);
1482         opendata->state = state;
1483         atomic_inc(&state->count);
1484         return opendata;
1485 }
1486
1487 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1488 {
1489         struct nfs4_state *newstate;
1490         int ret;
1491
1492         opendata->o_arg.open_flags = 0;
1493         opendata->o_arg.fmode = fmode;
1494         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1495         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1496         nfs4_init_opendata_res(opendata);
1497         ret = _nfs4_recover_proc_open(opendata);
1498         if (ret != 0)
1499                 return ret;
1500         newstate = nfs4_opendata_to_nfs4_state(opendata);
1501         if (IS_ERR(newstate))
1502                 return PTR_ERR(newstate);
1503         nfs4_close_state(newstate, fmode);
1504         *res = newstate;
1505         return 0;
1506 }
1507
1508 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1509 {
1510         struct nfs4_state *newstate;
1511         int ret;
1512
1513         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1514         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1515         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1516         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1517         /* memory barrier prior to reading state->n_* */
1518         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1519         clear_bit(NFS_OPEN_STATE, &state->flags);
1520         smp_rmb();
1521         if (state->n_rdwr != 0) {
1522                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1523                 if (ret != 0)
1524                         return ret;
1525                 if (newstate != state)
1526                         return -ESTALE;
1527         }
1528         if (state->n_wronly != 0) {
1529                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1530                 if (ret != 0)
1531                         return ret;
1532                 if (newstate != state)
1533                         return -ESTALE;
1534         }
1535         if (state->n_rdonly != 0) {
1536                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1537                 if (ret != 0)
1538                         return ret;
1539                 if (newstate != state)
1540                         return -ESTALE;
1541         }
1542         /*
1543          * We may have performed cached opens for all three recoveries.
1544          * Check if we need to update the current stateid.
1545          */
1546         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1547             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1548                 write_seqlock(&state->seqlock);
1549                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1550                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1551                 write_sequnlock(&state->seqlock);
1552         }
1553         return 0;
1554 }
1555
1556 /*
1557  * OPEN_RECLAIM:
1558  *      reclaim state on the server after a reboot.
1559  */
1560 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1561 {
1562         struct nfs_delegation *delegation;
1563         struct nfs4_opendata *opendata;
1564         fmode_t delegation_type = 0;
1565         int status;
1566
1567         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1568                         NFS4_OPEN_CLAIM_PREVIOUS);
1569         if (IS_ERR(opendata))
1570                 return PTR_ERR(opendata);
1571         rcu_read_lock();
1572         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1573         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1574                 delegation_type = delegation->type;
1575         rcu_read_unlock();
1576         opendata->o_arg.u.delegation_type = delegation_type;
1577         status = nfs4_open_recover(opendata, state);
1578         nfs4_opendata_put(opendata);
1579         return status;
1580 }
1581
1582 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1583 {
1584         struct nfs_server *server = NFS_SERVER(state->inode);
1585         struct nfs4_exception exception = { };
1586         int err;
1587         do {
1588                 err = _nfs4_do_open_reclaim(ctx, state);
1589                 trace_nfs4_open_reclaim(ctx, 0, err);
1590                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1591                         continue;
1592                 if (err != -NFS4ERR_DELAY)
1593                         break;
1594                 nfs4_handle_exception(server, err, &exception);
1595         } while (exception.retry);
1596         return err;
1597 }
1598
1599 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1600 {
1601         struct nfs_open_context *ctx;
1602         int ret;
1603
1604         ctx = nfs4_state_find_open_context(state);
1605         if (IS_ERR(ctx))
1606                 return -EAGAIN;
1607         ret = nfs4_do_open_reclaim(ctx, state);
1608         put_nfs_open_context(ctx);
1609         return ret;
1610 }
1611
1612 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1613 {
1614         switch (err) {
1615                 default:
1616                         printk(KERN_ERR "NFS: %s: unhandled error "
1617                                         "%d.\n", __func__, err);
1618                 case 0:
1619                 case -ENOENT:
1620                 case -ESTALE:
1621                         break;
1622                 case -NFS4ERR_BADSESSION:
1623                 case -NFS4ERR_BADSLOT:
1624                 case -NFS4ERR_BAD_HIGH_SLOT:
1625                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1626                 case -NFS4ERR_DEADSESSION:
1627                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1628                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1629                         return -EAGAIN;
1630                 case -NFS4ERR_STALE_CLIENTID:
1631                 case -NFS4ERR_STALE_STATEID:
1632                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1633                 case -NFS4ERR_EXPIRED:
1634                         /* Don't recall a delegation if it was lost */
1635                         nfs4_schedule_lease_recovery(server->nfs_client);
1636                         return -EAGAIN;
1637                 case -NFS4ERR_MOVED:
1638                         nfs4_schedule_migration_recovery(server);
1639                         return -EAGAIN;
1640                 case -NFS4ERR_LEASE_MOVED:
1641                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1642                         return -EAGAIN;
1643                 case -NFS4ERR_DELEG_REVOKED:
1644                 case -NFS4ERR_ADMIN_REVOKED:
1645                 case -NFS4ERR_BAD_STATEID:
1646                 case -NFS4ERR_OPENMODE:
1647                         nfs_inode_find_state_and_recover(state->inode,
1648                                         stateid);
1649                         nfs4_schedule_stateid_recovery(server, state);
1650                         return 0;
1651                 case -NFS4ERR_DELAY:
1652                 case -NFS4ERR_GRACE:
1653                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1654                         ssleep(1);
1655                         return -EAGAIN;
1656                 case -ENOMEM:
1657                 case -NFS4ERR_DENIED:
1658                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1659                         return 0;
1660         }
1661         return err;
1662 }
1663
1664 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1665 {
1666         struct nfs_server *server = NFS_SERVER(state->inode);
1667         struct nfs4_opendata *opendata;
1668         int err;
1669
1670         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1671                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1672         if (IS_ERR(opendata))
1673                 return PTR_ERR(opendata);
1674         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1675         err = nfs4_open_recover(opendata, state);
1676         nfs4_opendata_put(opendata);
1677         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1678 }
1679
1680 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1681 {
1682         struct nfs4_opendata *data = calldata;
1683
1684         nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1685                                 &data->c_res.seq_res, task);
1686 }
1687
1688 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1689 {
1690         struct nfs4_opendata *data = calldata;
1691
1692         nfs40_sequence_done(task, &data->c_res.seq_res);
1693
1694         data->rpc_status = task->tk_status;
1695         if (data->rpc_status == 0) {
1696                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1697                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1698                 renew_lease(data->o_res.server, data->timestamp);
1699                 data->rpc_done = 1;
1700         }
1701 }
1702
1703 static void nfs4_open_confirm_release(void *calldata)
1704 {
1705         struct nfs4_opendata *data = calldata;
1706         struct nfs4_state *state = NULL;
1707
1708         /* If this request hasn't been cancelled, do nothing */
1709         if (data->cancelled == 0)
1710                 goto out_free;
1711         /* In case of error, no cleanup! */
1712         if (!data->rpc_done)
1713                 goto out_free;
1714         state = nfs4_opendata_to_nfs4_state(data);
1715         if (!IS_ERR(state))
1716                 nfs4_close_state(state, data->o_arg.fmode);
1717 out_free:
1718         nfs4_opendata_put(data);
1719 }
1720
1721 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1722         .rpc_call_prepare = nfs4_open_confirm_prepare,
1723         .rpc_call_done = nfs4_open_confirm_done,
1724         .rpc_release = nfs4_open_confirm_release,
1725 };
1726
1727 /*
1728  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1729  */
1730 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1731 {
1732         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1733         struct rpc_task *task;
1734         struct  rpc_message msg = {
1735                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1736                 .rpc_argp = &data->c_arg,
1737                 .rpc_resp = &data->c_res,
1738                 .rpc_cred = data->owner->so_cred,
1739         };
1740         struct rpc_task_setup task_setup_data = {
1741                 .rpc_client = server->client,
1742                 .rpc_message = &msg,
1743                 .callback_ops = &nfs4_open_confirm_ops,
1744                 .callback_data = data,
1745                 .workqueue = nfsiod_workqueue,
1746                 .flags = RPC_TASK_ASYNC,
1747         };
1748         int status;
1749
1750         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1751         kref_get(&data->kref);
1752         data->rpc_done = 0;
1753         data->rpc_status = 0;
1754         data->timestamp = jiffies;
1755         task = rpc_run_task(&task_setup_data);
1756         if (IS_ERR(task))
1757                 return PTR_ERR(task);
1758         status = nfs4_wait_for_completion_rpc_task(task);
1759         if (status != 0) {
1760                 data->cancelled = 1;
1761                 smp_wmb();
1762         } else
1763                 status = data->rpc_status;
1764         rpc_put_task(task);
1765         return status;
1766 }
1767
1768 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1769 {
1770         struct nfs4_opendata *data = calldata;
1771         struct nfs4_state_owner *sp = data->owner;
1772         struct nfs_client *clp = sp->so_server->nfs_client;
1773
1774         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1775                 goto out_wait;
1776         /*
1777          * Check if we still need to send an OPEN call, or if we can use
1778          * a delegation instead.
1779          */
1780         if (data->state != NULL) {
1781                 struct nfs_delegation *delegation;
1782
1783                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1784                         goto out_no_action;
1785                 rcu_read_lock();
1786                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1787                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1788                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1789                     can_open_delegated(delegation, data->o_arg.fmode))
1790                         goto unlock_no_action;
1791                 rcu_read_unlock();
1792         }
1793         /* Update client id. */
1794         data->o_arg.clientid = clp->cl_clientid;
1795         switch (data->o_arg.claim) {
1796         case NFS4_OPEN_CLAIM_PREVIOUS:
1797         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1798         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1799                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1800         case NFS4_OPEN_CLAIM_FH:
1801                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1802                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1803         }
1804         data->timestamp = jiffies;
1805         if (nfs4_setup_sequence(data->o_arg.server,
1806                                 &data->o_arg.seq_args,
1807                                 &data->o_res.seq_res,
1808                                 task) != 0)
1809                 nfs_release_seqid(data->o_arg.seqid);
1810
1811         /* Set the create mode (note dependency on the session type) */
1812         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1813         if (data->o_arg.open_flags & O_EXCL) {
1814                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1815                 if (nfs4_has_persistent_session(clp))
1816                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1817                 else if (clp->cl_mvops->minor_version > 0)
1818                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1819         }
1820         return;
1821 unlock_no_action:
1822         rcu_read_unlock();
1823 out_no_action:
1824         task->tk_action = NULL;
1825 out_wait:
1826         nfs4_sequence_done(task, &data->o_res.seq_res);
1827 }
1828
1829 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1830 {
1831         struct nfs4_opendata *data = calldata;
1832
1833         data->rpc_status = task->tk_status;
1834
1835         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1836                 return;
1837
1838         if (task->tk_status == 0) {
1839                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1840                         switch (data->o_res.f_attr->mode & S_IFMT) {
1841                         case S_IFREG:
1842                                 break;
1843                         case S_IFLNK:
1844                                 data->rpc_status = -ELOOP;
1845                                 break;
1846                         case S_IFDIR:
1847                                 data->rpc_status = -EISDIR;
1848                                 break;
1849                         default:
1850                                 data->rpc_status = -ENOTDIR;
1851                         }
1852                 }
1853                 renew_lease(data->o_res.server, data->timestamp);
1854                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1855                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1856         }
1857         data->rpc_done = 1;
1858 }
1859
1860 static void nfs4_open_release(void *calldata)
1861 {
1862         struct nfs4_opendata *data = calldata;
1863         struct nfs4_state *state = NULL;
1864
1865         /* If this request hasn't been cancelled, do nothing */
1866         if (data->cancelled == 0)
1867                 goto out_free;
1868         /* In case of error, no cleanup! */
1869         if (data->rpc_status != 0 || !data->rpc_done)
1870                 goto out_free;
1871         /* In case we need an open_confirm, no cleanup! */
1872         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1873                 goto out_free;
1874         state = nfs4_opendata_to_nfs4_state(data);
1875         if (!IS_ERR(state))
1876                 nfs4_close_state(state, data->o_arg.fmode);
1877 out_free:
1878         nfs4_opendata_put(data);
1879 }
1880
1881 static const struct rpc_call_ops nfs4_open_ops = {
1882         .rpc_call_prepare = nfs4_open_prepare,
1883         .rpc_call_done = nfs4_open_done,
1884         .rpc_release = nfs4_open_release,
1885 };
1886
1887 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1888 {
1889         struct inode *dir = data->dir->d_inode;
1890         struct nfs_server *server = NFS_SERVER(dir);
1891         struct nfs_openargs *o_arg = &data->o_arg;
1892         struct nfs_openres *o_res = &data->o_res;
1893         struct rpc_task *task;
1894         struct rpc_message msg = {
1895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1896                 .rpc_argp = o_arg,
1897                 .rpc_resp = o_res,
1898                 .rpc_cred = data->owner->so_cred,
1899         };
1900         struct rpc_task_setup task_setup_data = {
1901                 .rpc_client = server->client,
1902                 .rpc_message = &msg,
1903                 .callback_ops = &nfs4_open_ops,
1904                 .callback_data = data,
1905                 .workqueue = nfsiod_workqueue,
1906                 .flags = RPC_TASK_ASYNC,
1907         };
1908         int status;
1909
1910         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1911         kref_get(&data->kref);
1912         data->rpc_done = 0;
1913         data->rpc_status = 0;
1914         data->cancelled = 0;
1915         data->is_recover = 0;
1916         if (isrecover) {
1917                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1918                 data->is_recover = 1;
1919         }
1920         task = rpc_run_task(&task_setup_data);
1921         if (IS_ERR(task))
1922                 return PTR_ERR(task);
1923         status = nfs4_wait_for_completion_rpc_task(task);
1924         if (status != 0) {
1925                 data->cancelled = 1;
1926                 smp_wmb();
1927         } else
1928                 status = data->rpc_status;
1929         rpc_put_task(task);
1930
1931         return status;
1932 }
1933
1934 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1935 {
1936         struct inode *dir = data->dir->d_inode;
1937         struct nfs_openres *o_res = &data->o_res;
1938         int status;
1939
1940         status = nfs4_run_open_task(data, 1);
1941         if (status != 0 || !data->rpc_done)
1942                 return status;
1943
1944         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1945
1946         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1947                 status = _nfs4_proc_open_confirm(data);
1948                 if (status != 0)
1949                         return status;
1950         }
1951
1952         return status;
1953 }
1954
1955 /*
1956  * Additional permission checks in order to distinguish between an
1957  * open for read, and an open for execute. This works around the
1958  * fact that NFSv4 OPEN treats read and execute permissions as being
1959  * the same.
1960  * Note that in the non-execute case, we want to turn off permission
1961  * checking if we just created a new file (POSIX open() semantics).
1962  */
1963 static int nfs4_opendata_access(struct rpc_cred *cred,
1964                                 struct nfs4_opendata *opendata,
1965                                 struct nfs4_state *state, fmode_t fmode,
1966                                 int openflags)
1967 {
1968         struct nfs_access_entry cache;
1969         u32 mask;
1970
1971         /* access call failed or for some reason the server doesn't
1972          * support any access modes -- defer access call until later */
1973         if (opendata->o_res.access_supported == 0)
1974                 return 0;
1975
1976         mask = 0;
1977         /*
1978          * Use openflags to check for exec, because fmode won't
1979          * always have FMODE_EXEC set when file open for exec.
1980          */
1981         if (openflags & __FMODE_EXEC) {
1982                 /* ONLY check for exec rights */
1983                 mask = MAY_EXEC;
1984         } else if ((fmode & FMODE_READ) && !opendata->file_created)
1985                 mask = MAY_READ;
1986
1987         cache.cred = cred;
1988         cache.jiffies = jiffies;
1989         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1990         nfs_access_add_cache(state->inode, &cache);
1991
1992         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1993                 return 0;
1994
1995         /* even though OPEN succeeded, access is denied. Close the file */
1996         nfs4_close_state(state, fmode);
1997         return -EACCES;
1998 }
1999
2000 /*
2001  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2002  */
2003 static int _nfs4_proc_open(struct nfs4_opendata *data)
2004 {
2005         struct inode *dir = data->dir->d_inode;
2006         struct nfs_server *server = NFS_SERVER(dir);
2007         struct nfs_openargs *o_arg = &data->o_arg;
2008         struct nfs_openres *o_res = &data->o_res;
2009         int status;
2010
2011         status = nfs4_run_open_task(data, 0);
2012         if (!data->rpc_done)
2013                 return status;
2014         if (status != 0) {
2015                 if (status == -NFS4ERR_BADNAME &&
2016                                 !(o_arg->open_flags & O_CREAT))
2017                         return -ENOENT;
2018                 return status;
2019         }
2020
2021         nfs_fattr_map_and_free_names(server, &data->f_attr);
2022
2023         if (o_arg->open_flags & O_CREAT) {
2024                 update_changeattr(dir, &o_res->cinfo);
2025                 if (o_arg->open_flags & O_EXCL)
2026                         data->file_created = 1;
2027                 else if (o_res->cinfo.before != o_res->cinfo.after)
2028                         data->file_created = 1;
2029         }
2030         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2031                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2032         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2033                 status = _nfs4_proc_open_confirm(data);
2034                 if (status != 0)
2035                         return status;
2036         }
2037         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2038                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2039         return 0;
2040 }
2041
2042 static int nfs4_recover_expired_lease(struct nfs_server *server)
2043 {
2044         return nfs4_client_recover_expired_lease(server->nfs_client);
2045 }
2046
2047 /*
2048  * OPEN_EXPIRED:
2049  *      reclaim state on the server after a network partition.
2050  *      Assumes caller holds the appropriate lock
2051  */
2052 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2053 {
2054         struct nfs4_opendata *opendata;
2055         int ret;
2056
2057         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2058                         NFS4_OPEN_CLAIM_FH);
2059         if (IS_ERR(opendata))
2060                 return PTR_ERR(opendata);
2061         ret = nfs4_open_recover(opendata, state);
2062         if (ret == -ESTALE)
2063                 d_drop(ctx->dentry);
2064         nfs4_opendata_put(opendata);
2065         return ret;
2066 }
2067
2068 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2069 {
2070         struct nfs_server *server = NFS_SERVER(state->inode);
2071         struct nfs4_exception exception = { };
2072         int err;
2073
2074         do {
2075                 err = _nfs4_open_expired(ctx, state);
2076                 trace_nfs4_open_expired(ctx, 0, err);
2077                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2078                         continue;
2079                 switch (err) {
2080                 default:
2081                         goto out;
2082                 case -NFS4ERR_GRACE:
2083                 case -NFS4ERR_DELAY:
2084                         nfs4_handle_exception(server, err, &exception);
2085                         err = 0;
2086                 }
2087         } while (exception.retry);
2088 out:
2089         return err;
2090 }
2091
2092 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2093 {
2094         struct nfs_open_context *ctx;
2095         int ret;
2096
2097         ctx = nfs4_state_find_open_context(state);
2098         if (IS_ERR(ctx))
2099                 return -EAGAIN;
2100         ret = nfs4_do_open_expired(ctx, state);
2101         put_nfs_open_context(ctx);
2102         return ret;
2103 }
2104
2105 #if defined(CONFIG_NFS_V4_1)
2106 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
2107 {
2108         struct nfs_server *server = NFS_SERVER(state->inode);
2109         nfs4_stateid *stateid = &state->stateid;
2110         struct nfs_delegation *delegation;
2111         struct rpc_cred *cred = NULL;
2112         int status = -NFS4ERR_BAD_STATEID;
2113
2114         /* If a state reset has been done, test_stateid is unneeded */
2115         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2116                 return;
2117
2118         /* Get the delegation credential for use by test/free_stateid */
2119         rcu_read_lock();
2120         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2121         if (delegation != NULL &&
2122             nfs4_stateid_match(&delegation->stateid, stateid)) {
2123                 cred = get_rpccred(delegation->cred);
2124                 rcu_read_unlock();
2125                 status = nfs41_test_stateid(server, stateid, cred);
2126                 trace_nfs4_test_delegation_stateid(state, NULL, status);
2127         } else
2128                 rcu_read_unlock();
2129
2130         if (status != NFS_OK) {
2131                 /* Free the stateid unless the server explicitly
2132                  * informs us the stateid is unrecognized. */
2133                 if (status != -NFS4ERR_BAD_STATEID)
2134                         nfs41_free_stateid(server, stateid, cred);
2135                 nfs_remove_bad_delegation(state->inode);
2136
2137                 write_seqlock(&state->seqlock);
2138                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2139                 write_sequnlock(&state->seqlock);
2140                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
2141         }
2142
2143         if (cred != NULL)
2144                 put_rpccred(cred);
2145 }
2146
2147 /**
2148  * nfs41_check_open_stateid - possibly free an open stateid
2149  *
2150  * @state: NFSv4 state for an inode
2151  *
2152  * Returns NFS_OK if recovery for this stateid is now finished.
2153  * Otherwise a negative NFS4ERR value is returned.
2154  */
2155 static int nfs41_check_open_stateid(struct nfs4_state *state)
2156 {
2157         struct nfs_server *server = NFS_SERVER(state->inode);
2158         nfs4_stateid *stateid = &state->open_stateid;
2159         struct rpc_cred *cred = state->owner->so_cred;
2160         int status;
2161
2162         /* If a state reset has been done, test_stateid is unneeded */
2163         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2164             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2165             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2166                 return -NFS4ERR_BAD_STATEID;
2167
2168         status = nfs41_test_stateid(server, stateid, cred);
2169         trace_nfs4_test_open_stateid(state, NULL, status);
2170         if (status != NFS_OK) {
2171                 /* Free the stateid unless the server explicitly
2172                  * informs us the stateid is unrecognized. */
2173                 if (status != -NFS4ERR_BAD_STATEID)
2174                         nfs41_free_stateid(server, stateid, cred);
2175
2176                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2177                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2178                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2179                 clear_bit(NFS_OPEN_STATE, &state->flags);
2180         }
2181         return status;
2182 }
2183
2184 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2185 {
2186         int status;
2187
2188         nfs41_clear_delegation_stateid(state);
2189         status = nfs41_check_open_stateid(state);
2190         if (status != NFS_OK)
2191                 status = nfs4_open_expired(sp, state);
2192         return status;
2193 }
2194 #endif
2195
2196 /*
2197  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2198  * fields corresponding to attributes that were used to store the verifier.
2199  * Make sure we clobber those fields in the later setattr call
2200  */
2201 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2202 {
2203         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2204             !(sattr->ia_valid & ATTR_ATIME_SET))
2205                 sattr->ia_valid |= ATTR_ATIME;
2206
2207         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2208             !(sattr->ia_valid & ATTR_MTIME_SET))
2209                 sattr->ia_valid |= ATTR_MTIME;
2210 }
2211
2212 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2213                 fmode_t fmode,
2214                 int flags,
2215                 struct nfs_open_context *ctx)
2216 {
2217         struct nfs4_state_owner *sp = opendata->owner;
2218         struct nfs_server *server = sp->so_server;
2219         struct dentry *dentry;
2220         struct nfs4_state *state;
2221         unsigned int seq;
2222         int ret;
2223
2224         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2225
2226         ret = _nfs4_proc_open(opendata);
2227         if (ret != 0) {
2228                 if (ret == -ENOENT) {
2229                         d_drop(opendata->dentry);
2230                         d_add(opendata->dentry, NULL);
2231                         nfs_set_verifier(opendata->dentry,
2232                                          nfs_save_change_attribute(opendata->dir->d_inode));
2233                 }
2234                 goto out;
2235         }
2236
2237         state = nfs4_opendata_to_nfs4_state(opendata);
2238         ret = PTR_ERR(state);
2239         if (IS_ERR(state))
2240                 goto out;
2241         if (server->caps & NFS_CAP_POSIX_LOCK)
2242                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2243
2244         dentry = opendata->dentry;
2245         if (dentry->d_inode == NULL) {
2246                 /* FIXME: Is this d_drop() ever needed? */
2247                 d_drop(dentry);
2248                 dentry = d_add_unique(dentry, igrab(state->inode));
2249                 if (dentry == NULL) {
2250                         dentry = opendata->dentry;
2251                 } else if (dentry != ctx->dentry) {
2252                         dput(ctx->dentry);
2253                         ctx->dentry = dget(dentry);
2254                 }
2255                 nfs_set_verifier(dentry,
2256                                 nfs_save_change_attribute(opendata->dir->d_inode));
2257         }
2258
2259         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2260         if (ret != 0)
2261                 goto out;
2262
2263         ctx->state = state;
2264         if (dentry->d_inode == state->inode) {
2265                 nfs_inode_attach_open_context(ctx);
2266                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2267                         nfs4_schedule_stateid_recovery(server, state);
2268         }
2269 out:
2270         return ret;
2271 }
2272
2273 /*
2274  * Returns a referenced nfs4_state
2275  */
2276 static int _nfs4_do_open(struct inode *dir,
2277                         struct nfs_open_context *ctx,
2278                         int flags,
2279                         struct iattr *sattr,
2280                         struct nfs4_label *label,
2281                         int *opened)
2282 {
2283         struct nfs4_state_owner  *sp;
2284         struct nfs4_state     *state = NULL;
2285         struct nfs_server       *server = NFS_SERVER(dir);
2286         struct nfs4_opendata *opendata;
2287         struct dentry *dentry = ctx->dentry;
2288         struct rpc_cred *cred = ctx->cred;
2289         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2290         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2291         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2292         struct nfs4_label *olabel = NULL;
2293         int status;
2294
2295         /* Protect against reboot recovery conflicts */
2296         status = -ENOMEM;
2297         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2298         if (sp == NULL) {
2299                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2300                 goto out_err;
2301         }
2302         status = nfs4_recover_expired_lease(server);
2303         if (status != 0)
2304                 goto err_put_state_owner;
2305         if (dentry->d_inode != NULL)
2306                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2307         status = -ENOMEM;
2308         if (dentry->d_inode)
2309                 claim = NFS4_OPEN_CLAIM_FH;
2310         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2311                         label, claim, GFP_KERNEL);
2312         if (opendata == NULL)
2313                 goto err_put_state_owner;
2314
2315         if (label) {
2316                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2317                 if (IS_ERR(olabel)) {
2318                         status = PTR_ERR(olabel);
2319                         goto err_opendata_put;
2320                 }
2321         }
2322
2323         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2324                 if (!opendata->f_attr.mdsthreshold) {
2325                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2326                         if (!opendata->f_attr.mdsthreshold)
2327                                 goto err_free_label;
2328                 }
2329                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2330         }
2331         if (dentry->d_inode != NULL)
2332                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2333
2334         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2335         if (status != 0)
2336                 goto err_free_label;
2337         state = ctx->state;
2338
2339         if ((opendata->o_arg.open_flags & O_EXCL) &&
2340             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2341                 nfs4_exclusive_attrset(opendata, sattr);
2342
2343                 nfs_fattr_init(opendata->o_res.f_attr);
2344                 status = nfs4_do_setattr(state->inode, cred,
2345                                 opendata->o_res.f_attr, sattr,
2346                                 state, label, olabel);
2347                 if (status == 0) {
2348                         nfs_setattr_update_inode(state->inode, sattr);
2349                         nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2350                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2351                 }
2352         }
2353         if (opendata->file_created)
2354                 *opened |= FILE_CREATED;
2355
2356         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2357                 *ctx_th = opendata->f_attr.mdsthreshold;
2358                 opendata->f_attr.mdsthreshold = NULL;
2359         }
2360
2361         nfs4_label_free(olabel);
2362
2363         nfs4_opendata_put(opendata);
2364         nfs4_put_state_owner(sp);
2365         return 0;
2366 err_free_label:
2367         nfs4_label_free(olabel);
2368 err_opendata_put:
2369         nfs4_opendata_put(opendata);
2370 err_put_state_owner:
2371         nfs4_put_state_owner(sp);
2372 out_err:
2373         return status;
2374 }
2375
2376
2377 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2378                                         struct nfs_open_context *ctx,
2379                                         int flags,
2380                                         struct iattr *sattr,
2381                                         struct nfs4_label *label,
2382                                         int *opened)
2383 {
2384         struct nfs_server *server = NFS_SERVER(dir);
2385         struct nfs4_exception exception = { };
2386         struct nfs4_state *res;
2387         int status;
2388
2389         do {
2390                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2391                 res = ctx->state;
2392                 trace_nfs4_open_file(ctx, flags, status);
2393                 if (status == 0)
2394                         break;
2395                 /* NOTE: BAD_SEQID means the server and client disagree about the
2396                  * book-keeping w.r.t. state-changing operations
2397                  * (OPEN/CLOSE/LOCK/LOCKU...)
2398                  * It is actually a sign of a bug on the client or on the server.
2399                  *
2400                  * If we receive a BAD_SEQID error in the particular case of
2401                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2402                  * have unhashed the old state_owner for us, and that we can
2403                  * therefore safely retry using a new one. We should still warn
2404                  * the user though...
2405                  */
2406                 if (status == -NFS4ERR_BAD_SEQID) {
2407                         pr_warn_ratelimited("NFS: v4 server %s "
2408                                         " returned a bad sequence-id error!\n",
2409                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2410                         exception.retry = 1;
2411                         continue;
2412                 }
2413                 /*
2414                  * BAD_STATEID on OPEN means that the server cancelled our
2415                  * state before it received the OPEN_CONFIRM.
2416                  * Recover by retrying the request as per the discussion
2417                  * on Page 181 of RFC3530.
2418                  */
2419                 if (status == -NFS4ERR_BAD_STATEID) {
2420                         exception.retry = 1;
2421                         continue;
2422                 }
2423                 if (status == -EAGAIN) {
2424                         /* We must have found a delegation */
2425                         exception.retry = 1;
2426                         continue;
2427                 }
2428                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2429                         continue;
2430                 res = ERR_PTR(nfs4_handle_exception(server,
2431                                         status, &exception));
2432         } while (exception.retry);
2433         return res;
2434 }
2435
2436 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2437                             struct nfs_fattr *fattr, struct iattr *sattr,
2438                             struct nfs4_state *state, struct nfs4_label *ilabel,
2439                             struct nfs4_label *olabel)
2440 {
2441         struct nfs_server *server = NFS_SERVER(inode);
2442         struct nfs_setattrargs  arg = {
2443                 .fh             = NFS_FH(inode),
2444                 .iap            = sattr,
2445                 .server         = server,
2446                 .bitmask = server->attr_bitmask,
2447                 .label          = ilabel,
2448         };
2449         struct nfs_setattrres  res = {
2450                 .fattr          = fattr,
2451                 .label          = olabel,
2452                 .server         = server,
2453         };
2454         struct rpc_message msg = {
2455                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2456                 .rpc_argp       = &arg,
2457                 .rpc_resp       = &res,
2458                 .rpc_cred       = cred,
2459         };
2460         unsigned long timestamp = jiffies;
2461         fmode_t fmode;
2462         bool truncate;
2463         int status;
2464
2465         arg.bitmask = nfs4_bitmask(server, ilabel);
2466         if (ilabel)
2467                 arg.bitmask = nfs4_bitmask(server, olabel);
2468
2469         nfs_fattr_init(fattr);
2470
2471         /* Servers should only apply open mode checks for file size changes */
2472         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2473         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2474
2475         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2476                 /* Use that stateid */
2477         } else if (truncate && state != NULL) {
2478                 struct nfs_lockowner lockowner = {
2479                         .l_owner = current->files,
2480                         .l_pid = current->tgid,
2481                 };
2482                 if (!nfs4_valid_open_stateid(state))
2483                         return -EBADF;
2484                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2485                                 &lockowner) == -EIO)
2486                         return -EBADF;
2487         } else
2488                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2489
2490         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2491         if (status == 0 && state != NULL)
2492                 renew_lease(server, timestamp);
2493         return status;
2494 }
2495
2496 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2497                            struct nfs_fattr *fattr, struct iattr *sattr,
2498                            struct nfs4_state *state, struct nfs4_label *ilabel,
2499                            struct nfs4_label *olabel)
2500 {
2501         struct nfs_server *server = NFS_SERVER(inode);
2502         struct nfs4_exception exception = {
2503                 .state = state,
2504                 .inode = inode,
2505         };
2506         int err;
2507         do {
2508                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2509                 trace_nfs4_setattr(inode, err);
2510                 switch (err) {
2511                 case -NFS4ERR_OPENMODE:
2512                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2513                                 pr_warn_once("NFSv4: server %s is incorrectly "
2514                                                 "applying open mode checks to "
2515                                                 "a SETATTR that is not "
2516                                                 "changing file size.\n",
2517                                                 server->nfs_client->cl_hostname);
2518                         }
2519                         if (state && !(state->state & FMODE_WRITE)) {
2520                                 err = -EBADF;
2521                                 if (sattr->ia_valid & ATTR_OPEN)
2522                                         err = -EACCES;
2523                                 goto out;
2524                         }
2525                 }
2526                 err = nfs4_handle_exception(server, err, &exception);
2527         } while (exception.retry);
2528 out:
2529         return err;
2530 }
2531
2532 struct nfs4_closedata {
2533         struct inode *inode;
2534         struct nfs4_state *state;
2535         struct nfs_closeargs arg;
2536         struct nfs_closeres res;
2537         struct nfs_fattr fattr;
2538         unsigned long timestamp;
2539         bool roc;
2540         u32 roc_barrier;
2541 };
2542
2543 static void nfs4_free_closedata(void *data)
2544 {
2545         struct nfs4_closedata *calldata = data;
2546         struct nfs4_state_owner *sp = calldata->state->owner;
2547         struct super_block *sb = calldata->state->inode->i_sb;
2548
2549         if (calldata->roc)
2550                 pnfs_roc_release(calldata->state->inode);
2551         nfs4_put_open_state(calldata->state);
2552         nfs_free_seqid(calldata->arg.seqid);
2553         nfs4_put_state_owner(sp);
2554         nfs_sb_deactive(sb);
2555         kfree(calldata);
2556 }
2557
2558 static void nfs4_close_done(struct rpc_task *task, void *data)
2559 {
2560         struct nfs4_closedata *calldata = data;
2561         struct nfs4_state *state = calldata->state;
2562         struct nfs_server *server = NFS_SERVER(calldata->inode);
2563
2564         dprintk("%s: begin!\n", __func__);
2565         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2566                 return;
2567         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2568         /* hmm. we are done with the inode, and in the process of freeing
2569          * the state_owner. we keep this around to process errors
2570          */
2571         switch (task->tk_status) {
2572                 case 0:
2573                         if (calldata->roc)
2574                                 pnfs_roc_set_barrier(state->inode,
2575                                                      calldata->roc_barrier);
2576                         nfs_clear_open_stateid(state, &calldata->res.stateid, 0);
2577                         renew_lease(server, calldata->timestamp);
2578                         goto out_release;
2579                 case -NFS4ERR_ADMIN_REVOKED:
2580                 case -NFS4ERR_STALE_STATEID:
2581                 case -NFS4ERR_OLD_STATEID:
2582                 case -NFS4ERR_BAD_STATEID:
2583                 case -NFS4ERR_EXPIRED:
2584                         if (calldata->arg.fmode == 0)
2585                                 break;
2586                 default:
2587                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
2588                                 rpc_restart_call_prepare(task);
2589                                 goto out_release;
2590                         }
2591         }
2592         nfs_clear_open_stateid(state, NULL, calldata->arg.fmode);
2593 out_release:
2594         nfs_release_seqid(calldata->arg.seqid);
2595         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2596         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2597 }
2598
2599 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2600 {
2601         struct nfs4_closedata *calldata = data;
2602         struct nfs4_state *state = calldata->state;
2603         struct inode *inode = calldata->inode;
2604         bool is_rdonly, is_wronly, is_rdwr;
2605         int call_close = 0;
2606
2607         dprintk("%s: begin!\n", __func__);
2608         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2609                 goto out_wait;
2610
2611         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2612         spin_lock(&state->owner->so_lock);
2613         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2614         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2615         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2616         /* Calculate the current open share mode */
2617         calldata->arg.fmode = 0;
2618         if (is_rdonly || is_rdwr)
2619                 calldata->arg.fmode |= FMODE_READ;
2620         if (is_wronly || is_rdwr)
2621                 calldata->arg.fmode |= FMODE_WRITE;
2622         /* Calculate the change in open mode */
2623         if (state->n_rdwr == 0) {
2624                 if (state->n_rdonly == 0) {
2625                         call_close |= is_rdonly || is_rdwr;
2626                         calldata->arg.fmode &= ~FMODE_READ;
2627                 }
2628                 if (state->n_wronly == 0) {
2629                         call_close |= is_wronly || is_rdwr;
2630                         calldata->arg.fmode &= ~FMODE_WRITE;
2631                 }
2632         }
2633         if (!nfs4_valid_open_stateid(state))
2634                 call_close = 0;
2635         spin_unlock(&state->owner->so_lock);
2636
2637         if (!call_close) {
2638                 /* Note: exit _without_ calling nfs4_close_done */
2639                 goto out_no_action;
2640         }
2641
2642         if (calldata->arg.fmode == 0) {
2643                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2644                 if (calldata->roc &&
2645                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2646                         nfs_release_seqid(calldata->arg.seqid);
2647                         goto out_wait;
2648                     }
2649         }
2650
2651         nfs_fattr_init(calldata->res.fattr);
2652         calldata->timestamp = jiffies;
2653         if (nfs4_setup_sequence(NFS_SERVER(inode),
2654                                 &calldata->arg.seq_args,
2655                                 &calldata->res.seq_res,
2656                                 task) != 0)
2657                 nfs_release_seqid(calldata->arg.seqid);
2658         dprintk("%s: done!\n", __func__);
2659         return;
2660 out_no_action:
2661         task->tk_action = NULL;
2662 out_wait:
2663         nfs4_sequence_done(task, &calldata->res.seq_res);
2664 }
2665
2666 static const struct rpc_call_ops nfs4_close_ops = {
2667         .rpc_call_prepare = nfs4_close_prepare,
2668         .rpc_call_done = nfs4_close_done,
2669         .rpc_release = nfs4_free_closedata,
2670 };
2671
2672 static bool nfs4_state_has_opener(struct nfs4_state *state)
2673 {
2674         /* first check existing openers */
2675         if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0 &&
2676             state->n_rdonly != 0)
2677                 return true;
2678
2679         if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0 &&
2680             state->n_wronly != 0)
2681                 return true;
2682
2683         if (test_bit(NFS_O_RDWR_STATE, &state->flags) != 0 &&
2684             state->n_rdwr != 0)
2685                 return true;
2686
2687         return false;
2688 }
2689
2690 static bool nfs4_roc(struct inode *inode)
2691 {
2692         struct nfs_inode *nfsi = NFS_I(inode);
2693         struct nfs_open_context *ctx;
2694         struct nfs4_state *state;
2695
2696         spin_lock(&inode->i_lock);
2697         list_for_each_entry(ctx, &nfsi->open_files, list) {
2698                 state = ctx->state;
2699                 if (state == NULL)
2700                         continue;
2701                 if (nfs4_state_has_opener(state)) {
2702                         spin_unlock(&inode->i_lock);
2703                         return false;
2704                 }
2705         }
2706         spin_unlock(&inode->i_lock);
2707
2708         if (nfs4_check_delegation(inode, FMODE_READ))
2709                 return false;
2710
2711         return pnfs_roc(inode);
2712 }
2713
2714 /*
2715  * It is possible for data to be read/written from a mem-mapped file
2716  * after the sys_close call (which hits the vfs layer as a flush).
2717  * This means that we can't safely call nfsv4 close on a file until
2718  * the inode is cleared. This in turn means that we are not good
2719  * NFSv4 citizens - we do not indicate to the server to update the file's
2720  * share state even when we are done with one of the three share
2721  * stateid's in the inode.
2722  *
2723  * NOTE: Caller must be holding the sp->so_owner semaphore!
2724  */
2725 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2726 {
2727         struct nfs_server *server = NFS_SERVER(state->inode);
2728         struct nfs4_closedata *calldata;
2729         struct nfs4_state_owner *sp = state->owner;
2730         struct rpc_task *task;
2731         struct rpc_message msg = {
2732                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2733                 .rpc_cred = state->owner->so_cred,
2734         };
2735         struct rpc_task_setup task_setup_data = {
2736                 .rpc_client = server->client,
2737                 .rpc_message = &msg,
2738                 .callback_ops = &nfs4_close_ops,
2739                 .workqueue = nfsiod_workqueue,
2740                 .flags = RPC_TASK_ASYNC,
2741         };
2742         int status = -ENOMEM;
2743
2744         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2745                 &task_setup_data.rpc_client, &msg);
2746
2747         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2748         if (calldata == NULL)
2749                 goto out;
2750         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2751         calldata->inode = state->inode;
2752         calldata->state = state;
2753         calldata->arg.fh = NFS_FH(state->inode);
2754         calldata->arg.stateid = &state->open_stateid;
2755         /* Serialization for the sequence id */
2756         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2757         if (calldata->arg.seqid == NULL)
2758                 goto out_free_calldata;
2759         calldata->arg.fmode = 0;
2760         calldata->arg.bitmask = server->cache_consistency_bitmask;
2761         calldata->res.fattr = &calldata->fattr;
2762         calldata->res.seqid = calldata->arg.seqid;
2763         calldata->res.server = server;
2764         calldata->roc = nfs4_roc(state->inode);
2765         nfs_sb_active(calldata->inode->i_sb);
2766
2767         msg.rpc_argp = &calldata->arg;
2768         msg.rpc_resp = &calldata->res;
2769         task_setup_data.callback_data = calldata;
2770         task = rpc_run_task(&task_setup_data);
2771         if (IS_ERR(task))
2772                 return PTR_ERR(task);
2773         status = 0;
2774         if (wait)
2775                 status = rpc_wait_for_completion_task(task);
2776         rpc_put_task(task);
2777         return status;
2778 out_free_calldata:
2779         kfree(calldata);
2780 out:
2781         nfs4_put_open_state(state);
2782         nfs4_put_state_owner(sp);
2783         return status;
2784 }
2785
2786 static struct inode *
2787 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2788                 int open_flags, struct iattr *attr, int *opened)
2789 {
2790         struct nfs4_state *state;
2791         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2792
2793         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2794
2795         /* Protect against concurrent sillydeletes */
2796         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2797
2798         nfs4_label_release_security(label);
2799
2800         if (IS_ERR(state))
2801                 return ERR_CAST(state);
2802         return state->inode;
2803 }
2804
2805 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2806 {
2807         if (ctx->state == NULL)
2808                 return;
2809         if (is_sync)
2810                 nfs4_close_sync(ctx->state, ctx->mode);
2811         else
2812                 nfs4_close_state(ctx->state, ctx->mode);
2813 }
2814
2815 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2816 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2817 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2818
2819 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2820 {
2821         struct nfs4_server_caps_arg args = {
2822                 .fhandle = fhandle,
2823         };
2824         struct nfs4_server_caps_res res = {};
2825         struct rpc_message msg = {
2826                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2827                 .rpc_argp = &args,
2828                 .rpc_resp = &res,
2829         };
2830         int status;
2831
2832         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2833         if (status == 0) {
2834                 /* Sanity check the server answers */
2835                 switch (server->nfs_client->cl_minorversion) {
2836                 case 0:
2837                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2838                         res.attr_bitmask[2] = 0;
2839                         break;
2840                 case 1:
2841                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2842                         break;
2843                 case 2:
2844                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2845                 }
2846                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2847                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2848                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2849                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2850                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2851                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2852                                 NFS_CAP_SECURITY_LABEL);
2853                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2854                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2855                         server->caps |= NFS_CAP_ACLS;
2856                 if (res.has_links != 0)
2857                         server->caps |= NFS_CAP_HARDLINKS;
2858                 if (res.has_symlinks != 0)
2859                         server->caps |= NFS_CAP_SYMLINKS;
2860                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2861                         server->caps |= NFS_CAP_FILEID;
2862                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2863                         server->caps |= NFS_CAP_MODE;
2864                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2865                         server->caps |= NFS_CAP_NLINK;
2866                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2867                         server->caps |= NFS_CAP_OWNER;
2868                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2869                         server->caps |= NFS_CAP_OWNER_GROUP;
2870                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2871                         server->caps |= NFS_CAP_ATIME;
2872                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2873                         server->caps |= NFS_CAP_CTIME;
2874                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2875                         server->caps |= NFS_CAP_MTIME;
2876 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2877                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2878                         server->caps |= NFS_CAP_SECURITY_LABEL;
2879 #endif
2880                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2881                                 sizeof(server->attr_bitmask));
2882                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2883
2884                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2885                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2886                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2887                 server->cache_consistency_bitmask[2] = 0;
2888                 server->acl_bitmask = res.acl_bitmask;
2889                 server->fh_expire_type = res.fh_expire_type;
2890         }
2891
2892         return status;
2893 }
2894
2895 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2896 {
2897         struct nfs4_exception exception = { };
2898         int err;
2899         do {
2900                 err = nfs4_handle_exception(server,
2901                                 _nfs4_server_capabilities(server, fhandle),
2902                                 &exception);
2903         } while (exception.retry);
2904         return err;
2905 }
2906
2907 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2908                 struct nfs_fsinfo *info)
2909 {
2910         u32 bitmask[3];
2911         struct nfs4_lookup_root_arg args = {
2912                 .bitmask = bitmask,
2913         };
2914         struct nfs4_lookup_res res = {
2915                 .server = server,
2916                 .fattr = info->fattr,
2917                 .fh = fhandle,
2918         };
2919         struct rpc_message msg = {
2920                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2921                 .rpc_argp = &args,
2922                 .rpc_resp = &res,
2923         };
2924
2925         bitmask[0] = nfs4_fattr_bitmap[0];
2926         bitmask[1] = nfs4_fattr_bitmap[1];
2927         /*
2928          * Process the label in the upcoming getfattr
2929          */
2930         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2931
2932         nfs_fattr_init(info->fattr);
2933         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2934 }
2935
2936 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2937                 struct nfs_fsinfo *info)
2938 {
2939         struct nfs4_exception exception = { };
2940         int err;
2941         do {
2942                 err = _nfs4_lookup_root(server, fhandle, info);
2943                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2944                 switch (err) {
2945                 case 0:
2946                 case -NFS4ERR_WRONGSEC:
2947                         goto out;
2948                 default:
2949                         err = nfs4_handle_exception(server, err, &exception);
2950                 }
2951         } while (exception.retry);
2952 out:
2953         return err;
2954 }
2955
2956 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2957                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2958 {
2959         struct rpc_auth_create_args auth_args = {
2960                 .pseudoflavor = flavor,
2961         };
2962         struct rpc_auth *auth;
2963         int ret;
2964
2965         auth = rpcauth_create(&auth_args, server->client);
2966         if (IS_ERR(auth)) {
2967                 ret = -EACCES;
2968                 goto out;
2969         }
2970         ret = nfs4_lookup_root(server, fhandle, info);
2971 out:
2972         return ret;
2973 }
2974
2975 /*
2976  * Retry pseudoroot lookup with various security flavors.  We do this when:
2977  *
2978  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2979  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2980  *
2981  * Returns zero on success, or a negative NFS4ERR value, or a
2982  * negative errno value.
2983  */
2984 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2985                               struct nfs_fsinfo *info)
2986 {
2987         /* Per 3530bis 15.33.5 */
2988         static const rpc_authflavor_t flav_array[] = {
2989                 RPC_AUTH_GSS_KRB5P,
2990                 RPC_AUTH_GSS_KRB5I,
2991                 RPC_AUTH_GSS_KRB5,
2992                 RPC_AUTH_UNIX,                  /* courtesy */
2993                 RPC_AUTH_NULL,
2994         };
2995         int status = -EPERM;
2996         size_t i;
2997
2998         if (server->auth_info.flavor_len > 0) {
2999                 /* try each flavor specified by user */
3000                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3001                         status = nfs4_lookup_root_sec(server, fhandle, info,
3002                                                 server->auth_info.flavors[i]);
3003                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3004                                 continue;
3005                         break;
3006                 }
3007         } else {
3008                 /* no flavors specified by user, try default list */
3009                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3010                         status = nfs4_lookup_root_sec(server, fhandle, info,
3011                                                       flav_array[i]);
3012                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3013                                 continue;
3014                         break;
3015                 }
3016         }
3017
3018         /*
3019          * -EACCESS could mean that the user doesn't have correct permissions
3020          * to access the mount.  It could also mean that we tried to mount
3021          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3022          * existing mount programs don't handle -EACCES very well so it should
3023          * be mapped to -EPERM instead.
3024          */
3025         if (status == -EACCES)
3026                 status = -EPERM;
3027         return status;
3028 }
3029
3030 static int nfs4_do_find_root_sec(struct nfs_server *server,
3031                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3032 {
3033         int mv = server->nfs_client->cl_minorversion;
3034         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3035 }
3036
3037 /**
3038  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3039  * @server: initialized nfs_server handle
3040  * @fhandle: we fill in the pseudo-fs root file handle
3041  * @info: we fill in an FSINFO struct
3042  * @auth_probe: probe the auth flavours
3043  *
3044  * Returns zero on success, or a negative errno.
3045  */
3046 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3047                          struct nfs_fsinfo *info,
3048                          bool auth_probe)
3049 {
3050         int status;
3051
3052         switch (auth_probe) {
3053         case false:
3054                 status = nfs4_lookup_root(server, fhandle, info);
3055                 if (status != -NFS4ERR_WRONGSEC)
3056                         break;
3057         default:
3058                 status = nfs4_do_find_root_sec(server, fhandle, info);
3059         }
3060
3061         if (status == 0)
3062                 status = nfs4_server_capabilities(server, fhandle);
3063         if (status == 0)
3064                 status = nfs4_do_fsinfo(server, fhandle, info);
3065
3066         return nfs4_map_errors(status);
3067 }
3068
3069 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3070                               struct nfs_fsinfo *info)
3071 {
3072         int error;
3073         struct nfs_fattr *fattr = info->fattr;
3074         struct nfs4_label *label = NULL;
3075
3076         error = nfs4_server_capabilities(server, mntfh);
3077         if (error < 0) {
3078                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3079                 return error;
3080         }
3081
3082         label = nfs4_label_alloc(server, GFP_KERNEL);
3083         if (IS_ERR(label))
3084                 return PTR_ERR(label);
3085
3086         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3087         if (error < 0) {
3088                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3089                 goto err_free_label;
3090         }
3091
3092         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3093             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3094                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3095
3096 err_free_label:
3097         nfs4_label_free(label);
3098
3099         return error;
3100 }
3101
3102 /*
3103  * Get locations and (maybe) other attributes of a referral.
3104  * Note that we'll actually follow the referral later when
3105  * we detect fsid mismatch in inode revalidation
3106  */
3107 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3108                              const struct qstr *name, struct nfs_fattr *fattr,
3109                              struct nfs_fh *fhandle)
3110 {
3111         int status = -ENOMEM;
3112         struct page *page = NULL;
3113         struct nfs4_fs_locations *locations = NULL;
3114
3115         page = alloc_page(GFP_KERNEL);
3116         if (page == NULL)
3117                 goto out;
3118         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3119         if (locations == NULL)
3120                 goto out;
3121
3122         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3123         if (status != 0)
3124                 goto out;
3125
3126         /*
3127          * If the fsid didn't change, this is a migration event, not a
3128          * referral.  Cause us to drop into the exception handler, which
3129          * will kick off migration recovery.
3130          */
3131         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3132                 dprintk("%s: server did not return a different fsid for"
3133                         " a referral at %s\n", __func__, name->name);
3134                 status = -NFS4ERR_MOVED;
3135                 goto out;
3136         }
3137         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3138         nfs_fixup_referral_attributes(&locations->fattr);
3139
3140         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3141         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3142         memset(fhandle, 0, sizeof(struct nfs_fh));
3143 out:
3144         if (page)
3145                 __free_page(page);
3146         kfree(locations);
3147         return status;
3148 }
3149
3150 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3151                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3152 {
3153         struct nfs4_getattr_arg args = {
3154                 .fh = fhandle,
3155                 .bitmask = server->attr_bitmask,
3156         };
3157         struct nfs4_getattr_res res = {
3158                 .fattr = fattr,
3159                 .label = label,
3160                 .server = server,
3161         };
3162         struct rpc_message msg = {
3163                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3164                 .rpc_argp = &args,
3165                 .rpc_resp = &res,
3166         };
3167
3168         args.bitmask = nfs4_bitmask(server, label);
3169
3170         nfs_fattr_init(fattr);
3171         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3172 }
3173
3174 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3175                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3176 {
3177         struct nfs4_exception exception = { };
3178         int err;
3179         do {
3180                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3181                 trace_nfs4_getattr(server, fhandle, fattr, err);
3182                 err = nfs4_handle_exception(server, err,
3183                                 &exception);
3184         } while (exception.retry);
3185         return err;
3186 }
3187
3188 /*
3189  * The file is not closed if it is opened due to the a request to change
3190  * the size of the file. The open call will not be needed once the
3191  * VFS layer lookup-intents are implemented.
3192  *
3193  * Close is called when the inode is destroyed.
3194  * If we haven't opened the file for O_WRONLY, we
3195  * need to in the size_change case to obtain a stateid.
3196  *
3197  * Got race?
3198  * Because OPEN is always done by name in nfsv4, it is
3199  * possible that we opened a different file by the same
3200  * name.  We can recognize this race condition, but we
3201  * can't do anything about it besides returning an error.
3202  *
3203  * This will be fixed with VFS changes (lookup-intent).
3204  */
3205 static int
3206 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3207                   struct iattr *sattr)
3208 {
3209         struct inode *inode = dentry->d_inode;
3210         struct rpc_cred *cred = NULL;
3211         struct nfs4_state *state = NULL;
3212         struct nfs4_label *label = NULL;
3213         int status;
3214
3215         if (pnfs_ld_layoutret_on_setattr(inode))
3216                 pnfs_commit_and_return_layout(inode);
3217
3218         nfs_fattr_init(fattr);
3219
3220         /* Deal with open(O_TRUNC) */
3221         if (sattr->ia_valid & ATTR_OPEN)
3222                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3223
3224         /* Optimization: if the end result is no change, don't RPC */
3225         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3226                 return 0;
3227
3228         /* Search for an existing open(O_WRITE) file */
3229         if (sattr->ia_valid & ATTR_FILE) {
3230                 struct nfs_open_context *ctx;
3231
3232                 ctx = nfs_file_open_context(sattr->ia_file);
3233                 if (ctx) {
3234                         cred = ctx->cred;
3235                         state = ctx->state;
3236                 }
3237         }
3238
3239         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3240         if (IS_ERR(label))
3241                 return PTR_ERR(label);
3242
3243         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3244         if (status == 0) {
3245                 nfs_setattr_update_inode(inode, sattr);
3246                 nfs_setsecurity(inode, fattr, label);
3247         }
3248         nfs4_label_free(label);
3249         return status;
3250 }
3251
3252 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3253                 const struct qstr *name, struct nfs_fh *fhandle,
3254                 struct nfs_fattr *fattr, struct nfs4_label *label)
3255 {
3256         struct nfs_server *server = NFS_SERVER(dir);
3257         int                    status;
3258         struct nfs4_lookup_arg args = {
3259                 .bitmask = server->attr_bitmask,
3260                 .dir_fh = NFS_FH(dir),
3261                 .name = name,
3262         };
3263         struct nfs4_lookup_res res = {
3264                 .server = server,
3265                 .fattr = fattr,
3266                 .label = label,
3267                 .fh = fhandle,
3268         };
3269         struct rpc_message msg = {
3270                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3271                 .rpc_argp = &args,
3272                 .rpc_resp = &res,
3273         };
3274
3275         args.bitmask = nfs4_bitmask(server, label);
3276
3277         nfs_fattr_init(fattr);
3278
3279         dprintk("NFS call  lookup %s\n", name->name);
3280         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3281         dprintk("NFS reply lookup: %d\n", status);
3282         return status;
3283 }
3284
3285 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3286 {
3287         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3288                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3289         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3290         fattr->nlink = 2;
3291 }
3292
3293 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3294                                    struct qstr *name, struct nfs_fh *fhandle,
3295                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3296 {
3297         struct nfs4_exception exception = { };
3298         struct rpc_clnt *client = *clnt;
3299         int err;
3300         do {
3301                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3302                 trace_nfs4_lookup(dir, name, err);
3303                 switch (err) {
3304                 case -NFS4ERR_BADNAME:
3305                         err = -ENOENT;
3306                         goto out;
3307                 case -NFS4ERR_MOVED:
3308                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3309                         goto out;
3310                 case -NFS4ERR_WRONGSEC:
3311                         err = -EPERM;
3312                         if (client != *clnt)
3313                                 goto out;
3314                         client = nfs4_negotiate_security(client, dir, name);
3315                         if (IS_ERR(client))
3316                                 return PTR_ERR(client);
3317
3318                         exception.retry = 1;
3319                         break;
3320                 default:
3321                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3322                 }
3323         } while (exception.retry);
3324
3325 out:
3326         if (err == 0)
3327                 *clnt = client;
3328         else if (client != *clnt)
3329                 rpc_shutdown_client(client);
3330
3331         return err;
3332 }
3333
3334 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3335                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3336                             struct nfs4_label *label)
3337 {
3338         int status;
3339         struct rpc_clnt *client = NFS_CLIENT(dir);
3340
3341         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3342         if (client != NFS_CLIENT(dir)) {
3343                 rpc_shutdown_client(client);
3344                 nfs_fixup_secinfo_attributes(fattr);
3345         }
3346         return status;
3347 }
3348
3349 struct rpc_clnt *
3350 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3351                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3352 {
3353         struct rpc_clnt *client = NFS_CLIENT(dir);
3354         int status;
3355
3356         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3357         if (status < 0)
3358                 return ERR_PTR(status);
3359         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3360 }
3361
3362 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3363 {
3364         struct nfs_server *server = NFS_SERVER(inode);
3365         struct nfs4_accessargs args = {
3366                 .fh = NFS_FH(inode),
3367                 .bitmask = server->cache_consistency_bitmask,
3368         };
3369         struct nfs4_accessres res = {
3370                 .server = server,
3371         };
3372         struct rpc_message msg = {
3373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3374                 .rpc_argp = &args,
3375                 .rpc_resp = &res,
3376                 .rpc_cred = entry->cred,
3377         };
3378         int mode = entry->mask;
3379         int status = 0;
3380
3381         /*
3382          * Determine which access bits we want to ask for...
3383          */
3384         if (mode & MAY_READ)
3385                 args.access |= NFS4_ACCESS_READ;
3386         if (S_ISDIR(inode->i_mode)) {
3387                 if (mode & MAY_WRITE)
3388                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3389                 if (mode & MAY_EXEC)
3390                         args.access |= NFS4_ACCESS_LOOKUP;
3391         } else {
3392                 if (mode & MAY_WRITE)
3393                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3394                 if (mode & MAY_EXEC)
3395                         args.access |= NFS4_ACCESS_EXECUTE;
3396         }
3397
3398         res.fattr = nfs_alloc_fattr();
3399         if (res.fattr == NULL)
3400                 return -ENOMEM;
3401
3402         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3403         if (!status) {
3404                 nfs_access_set_mask(entry, res.access);
3405                 nfs_refresh_inode(inode, res.fattr);
3406         }
3407         nfs_free_fattr(res.fattr);
3408         return status;
3409 }
3410
3411 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3412 {
3413         struct nfs4_exception exception = { };
3414         int err;
3415         do {
3416                 err = _nfs4_proc_access(inode, entry);
3417                 trace_nfs4_access(inode, err);
3418                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3419                                 &exception);
3420         } while (exception.retry);
3421         return err;
3422 }
3423
3424 /*
3425  * TODO: For the time being, we don't try to get any attributes
3426  * along with any of the zero-copy operations READ, READDIR,
3427  * READLINK, WRITE.
3428  *
3429  * In the case of the first three, we want to put the GETATTR
3430  * after the read-type operation -- this is because it is hard
3431  * to predict the length of a GETATTR response in v4, and thus
3432  * align the READ data correctly.  This means that the GETATTR
3433  * may end up partially falling into the page cache, and we should
3434  * shift it into the 'tail' of the xdr_buf before processing.
3435  * To do this efficiently, we need to know the total length
3436  * of data received, which doesn't seem to be available outside
3437  * of the RPC layer.
3438  *
3439  * In the case of WRITE, we also want to put the GETATTR after
3440  * the operation -- in this case because we want to make sure
3441  * we get the post-operation mtime and size.
3442  *
3443  * Both of these changes to the XDR layer would in fact be quite
3444  * minor, but I decided to leave them for a subsequent patch.
3445  */
3446 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3447                 unsigned int pgbase, unsigned int pglen)
3448 {
3449         struct nfs4_readlink args = {
3450                 .fh       = NFS_FH(inode),
3451                 .pgbase   = pgbase,
3452                 .pglen    = pglen,
3453                 .pages    = &page,
3454         };
3455         struct nfs4_readlink_res res;
3456         struct rpc_message msg = {
3457                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3458                 .rpc_argp = &args,
3459                 .rpc_resp = &res,
3460         };
3461
3462         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3463 }
3464
3465 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3466                 unsigned int pgbase, unsigned int pglen)
3467 {
3468         struct nfs4_exception exception = { };
3469         int err;
3470         do {
3471                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3472                 trace_nfs4_readlink(inode, err);
3473                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3474                                 &exception);
3475         } while (exception.retry);
3476         return err;
3477 }
3478
3479 /*
3480  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3481  */
3482 static int
3483 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3484                  int flags)
3485 {
3486         struct nfs4_label l, *ilabel = NULL;
3487         struct nfs_open_context *ctx;
3488         struct nfs4_state *state;
3489         int opened = 0;
3490         int status = 0;
3491
3492         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3493         if (IS_ERR(ctx))
3494                 return PTR_ERR(ctx);
3495
3496         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3497
3498         sattr->ia_mode &= ~current_umask();
3499         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3500         if (IS_ERR(state)) {
3501                 status = PTR_ERR(state);
3502                 goto out;
3503         }
3504 out:
3505         nfs4_label_release_security(ilabel);
3506         put_nfs_open_context(ctx);
3507         return status;
3508 }
3509
3510 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3511 {
3512         struct nfs_server *server = NFS_SERVER(dir);
3513         struct nfs_removeargs args = {
3514                 .fh = NFS_FH(dir),
3515                 .name = *name,
3516         };
3517         struct nfs_removeres res = {
3518                 .server = server,
3519         };
3520         struct rpc_message msg = {
3521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3522                 .rpc_argp = &args,
3523                 .rpc_resp = &res,
3524         };
3525         int status;
3526
3527         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3528         if (status == 0)
3529                 update_changeattr(dir, &res.cinfo);
3530         return status;
3531 }
3532
3533 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3534 {
3535         struct nfs4_exception exception = { };
3536         int err;
3537         do {
3538                 err = _nfs4_proc_remove(dir, name);
3539                 trace_nfs4_remove(dir, name, err);
3540                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3541                                 &exception);
3542         } while (exception.retry);
3543         return err;
3544 }
3545
3546 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3547 {
3548         struct nfs_server *server = NFS_SERVER(dir);
3549         struct nfs_removeargs *args = msg->rpc_argp;
3550         struct nfs_removeres *res = msg->rpc_resp;
3551
3552         res->server = server;
3553         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3554         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3555
3556         nfs_fattr_init(res->dir_attr);
3557 }
3558
3559 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3560 {
3561         nfs4_setup_sequence(NFS_SERVER(data->dir),
3562                         &data->args.seq_args,
3563                         &data->res.seq_res,
3564                         task);
3565 }
3566
3567 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3568 {
3569         struct nfs_unlinkdata *data = task->tk_calldata;
3570         struct nfs_removeres *res = &data->res;
3571
3572         if (!nfs4_sequence_done(task, &res->seq_res))
3573                 return 0;
3574         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3575                 return 0;
3576         update_changeattr(dir, &res->cinfo);
3577         return 1;
3578 }
3579
3580 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3581 {
3582         struct nfs_server *server = NFS_SERVER(dir);
3583         struct nfs_renameargs *arg = msg->rpc_argp;
3584         struct nfs_renameres *res = msg->rpc_resp;
3585
3586         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3587         res->server = server;
3588         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3589 }
3590
3591 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3592 {
3593         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3594                         &data->args.seq_args,
3595                         &data->res.seq_res,
3596                         task);
3597 }
3598
3599 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3600                                  struct inode *new_dir)
3601 {
3602         struct nfs_renamedata *data = task->tk_calldata;
3603         struct nfs_renameres *res = &data->res;
3604
3605         if (!nfs4_sequence_done(task, &res->seq_res))
3606                 return 0;
3607         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3608                 return 0;
3609
3610         update_changeattr(old_dir, &res->old_cinfo);
3611         update_changeattr(new_dir, &res->new_cinfo);
3612         return 1;
3613 }
3614
3615 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3616 {
3617         struct nfs_server *server = NFS_SERVER(inode);
3618         struct nfs4_link_arg arg = {
3619                 .fh     = NFS_FH(inode),
3620                 .dir_fh = NFS_FH(dir),
3621                 .name   = name,
3622                 .bitmask = server->attr_bitmask,
3623         };
3624         struct nfs4_link_res res = {
3625                 .server = server,
3626                 .label = NULL,
3627         };
3628         struct rpc_message msg = {
3629                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3630                 .rpc_argp = &arg,
3631                 .rpc_resp = &res,
3632         };
3633         int status = -ENOMEM;
3634
3635         res.fattr = nfs_alloc_fattr();
3636         if (res.fattr == NULL)
3637                 goto out;
3638
3639         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3640         if (IS_ERR(res.label)) {
3641                 status = PTR_ERR(res.label);
3642                 goto out;
3643         }
3644         arg.bitmask = nfs4_bitmask(server, res.label);
3645
3646         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3647         if (!status) {
3648                 update_changeattr(dir, &res.cinfo);
3649                 status = nfs_post_op_update_inode(inode, res.fattr);
3650                 if (!status)
3651                         nfs_setsecurity(inode, res.fattr, res.label);
3652         }
3653
3654
3655         nfs4_label_free(res.label);
3656
3657 out:
3658         nfs_free_fattr(res.fattr);
3659         return status;
3660 }
3661
3662 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3663 {
3664         struct nfs4_exception exception = { };
3665         int err;
3666         do {
3667                 err = nfs4_handle_exception(NFS_SERVER(inode),
3668                                 _nfs4_proc_link(inode, dir, name),
3669                                 &exception);
3670         } while (exception.retry);
3671         return err;
3672 }
3673
3674 struct nfs4_createdata {
3675         struct rpc_message msg;
3676         struct nfs4_create_arg arg;
3677         struct nfs4_create_res res;
3678         struct nfs_fh fh;
3679         struct nfs_fattr fattr;
3680         struct nfs4_label *label;
3681 };
3682
3683 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3684                 struct qstr *name, struct iattr *sattr, u32 ftype)
3685 {
3686         struct nfs4_createdata *data;
3687
3688         data = kzalloc(sizeof(*data), GFP_KERNEL);
3689         if (data != NULL) {
3690                 struct nfs_server *server = NFS_SERVER(dir);
3691
3692                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3693                 if (IS_ERR(data->label))
3694                         goto out_free;
3695
3696                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3697                 data->msg.rpc_argp = &data->arg;
3698                 data->msg.rpc_resp = &data->res;
3699                 data->arg.dir_fh = NFS_FH(dir);
3700                 data->arg.server = server;
3701                 data->arg.name = name;
3702                 data->arg.attrs = sattr;
3703                 data->arg.ftype = ftype;
3704                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3705                 data->res.server = server;
3706                 data->res.fh = &data->fh;
3707                 data->res.fattr = &data->fattr;
3708                 data->res.label = data->label;
3709                 nfs_fattr_init(data->res.fattr);
3710         }
3711         return data;
3712 out_free:
3713         kfree(data);
3714         return NULL;
3715 }
3716
3717 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3718 {
3719         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3720                                     &data->arg.seq_args, &data->res.seq_res, 1);
3721         if (status == 0) {
3722                 update_changeattr(dir, &data->res.dir_cinfo);
3723                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3724         }
3725         return status;
3726 }
3727
3728 static void nfs4_free_createdata(struct nfs4_createdata *data)
3729 {
3730         nfs4_label_free(data->label);
3731         kfree(data);
3732 }
3733
3734 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3735                 struct page *page, unsigned int len, struct iattr *sattr,
3736                 struct nfs4_label *label)
3737 {
3738         struct nfs4_createdata *data;
3739         int status = -ENAMETOOLONG;
3740
3741         if (len > NFS4_MAXPATHLEN)
3742                 goto out;
3743
3744         status = -ENOMEM;
3745         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3746         if (data == NULL)
3747                 goto out;
3748
3749         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3750         data->arg.u.symlink.pages = &page;
3751         data->arg.u.symlink.len = len;
3752         data->arg.label = label;
3753
3754         status = nfs4_do_create(dir, dentry, data);
3755
3756         nfs4_free_createdata(data);
3757 out:
3758         return status;
3759 }
3760
3761 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3762                 struct page *page, unsigned int len, struct iattr *sattr)
3763 {
3764         struct nfs4_exception exception = { };
3765         struct nfs4_label l, *label = NULL;
3766         int err;
3767
3768         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3769
3770         do {
3771                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3772                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3773                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3774                                 &exception);
3775         } while (exception.retry);
3776
3777         nfs4_label_release_security(label);
3778         return err;
3779 }
3780
3781 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3782                 struct iattr *sattr, struct nfs4_label *label)
3783 {
3784         struct nfs4_createdata *data;
3785         int status = -ENOMEM;
3786
3787         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3788         if (data == NULL)
3789                 goto out;
3790
3791         data->arg.label = label;
3792         status = nfs4_do_create(dir, dentry, data);
3793
3794         nfs4_free_createdata(data);
3795 out:
3796         return status;
3797 }
3798
3799 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3800                 struct iattr *sattr)
3801 {
3802         struct nfs4_exception exception = { };
3803         struct nfs4_label l, *label = NULL;
3804         int err;
3805
3806         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3807
3808         sattr->ia_mode &= ~current_umask();
3809         do {
3810                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3811                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3812                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3813                                 &exception);
3814         } while (exception.retry);
3815         nfs4_label_release_security(label);
3816
3817         return err;
3818 }
3819
3820 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3821                 u64 cookie, struct page **pages, unsigned int count, int plus)
3822 {
3823         struct inode            *dir = dentry->d_inode;
3824         struct nfs4_readdir_arg args = {
3825                 .fh = NFS_FH(dir),
3826                 .pages = pages,
3827                 .pgbase = 0,
3828                 .count = count,
3829                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3830                 .plus = plus,
3831         };
3832         struct nfs4_readdir_res res;
3833         struct rpc_message msg = {
3834                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3835                 .rpc_argp = &args,
3836                 .rpc_resp = &res,
3837                 .rpc_cred = cred,
3838         };
3839         int                     status;
3840
3841         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3842                         dentry,
3843                         (unsigned long long)cookie);
3844         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3845         res.pgbase = args.pgbase;
3846         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3847         if (status >= 0) {
3848                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3849                 status += args.pgbase;
3850         }
3851
3852         nfs_invalidate_atime(dir);
3853
3854         dprintk("%s: returns %d\n", __func__, status);
3855         return status;
3856 }
3857
3858 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3859                 u64 cookie, struct page **pages, unsigned int count, int plus)
3860 {
3861         struct nfs4_exception exception = { };
3862         int err;
3863         do {
3864                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3865                                 pages, count, plus);
3866                 trace_nfs4_readdir(dentry->d_inode, err);
3867                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3868                                 &exception);
3869         } while (exception.retry);
3870         return err;
3871 }
3872
3873 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3874                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3875 {
3876         struct nfs4_createdata *data;
3877         int mode = sattr->ia_mode;
3878         int status = -ENOMEM;
3879
3880         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3881         if (data == NULL)
3882                 goto out;
3883
3884         if (S_ISFIFO(mode))
3885                 data->arg.ftype = NF4FIFO;
3886         else if (S_ISBLK(mode)) {
3887                 data->arg.ftype = NF4BLK;
3888                 data->arg.u.device.specdata1 = MAJOR(rdev);
3889                 data->arg.u.device.specdata2 = MINOR(rdev);
3890         }
3891         else if (S_ISCHR(mode)) {
3892                 data->arg.ftype = NF4CHR;
3893                 data->arg.u.device.specdata1 = MAJOR(rdev);
3894                 data->arg.u.device.specdata2 = MINOR(rdev);
3895         } else if (!S_ISSOCK(mode)) {
3896                 status = -EINVAL;
3897                 goto out_free;
3898         }
3899
3900         data->arg.label = label;
3901         status = nfs4_do_create(dir, dentry, data);
3902 out_free:
3903         nfs4_free_createdata(data);
3904 out:
3905         return status;
3906 }
3907
3908 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3909                 struct iattr *sattr, dev_t rdev)
3910 {
3911         struct nfs4_exception exception = { };
3912         struct nfs4_label l, *label = NULL;
3913         int err;
3914
3915         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3916
3917         sattr->ia_mode &= ~current_umask();
3918         do {
3919                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3920                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3921                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3922                                 &exception);
3923         } while (exception.retry);
3924
3925         nfs4_label_release_security(label);
3926
3927         return err;
3928 }
3929
3930 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3931                  struct nfs_fsstat *fsstat)
3932 {
3933         struct nfs4_statfs_arg args = {
3934                 .fh = fhandle,
3935                 .bitmask = server->attr_bitmask,
3936         };
3937         struct nfs4_statfs_res res = {
3938                 .fsstat = fsstat,
3939         };
3940         struct rpc_message msg = {
3941                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3942                 .rpc_argp = &args,
3943                 .rpc_resp = &res,
3944         };
3945
3946         nfs_fattr_init(fsstat->fattr);
3947         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3948 }
3949
3950 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3951 {
3952         struct nfs4_exception exception = { };
3953         int err;
3954         do {
3955                 err = nfs4_handle_exception(server,
3956                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3957                                 &exception);
3958         } while (exception.retry);
3959         return err;
3960 }
3961
3962 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3963                 struct nfs_fsinfo *fsinfo)
3964 {
3965         struct nfs4_fsinfo_arg args = {
3966                 .fh = fhandle,
3967                 .bitmask = server->attr_bitmask,
3968         };
3969         struct nfs4_fsinfo_res res = {
3970                 .fsinfo = fsinfo,
3971         };
3972         struct rpc_message msg = {
3973                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3974                 .rpc_argp = &args,
3975                 .rpc_resp = &res,
3976         };
3977
3978         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3979 }
3980
3981 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3982 {
3983         struct nfs4_exception exception = { };
3984         unsigned long now = jiffies;
3985         int err;
3986
3987         do {
3988                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3989                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
3990                 if (err == 0) {
3991                         struct nfs_client *clp = server->nfs_client;
3992
3993                         spin_lock(&clp->cl_lock);
3994                         clp->cl_lease_time = fsinfo->lease_time * HZ;
3995                         clp->cl_last_renewal = now;
3996                         spin_unlock(&clp->cl_lock);
3997                         break;
3998                 }
3999                 err = nfs4_handle_exception(server, err, &exception);
4000         } while (exception.retry);
4001         return err;
4002 }
4003
4004 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4005 {
4006         int error;
4007
4008         nfs_fattr_init(fsinfo->fattr);
4009         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4010         if (error == 0) {
4011                 /* block layout checks this! */
4012                 server->pnfs_blksize = fsinfo->blksize;
4013                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4014         }
4015
4016         return error;
4017 }
4018
4019 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4020                 struct nfs_pathconf *pathconf)
4021 {
4022         struct nfs4_pathconf_arg args = {
4023                 .fh = fhandle,
4024                 .bitmask = server->attr_bitmask,
4025         };
4026         struct nfs4_pathconf_res res = {
4027                 .pathconf = pathconf,
4028         };
4029         struct rpc_message msg = {
4030                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4031                 .rpc_argp = &args,
4032                 .rpc_resp = &res,
4033         };
4034
4035         /* None of the pathconf attributes are mandatory to implement */
4036         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4037                 memset(pathconf, 0, sizeof(*pathconf));
4038                 return 0;
4039         }
4040
4041         nfs_fattr_init(pathconf->fattr);
4042         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4043 }
4044
4045 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4046                 struct nfs_pathconf *pathconf)
4047 {
4048         struct nfs4_exception exception = { };
4049         int err;
4050
4051         do {
4052                 err = nfs4_handle_exception(server,
4053                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4054                                 &exception);
4055         } while (exception.retry);
4056         return err;
4057 }
4058
4059 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4060                 const struct nfs_open_context *ctx,
4061                 const struct nfs_lock_context *l_ctx,
4062                 fmode_t fmode)
4063 {
4064         const struct nfs_lockowner *lockowner = NULL;
4065
4066         if (l_ctx != NULL)
4067                 lockowner = &l_ctx->lockowner;
4068         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4069 }
4070 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4071
4072 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4073                 const struct nfs_open_context *ctx,
4074                 const struct nfs_lock_context *l_ctx,
4075                 fmode_t fmode)
4076 {
4077         nfs4_stateid current_stateid;
4078
4079         /* If the current stateid represents a lost lock, then exit */
4080         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4081                 return true;
4082         return nfs4_stateid_match(stateid, &current_stateid);
4083 }
4084
4085 static bool nfs4_error_stateid_expired(int err)
4086 {
4087         switch (err) {
4088         case -NFS4ERR_DELEG_REVOKED:
4089         case -NFS4ERR_ADMIN_REVOKED:
4090         case -NFS4ERR_BAD_STATEID:
4091         case -NFS4ERR_STALE_STATEID:
4092         case -NFS4ERR_OLD_STATEID:
4093         case -NFS4ERR_OPENMODE:
4094         case -NFS4ERR_EXPIRED:
4095                 return true;
4096         }
4097         return false;
4098 }
4099
4100 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4101 {
4102         nfs_invalidate_atime(hdr->inode);
4103 }
4104
4105 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4106 {
4107         struct nfs_server *server = NFS_SERVER(hdr->inode);
4108
4109         trace_nfs4_read(hdr, task->tk_status);
4110         if (nfs4_async_handle_error(task, server,
4111                                     hdr->args.context->state) == -EAGAIN) {
4112                 rpc_restart_call_prepare(task);
4113                 return -EAGAIN;
4114         }
4115
4116         __nfs4_read_done_cb(hdr);
4117         if (task->tk_status > 0)
4118                 renew_lease(server, hdr->timestamp);
4119         return 0;
4120 }
4121
4122 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4123                 struct nfs_pgio_args *args)
4124 {
4125
4126         if (!nfs4_error_stateid_expired(task->tk_status) ||
4127                 nfs4_stateid_is_current(&args->stateid,
4128                                 args->context,
4129                                 args->lock_context,
4130                                 FMODE_READ))
4131                 return false;
4132         rpc_restart_call_prepare(task);
4133         return true;
4134 }
4135
4136 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4137 {
4138
4139         dprintk("--> %s\n", __func__);
4140
4141         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4142                 return -EAGAIN;
4143         if (nfs4_read_stateid_changed(task, &hdr->args))
4144                 return -EAGAIN;
4145         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4146                                     nfs4_read_done_cb(task, hdr);
4147 }
4148
4149 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4150                                  struct rpc_message *msg)
4151 {
4152         hdr->timestamp   = jiffies;
4153         hdr->pgio_done_cb = nfs4_read_done_cb;
4154         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4155         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4156 }
4157
4158 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4159                                       struct nfs_pgio_header *hdr)
4160 {
4161         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4162                         &hdr->args.seq_args,
4163                         &hdr->res.seq_res,
4164                         task))
4165                 return 0;
4166         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4167                                 hdr->args.lock_context,
4168                                 hdr->rw_ops->rw_mode) == -EIO)
4169                 return -EIO;
4170         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4171                 return -EIO;
4172         return 0;
4173 }
4174
4175 static int nfs4_write_done_cb(struct rpc_task *task,
4176                               struct nfs_pgio_header *hdr)
4177 {
4178         struct inode *inode = hdr->inode;
4179
4180         trace_nfs4_write(hdr, task->tk_status);
4181         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4182                                     hdr->args.context->state) == -EAGAIN) {
4183                 rpc_restart_call_prepare(task);
4184                 return -EAGAIN;
4185         }
4186         if (task->tk_status >= 0) {
4187                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4188                 nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4189         }
4190         return 0;
4191 }
4192
4193 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4194                 struct nfs_pgio_args *args)
4195 {
4196
4197         if (!nfs4_error_stateid_expired(task->tk_status) ||
4198                 nfs4_stateid_is_current(&args->stateid,
4199                                 args->context,
4200                                 args->lock_context,
4201                                 FMODE_WRITE))
4202                 return false;
4203         rpc_restart_call_prepare(task);
4204         return true;
4205 }
4206
4207 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4208 {
4209         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4210                 return -EAGAIN;
4211         if (nfs4_write_stateid_changed(task, &hdr->args))
4212                 return -EAGAIN;
4213         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4214                 nfs4_write_done_cb(task, hdr);
4215 }
4216
4217 static
4218 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4219 {
4220         /* Don't request attributes for pNFS or O_DIRECT writes */
4221         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4222                 return false;
4223         /* Otherwise, request attributes if and only if we don't hold
4224          * a delegation
4225          */
4226         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4227 }
4228
4229 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4230                                   struct rpc_message *msg)
4231 {
4232         struct nfs_server *server = NFS_SERVER(hdr->inode);
4233
4234         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4235                 hdr->args.bitmask = NULL;
4236                 hdr->res.fattr = NULL;
4237         } else
4238                 hdr->args.bitmask = server->cache_consistency_bitmask;
4239
4240         if (!hdr->pgio_done_cb)
4241                 hdr->pgio_done_cb = nfs4_write_done_cb;
4242         hdr->res.server = server;
4243         hdr->timestamp   = jiffies;
4244
4245         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4246         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4247 }
4248
4249 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4250 {
4251         nfs4_setup_sequence(NFS_SERVER(data->inode),
4252                         &data->args.seq_args,
4253                         &data->res.seq_res,
4254                         task);
4255 }
4256
4257 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4258 {
4259         struct inode *inode = data->inode;
4260
4261         trace_nfs4_commit(data, task->tk_status);
4262         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4263                 rpc_restart_call_prepare(task);
4264                 return -EAGAIN;
4265         }
4266         return 0;
4267 }
4268
4269 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4270 {
4271         if (!nfs4_sequence_done(task, &data->res.seq_res))
4272                 return -EAGAIN;
4273         return data->commit_done_cb(task, data);
4274 }
4275
4276 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4277 {
4278         struct nfs_server *server = NFS_SERVER(data->inode);
4279
4280         if (data->commit_done_cb == NULL)
4281                 data->commit_done_cb = nfs4_commit_done_cb;
4282         data->res.server = server;
4283         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4284         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4285 }
4286
4287 struct nfs4_renewdata {
4288         struct nfs_client       *client;
4289         unsigned long           timestamp;
4290 };
4291
4292 /*
4293  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4294  * standalone procedure for queueing an asynchronous RENEW.
4295  */
4296 static void nfs4_renew_release(void *calldata)
4297 {
4298         struct nfs4_renewdata *data = calldata;
4299         struct nfs_client *clp = data->client;
4300
4301         if (atomic_read(&clp->cl_count) > 1)
4302                 nfs4_schedule_state_renewal(clp);
4303         nfs_put_client(clp);
4304         kfree(data);
4305 }
4306
4307 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4308 {
4309         struct nfs4_renewdata *data = calldata;
4310         struct nfs_client *clp = data->client;
4311         unsigned long timestamp = data->timestamp;
4312
4313         trace_nfs4_renew_async(clp, task->tk_status);
4314         switch (task->tk_status) {
4315         case 0:
4316                 break;
4317         case -NFS4ERR_LEASE_MOVED:
4318                 nfs4_schedule_lease_moved_recovery(clp);
4319                 break;
4320         default:
4321                 /* Unless we're shutting down, schedule state recovery! */
4322                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4323                         return;
4324                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4325                         nfs4_schedule_lease_recovery(clp);
4326                         return;
4327                 }
4328                 nfs4_schedule_path_down_recovery(clp);
4329         }
4330         do_renew_lease(clp, timestamp);
4331 }
4332
4333 static const struct rpc_call_ops nfs4_renew_ops = {
4334         .rpc_call_done = nfs4_renew_done,
4335         .rpc_release = nfs4_renew_release,
4336 };
4337
4338 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4339 {
4340         struct rpc_message msg = {
4341                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4342                 .rpc_argp       = clp,
4343                 .rpc_cred       = cred,
4344         };
4345         struct nfs4_renewdata *data;
4346
4347         if (renew_flags == 0)
4348                 return 0;
4349         if (!atomic_inc_not_zero(&clp->cl_count))
4350                 return -EIO;
4351         data = kmalloc(sizeof(*data), GFP_NOFS);
4352         if (data == NULL)
4353                 return -ENOMEM;
4354         data->client = clp;
4355         data->timestamp = jiffies;
4356         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4357                         &nfs4_renew_ops, data);
4358 }
4359
4360 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4361 {
4362         struct rpc_message msg = {
4363                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4364                 .rpc_argp       = clp,
4365                 .rpc_cred       = cred,
4366         };
4367         unsigned long now = jiffies;
4368         int status;
4369
4370         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4371         if (status < 0)
4372                 return status;
4373         do_renew_lease(clp, now);
4374         return 0;
4375 }
4376
4377 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4378 {
4379         return server->caps & NFS_CAP_ACLS;
4380 }
4381
4382 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4383  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4384  * the stack.
4385  */
4386 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4387
4388 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4389                 struct page **pages, unsigned int *pgbase)
4390 {
4391         struct page *newpage, **spages;
4392         int rc = 0;
4393         size_t len;
4394         spages = pages;
4395
4396         do {
4397                 len = min_t(size_t, PAGE_SIZE, buflen);
4398                 newpage = alloc_page(GFP_KERNEL);
4399
4400                 if (newpage == NULL)
4401                         goto unwind;
4402                 memcpy(page_address(newpage), buf, len);
4403                 buf += len;
4404                 buflen -= len;
4405                 *pages++ = newpage;
4406                 rc++;
4407         } while (buflen != 0);
4408
4409         return rc;
4410
4411 unwind:
4412         for(; rc > 0; rc--)
4413                 __free_page(spages[rc-1]);
4414         return -ENOMEM;
4415 }
4416
4417 struct nfs4_cached_acl {
4418         int cached;
4419         size_t len;
4420         char data[0];
4421 };
4422
4423 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4424 {
4425         struct nfs_inode *nfsi = NFS_I(inode);
4426
4427         spin_lock(&inode->i_lock);
4428         kfree(nfsi->nfs4_acl);
4429         nfsi->nfs4_acl = acl;
4430         spin_unlock(&inode->i_lock);
4431 }
4432
4433 static void nfs4_zap_acl_attr(struct inode *inode)
4434 {
4435         nfs4_set_cached_acl(inode, NULL);
4436 }
4437
4438 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4439 {
4440         struct nfs_inode *nfsi = NFS_I(inode);
4441         struct nfs4_cached_acl *acl;
4442         int ret = -ENOENT;
4443
4444         spin_lock(&inode->i_lock);
4445         acl = nfsi->nfs4_acl;
4446         if (acl == NULL)
4447                 goto out;
4448         if (buf == NULL) /* user is just asking for length */
4449                 goto out_len;
4450         if (acl->cached == 0)
4451                 goto out;
4452         ret = -ERANGE; /* see getxattr(2) man page */
4453         if (acl->len > buflen)
4454                 goto out;
4455         memcpy(buf, acl->data, acl->len);
4456 out_len:
4457         ret = acl->len;
4458 out:
4459         spin_unlock(&inode->i_lock);
4460         return ret;
4461 }
4462
4463 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4464 {
4465         struct nfs4_cached_acl *acl;
4466         size_t buflen = sizeof(*acl) + acl_len;
4467
4468         if (buflen <= PAGE_SIZE) {
4469                 acl = kmalloc(buflen, GFP_KERNEL);
4470                 if (acl == NULL)
4471                         goto out;
4472                 acl->cached = 1;
4473                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4474         } else {
4475                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4476                 if (acl == NULL)
4477                         goto out;
4478                 acl->cached = 0;
4479         }
4480         acl->len = acl_len;
4481 out:
4482         nfs4_set_cached_acl(inode, acl);
4483 }
4484
4485 /*
4486  * The getxattr API returns the required buffer length when called with a
4487  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4488  * the required buf.  On a NULL buf, we send a page of data to the server
4489  * guessing that the ACL request can be serviced by a page. If so, we cache
4490  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4491  * the cache. If not so, we throw away the page, and cache the required
4492  * length. The next getxattr call will then produce another round trip to
4493  * the server, this time with the input buf of the required size.
4494  */
4495 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4496 {
4497         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4498         struct nfs_getaclargs args = {
4499                 .fh = NFS_FH(inode),
4500                 .acl_pages = pages,
4501                 .acl_len = buflen,
4502         };
4503         struct nfs_getaclres res = {
4504                 .acl_len = buflen,
4505         };
4506         struct rpc_message msg = {
4507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4508                 .rpc_argp = &args,
4509                 .rpc_resp = &res,
4510         };
4511         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4512         int ret = -ENOMEM, i;
4513
4514         /* As long as we're doing a round trip to the server anyway,
4515          * let's be prepared for a page of acl data. */
4516         if (npages == 0)
4517                 npages = 1;
4518         if (npages > ARRAY_SIZE(pages))
4519                 return -ERANGE;
4520
4521         for (i = 0; i < npages; i++) {
4522                 pages[i] = alloc_page(GFP_KERNEL);
4523                 if (!pages[i])
4524                         goto out_free;
4525         }
4526
4527         /* for decoding across pages */
4528         res.acl_scratch = alloc_page(GFP_KERNEL);
4529         if (!res.acl_scratch)
4530                 goto out_free;
4531
4532         args.acl_len = npages * PAGE_SIZE;
4533         args.acl_pgbase = 0;
4534
4535         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4536                 __func__, buf, buflen, npages, args.acl_len);
4537         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4538                              &msg, &args.seq_args, &res.seq_res, 0);
4539         if (ret)
4540                 goto out_free;
4541
4542         /* Handle the case where the passed-in buffer is too short */
4543         if (res.acl_flags & NFS4_ACL_TRUNC) {
4544                 /* Did the user only issue a request for the acl length? */
4545                 if (buf == NULL)
4546                         goto out_ok;
4547                 ret = -ERANGE;
4548                 goto out_free;
4549         }
4550         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4551         if (buf) {
4552                 if (res.acl_len > buflen) {
4553                         ret = -ERANGE;
4554                         goto out_free;
4555                 }
4556                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4557         }
4558 out_ok:
4559         ret = res.acl_len;
4560 out_free:
4561         for (i = 0; i < npages; i++)
4562                 if (pages[i])
4563                         __free_page(pages[i]);
4564         if (res.acl_scratch)
4565                 __free_page(res.acl_scratch);
4566         return ret;
4567 }
4568
4569 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4570 {
4571         struct nfs4_exception exception = { };
4572         ssize_t ret;
4573         do {
4574                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4575                 trace_nfs4_get_acl(inode, ret);
4576                 if (ret >= 0)
4577                         break;
4578                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4579         } while (exception.retry);
4580         return ret;
4581 }
4582
4583 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4584 {
4585         struct nfs_server *server = NFS_SERVER(inode);
4586         int ret;
4587
4588         if (!nfs4_server_supports_acls(server))
4589                 return -EOPNOTSUPP;
4590         ret = nfs_revalidate_inode(server, inode);
4591         if (ret < 0)
4592                 return ret;
4593         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4594                 nfs_zap_acl_cache(inode);
4595         ret = nfs4_read_cached_acl(inode, buf, buflen);
4596         if (ret != -ENOENT)
4597                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4598                  * but no cached acl data, just the acl length */
4599                 return ret;
4600         return nfs4_get_acl_uncached(inode, buf, buflen);
4601 }
4602
4603 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4604 {
4605         struct nfs_server *server = NFS_SERVER(inode);
4606         struct page *pages[NFS4ACL_MAXPAGES];
4607         struct nfs_setaclargs arg = {
4608                 .fh             = NFS_FH(inode),
4609                 .acl_pages      = pages,
4610                 .acl_len        = buflen,
4611         };
4612         struct nfs_setaclres res;
4613         struct rpc_message msg = {
4614                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4615                 .rpc_argp       = &arg,
4616                 .rpc_resp       = &res,
4617         };
4618         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4619         int ret, i;
4620
4621         if (!nfs4_server_supports_acls(server))
4622                 return -EOPNOTSUPP;
4623         if (npages > ARRAY_SIZE(pages))
4624                 return -ERANGE;
4625         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4626         if (i < 0)
4627                 return i;
4628         nfs4_inode_return_delegation(inode);
4629         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4630
4631         /*
4632          * Free each page after tx, so the only ref left is
4633          * held by the network stack
4634          */
4635         for (; i > 0; i--)
4636                 put_page(pages[i-1]);
4637
4638         /*
4639          * Acl update can result in inode attribute update.
4640          * so mark the attribute cache invalid.
4641          */
4642         spin_lock(&inode->i_lock);
4643         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4644         spin_unlock(&inode->i_lock);
4645         nfs_access_zap_cache(inode);
4646         nfs_zap_acl_cache(inode);
4647         return ret;
4648 }
4649
4650 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4651 {
4652         struct nfs4_exception exception = { };
4653         int err;
4654         do {
4655                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4656                 trace_nfs4_set_acl(inode, err);
4657                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4658                                 &exception);
4659         } while (exception.retry);
4660         return err;
4661 }
4662
4663 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4664 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4665                                         size_t buflen)
4666 {
4667         struct nfs_server *server = NFS_SERVER(inode);
4668         struct nfs_fattr fattr;
4669         struct nfs4_label label = {0, 0, buflen, buf};
4670
4671         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4672         struct nfs4_getattr_arg arg = {
4673                 .fh             = NFS_FH(inode),
4674                 .bitmask        = bitmask,
4675         };
4676         struct nfs4_getattr_res res = {
4677                 .fattr          = &fattr,
4678                 .label          = &label,
4679                 .server         = server,
4680         };
4681         struct rpc_message msg = {
4682                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4683                 .rpc_argp       = &arg,
4684                 .rpc_resp       = &res,
4685         };
4686         int ret;
4687
4688         nfs_fattr_init(&fattr);
4689
4690         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4691         if (ret)
4692                 return ret;
4693         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4694                 return -ENOENT;
4695         if (buflen < label.len)
4696                 return -ERANGE;
4697         return 0;
4698 }
4699
4700 static int nfs4_get_security_label(struct inode *inode, void *buf,
4701                                         size_t buflen)
4702 {
4703         struct nfs4_exception exception = { };
4704         int err;
4705
4706         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4707                 return -EOPNOTSUPP;
4708
4709         do {
4710                 err = _nfs4_get_security_label(inode, buf, buflen);
4711                 trace_nfs4_get_security_label(inode, err);
4712                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4713                                 &exception);
4714         } while (exception.retry);
4715         return err;
4716 }
4717
4718 static int _nfs4_do_set_security_label(struct inode *inode,
4719                 struct nfs4_label *ilabel,
4720                 struct nfs_fattr *fattr,
4721                 struct nfs4_label *olabel)
4722 {
4723
4724         struct iattr sattr = {0};
4725         struct nfs_server *server = NFS_SERVER(inode);
4726         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4727         struct nfs_setattrargs arg = {
4728                 .fh             = NFS_FH(inode),
4729                 .iap            = &sattr,
4730                 .server         = server,
4731                 .bitmask        = bitmask,
4732                 .label          = ilabel,
4733         };
4734         struct nfs_setattrres res = {
4735                 .fattr          = fattr,
4736                 .label          = olabel,
4737                 .server         = server,
4738         };
4739         struct rpc_message msg = {
4740                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4741                 .rpc_argp       = &arg,
4742                 .rpc_resp       = &res,
4743         };
4744         int status;
4745
4746         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4747
4748         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4749         if (status)
4750                 dprintk("%s failed: %d\n", __func__, status);
4751
4752         return status;
4753 }
4754
4755 static int nfs4_do_set_security_label(struct inode *inode,
4756                 struct nfs4_label *ilabel,
4757                 struct nfs_fattr *fattr,
4758                 struct nfs4_label *olabel)
4759 {
4760         struct nfs4_exception exception = { };
4761         int err;
4762
4763         do {
4764                 err = _nfs4_do_set_security_label(inode, ilabel,
4765                                 fattr, olabel);
4766                 trace_nfs4_set_security_label(inode, err);
4767                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4768                                 &exception);
4769         } while (exception.retry);
4770         return err;
4771 }
4772
4773 static int
4774 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4775 {
4776         struct nfs4_label ilabel, *olabel = NULL;
4777         struct nfs_fattr fattr;
4778         struct rpc_cred *cred;
4779         struct inode *inode = dentry->d_inode;
4780         int status;
4781
4782         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4783                 return -EOPNOTSUPP;
4784
4785         nfs_fattr_init(&fattr);
4786
4787         ilabel.pi = 0;
4788         ilabel.lfs = 0;
4789         ilabel.label = (char *)buf;
4790         ilabel.len = buflen;
4791
4792         cred = rpc_lookup_cred();
4793         if (IS_ERR(cred))
4794                 return PTR_ERR(cred);
4795
4796         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4797         if (IS_ERR(olabel)) {
4798                 status = -PTR_ERR(olabel);
4799                 goto out;
4800         }
4801
4802         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4803         if (status == 0)
4804                 nfs_setsecurity(inode, &fattr, olabel);
4805
4806         nfs4_label_free(olabel);
4807 out:
4808         put_rpccred(cred);
4809         return status;
4810 }
4811 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4812
4813
4814 static int
4815 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4816 {
4817         struct nfs_client *clp = server->nfs_client;
4818
4819         if (task->tk_status >= 0)
4820                 return 0;
4821         switch(task->tk_status) {
4822                 case -NFS4ERR_DELEG_REVOKED:
4823                 case -NFS4ERR_ADMIN_REVOKED:
4824                 case -NFS4ERR_BAD_STATEID:
4825                         if (state == NULL)
4826                                 break;
4827                         nfs_remove_bad_delegation(state->inode);
4828                 case -NFS4ERR_OPENMODE:
4829                         if (state == NULL)
4830                                 break;
4831                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4832                                 goto recovery_failed;
4833                         goto wait_on_recovery;
4834                 case -NFS4ERR_EXPIRED:
4835                         if (state != NULL) {
4836                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4837                                         goto recovery_failed;
4838                         }
4839                 case -NFS4ERR_STALE_STATEID:
4840                 case -NFS4ERR_STALE_CLIENTID:
4841                         nfs4_schedule_lease_recovery(clp);
4842                         goto wait_on_recovery;
4843                 case -NFS4ERR_MOVED:
4844                         if (nfs4_schedule_migration_recovery(server) < 0)
4845                                 goto recovery_failed;
4846                         goto wait_on_recovery;
4847                 case -NFS4ERR_LEASE_MOVED:
4848                         nfs4_schedule_lease_moved_recovery(clp);
4849                         goto wait_on_recovery;
4850 #if defined(CONFIG_NFS_V4_1)
4851                 case -NFS4ERR_BADSESSION:
4852                 case -NFS4ERR_BADSLOT:
4853                 case -NFS4ERR_BAD_HIGH_SLOT:
4854                 case -NFS4ERR_DEADSESSION:
4855                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4856                 case -NFS4ERR_SEQ_FALSE_RETRY:
4857                 case -NFS4ERR_SEQ_MISORDERED:
4858                         dprintk("%s ERROR %d, Reset session\n", __func__,
4859                                 task->tk_status);
4860                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4861                         goto wait_on_recovery;
4862 #endif /* CONFIG_NFS_V4_1 */
4863                 case -NFS4ERR_DELAY:
4864                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4865                 case -NFS4ERR_GRACE:
4866                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4867                 case -NFS4ERR_RETRY_UNCACHED_REP:
4868                 case -NFS4ERR_OLD_STATEID:
4869                         goto restart_call;
4870         }
4871         task->tk_status = nfs4_map_errors(task->tk_status);
4872         return 0;
4873 recovery_failed:
4874         task->tk_status = -EIO;
4875         return 0;
4876 wait_on_recovery:
4877         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4878         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4879                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4880         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4881                 goto recovery_failed;
4882 restart_call:
4883         task->tk_status = 0;
4884         return -EAGAIN;
4885 }
4886
4887 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4888                                     nfs4_verifier *bootverf)
4889 {
4890         __be32 verf[2];
4891
4892         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4893                 /* An impossible timestamp guarantees this value
4894                  * will never match a generated boot time. */
4895                 verf[0] = 0;
4896                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4897         } else {
4898                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4899                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4900                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4901         }
4902         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4903 }
4904
4905 static unsigned int
4906 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4907                                    char *buf, size_t len)
4908 {
4909         unsigned int result;
4910
4911         rcu_read_lock();
4912         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4913                                 clp->cl_ipaddr,
4914                                 rpc_peeraddr2str(clp->cl_rpcclient,
4915                                                         RPC_DISPLAY_ADDR),
4916                                 rpc_peeraddr2str(clp->cl_rpcclient,
4917                                                         RPC_DISPLAY_PROTO));
4918         rcu_read_unlock();
4919         return result;
4920 }
4921
4922 static unsigned int
4923 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4924                                 char *buf, size_t len)
4925 {
4926         const char *nodename = clp->cl_rpcclient->cl_nodename;
4927
4928         if (nfs4_client_id_uniquifier[0] != '\0')
4929                 return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4930                                 clp->rpc_ops->version,
4931                                 clp->cl_minorversion,
4932                                 nfs4_client_id_uniquifier,
4933                                 nodename);
4934         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4935                                 clp->rpc_ops->version, clp->cl_minorversion,
4936                                 nodename);
4937 }
4938
4939 /*
4940  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4941  * services.  Advertise one based on the address family of the
4942  * clientaddr.
4943  */
4944 static unsigned int
4945 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4946 {
4947         if (strchr(clp->cl_ipaddr, ':') != NULL)
4948                 return scnprintf(buf, len, "tcp6");
4949         else
4950                 return scnprintf(buf, len, "tcp");
4951 }
4952
4953 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
4954 {
4955         struct nfs4_setclientid *sc = calldata;
4956
4957         if (task->tk_status == 0)
4958                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
4959 }
4960
4961 static const struct rpc_call_ops nfs4_setclientid_ops = {
4962         .rpc_call_done = nfs4_setclientid_done,
4963 };
4964
4965 /**
4966  * nfs4_proc_setclientid - Negotiate client ID
4967  * @clp: state data structure
4968  * @program: RPC program for NFSv4 callback service
4969  * @port: IP port number for NFS4 callback service
4970  * @cred: RPC credential to use for this call
4971  * @res: where to place the result
4972  *
4973  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4974  */
4975 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4976                 unsigned short port, struct rpc_cred *cred,
4977                 struct nfs4_setclientid_res *res)
4978 {
4979         nfs4_verifier sc_verifier;
4980         struct nfs4_setclientid setclientid = {
4981                 .sc_verifier = &sc_verifier,
4982                 .sc_prog = program,
4983                 .sc_cb_ident = clp->cl_cb_ident,
4984         };
4985         struct rpc_message msg = {
4986                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4987                 .rpc_argp = &setclientid,
4988                 .rpc_resp = res,
4989                 .rpc_cred = cred,
4990         };
4991         struct rpc_task *task;
4992         struct rpc_task_setup task_setup_data = {
4993                 .rpc_client = clp->cl_rpcclient,
4994                 .rpc_message = &msg,
4995                 .callback_ops = &nfs4_setclientid_ops,
4996                 .callback_data = &setclientid,
4997                 .flags = RPC_TASK_TIMEOUT,
4998         };
4999         int status;
5000
5001         /* nfs_client_id4 */
5002         nfs4_init_boot_verifier(clp, &sc_verifier);
5003         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5004                 setclientid.sc_name_len =
5005                                 nfs4_init_uniform_client_string(clp,
5006                                                 setclientid.sc_name,
5007                                                 sizeof(setclientid.sc_name));
5008         else
5009                 setclientid.sc_name_len =
5010                                 nfs4_init_nonuniform_client_string(clp,
5011                                                 setclientid.sc_name,
5012                                                 sizeof(setclientid.sc_name));
5013         /* cb_client4 */
5014         setclientid.sc_netid_len =
5015                                 nfs4_init_callback_netid(clp,
5016                                                 setclientid.sc_netid,
5017                                                 sizeof(setclientid.sc_netid));
5018         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5019                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5020                                 clp->cl_ipaddr, port >> 8, port & 255);
5021
5022         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5023                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5024                 setclientid.sc_name_len, setclientid.sc_name);
5025         task = rpc_run_task(&task_setup_data);
5026         if (IS_ERR(task)) {
5027                 status = PTR_ERR(task);
5028                 goto out;
5029         }
5030         status = task->tk_status;
5031         if (setclientid.sc_cred) {
5032                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5033                 put_rpccred(setclientid.sc_cred);
5034         }
5035         rpc_put_task(task);
5036 out:
5037         trace_nfs4_setclientid(clp, status);
5038         dprintk("NFS reply setclientid: %d\n", status);
5039         return status;
5040 }
5041
5042 /**
5043  * nfs4_proc_setclientid_confirm - Confirm client ID
5044  * @clp: state data structure
5045  * @res: result of a previous SETCLIENTID
5046  * @cred: RPC credential to use for this call
5047  *
5048  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5049  */
5050 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5051                 struct nfs4_setclientid_res *arg,
5052                 struct rpc_cred *cred)
5053 {
5054         struct rpc_message msg = {
5055                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5056                 .rpc_argp = arg,
5057                 .rpc_cred = cred,
5058         };
5059         int status;
5060
5061         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5062                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5063                 clp->cl_clientid);
5064         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5065         trace_nfs4_setclientid_confirm(clp, status);
5066         dprintk("NFS reply setclientid_confirm: %d\n", status);
5067         return status;
5068 }
5069
5070 struct nfs4_delegreturndata {
5071         struct nfs4_delegreturnargs args;
5072         struct nfs4_delegreturnres res;
5073         struct nfs_fh fh;
5074         nfs4_stateid stateid;
5075         unsigned long timestamp;
5076         struct nfs_fattr fattr;
5077         int rpc_status;
5078         struct inode *inode;
5079         bool roc;
5080         u32 roc_barrier;
5081 };
5082
5083 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5084 {
5085         struct nfs4_delegreturndata *data = calldata;
5086
5087         if (!nfs4_sequence_done(task, &data->res.seq_res))
5088                 return;
5089
5090         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5091         switch (task->tk_status) {
5092         case 0:
5093                 renew_lease(data->res.server, data->timestamp);
5094         case -NFS4ERR_ADMIN_REVOKED:
5095         case -NFS4ERR_DELEG_REVOKED:
5096         case -NFS4ERR_BAD_STATEID:
5097         case -NFS4ERR_OLD_STATEID:
5098         case -NFS4ERR_STALE_STATEID:
5099         case -NFS4ERR_EXPIRED:
5100                 task->tk_status = 0;
5101                 if (data->roc)
5102                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5103                 break;
5104         default:
5105                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
5106                                 -EAGAIN) {
5107                         rpc_restart_call_prepare(task);
5108                         return;
5109                 }
5110         }
5111         data->rpc_status = task->tk_status;
5112 }
5113
5114 static void nfs4_delegreturn_release(void *calldata)
5115 {
5116         struct nfs4_delegreturndata *data = calldata;
5117
5118         if (data->roc)
5119                 pnfs_roc_release(data->inode);
5120         kfree(calldata);
5121 }
5122
5123 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5124 {
5125         struct nfs4_delegreturndata *d_data;
5126
5127         d_data = (struct nfs4_delegreturndata *)data;
5128
5129         if (d_data->roc &&
5130             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5131                 return;
5132
5133         nfs4_setup_sequence(d_data->res.server,
5134                         &d_data->args.seq_args,
5135                         &d_data->res.seq_res,
5136                         task);
5137 }
5138
5139 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5140         .rpc_call_prepare = nfs4_delegreturn_prepare,
5141         .rpc_call_done = nfs4_delegreturn_done,
5142         .rpc_release = nfs4_delegreturn_release,
5143 };
5144
5145 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5146 {
5147         struct nfs4_delegreturndata *data;
5148         struct nfs_server *server = NFS_SERVER(inode);
5149         struct rpc_task *task;
5150         struct rpc_message msg = {
5151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5152                 .rpc_cred = cred,
5153         };
5154         struct rpc_task_setup task_setup_data = {
5155                 .rpc_client = server->client,
5156                 .rpc_message = &msg,
5157                 .callback_ops = &nfs4_delegreturn_ops,
5158                 .flags = RPC_TASK_ASYNC,
5159         };
5160         int status = 0;
5161
5162         data = kzalloc(sizeof(*data), GFP_NOFS);
5163         if (data == NULL)
5164                 return -ENOMEM;
5165         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5166         data->args.fhandle = &data->fh;
5167         data->args.stateid = &data->stateid;
5168         data->args.bitmask = server->cache_consistency_bitmask;
5169         nfs_copy_fh(&data->fh, NFS_FH(inode));
5170         nfs4_stateid_copy(&data->stateid, stateid);
5171         data->res.fattr = &data->fattr;
5172         data->res.server = server;
5173         nfs_fattr_init(data->res.fattr);
5174         data->timestamp = jiffies;
5175         data->rpc_status = 0;
5176         data->inode = inode;
5177         data->roc = list_empty(&NFS_I(inode)->open_files) ?
5178                     pnfs_roc(inode) : false;
5179
5180         task_setup_data.callback_data = data;
5181         msg.rpc_argp = &data->args;
5182         msg.rpc_resp = &data->res;
5183         task = rpc_run_task(&task_setup_data);
5184         if (IS_ERR(task))
5185                 return PTR_ERR(task);
5186         if (!issync)
5187                 goto out;
5188         status = nfs4_wait_for_completion_rpc_task(task);
5189         if (status != 0)
5190                 goto out;
5191         status = data->rpc_status;
5192         if (status == 0)
5193                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5194         else
5195                 nfs_refresh_inode(inode, &data->fattr);
5196 out:
5197         rpc_put_task(task);
5198         return status;
5199 }
5200
5201 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5202 {
5203         struct nfs_server *server = NFS_SERVER(inode);
5204         struct nfs4_exception exception = { };
5205         int err;
5206         do {
5207                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5208                 trace_nfs4_delegreturn(inode, err);
5209                 switch (err) {
5210                         case -NFS4ERR_STALE_STATEID:
5211                         case -NFS4ERR_EXPIRED:
5212                         case 0:
5213                                 return 0;
5214                 }
5215                 err = nfs4_handle_exception(server, err, &exception);
5216         } while (exception.retry);
5217         return err;
5218 }
5219
5220 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5221 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5222
5223 /*
5224  * sleep, with exponential backoff, and retry the LOCK operation.
5225  */
5226 static unsigned long
5227 nfs4_set_lock_task_retry(unsigned long timeout)
5228 {
5229         freezable_schedule_timeout_killable_unsafe(timeout);
5230         timeout <<= 1;
5231         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5232                 return NFS4_LOCK_MAXTIMEOUT;
5233         return timeout;
5234 }
5235
5236 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5237 {
5238         struct inode *inode = state->inode;
5239         struct nfs_server *server = NFS_SERVER(inode);
5240         struct nfs_client *clp = server->nfs_client;
5241         struct nfs_lockt_args arg = {
5242                 .fh = NFS_FH(inode),
5243                 .fl = request,
5244         };
5245         struct nfs_lockt_res res = {
5246                 .denied = request,
5247         };
5248         struct rpc_message msg = {
5249                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5250                 .rpc_argp       = &arg,
5251                 .rpc_resp       = &res,
5252                 .rpc_cred       = state->owner->so_cred,
5253         };
5254         struct nfs4_lock_state *lsp;
5255         int status;
5256
5257         arg.lock_owner.clientid = clp->cl_clientid;
5258         status = nfs4_set_lock_state(state, request);
5259         if (status != 0)
5260                 goto out;
5261         lsp = request->fl_u.nfs4_fl.owner;
5262         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5263         arg.lock_owner.s_dev = server->s_dev;
5264         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5265         switch (status) {
5266                 case 0:
5267                         request->fl_type = F_UNLCK;
5268                         break;
5269                 case -NFS4ERR_DENIED:
5270                         status = 0;
5271         }
5272         request->fl_ops->fl_release_private(request);
5273         request->fl_ops = NULL;
5274 out:
5275         return status;
5276 }
5277
5278 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5279 {
5280         struct nfs4_exception exception = { };
5281         int err;
5282
5283         do {
5284                 err = _nfs4_proc_getlk(state, cmd, request);
5285                 trace_nfs4_get_lock(request, state, cmd, err);
5286                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5287                                 &exception);
5288         } while (exception.retry);
5289         return err;
5290 }
5291
5292 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5293 {
5294         int res = 0;
5295         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5296                 case FL_POSIX:
5297                         res = posix_lock_file_wait(file, fl);
5298                         break;
5299                 case FL_FLOCK:
5300                         res = flock_lock_file_wait(file, fl);
5301                         break;
5302                 default:
5303                         BUG();
5304         }
5305         return res;
5306 }
5307
5308 struct nfs4_unlockdata {
5309         struct nfs_locku_args arg;
5310         struct nfs_locku_res res;
5311         struct nfs4_lock_state *lsp;
5312         struct nfs_open_context *ctx;
5313         struct file_lock fl;
5314         const struct nfs_server *server;
5315         unsigned long timestamp;
5316 };
5317
5318 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5319                 struct nfs_open_context *ctx,
5320                 struct nfs4_lock_state *lsp,
5321                 struct nfs_seqid *seqid)
5322 {
5323         struct nfs4_unlockdata *p;
5324         struct inode *inode = lsp->ls_state->inode;
5325
5326         p = kzalloc(sizeof(*p), GFP_NOFS);
5327         if (p == NULL)
5328                 return NULL;
5329         p->arg.fh = NFS_FH(inode);
5330         p->arg.fl = &p->fl;
5331         p->arg.seqid = seqid;
5332         p->res.seqid = seqid;
5333         p->arg.stateid = &lsp->ls_stateid;
5334         p->lsp = lsp;
5335         atomic_inc(&lsp->ls_count);
5336         /* Ensure we don't close file until we're done freeing locks! */
5337         p->ctx = get_nfs_open_context(ctx);
5338         memcpy(&p->fl, fl, sizeof(p->fl));
5339         p->server = NFS_SERVER(inode);
5340         return p;
5341 }
5342
5343 static void nfs4_locku_release_calldata(void *data)
5344 {
5345         struct nfs4_unlockdata *calldata = data;
5346         nfs_free_seqid(calldata->arg.seqid);
5347         nfs4_put_lock_state(calldata->lsp);
5348         put_nfs_open_context(calldata->ctx);
5349         kfree(calldata);
5350 }
5351
5352 static void nfs4_locku_done(struct rpc_task *task, void *data)
5353 {
5354         struct nfs4_unlockdata *calldata = data;
5355
5356         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5357                 return;
5358         switch (task->tk_status) {
5359                 case 0:
5360                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5361                                         &calldata->res.stateid);
5362                         renew_lease(calldata->server, calldata->timestamp);
5363                         break;
5364                 case -NFS4ERR_BAD_STATEID:
5365                 case -NFS4ERR_OLD_STATEID:
5366                 case -NFS4ERR_STALE_STATEID:
5367                 case -NFS4ERR_EXPIRED:
5368                         break;
5369                 default:
5370                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5371                                 rpc_restart_call_prepare(task);
5372         }
5373         nfs_release_seqid(calldata->arg.seqid);
5374 }
5375
5376 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5377 {
5378         struct nfs4_unlockdata *calldata = data;
5379
5380         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5381                 goto out_wait;
5382         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5383                 /* Note: exit _without_ running nfs4_locku_done */
5384                 goto out_no_action;
5385         }
5386         calldata->timestamp = jiffies;
5387         if (nfs4_setup_sequence(calldata->server,
5388                                 &calldata->arg.seq_args,
5389                                 &calldata->res.seq_res,
5390                                 task) != 0)
5391                 nfs_release_seqid(calldata->arg.seqid);
5392         return;
5393 out_no_action:
5394         task->tk_action = NULL;
5395 out_wait:
5396         nfs4_sequence_done(task, &calldata->res.seq_res);
5397 }
5398
5399 static const struct rpc_call_ops nfs4_locku_ops = {
5400         .rpc_call_prepare = nfs4_locku_prepare,
5401         .rpc_call_done = nfs4_locku_done,
5402         .rpc_release = nfs4_locku_release_calldata,
5403 };
5404
5405 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5406                 struct nfs_open_context *ctx,
5407                 struct nfs4_lock_state *lsp,
5408                 struct nfs_seqid *seqid)
5409 {
5410         struct nfs4_unlockdata *data;
5411         struct rpc_message msg = {
5412                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5413                 .rpc_cred = ctx->cred,
5414         };
5415         struct rpc_task_setup task_setup_data = {
5416                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5417                 .rpc_message = &msg,
5418                 .callback_ops = &nfs4_locku_ops,
5419                 .workqueue = nfsiod_workqueue,
5420                 .flags = RPC_TASK_ASYNC,
5421         };
5422
5423         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5424                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5425
5426         /* Ensure this is an unlock - when canceling a lock, the
5427          * canceled lock is passed in, and it won't be an unlock.
5428          */
5429         fl->fl_type = F_UNLCK;
5430
5431         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5432         if (data == NULL) {
5433                 nfs_free_seqid(seqid);
5434                 return ERR_PTR(-ENOMEM);
5435         }
5436
5437         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5438         msg.rpc_argp = &data->arg;
5439         msg.rpc_resp = &data->res;
5440         task_setup_data.callback_data = data;
5441         return rpc_run_task(&task_setup_data);
5442 }
5443
5444 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5445 {
5446         struct inode *inode = state->inode;
5447         struct nfs4_state_owner *sp = state->owner;
5448         struct nfs_inode *nfsi = NFS_I(inode);
5449         struct nfs_seqid *seqid;
5450         struct nfs4_lock_state *lsp;
5451         struct rpc_task *task;
5452         int status = 0;
5453         unsigned char fl_flags = request->fl_flags;
5454
5455         status = nfs4_set_lock_state(state, request);
5456         /* Unlock _before_ we do the RPC call */
5457         request->fl_flags |= FL_EXISTS;
5458         /* Exclude nfs_delegation_claim_locks() */
5459         mutex_lock(&sp->so_delegreturn_mutex);
5460         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5461         down_read(&nfsi->rwsem);
5462         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5463                 up_read(&nfsi->rwsem);
5464                 mutex_unlock(&sp->so_delegreturn_mutex);
5465                 goto out;
5466         }
5467         up_read(&nfsi->rwsem);
5468         mutex_unlock(&sp->so_delegreturn_mutex);
5469         if (status != 0)
5470                 goto out;
5471         /* Is this a delegated lock? */
5472         lsp = request->fl_u.nfs4_fl.owner;
5473         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5474                 goto out;
5475         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5476         status = -ENOMEM;
5477         if (seqid == NULL)
5478                 goto out;
5479         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5480         status = PTR_ERR(task);
5481         if (IS_ERR(task))
5482                 goto out;
5483         status = nfs4_wait_for_completion_rpc_task(task);
5484         rpc_put_task(task);
5485 out:
5486         request->fl_flags = fl_flags;
5487         trace_nfs4_unlock(request, state, F_SETLK, status);
5488         return status;
5489 }
5490
5491 struct nfs4_lockdata {
5492         struct nfs_lock_args arg;
5493         struct nfs_lock_res res;
5494         struct nfs4_lock_state *lsp;
5495         struct nfs_open_context *ctx;
5496         struct file_lock fl;
5497         unsigned long timestamp;
5498         int rpc_status;
5499         int cancelled;
5500         struct nfs_server *server;
5501 };
5502
5503 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5504                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5505                 gfp_t gfp_mask)
5506 {
5507         struct nfs4_lockdata *p;
5508         struct inode *inode = lsp->ls_state->inode;
5509         struct nfs_server *server = NFS_SERVER(inode);
5510
5511         p = kzalloc(sizeof(*p), gfp_mask);
5512         if (p == NULL)
5513                 return NULL;
5514
5515         p->arg.fh = NFS_FH(inode);
5516         p->arg.fl = &p->fl;
5517         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5518         if (p->arg.open_seqid == NULL)
5519                 goto out_free;
5520         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5521         if (p->arg.lock_seqid == NULL)
5522                 goto out_free_seqid;
5523         p->arg.lock_stateid = &lsp->ls_stateid;
5524         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5525         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5526         p->arg.lock_owner.s_dev = server->s_dev;
5527         p->res.lock_seqid = p->arg.lock_seqid;
5528         p->lsp = lsp;
5529         p->server = server;
5530         atomic_inc(&lsp->ls_count);
5531         p->ctx = get_nfs_open_context(ctx);
5532         memcpy(&p->fl, fl, sizeof(p->fl));
5533         return p;
5534 out_free_seqid:
5535         nfs_free_seqid(p->arg.open_seqid);
5536 out_free:
5537         kfree(p);
5538         return NULL;
5539 }
5540
5541 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5542 {
5543         struct nfs4_lockdata *data = calldata;
5544         struct nfs4_state *state = data->lsp->ls_state;
5545
5546         dprintk("%s: begin!\n", __func__);
5547         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5548                 goto out_wait;
5549         /* Do we need to do an open_to_lock_owner? */
5550         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5551                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5552                         goto out_release_lock_seqid;
5553                 }
5554                 data->arg.open_stateid = &state->open_stateid;
5555                 data->arg.new_lock_owner = 1;
5556                 data->res.open_seqid = data->arg.open_seqid;
5557         } else
5558                 data->arg.new_lock_owner = 0;
5559         if (!nfs4_valid_open_stateid(state)) {
5560                 data->rpc_status = -EBADF;
5561                 task->tk_action = NULL;
5562                 goto out_release_open_seqid;
5563         }
5564         data->timestamp = jiffies;
5565         if (nfs4_setup_sequence(data->server,
5566                                 &data->arg.seq_args,
5567                                 &data->res.seq_res,
5568                                 task) == 0)
5569                 return;
5570 out_release_open_seqid:
5571         nfs_release_seqid(data->arg.open_seqid);
5572 out_release_lock_seqid:
5573         nfs_release_seqid(data->arg.lock_seqid);
5574 out_wait:
5575         nfs4_sequence_done(task, &data->res.seq_res);
5576         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5577 }
5578
5579 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5580 {
5581         struct nfs4_lockdata *data = calldata;
5582
5583         dprintk("%s: begin!\n", __func__);
5584
5585         if (!nfs4_sequence_done(task, &data->res.seq_res))
5586                 return;
5587
5588         data->rpc_status = task->tk_status;
5589         if (data->arg.new_lock_owner != 0) {
5590                 if (data->rpc_status == 0)
5591                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5592                 else
5593                         goto out;
5594         }
5595         if (data->rpc_status == 0) {
5596                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5597                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5598                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5599         }
5600 out:
5601         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5602 }
5603
5604 static void nfs4_lock_release(void *calldata)
5605 {
5606         struct nfs4_lockdata *data = calldata;
5607
5608         dprintk("%s: begin!\n", __func__);
5609         nfs_free_seqid(data->arg.open_seqid);
5610         if (data->cancelled != 0) {
5611                 struct rpc_task *task;
5612                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5613                                 data->arg.lock_seqid);
5614                 if (!IS_ERR(task))
5615                         rpc_put_task_async(task);
5616                 dprintk("%s: cancelling lock!\n", __func__);
5617         } else
5618                 nfs_free_seqid(data->arg.lock_seqid);
5619         nfs4_put_lock_state(data->lsp);
5620         put_nfs_open_context(data->ctx);
5621         kfree(data);
5622         dprintk("%s: done!\n", __func__);
5623 }
5624
5625 static const struct rpc_call_ops nfs4_lock_ops = {
5626         .rpc_call_prepare = nfs4_lock_prepare,
5627         .rpc_call_done = nfs4_lock_done,
5628         .rpc_release = nfs4_lock_release,
5629 };
5630
5631 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5632 {
5633         switch (error) {
5634         case -NFS4ERR_ADMIN_REVOKED:
5635         case -NFS4ERR_BAD_STATEID:
5636                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5637                 if (new_lock_owner != 0 ||
5638                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5639                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5640                 break;
5641         case -NFS4ERR_STALE_STATEID:
5642                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5643         case -NFS4ERR_EXPIRED:
5644                 nfs4_schedule_lease_recovery(server->nfs_client);
5645         };
5646 }
5647
5648 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5649 {
5650         struct nfs4_lockdata *data;
5651         struct rpc_task *task;
5652         struct rpc_message msg = {
5653                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5654                 .rpc_cred = state->owner->so_cred,
5655         };
5656         struct rpc_task_setup task_setup_data = {
5657                 .rpc_client = NFS_CLIENT(state->inode),
5658                 .rpc_message = &msg,
5659                 .callback_ops = &nfs4_lock_ops,
5660                 .workqueue = nfsiod_workqueue,
5661                 .flags = RPC_TASK_ASYNC,
5662         };
5663         int ret;
5664
5665         dprintk("%s: begin!\n", __func__);
5666         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5667                         fl->fl_u.nfs4_fl.owner,
5668                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5669         if (data == NULL)
5670                 return -ENOMEM;
5671         if (IS_SETLKW(cmd))
5672                 data->arg.block = 1;
5673         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5674         msg.rpc_argp = &data->arg;
5675         msg.rpc_resp = &data->res;
5676         task_setup_data.callback_data = data;
5677         if (recovery_type > NFS_LOCK_NEW) {
5678                 if (recovery_type == NFS_LOCK_RECLAIM)
5679                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5680                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5681         }
5682         task = rpc_run_task(&task_setup_data);
5683         if (IS_ERR(task))
5684                 return PTR_ERR(task);
5685         ret = nfs4_wait_for_completion_rpc_task(task);
5686         if (ret == 0) {
5687                 ret = data->rpc_status;
5688                 if (ret)
5689                         nfs4_handle_setlk_error(data->server, data->lsp,
5690                                         data->arg.new_lock_owner, ret);
5691         } else
5692                 data->cancelled = 1;
5693         rpc_put_task(task);
5694         dprintk("%s: done, ret = %d!\n", __func__, ret);
5695         return ret;
5696 }
5697
5698 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5699 {
5700         struct nfs_server *server = NFS_SERVER(state->inode);
5701         struct nfs4_exception exception = {
5702                 .inode = state->inode,
5703         };
5704         int err;
5705
5706         do {
5707                 /* Cache the lock if possible... */
5708                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5709                         return 0;
5710                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5711                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5712                 if (err != -NFS4ERR_DELAY)
5713                         break;
5714                 nfs4_handle_exception(server, err, &exception);
5715         } while (exception.retry);
5716         return err;
5717 }
5718
5719 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5720 {
5721         struct nfs_server *server = NFS_SERVER(state->inode);
5722         struct nfs4_exception exception = {
5723                 .inode = state->inode,
5724         };
5725         int err;
5726
5727         err = nfs4_set_lock_state(state, request);
5728         if (err != 0)
5729                 return err;
5730         if (!recover_lost_locks) {
5731                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5732                 return 0;
5733         }
5734         do {
5735                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5736                         return 0;
5737                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5738                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5739                 switch (err) {
5740                 default:
5741                         goto out;
5742                 case -NFS4ERR_GRACE:
5743                 case -NFS4ERR_DELAY:
5744                         nfs4_handle_exception(server, err, &exception);
5745                         err = 0;
5746                 }
5747         } while (exception.retry);
5748 out:
5749         return err;
5750 }
5751
5752 #if defined(CONFIG_NFS_V4_1)
5753 /**
5754  * nfs41_check_expired_locks - possibly free a lock stateid
5755  *
5756  * @state: NFSv4 state for an inode
5757  *
5758  * Returns NFS_OK if recovery for this stateid is now finished.
5759  * Otherwise a negative NFS4ERR value is returned.
5760  */
5761 static int nfs41_check_expired_locks(struct nfs4_state *state)
5762 {
5763         int status, ret = -NFS4ERR_BAD_STATEID;
5764         struct nfs4_lock_state *lsp;
5765         struct nfs_server *server = NFS_SERVER(state->inode);
5766
5767         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5768                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5769                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5770
5771                         status = nfs41_test_stateid(server,
5772                                         &lsp->ls_stateid,
5773                                         cred);
5774                         trace_nfs4_test_lock_stateid(state, lsp, status);
5775                         if (status != NFS_OK) {
5776                                 /* Free the stateid unless the server
5777                                  * informs us the stateid is unrecognized. */
5778                                 if (status != -NFS4ERR_BAD_STATEID)
5779                                         nfs41_free_stateid(server,
5780                                                         &lsp->ls_stateid,
5781                                                         cred);
5782                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5783                                 ret = status;
5784                         }
5785                 }
5786         };
5787
5788         return ret;
5789 }
5790
5791 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5792 {
5793         int status = NFS_OK;
5794
5795         if (test_bit(LK_STATE_IN_USE, &state->flags))
5796                 status = nfs41_check_expired_locks(state);
5797         if (status != NFS_OK)
5798                 status = nfs4_lock_expired(state, request);
5799         return status;
5800 }
5801 #endif
5802
5803 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5804 {
5805         struct nfs4_state_owner *sp = state->owner;
5806         struct nfs_inode *nfsi = NFS_I(state->inode);
5807         unsigned char fl_flags = request->fl_flags;
5808         unsigned int seq;
5809         int status = -ENOLCK;
5810
5811         if ((fl_flags & FL_POSIX) &&
5812                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5813                 goto out;
5814         /* Is this a delegated open? */
5815         status = nfs4_set_lock_state(state, request);
5816         if (status != 0)
5817                 goto out;
5818         request->fl_flags |= FL_ACCESS;
5819         status = do_vfs_lock(request->fl_file, request);
5820         if (status < 0)
5821                 goto out;
5822         down_read(&nfsi->rwsem);
5823         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5824                 /* Yes: cache locks! */
5825                 /* ...but avoid races with delegation recall... */
5826                 request->fl_flags = fl_flags & ~FL_SLEEP;
5827                 status = do_vfs_lock(request->fl_file, request);
5828                 goto out_unlock;
5829         }
5830         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5831         up_read(&nfsi->rwsem);
5832         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5833         if (status != 0)
5834                 goto out;
5835         down_read(&nfsi->rwsem);
5836         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5837                 status = -NFS4ERR_DELAY;
5838                 goto out_unlock;
5839         }
5840         /* Note: we always want to sleep here! */
5841         request->fl_flags = fl_flags | FL_SLEEP;
5842         if (do_vfs_lock(request->fl_file, request) < 0)
5843                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5844                         "manager!\n", __func__);
5845 out_unlock:
5846         up_read(&nfsi->rwsem);
5847 out:
5848         request->fl_flags = fl_flags;
5849         return status;
5850 }
5851
5852 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5853 {
5854         struct nfs4_exception exception = {
5855                 .state = state,
5856                 .inode = state->inode,
5857         };
5858         int err;
5859
5860         do {
5861                 err = _nfs4_proc_setlk(state, cmd, request);
5862                 trace_nfs4_set_lock(request, state, cmd, err);
5863                 if (err == -NFS4ERR_DENIED)
5864                         err = -EAGAIN;
5865                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5866                                 err, &exception);
5867         } while (exception.retry);
5868         return err;
5869 }
5870
5871 static int
5872 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5873 {
5874         struct nfs_open_context *ctx;
5875         struct nfs4_state *state;
5876         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5877         int status;
5878
5879         /* verify open state */
5880         ctx = nfs_file_open_context(filp);
5881         state = ctx->state;
5882
5883         if (request->fl_start < 0 || request->fl_end < 0)
5884                 return -EINVAL;
5885
5886         if (IS_GETLK(cmd)) {
5887                 if (state != NULL)
5888                         return nfs4_proc_getlk(state, F_GETLK, request);
5889                 return 0;
5890         }
5891
5892         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5893                 return -EINVAL;
5894
5895         if (request->fl_type == F_UNLCK) {
5896                 if (state != NULL)
5897                         return nfs4_proc_unlck(state, cmd, request);
5898                 return 0;
5899         }
5900
5901         if (state == NULL)
5902                 return -ENOLCK;
5903         /*
5904          * Don't rely on the VFS having checked the file open mode,
5905          * since it won't do this for flock() locks.
5906          */
5907         switch (request->fl_type) {
5908         case F_RDLCK:
5909                 if (!(filp->f_mode & FMODE_READ))
5910                         return -EBADF;
5911                 break;
5912         case F_WRLCK:
5913                 if (!(filp->f_mode & FMODE_WRITE))
5914                         return -EBADF;
5915         }
5916
5917         do {
5918                 status = nfs4_proc_setlk(state, cmd, request);
5919                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5920                         break;
5921                 timeout = nfs4_set_lock_task_retry(timeout);
5922                 status = -ERESTARTSYS;
5923                 if (signalled())
5924                         break;
5925         } while(status < 0);
5926         return status;
5927 }
5928
5929 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5930 {
5931         struct nfs_server *server = NFS_SERVER(state->inode);
5932         int err;
5933
5934         err = nfs4_set_lock_state(state, fl);
5935         if (err != 0)
5936                 return err;
5937         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5938         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5939 }
5940
5941 struct nfs_release_lockowner_data {
5942         struct nfs4_lock_state *lsp;
5943         struct nfs_server *server;
5944         struct nfs_release_lockowner_args args;
5945         struct nfs_release_lockowner_res res;
5946         unsigned long timestamp;
5947 };
5948
5949 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5950 {
5951         struct nfs_release_lockowner_data *data = calldata;
5952         struct nfs_server *server = data->server;
5953         nfs40_setup_sequence(server, &data->args.seq_args,
5954                                 &data->res.seq_res, task);
5955         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5956         data->timestamp = jiffies;
5957 }
5958
5959 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5960 {
5961         struct nfs_release_lockowner_data *data = calldata;
5962         struct nfs_server *server = data->server;
5963
5964         nfs40_sequence_done(task, &data->res.seq_res);
5965
5966         switch (task->tk_status) {
5967         case 0:
5968                 renew_lease(server, data->timestamp);
5969                 break;
5970         case -NFS4ERR_STALE_CLIENTID:
5971         case -NFS4ERR_EXPIRED:
5972                 nfs4_schedule_lease_recovery(server->nfs_client);
5973                 break;
5974         case -NFS4ERR_LEASE_MOVED:
5975         case -NFS4ERR_DELAY:
5976                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN)
5977                         rpc_restart_call_prepare(task);
5978         }
5979 }
5980
5981 static void nfs4_release_lockowner_release(void *calldata)
5982 {
5983         struct nfs_release_lockowner_data *data = calldata;
5984         nfs4_free_lock_state(data->server, data->lsp);
5985         kfree(calldata);
5986 }
5987
5988 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5989         .rpc_call_prepare = nfs4_release_lockowner_prepare,
5990         .rpc_call_done = nfs4_release_lockowner_done,
5991         .rpc_release = nfs4_release_lockowner_release,
5992 };
5993
5994 static void
5995 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5996 {
5997         struct nfs_release_lockowner_data *data;
5998         struct rpc_message msg = {
5999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6000         };
6001
6002         if (server->nfs_client->cl_mvops->minor_version != 0)
6003                 return;
6004
6005         data = kmalloc(sizeof(*data), GFP_NOFS);
6006         if (!data)
6007                 return;
6008         data->lsp = lsp;
6009         data->server = server;
6010         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6011         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6012         data->args.lock_owner.s_dev = server->s_dev;
6013
6014         msg.rpc_argp = &data->args;
6015         msg.rpc_resp = &data->res;
6016         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6017         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6018 }
6019
6020 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6021
6022 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6023                                    const void *buf, size_t buflen,
6024                                    int flags, int type)
6025 {
6026         if (strcmp(key, "") != 0)
6027                 return -EINVAL;
6028
6029         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6030 }
6031
6032 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6033                                    void *buf, size_t buflen, int type)
6034 {
6035         if (strcmp(key, "") != 0)
6036                 return -EINVAL;
6037
6038         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6039 }
6040
6041 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6042                                        size_t list_len, const char *name,
6043                                        size_t name_len, int type)
6044 {
6045         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6046
6047         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6048                 return 0;
6049
6050         if (list && len <= list_len)
6051                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6052         return len;
6053 }
6054
6055 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6056 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6057 {
6058         return server->caps & NFS_CAP_SECURITY_LABEL;
6059 }
6060
6061 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6062                                    const void *buf, size_t buflen,
6063                                    int flags, int type)
6064 {
6065         if (security_ismaclabel(key))
6066                 return nfs4_set_security_label(dentry, buf, buflen);
6067
6068         return -EOPNOTSUPP;
6069 }
6070
6071 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6072                                    void *buf, size_t buflen, int type)
6073 {
6074         if (security_ismaclabel(key))
6075                 return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6076         return -EOPNOTSUPP;
6077 }
6078
6079 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6080                                        size_t list_len, const char *name,
6081                                        size_t name_len, int type)
6082 {
6083         size_t len = 0;
6084
6085         if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6086                 len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6087                 if (list && len <= list_len)
6088                         security_inode_listsecurity(dentry->d_inode, list, len);
6089         }
6090         return len;
6091 }
6092
6093 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6094         .prefix = XATTR_SECURITY_PREFIX,
6095         .list   = nfs4_xattr_list_nfs4_label,
6096         .get    = nfs4_xattr_get_nfs4_label,
6097         .set    = nfs4_xattr_set_nfs4_label,
6098 };
6099 #endif
6100
6101
6102 /*
6103  * nfs_fhget will use either the mounted_on_fileid or the fileid
6104  */
6105 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6106 {
6107         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6108                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6109               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6110               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6111                 return;
6112
6113         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6114                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6115         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6116         fattr->nlink = 2;
6117 }
6118
6119 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6120                                    const struct qstr *name,
6121                                    struct nfs4_fs_locations *fs_locations,
6122                                    struct page *page)
6123 {
6124         struct nfs_server *server = NFS_SERVER(dir);
6125         u32 bitmask[3] = {
6126                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6127         };
6128         struct nfs4_fs_locations_arg args = {
6129                 .dir_fh = NFS_FH(dir),
6130                 .name = name,
6131                 .page = page,
6132                 .bitmask = bitmask,
6133         };
6134         struct nfs4_fs_locations_res res = {
6135                 .fs_locations = fs_locations,
6136         };
6137         struct rpc_message msg = {
6138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6139                 .rpc_argp = &args,
6140                 .rpc_resp = &res,
6141         };
6142         int status;
6143
6144         dprintk("%s: start\n", __func__);
6145
6146         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6147          * is not supported */
6148         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6149                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6150         else
6151                 bitmask[0] |= FATTR4_WORD0_FILEID;
6152
6153         nfs_fattr_init(&fs_locations->fattr);
6154         fs_locations->server = server;
6155         fs_locations->nlocations = 0;
6156         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6157         dprintk("%s: returned status = %d\n", __func__, status);
6158         return status;
6159 }
6160
6161 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6162                            const struct qstr *name,
6163                            struct nfs4_fs_locations *fs_locations,
6164                            struct page *page)
6165 {
6166         struct nfs4_exception exception = { };
6167         int err;
6168         do {
6169                 err = _nfs4_proc_fs_locations(client, dir, name,
6170                                 fs_locations, page);
6171                 trace_nfs4_get_fs_locations(dir, name, err);
6172                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6173                                 &exception);
6174         } while (exception.retry);
6175         return err;
6176 }
6177
6178 /*
6179  * This operation also signals the server that this client is
6180  * performing migration recovery.  The server can stop returning
6181  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6182  * appended to this compound to identify the client ID which is
6183  * performing recovery.
6184  */
6185 static int _nfs40_proc_get_locations(struct inode *inode,
6186                                      struct nfs4_fs_locations *locations,
6187                                      struct page *page, struct rpc_cred *cred)
6188 {
6189         struct nfs_server *server = NFS_SERVER(inode);
6190         struct rpc_clnt *clnt = server->client;
6191         u32 bitmask[2] = {
6192                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6193         };
6194         struct nfs4_fs_locations_arg args = {
6195                 .clientid       = server->nfs_client->cl_clientid,
6196                 .fh             = NFS_FH(inode),
6197                 .page           = page,
6198                 .bitmask        = bitmask,
6199                 .migration      = 1,            /* skip LOOKUP */
6200                 .renew          = 1,            /* append RENEW */
6201         };
6202         struct nfs4_fs_locations_res res = {
6203                 .fs_locations   = locations,
6204                 .migration      = 1,
6205                 .renew          = 1,
6206         };
6207         struct rpc_message msg = {
6208                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6209                 .rpc_argp       = &args,
6210                 .rpc_resp       = &res,
6211                 .rpc_cred       = cred,
6212         };
6213         unsigned long now = jiffies;
6214         int status;
6215
6216         nfs_fattr_init(&locations->fattr);
6217         locations->server = server;
6218         locations->nlocations = 0;
6219
6220         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6221         nfs4_set_sequence_privileged(&args.seq_args);
6222         status = nfs4_call_sync_sequence(clnt, server, &msg,
6223                                         &args.seq_args, &res.seq_res);
6224         if (status)
6225                 return status;
6226
6227         renew_lease(server, now);
6228         return 0;
6229 }
6230
6231 #ifdef CONFIG_NFS_V4_1
6232
6233 /*
6234  * This operation also signals the server that this client is
6235  * performing migration recovery.  The server can stop asserting
6236  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6237  * performing this operation is identified in the SEQUENCE
6238  * operation in this compound.
6239  *
6240  * When the client supports GETATTR(fs_locations_info), it can
6241  * be plumbed in here.
6242  */
6243 static int _nfs41_proc_get_locations(struct inode *inode,
6244                                      struct nfs4_fs_locations *locations,
6245                                      struct page *page, struct rpc_cred *cred)
6246 {
6247         struct nfs_server *server = NFS_SERVER(inode);
6248         struct rpc_clnt *clnt = server->client;
6249         u32 bitmask[2] = {
6250                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6251         };
6252         struct nfs4_fs_locations_arg args = {
6253                 .fh             = NFS_FH(inode),
6254                 .page           = page,
6255                 .bitmask        = bitmask,
6256                 .migration      = 1,            /* skip LOOKUP */
6257         };
6258         struct nfs4_fs_locations_res res = {
6259                 .fs_locations   = locations,
6260                 .migration      = 1,
6261         };
6262         struct rpc_message msg = {
6263                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6264                 .rpc_argp       = &args,
6265                 .rpc_resp       = &res,
6266                 .rpc_cred       = cred,
6267         };
6268         int status;
6269
6270         nfs_fattr_init(&locations->fattr);
6271         locations->server = server;
6272         locations->nlocations = 0;
6273
6274         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6275         nfs4_set_sequence_privileged(&args.seq_args);
6276         status = nfs4_call_sync_sequence(clnt, server, &msg,
6277                                         &args.seq_args, &res.seq_res);
6278         if (status == NFS4_OK &&
6279             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6280                 status = -NFS4ERR_LEASE_MOVED;
6281         return status;
6282 }
6283
6284 #endif  /* CONFIG_NFS_V4_1 */
6285
6286 /**
6287  * nfs4_proc_get_locations - discover locations for a migrated FSID
6288  * @inode: inode on FSID that is migrating
6289  * @locations: result of query
6290  * @page: buffer
6291  * @cred: credential to use for this operation
6292  *
6293  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6294  * operation failed, or a negative errno if a local error occurred.
6295  *
6296  * On success, "locations" is filled in, but if the server has
6297  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6298  * asserted.
6299  *
6300  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6301  * from this client that require migration recovery.
6302  */
6303 int nfs4_proc_get_locations(struct inode *inode,
6304                             struct nfs4_fs_locations *locations,
6305                             struct page *page, struct rpc_cred *cred)
6306 {
6307         struct nfs_server *server = NFS_SERVER(inode);
6308         struct nfs_client *clp = server->nfs_client;
6309         const struct nfs4_mig_recovery_ops *ops =
6310                                         clp->cl_mvops->mig_recovery_ops;
6311         struct nfs4_exception exception = { };
6312         int status;
6313
6314         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6315                 (unsigned long long)server->fsid.major,
6316                 (unsigned long long)server->fsid.minor,
6317                 clp->cl_hostname);
6318         nfs_display_fhandle(NFS_FH(inode), __func__);
6319
6320         do {
6321                 status = ops->get_locations(inode, locations, page, cred);
6322                 if (status != -NFS4ERR_DELAY)
6323                         break;
6324                 nfs4_handle_exception(server, status, &exception);
6325         } while (exception.retry);
6326         return status;
6327 }
6328
6329 /*
6330  * This operation also signals the server that this client is
6331  * performing "lease moved" recovery.  The server can stop
6332  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6333  * is appended to this compound to identify the client ID which is
6334  * performing recovery.
6335  */
6336 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6337 {
6338         struct nfs_server *server = NFS_SERVER(inode);
6339         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6340         struct rpc_clnt *clnt = server->client;
6341         struct nfs4_fsid_present_arg args = {
6342                 .fh             = NFS_FH(inode),
6343                 .clientid       = clp->cl_clientid,
6344                 .renew          = 1,            /* append RENEW */
6345         };
6346         struct nfs4_fsid_present_res res = {
6347                 .renew          = 1,
6348         };
6349         struct rpc_message msg = {
6350                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6351                 .rpc_argp       = &args,
6352                 .rpc_resp       = &res,
6353                 .rpc_cred       = cred,
6354         };
6355         unsigned long now = jiffies;
6356         int status;
6357
6358         res.fh = nfs_alloc_fhandle();
6359         if (res.fh == NULL)
6360                 return -ENOMEM;
6361
6362         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6363         nfs4_set_sequence_privileged(&args.seq_args);
6364         status = nfs4_call_sync_sequence(clnt, server, &msg,
6365                                                 &args.seq_args, &res.seq_res);
6366         nfs_free_fhandle(res.fh);
6367         if (status)
6368                 return status;
6369
6370         do_renew_lease(clp, now);
6371         return 0;
6372 }
6373
6374 #ifdef CONFIG_NFS_V4_1
6375
6376 /*
6377  * This operation also signals the server that this client is
6378  * performing "lease moved" recovery.  The server can stop asserting
6379  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6380  * this operation is identified in the SEQUENCE operation in this
6381  * compound.
6382  */
6383 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6384 {
6385         struct nfs_server *server = NFS_SERVER(inode);
6386         struct rpc_clnt *clnt = server->client;
6387         struct nfs4_fsid_present_arg args = {
6388                 .fh             = NFS_FH(inode),
6389         };
6390         struct nfs4_fsid_present_res res = {
6391         };
6392         struct rpc_message msg = {
6393                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6394                 .rpc_argp       = &args,
6395                 .rpc_resp       = &res,
6396                 .rpc_cred       = cred,
6397         };
6398         int status;
6399
6400         res.fh = nfs_alloc_fhandle();
6401         if (res.fh == NULL)
6402                 return -ENOMEM;
6403
6404         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6405         nfs4_set_sequence_privileged(&args.seq_args);
6406         status = nfs4_call_sync_sequence(clnt, server, &msg,
6407                                                 &args.seq_args, &res.seq_res);
6408         nfs_free_fhandle(res.fh);
6409         if (status == NFS4_OK &&
6410             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6411                 status = -NFS4ERR_LEASE_MOVED;
6412         return status;
6413 }
6414
6415 #endif  /* CONFIG_NFS_V4_1 */
6416
6417 /**
6418  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6419  * @inode: inode on FSID to check
6420  * @cred: credential to use for this operation
6421  *
6422  * Server indicates whether the FSID is present, moved, or not
6423  * recognized.  This operation is necessary to clear a LEASE_MOVED
6424  * condition for this client ID.
6425  *
6426  * Returns NFS4_OK if the FSID is present on this server,
6427  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6428  *  NFS4ERR code if some error occurred on the server, or a
6429  *  negative errno if a local failure occurred.
6430  */
6431 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6432 {
6433         struct nfs_server *server = NFS_SERVER(inode);
6434         struct nfs_client *clp = server->nfs_client;
6435         const struct nfs4_mig_recovery_ops *ops =
6436                                         clp->cl_mvops->mig_recovery_ops;
6437         struct nfs4_exception exception = { };
6438         int status;
6439
6440         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6441                 (unsigned long long)server->fsid.major,
6442                 (unsigned long long)server->fsid.minor,
6443                 clp->cl_hostname);
6444         nfs_display_fhandle(NFS_FH(inode), __func__);
6445
6446         do {
6447                 status = ops->fsid_present(inode, cred);
6448                 if (status != -NFS4ERR_DELAY)
6449                         break;
6450                 nfs4_handle_exception(server, status, &exception);
6451         } while (exception.retry);
6452         return status;
6453 }
6454
6455 /**
6456  * If 'use_integrity' is true and the state managment nfs_client
6457  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6458  * and the machine credential as per RFC3530bis and RFC5661 Security
6459  * Considerations sections. Otherwise, just use the user cred with the
6460  * filesystem's rpc_client.
6461  */
6462 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6463 {
6464         int status;
6465         struct nfs4_secinfo_arg args = {
6466                 .dir_fh = NFS_FH(dir),
6467                 .name   = name,
6468         };
6469         struct nfs4_secinfo_res res = {
6470                 .flavors     = flavors,
6471         };
6472         struct rpc_message msg = {
6473                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6474                 .rpc_argp = &args,
6475                 .rpc_resp = &res,
6476         };
6477         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6478         struct rpc_cred *cred = NULL;
6479
6480         if (use_integrity) {
6481                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6482                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6483                 msg.rpc_cred = cred;
6484         }
6485
6486         dprintk("NFS call  secinfo %s\n", name->name);
6487
6488         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6489                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6490
6491         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6492                                 &res.seq_res, 0);
6493         dprintk("NFS reply  secinfo: %d\n", status);
6494
6495         if (cred)
6496                 put_rpccred(cred);
6497
6498         return status;
6499 }
6500
6501 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6502                       struct nfs4_secinfo_flavors *flavors)
6503 {
6504         struct nfs4_exception exception = { };
6505         int err;
6506         do {
6507                 err = -NFS4ERR_WRONGSEC;
6508
6509                 /* try to use integrity protection with machine cred */
6510                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6511                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6512
6513                 /*
6514                  * if unable to use integrity protection, or SECINFO with
6515                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6516                  * disallowed by spec, but exists in deployed servers) use
6517                  * the current filesystem's rpc_client and the user cred.
6518                  */
6519                 if (err == -NFS4ERR_WRONGSEC)
6520                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6521
6522                 trace_nfs4_secinfo(dir, name, err);
6523                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6524                                 &exception);
6525         } while (exception.retry);
6526         return err;
6527 }
6528
6529 #ifdef CONFIG_NFS_V4_1
6530 /*
6531  * Check the exchange flags returned by the server for invalid flags, having
6532  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6533  * DS flags set.
6534  */
6535 static int nfs4_check_cl_exchange_flags(u32 flags)
6536 {
6537         if (flags & ~EXCHGID4_FLAG_MASK_R)
6538                 goto out_inval;
6539         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6540             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6541                 goto out_inval;
6542         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6543                 goto out_inval;
6544         return NFS_OK;
6545 out_inval:
6546         return -NFS4ERR_INVAL;
6547 }
6548
6549 static bool
6550 nfs41_same_server_scope(struct nfs41_server_scope *a,
6551                         struct nfs41_server_scope *b)
6552 {
6553         if (a->server_scope_sz == b->server_scope_sz &&
6554             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6555                 return true;
6556
6557         return false;
6558 }
6559
6560 /*
6561  * nfs4_proc_bind_conn_to_session()
6562  *
6563  * The 4.1 client currently uses the same TCP connection for the
6564  * fore and backchannel.
6565  */
6566 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6567 {
6568         int status;
6569         struct nfs41_bind_conn_to_session_res res;
6570         struct rpc_message msg = {
6571                 .rpc_proc =
6572                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6573                 .rpc_argp = clp,
6574                 .rpc_resp = &res,
6575                 .rpc_cred = cred,
6576         };
6577
6578         dprintk("--> %s\n", __func__);
6579
6580         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6581         if (unlikely(res.session == NULL)) {
6582                 status = -ENOMEM;
6583                 goto out;
6584         }
6585
6586         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6587         trace_nfs4_bind_conn_to_session(clp, status);
6588         if (status == 0) {
6589                 if (memcmp(res.session->sess_id.data,
6590                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6591                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6592                         status = -EIO;
6593                         goto out_session;
6594                 }
6595                 if (res.dir != NFS4_CDFS4_BOTH) {
6596                         dprintk("NFS: %s: Unexpected direction from server\n",
6597                                 __func__);
6598                         status = -EIO;
6599                         goto out_session;
6600                 }
6601                 if (res.use_conn_in_rdma_mode) {
6602                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6603                                 __func__);
6604                         status = -EIO;
6605                         goto out_session;
6606                 }
6607         }
6608 out_session:
6609         kfree(res.session);
6610 out:
6611         dprintk("<-- %s status= %d\n", __func__, status);
6612         return status;
6613 }
6614
6615 /*
6616  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6617  * and operations we'd like to see to enable certain features in the allow map
6618  */
6619 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6620         .how = SP4_MACH_CRED,
6621         .enforce.u.words = {
6622                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6623                       1 << (OP_EXCHANGE_ID - 32) |
6624                       1 << (OP_CREATE_SESSION - 32) |
6625                       1 << (OP_DESTROY_SESSION - 32) |
6626                       1 << (OP_DESTROY_CLIENTID - 32)
6627         },
6628         .allow.u.words = {
6629                 [0] = 1 << (OP_CLOSE) |
6630                       1 << (OP_LOCKU) |
6631                       1 << (OP_COMMIT),
6632                 [1] = 1 << (OP_SECINFO - 32) |
6633                       1 << (OP_SECINFO_NO_NAME - 32) |
6634                       1 << (OP_TEST_STATEID - 32) |
6635                       1 << (OP_FREE_STATEID - 32) |
6636                       1 << (OP_WRITE - 32)
6637         }
6638 };
6639
6640 /*
6641  * Select the state protection mode for client `clp' given the server results
6642  * from exchange_id in `sp'.
6643  *
6644  * Returns 0 on success, negative errno otherwise.
6645  */
6646 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6647                                  struct nfs41_state_protection *sp)
6648 {
6649         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6650                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6651                       1 << (OP_EXCHANGE_ID - 32) |
6652                       1 << (OP_CREATE_SESSION - 32) |
6653                       1 << (OP_DESTROY_SESSION - 32) |
6654                       1 << (OP_DESTROY_CLIENTID - 32)
6655         };
6656         unsigned int i;
6657
6658         if (sp->how == SP4_MACH_CRED) {
6659                 /* Print state protect result */
6660                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6661                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6662                         if (test_bit(i, sp->enforce.u.longs))
6663                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6664                         if (test_bit(i, sp->allow.u.longs))
6665                                 dfprintk(MOUNT, "  allow op %d\n", i);
6666                 }
6667
6668                 /* make sure nothing is on enforce list that isn't supported */
6669                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6670                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6671                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6672                                 return -EINVAL;
6673                         }
6674                 }
6675
6676                 /*
6677                  * Minimal mode - state operations are allowed to use machine
6678                  * credential.  Note this already happens by default, so the
6679                  * client doesn't have to do anything more than the negotiation.
6680                  *
6681                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6682                  *       we're already using the machine cred for exchange_id
6683                  *       and will never use a different cred.
6684                  */
6685                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6686                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6687                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6688                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6689                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6690                         dfprintk(MOUNT, "  minimal mode enabled\n");
6691                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6692                 } else {
6693                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6694                         return -EINVAL;
6695                 }
6696
6697                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6698                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6699                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6700                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6701                 }
6702
6703                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6704                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6705                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6706                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6707                 }
6708
6709                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6710                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6711                         dfprintk(MOUNT, "  stateid mode enabled\n");
6712                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6713                 }
6714
6715                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6716                         dfprintk(MOUNT, "  write mode enabled\n");
6717                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6718                 }
6719
6720                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6721                         dfprintk(MOUNT, "  commit mode enabled\n");
6722                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6723                 }
6724         }
6725
6726         return 0;
6727 }
6728
6729 /*
6730  * _nfs4_proc_exchange_id()
6731  *
6732  * Wrapper for EXCHANGE_ID operation.
6733  */
6734 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6735         u32 sp4_how)
6736 {
6737         nfs4_verifier verifier;
6738         struct nfs41_exchange_id_args args = {
6739                 .verifier = &verifier,
6740                 .client = clp,
6741 #ifdef CONFIG_NFS_V4_1_MIGRATION
6742                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6743                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6744                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6745 #else
6746                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6747                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6748 #endif
6749         };
6750         struct nfs41_exchange_id_res res = {
6751                 0
6752         };
6753         int status;
6754         struct rpc_message msg = {
6755                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6756                 .rpc_argp = &args,
6757                 .rpc_resp = &res,
6758                 .rpc_cred = cred,
6759         };
6760
6761         nfs4_init_boot_verifier(clp, &verifier);
6762         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6763                                                         sizeof(args.id));
6764         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6765                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6766                 args.id_len, args.id);
6767
6768         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6769                                         GFP_NOFS);
6770         if (unlikely(res.server_owner == NULL)) {
6771                 status = -ENOMEM;
6772                 goto out;
6773         }
6774
6775         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6776                                         GFP_NOFS);
6777         if (unlikely(res.server_scope == NULL)) {
6778                 status = -ENOMEM;
6779                 goto out_server_owner;
6780         }
6781
6782         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6783         if (unlikely(res.impl_id == NULL)) {
6784                 status = -ENOMEM;
6785                 goto out_server_scope;
6786         }
6787
6788         switch (sp4_how) {
6789         case SP4_NONE:
6790                 args.state_protect.how = SP4_NONE;
6791                 break;
6792
6793         case SP4_MACH_CRED:
6794                 args.state_protect = nfs4_sp4_mach_cred_request;
6795                 break;
6796
6797         default:
6798                 /* unsupported! */
6799                 WARN_ON_ONCE(1);
6800                 status = -EINVAL;
6801                 goto out_server_scope;
6802         }
6803
6804         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6805         trace_nfs4_exchange_id(clp, status);
6806         if (status == 0)
6807                 status = nfs4_check_cl_exchange_flags(res.flags);
6808
6809         if (status == 0)
6810                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6811
6812         if (status == 0) {
6813                 clp->cl_clientid = res.clientid;
6814                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6815                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6816                         clp->cl_seqid = res.seqid;
6817
6818                 kfree(clp->cl_serverowner);
6819                 clp->cl_serverowner = res.server_owner;
6820                 res.server_owner = NULL;
6821
6822                 /* use the most recent implementation id */
6823                 kfree(clp->cl_implid);
6824                 clp->cl_implid = res.impl_id;
6825
6826                 if (clp->cl_serverscope != NULL &&
6827                     !nfs41_same_server_scope(clp->cl_serverscope,
6828                                              res.server_scope)) {
6829                         dprintk("%s: server_scope mismatch detected\n",
6830                                 __func__);
6831                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6832                         kfree(clp->cl_serverscope);
6833                         clp->cl_serverscope = NULL;
6834                 }
6835
6836                 if (clp->cl_serverscope == NULL) {
6837                         clp->cl_serverscope = res.server_scope;
6838                         goto out;
6839                 }
6840         } else
6841                 kfree(res.impl_id);
6842
6843 out_server_owner:
6844         kfree(res.server_owner);
6845 out_server_scope:
6846         kfree(res.server_scope);
6847 out:
6848         if (clp->cl_implid != NULL)
6849                 dprintk("NFS reply exchange_id: Server Implementation ID: "
6850                         "domain: %s, name: %s, date: %llu,%u\n",
6851                         clp->cl_implid->domain, clp->cl_implid->name,
6852                         clp->cl_implid->date.seconds,
6853                         clp->cl_implid->date.nseconds);
6854         dprintk("NFS reply exchange_id: %d\n", status);
6855         return status;
6856 }
6857
6858 /*
6859  * nfs4_proc_exchange_id()
6860  *
6861  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6862  *
6863  * Since the clientid has expired, all compounds using sessions
6864  * associated with the stale clientid will be returning
6865  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6866  * be in some phase of session reset.
6867  *
6868  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6869  */
6870 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6871 {
6872         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6873         int status;
6874
6875         /* try SP4_MACH_CRED if krb5i/p */
6876         if (authflavor == RPC_AUTH_GSS_KRB5I ||
6877             authflavor == RPC_AUTH_GSS_KRB5P) {
6878                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6879                 if (!status)
6880                         return 0;
6881         }
6882
6883         /* try SP4_NONE */
6884         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6885 }
6886
6887 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6888                 struct rpc_cred *cred)
6889 {
6890         struct rpc_message msg = {
6891                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6892                 .rpc_argp = clp,
6893                 .rpc_cred = cred,
6894         };
6895         int status;
6896
6897         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6898         trace_nfs4_destroy_clientid(clp, status);
6899         if (status)
6900                 dprintk("NFS: Got error %d from the server %s on "
6901                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
6902         return status;
6903 }
6904
6905 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6906                 struct rpc_cred *cred)
6907 {
6908         unsigned int loop;
6909         int ret;
6910
6911         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6912                 ret = _nfs4_proc_destroy_clientid(clp, cred);
6913                 switch (ret) {
6914                 case -NFS4ERR_DELAY:
6915                 case -NFS4ERR_CLIENTID_BUSY:
6916                         ssleep(1);
6917                         break;
6918                 default:
6919                         return ret;
6920                 }
6921         }
6922         return 0;
6923 }
6924
6925 int nfs4_destroy_clientid(struct nfs_client *clp)
6926 {
6927         struct rpc_cred *cred;
6928         int ret = 0;
6929
6930         if (clp->cl_mvops->minor_version < 1)
6931                 goto out;
6932         if (clp->cl_exchange_flags == 0)
6933                 goto out;
6934         if (clp->cl_preserve_clid)
6935                 goto out;
6936         cred = nfs4_get_clid_cred(clp);
6937         ret = nfs4_proc_destroy_clientid(clp, cred);
6938         if (cred)
6939                 put_rpccred(cred);
6940         switch (ret) {
6941         case 0:
6942         case -NFS4ERR_STALE_CLIENTID:
6943                 clp->cl_exchange_flags = 0;
6944         }
6945 out:
6946         return ret;
6947 }
6948
6949 struct nfs4_get_lease_time_data {
6950         struct nfs4_get_lease_time_args *args;
6951         struct nfs4_get_lease_time_res *res;
6952         struct nfs_client *clp;
6953 };
6954
6955 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6956                                         void *calldata)
6957 {
6958         struct nfs4_get_lease_time_data *data =
6959                         (struct nfs4_get_lease_time_data *)calldata;
6960
6961         dprintk("--> %s\n", __func__);
6962         /* just setup sequence, do not trigger session recovery
6963            since we're invoked within one */
6964         nfs41_setup_sequence(data->clp->cl_session,
6965                         &data->args->la_seq_args,
6966                         &data->res->lr_seq_res,
6967                         task);
6968         dprintk("<-- %s\n", __func__);
6969 }
6970
6971 /*
6972  * Called from nfs4_state_manager thread for session setup, so don't recover
6973  * from sequence operation or clientid errors.
6974  */
6975 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6976 {
6977         struct nfs4_get_lease_time_data *data =
6978                         (struct nfs4_get_lease_time_data *)calldata;
6979
6980         dprintk("--> %s\n", __func__);
6981         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6982                 return;
6983         switch (task->tk_status) {
6984         case -NFS4ERR_DELAY:
6985         case -NFS4ERR_GRACE:
6986                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6987                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
6988                 task->tk_status = 0;
6989                 /* fall through */
6990         case -NFS4ERR_RETRY_UNCACHED_REP:
6991                 rpc_restart_call_prepare(task);
6992                 return;
6993         }
6994         dprintk("<-- %s\n", __func__);
6995 }
6996
6997 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
6998         .rpc_call_prepare = nfs4_get_lease_time_prepare,
6999         .rpc_call_done = nfs4_get_lease_time_done,
7000 };
7001
7002 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7003 {
7004         struct rpc_task *task;
7005         struct nfs4_get_lease_time_args args;
7006         struct nfs4_get_lease_time_res res = {
7007                 .lr_fsinfo = fsinfo,
7008         };
7009         struct nfs4_get_lease_time_data data = {
7010                 .args = &args,
7011                 .res = &res,
7012                 .clp = clp,
7013         };
7014         struct rpc_message msg = {
7015                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7016                 .rpc_argp = &args,
7017                 .rpc_resp = &res,
7018         };
7019         struct rpc_task_setup task_setup = {
7020                 .rpc_client = clp->cl_rpcclient,
7021                 .rpc_message = &msg,
7022                 .callback_ops = &nfs4_get_lease_time_ops,
7023                 .callback_data = &data,
7024                 .flags = RPC_TASK_TIMEOUT,
7025         };
7026         int status;
7027
7028         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7029         nfs4_set_sequence_privileged(&args.la_seq_args);
7030         dprintk("--> %s\n", __func__);
7031         task = rpc_run_task(&task_setup);
7032
7033         if (IS_ERR(task))
7034                 status = PTR_ERR(task);
7035         else {
7036                 status = task->tk_status;
7037                 rpc_put_task(task);
7038         }
7039         dprintk("<-- %s return %d\n", __func__, status);
7040
7041         return status;
7042 }
7043
7044 /*
7045  * Initialize the values to be used by the client in CREATE_SESSION
7046  * If nfs4_init_session set the fore channel request and response sizes,
7047  * use them.
7048  *
7049  * Set the back channel max_resp_sz_cached to zero to force the client to
7050  * always set csa_cachethis to FALSE because the current implementation
7051  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7052  */
7053 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7054 {
7055         unsigned int max_rqst_sz, max_resp_sz;
7056
7057         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7058         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7059
7060         /* Fore channel attributes */
7061         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7062         args->fc_attrs.max_resp_sz = max_resp_sz;
7063         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7064         args->fc_attrs.max_reqs = max_session_slots;
7065
7066         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7067                 "max_ops=%u max_reqs=%u\n",
7068                 __func__,
7069                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7070                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7071
7072         /* Back channel attributes */
7073         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7074         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7075         args->bc_attrs.max_resp_sz_cached = 0;
7076         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7077         args->bc_attrs.max_reqs = 1;
7078
7079         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7080                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7081                 __func__,
7082                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7083                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7084                 args->bc_attrs.max_reqs);
7085 }
7086
7087 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7088 {
7089         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7090         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7091
7092         if (rcvd->max_resp_sz > sent->max_resp_sz)
7093                 return -EINVAL;
7094         /*
7095          * Our requested max_ops is the minimum we need; we're not
7096          * prepared to break up compounds into smaller pieces than that.
7097          * So, no point even trying to continue if the server won't
7098          * cooperate:
7099          */
7100         if (rcvd->max_ops < sent->max_ops)
7101                 return -EINVAL;
7102         if (rcvd->max_reqs == 0)
7103                 return -EINVAL;
7104         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7105                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7106         return 0;
7107 }
7108
7109 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7110 {
7111         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7112         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7113
7114         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7115                 return -EINVAL;
7116         if (rcvd->max_resp_sz < sent->max_resp_sz)
7117                 return -EINVAL;
7118         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7119                 return -EINVAL;
7120         /* These would render the backchannel useless: */
7121         if (rcvd->max_ops != sent->max_ops)
7122                 return -EINVAL;
7123         if (rcvd->max_reqs != sent->max_reqs)
7124                 return -EINVAL;
7125         return 0;
7126 }
7127
7128 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7129                                      struct nfs4_session *session)
7130 {
7131         int ret;
7132
7133         ret = nfs4_verify_fore_channel_attrs(args, session);
7134         if (ret)
7135                 return ret;
7136         return nfs4_verify_back_channel_attrs(args, session);
7137 }
7138
7139 static int _nfs4_proc_create_session(struct nfs_client *clp,
7140                 struct rpc_cred *cred)
7141 {
7142         struct nfs4_session *session = clp->cl_session;
7143         struct nfs41_create_session_args args = {
7144                 .client = clp,
7145                 .cb_program = NFS4_CALLBACK,
7146         };
7147         struct nfs41_create_session_res res = {
7148                 .client = clp,
7149         };
7150         struct rpc_message msg = {
7151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7152                 .rpc_argp = &args,
7153                 .rpc_resp = &res,
7154                 .rpc_cred = cred,
7155         };
7156         int status;
7157
7158         nfs4_init_channel_attrs(&args);
7159         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7160
7161         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7162         trace_nfs4_create_session(clp, status);
7163
7164         if (!status) {
7165                 /* Verify the session's negotiated channel_attrs values */
7166                 status = nfs4_verify_channel_attrs(&args, session);
7167                 /* Increment the clientid slot sequence id */
7168                 clp->cl_seqid++;
7169         }
7170
7171         return status;
7172 }
7173
7174 /*
7175  * Issues a CREATE_SESSION operation to the server.
7176  * It is the responsibility of the caller to verify the session is
7177  * expired before calling this routine.
7178  */
7179 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7180 {
7181         int status;
7182         unsigned *ptr;
7183         struct nfs4_session *session = clp->cl_session;
7184
7185         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7186
7187         status = _nfs4_proc_create_session(clp, cred);
7188         if (status)
7189                 goto out;
7190
7191         /* Init or reset the session slot tables */
7192         status = nfs4_setup_session_slot_tables(session);
7193         dprintk("slot table setup returned %d\n", status);
7194         if (status)
7195                 goto out;
7196
7197         ptr = (unsigned *)&session->sess_id.data[0];
7198         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7199                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7200 out:
7201         dprintk("<-- %s\n", __func__);
7202         return status;
7203 }
7204
7205 /*
7206  * Issue the over-the-wire RPC DESTROY_SESSION.
7207  * The caller must serialize access to this routine.
7208  */
7209 int nfs4_proc_destroy_session(struct nfs4_session *session,
7210                 struct rpc_cred *cred)
7211 {
7212         struct rpc_message msg = {
7213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7214                 .rpc_argp = session,
7215                 .rpc_cred = cred,
7216         };
7217         int status = 0;
7218
7219         dprintk("--> nfs4_proc_destroy_session\n");
7220
7221         /* session is still being setup */
7222         if (session->clp->cl_cons_state != NFS_CS_READY)
7223                 return status;
7224
7225         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7226         trace_nfs4_destroy_session(session->clp, status);
7227
7228         if (status)
7229                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7230                         "Session has been destroyed regardless...\n", status);
7231
7232         dprintk("<-- nfs4_proc_destroy_session\n");
7233         return status;
7234 }
7235
7236 /*
7237  * Renew the cl_session lease.
7238  */
7239 struct nfs4_sequence_data {
7240         struct nfs_client *clp;
7241         struct nfs4_sequence_args args;
7242         struct nfs4_sequence_res res;
7243 };
7244
7245 static void nfs41_sequence_release(void *data)
7246 {
7247         struct nfs4_sequence_data *calldata = data;
7248         struct nfs_client *clp = calldata->clp;
7249
7250         if (atomic_read(&clp->cl_count) > 1)
7251                 nfs4_schedule_state_renewal(clp);
7252         nfs_put_client(clp);
7253         kfree(calldata);
7254 }
7255
7256 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7257 {
7258         switch(task->tk_status) {
7259         case -NFS4ERR_DELAY:
7260                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7261                 return -EAGAIN;
7262         default:
7263                 nfs4_schedule_lease_recovery(clp);
7264         }
7265         return 0;
7266 }
7267
7268 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7269 {
7270         struct nfs4_sequence_data *calldata = data;
7271         struct nfs_client *clp = calldata->clp;
7272
7273         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7274                 return;
7275
7276         trace_nfs4_sequence(clp, task->tk_status);
7277         if (task->tk_status < 0) {
7278                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7279                 if (atomic_read(&clp->cl_count) == 1)
7280                         goto out;
7281
7282                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7283                         rpc_restart_call_prepare(task);
7284                         return;
7285                 }
7286         }
7287         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7288 out:
7289         dprintk("<-- %s\n", __func__);
7290 }
7291
7292 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7293 {
7294         struct nfs4_sequence_data *calldata = data;
7295         struct nfs_client *clp = calldata->clp;
7296         struct nfs4_sequence_args *args;
7297         struct nfs4_sequence_res *res;
7298
7299         args = task->tk_msg.rpc_argp;
7300         res = task->tk_msg.rpc_resp;
7301
7302         nfs41_setup_sequence(clp->cl_session, args, res, task);
7303 }
7304
7305 static const struct rpc_call_ops nfs41_sequence_ops = {
7306         .rpc_call_done = nfs41_sequence_call_done,
7307         .rpc_call_prepare = nfs41_sequence_prepare,
7308         .rpc_release = nfs41_sequence_release,
7309 };
7310
7311 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7312                 struct rpc_cred *cred,
7313                 bool is_privileged)
7314 {
7315         struct nfs4_sequence_data *calldata;
7316         struct rpc_message msg = {
7317                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7318                 .rpc_cred = cred,
7319         };
7320         struct rpc_task_setup task_setup_data = {
7321                 .rpc_client = clp->cl_rpcclient,
7322                 .rpc_message = &msg,
7323                 .callback_ops = &nfs41_sequence_ops,
7324                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7325         };
7326
7327         if (!atomic_inc_not_zero(&clp->cl_count))
7328                 return ERR_PTR(-EIO);
7329         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7330         if (calldata == NULL) {
7331                 nfs_put_client(clp);
7332                 return ERR_PTR(-ENOMEM);
7333         }
7334         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7335         if (is_privileged)
7336                 nfs4_set_sequence_privileged(&calldata->args);
7337         msg.rpc_argp = &calldata->args;
7338         msg.rpc_resp = &calldata->res;
7339         calldata->clp = clp;
7340         task_setup_data.callback_data = calldata;
7341
7342         return rpc_run_task(&task_setup_data);
7343 }
7344
7345 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7346 {
7347         struct rpc_task *task;
7348         int ret = 0;
7349
7350         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7351                 return 0;
7352         task = _nfs41_proc_sequence(clp, cred, false);
7353         if (IS_ERR(task))
7354                 ret = PTR_ERR(task);
7355         else
7356                 rpc_put_task_async(task);
7357         dprintk("<-- %s status=%d\n", __func__, ret);
7358         return ret;
7359 }
7360
7361 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7362 {
7363         struct rpc_task *task;
7364         int ret;
7365
7366         task = _nfs41_proc_sequence(clp, cred, true);
7367         if (IS_ERR(task)) {
7368                 ret = PTR_ERR(task);
7369                 goto out;
7370         }
7371         ret = rpc_wait_for_completion_task(task);
7372         if (!ret) {
7373                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7374
7375                 if (task->tk_status == 0)
7376                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7377                 ret = task->tk_status;
7378         }
7379         rpc_put_task(task);
7380 out:
7381         dprintk("<-- %s status=%d\n", __func__, ret);
7382         return ret;
7383 }
7384
7385 struct nfs4_reclaim_complete_data {
7386         struct nfs_client *clp;
7387         struct nfs41_reclaim_complete_args arg;
7388         struct nfs41_reclaim_complete_res res;
7389 };
7390
7391 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7392 {
7393         struct nfs4_reclaim_complete_data *calldata = data;
7394
7395         nfs41_setup_sequence(calldata->clp->cl_session,
7396                         &calldata->arg.seq_args,
7397                         &calldata->res.seq_res,
7398                         task);
7399 }
7400
7401 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7402 {
7403         switch(task->tk_status) {
7404         case 0:
7405         case -NFS4ERR_COMPLETE_ALREADY:
7406         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7407                 break;
7408         case -NFS4ERR_DELAY:
7409                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7410                 /* fall through */
7411         case -NFS4ERR_RETRY_UNCACHED_REP:
7412                 return -EAGAIN;
7413         default:
7414                 nfs4_schedule_lease_recovery(clp);
7415         }
7416         return 0;
7417 }
7418
7419 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7420 {
7421         struct nfs4_reclaim_complete_data *calldata = data;
7422         struct nfs_client *clp = calldata->clp;
7423         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7424
7425         dprintk("--> %s\n", __func__);
7426         if (!nfs41_sequence_done(task, res))
7427                 return;
7428
7429         trace_nfs4_reclaim_complete(clp, task->tk_status);
7430         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7431                 rpc_restart_call_prepare(task);
7432                 return;
7433         }
7434         dprintk("<-- %s\n", __func__);
7435 }
7436
7437 static void nfs4_free_reclaim_complete_data(void *data)
7438 {
7439         struct nfs4_reclaim_complete_data *calldata = data;
7440
7441         kfree(calldata);
7442 }
7443
7444 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7445         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7446         .rpc_call_done = nfs4_reclaim_complete_done,
7447         .rpc_release = nfs4_free_reclaim_complete_data,
7448 };
7449
7450 /*
7451  * Issue a global reclaim complete.
7452  */
7453 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7454                 struct rpc_cred *cred)
7455 {
7456         struct nfs4_reclaim_complete_data *calldata;
7457         struct rpc_task *task;
7458         struct rpc_message msg = {
7459                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7460                 .rpc_cred = cred,
7461         };
7462         struct rpc_task_setup task_setup_data = {
7463                 .rpc_client = clp->cl_rpcclient,
7464                 .rpc_message = &msg,
7465                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7466                 .flags = RPC_TASK_ASYNC,
7467         };
7468         int status = -ENOMEM;
7469
7470         dprintk("--> %s\n", __func__);
7471         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7472         if (calldata == NULL)
7473                 goto out;
7474         calldata->clp = clp;
7475         calldata->arg.one_fs = 0;
7476
7477         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7478         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7479         msg.rpc_argp = &calldata->arg;
7480         msg.rpc_resp = &calldata->res;
7481         task_setup_data.callback_data = calldata;
7482         task = rpc_run_task(&task_setup_data);
7483         if (IS_ERR(task)) {
7484                 status = PTR_ERR(task);
7485                 goto out;
7486         }
7487         status = nfs4_wait_for_completion_rpc_task(task);
7488         if (status == 0)
7489                 status = task->tk_status;
7490         rpc_put_task(task);
7491         return 0;
7492 out:
7493         dprintk("<-- %s status=%d\n", __func__, status);
7494         return status;
7495 }
7496
7497 static void
7498 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7499 {
7500         struct nfs4_layoutget *lgp = calldata;
7501         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7502         struct nfs4_session *session = nfs4_get_session(server);
7503
7504         dprintk("--> %s\n", __func__);
7505         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7506          * right now covering the LAYOUTGET we are about to send.
7507          * However, that is not so catastrophic, and there seems
7508          * to be no way to prevent it completely.
7509          */
7510         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7511                                 &lgp->res.seq_res, task))
7512                 return;
7513         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7514                                           NFS_I(lgp->args.inode)->layout,
7515                                           lgp->args.ctx->state)) {
7516                 rpc_exit(task, NFS4_OK);
7517         }
7518 }
7519
7520 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7521 {
7522         struct nfs4_layoutget *lgp = calldata;
7523         struct inode *inode = lgp->args.inode;
7524         struct nfs_server *server = NFS_SERVER(inode);
7525         struct pnfs_layout_hdr *lo;
7526         struct nfs4_state *state = NULL;
7527         unsigned long timeo, now, giveup;
7528
7529         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7530
7531         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7532                 goto out;
7533
7534         switch (task->tk_status) {
7535         case 0:
7536                 goto out;
7537         /*
7538          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7539          * (or clients) writing to the same RAID stripe
7540          */
7541         case -NFS4ERR_LAYOUTTRYLATER:
7542         /*
7543          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7544          * existing layout before getting a new one).
7545          */
7546         case -NFS4ERR_RECALLCONFLICT:
7547                 timeo = rpc_get_timeout(task->tk_client);
7548                 giveup = lgp->args.timestamp + timeo;
7549                 now = jiffies;
7550                 if (time_after(giveup, now)) {
7551                         unsigned long delay;
7552
7553                         /* Delay for:
7554                          * - Not less then NFS4_POLL_RETRY_MIN.
7555                          * - One last time a jiffie before we give up
7556                          * - exponential backoff (time_now minus start_attempt)
7557                          */
7558                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7559                                     min((giveup - now - 1),
7560                                         now - lgp->args.timestamp));
7561
7562                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7563                                 __func__, delay);
7564                         rpc_delay(task, delay);
7565                         task->tk_status = 0;
7566                         rpc_restart_call_prepare(task);
7567                         goto out; /* Do not call nfs4_async_handle_error() */
7568                 }
7569                 break;
7570         case -NFS4ERR_EXPIRED:
7571         case -NFS4ERR_BAD_STATEID:
7572                 spin_lock(&inode->i_lock);
7573                 lo = NFS_I(inode)->layout;
7574                 if (!lo || list_empty(&lo->plh_segs)) {
7575                         spin_unlock(&inode->i_lock);
7576                         /* If the open stateid was bad, then recover it. */
7577                         state = lgp->args.ctx->state;
7578                 } else {
7579                         LIST_HEAD(head);
7580
7581                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7582                         spin_unlock(&inode->i_lock);
7583                         /* Mark the bad layout state as invalid, then
7584                          * retry using the open stateid. */
7585                         pnfs_free_lseg_list(&head);
7586                 }
7587         }
7588         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
7589                 rpc_restart_call_prepare(task);
7590 out:
7591         dprintk("<-- %s\n", __func__);
7592 }
7593
7594 static size_t max_response_pages(struct nfs_server *server)
7595 {
7596         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7597         return nfs_page_array_len(0, max_resp_sz);
7598 }
7599
7600 static void nfs4_free_pages(struct page **pages, size_t size)
7601 {
7602         int i;
7603
7604         if (!pages)
7605                 return;
7606
7607         for (i = 0; i < size; i++) {
7608                 if (!pages[i])
7609                         break;
7610                 __free_page(pages[i]);
7611         }
7612         kfree(pages);
7613 }
7614
7615 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7616 {
7617         struct page **pages;
7618         int i;
7619
7620         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7621         if (!pages) {
7622                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7623                 return NULL;
7624         }
7625
7626         for (i = 0; i < size; i++) {
7627                 pages[i] = alloc_page(gfp_flags);
7628                 if (!pages[i]) {
7629                         dprintk("%s: failed to allocate page\n", __func__);
7630                         nfs4_free_pages(pages, size);
7631                         return NULL;
7632                 }
7633         }
7634
7635         return pages;
7636 }
7637
7638 static void nfs4_layoutget_release(void *calldata)
7639 {
7640         struct nfs4_layoutget *lgp = calldata;
7641         struct inode *inode = lgp->args.inode;
7642         struct nfs_server *server = NFS_SERVER(inode);
7643         size_t max_pages = max_response_pages(server);
7644
7645         dprintk("--> %s\n", __func__);
7646         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7647         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7648         put_nfs_open_context(lgp->args.ctx);
7649         kfree(calldata);
7650         dprintk("<-- %s\n", __func__);
7651 }
7652
7653 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7654         .rpc_call_prepare = nfs4_layoutget_prepare,
7655         .rpc_call_done = nfs4_layoutget_done,
7656         .rpc_release = nfs4_layoutget_release,
7657 };
7658
7659 struct pnfs_layout_segment *
7660 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7661 {
7662         struct inode *inode = lgp->args.inode;
7663         struct nfs_server *server = NFS_SERVER(inode);
7664         size_t max_pages = max_response_pages(server);
7665         struct rpc_task *task;
7666         struct rpc_message msg = {
7667                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7668                 .rpc_argp = &lgp->args,
7669                 .rpc_resp = &lgp->res,
7670                 .rpc_cred = lgp->cred,
7671         };
7672         struct rpc_task_setup task_setup_data = {
7673                 .rpc_client = server->client,
7674                 .rpc_message = &msg,
7675                 .callback_ops = &nfs4_layoutget_call_ops,
7676                 .callback_data = lgp,
7677                 .flags = RPC_TASK_ASYNC,
7678         };
7679         struct pnfs_layout_segment *lseg = NULL;
7680         int status = 0;
7681
7682         dprintk("--> %s\n", __func__);
7683
7684         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7685         if (!lgp->args.layout.pages) {
7686                 nfs4_layoutget_release(lgp);
7687                 return ERR_PTR(-ENOMEM);
7688         }
7689         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7690         lgp->args.timestamp = jiffies;
7691
7692         lgp->res.layoutp = &lgp->args.layout;
7693         lgp->res.seq_res.sr_slot = NULL;
7694         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7695
7696         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7697         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7698
7699         task = rpc_run_task(&task_setup_data);
7700         if (IS_ERR(task))
7701                 return ERR_CAST(task);
7702         status = nfs4_wait_for_completion_rpc_task(task);
7703         if (status == 0)
7704                 status = task->tk_status;
7705         trace_nfs4_layoutget(lgp->args.ctx,
7706                         &lgp->args.range,
7707                         &lgp->res.range,
7708                         status);
7709         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7710         if (status == 0 && lgp->res.layoutp->len)
7711                 lseg = pnfs_layout_process(lgp);
7712         rpc_put_task(task);
7713         dprintk("<-- %s status=%d\n", __func__, status);
7714         if (status)
7715                 return ERR_PTR(status);
7716         return lseg;
7717 }
7718
7719 static void
7720 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7721 {
7722         struct nfs4_layoutreturn *lrp = calldata;
7723
7724         dprintk("--> %s\n", __func__);
7725         nfs41_setup_sequence(lrp->clp->cl_session,
7726                         &lrp->args.seq_args,
7727                         &lrp->res.seq_res,
7728                         task);
7729 }
7730
7731 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7732 {
7733         struct nfs4_layoutreturn *lrp = calldata;
7734         struct nfs_server *server;
7735
7736         dprintk("--> %s\n", __func__);
7737
7738         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7739                 return;
7740
7741         server = NFS_SERVER(lrp->args.inode);
7742         switch (task->tk_status) {
7743         default:
7744                 task->tk_status = 0;
7745         case 0:
7746                 break;
7747         case -NFS4ERR_DELAY:
7748                 if (nfs4_async_handle_error(task, server, NULL) != -EAGAIN)
7749                         break;
7750                 rpc_restart_call_prepare(task);
7751                 return;
7752         }
7753         dprintk("<-- %s\n", __func__);
7754 }
7755
7756 static void nfs4_layoutreturn_release(void *calldata)
7757 {
7758         struct nfs4_layoutreturn *lrp = calldata;
7759         struct pnfs_layout_hdr *lo = lrp->args.layout;
7760
7761         dprintk("--> %s\n", __func__);
7762         spin_lock(&lo->plh_inode->i_lock);
7763         if (lrp->res.lrs_present)
7764                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7765         lo->plh_block_lgets--;
7766         spin_unlock(&lo->plh_inode->i_lock);
7767         pnfs_put_layout_hdr(lrp->args.layout);
7768         kfree(calldata);
7769         dprintk("<-- %s\n", __func__);
7770 }
7771
7772 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7773         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7774         .rpc_call_done = nfs4_layoutreturn_done,
7775         .rpc_release = nfs4_layoutreturn_release,
7776 };
7777
7778 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7779 {
7780         struct rpc_task *task;
7781         struct rpc_message msg = {
7782                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7783                 .rpc_argp = &lrp->args,
7784                 .rpc_resp = &lrp->res,
7785                 .rpc_cred = lrp->cred,
7786         };
7787         struct rpc_task_setup task_setup_data = {
7788                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
7789                 .rpc_message = &msg,
7790                 .callback_ops = &nfs4_layoutreturn_call_ops,
7791                 .callback_data = lrp,
7792         };
7793         int status;
7794
7795         dprintk("--> %s\n", __func__);
7796         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7797         task = rpc_run_task(&task_setup_data);
7798         if (IS_ERR(task))
7799                 return PTR_ERR(task);
7800         status = task->tk_status;
7801         trace_nfs4_layoutreturn(lrp->args.inode, status);
7802         dprintk("<-- %s status=%d\n", __func__, status);
7803         rpc_put_task(task);
7804         return status;
7805 }
7806
7807 /*
7808  * Retrieve the list of Data Server devices from the MDS.
7809  */
7810 static int _nfs4_getdevicelist(struct nfs_server *server,
7811                                     const struct nfs_fh *fh,
7812                                     struct pnfs_devicelist *devlist)
7813 {
7814         struct nfs4_getdevicelist_args args = {
7815                 .fh = fh,
7816                 .layoutclass = server->pnfs_curr_ld->id,
7817         };
7818         struct nfs4_getdevicelist_res res = {
7819                 .devlist = devlist,
7820         };
7821         struct rpc_message msg = {
7822                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
7823                 .rpc_argp = &args,
7824                 .rpc_resp = &res,
7825         };
7826         int status;
7827
7828         dprintk("--> %s\n", __func__);
7829         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
7830                                 &res.seq_res, 0);
7831         dprintk("<-- %s status=%d\n", __func__, status);
7832         return status;
7833 }
7834
7835 int nfs4_proc_getdevicelist(struct nfs_server *server,
7836                             const struct nfs_fh *fh,
7837                             struct pnfs_devicelist *devlist)
7838 {
7839         struct nfs4_exception exception = { };
7840         int err;
7841
7842         do {
7843                 err = nfs4_handle_exception(server,
7844                                 _nfs4_getdevicelist(server, fh, devlist),
7845                                 &exception);
7846         } while (exception.retry);
7847
7848         dprintk("%s: err=%d, num_devs=%u\n", __func__,
7849                 err, devlist->num_devs);
7850
7851         return err;
7852 }
7853 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
7854
7855 static int
7856 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7857                 struct pnfs_device *pdev,
7858                 struct rpc_cred *cred)
7859 {
7860         struct nfs4_getdeviceinfo_args args = {
7861                 .pdev = pdev,
7862         };
7863         struct nfs4_getdeviceinfo_res res = {
7864                 .pdev = pdev,
7865         };
7866         struct rpc_message msg = {
7867                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7868                 .rpc_argp = &args,
7869                 .rpc_resp = &res,
7870                 .rpc_cred = cred,
7871         };
7872         int status;
7873
7874         dprintk("--> %s\n", __func__);
7875         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7876         dprintk("<-- %s status=%d\n", __func__, status);
7877
7878         return status;
7879 }
7880
7881 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7882                 struct pnfs_device *pdev,
7883                 struct rpc_cred *cred)
7884 {
7885         struct nfs4_exception exception = { };
7886         int err;
7887
7888         do {
7889                 err = nfs4_handle_exception(server,
7890                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
7891                                         &exception);
7892         } while (exception.retry);
7893         return err;
7894 }
7895 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7896
7897 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7898 {
7899         struct nfs4_layoutcommit_data *data = calldata;
7900         struct nfs_server *server = NFS_SERVER(data->args.inode);
7901         struct nfs4_session *session = nfs4_get_session(server);
7902
7903         nfs41_setup_sequence(session,
7904                         &data->args.seq_args,
7905                         &data->res.seq_res,
7906                         task);
7907 }
7908
7909 static void
7910 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7911 {
7912         struct nfs4_layoutcommit_data *data = calldata;
7913         struct nfs_server *server = NFS_SERVER(data->args.inode);
7914
7915         if (!nfs41_sequence_done(task, &data->res.seq_res))
7916                 return;
7917
7918         switch (task->tk_status) { /* Just ignore these failures */
7919         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7920         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7921         case -NFS4ERR_BADLAYOUT:     /* no layout */
7922         case -NFS4ERR_GRACE:        /* loca_recalim always false */
7923                 task->tk_status = 0;
7924         case 0:
7925                 break;
7926         default:
7927                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7928                         rpc_restart_call_prepare(task);
7929                         return;
7930                 }
7931         }
7932 }
7933
7934 static void nfs4_layoutcommit_release(void *calldata)
7935 {
7936         struct nfs4_layoutcommit_data *data = calldata;
7937
7938         pnfs_cleanup_layoutcommit(data);
7939         nfs_post_op_update_inode_force_wcc(data->args.inode,
7940                                            data->res.fattr);
7941         put_rpccred(data->cred);
7942         kfree(data);
7943 }
7944
7945 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7946         .rpc_call_prepare = nfs4_layoutcommit_prepare,
7947         .rpc_call_done = nfs4_layoutcommit_done,
7948         .rpc_release = nfs4_layoutcommit_release,
7949 };
7950
7951 int
7952 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7953 {
7954         struct rpc_message msg = {
7955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7956                 .rpc_argp = &data->args,
7957                 .rpc_resp = &data->res,
7958                 .rpc_cred = data->cred,
7959         };
7960         struct rpc_task_setup task_setup_data = {
7961                 .task = &data->task,
7962                 .rpc_client = NFS_CLIENT(data->args.inode),
7963                 .rpc_message = &msg,
7964                 .callback_ops = &nfs4_layoutcommit_ops,
7965                 .callback_data = data,
7966                 .flags = RPC_TASK_ASYNC,
7967         };
7968         struct rpc_task *task;
7969         int status = 0;
7970
7971         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7972                 "lbw: %llu inode %lu\n",
7973                 data->task.tk_pid, sync,
7974                 data->args.lastbytewritten,
7975                 data->args.inode->i_ino);
7976
7977         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7978         task = rpc_run_task(&task_setup_data);
7979         if (IS_ERR(task))
7980                 return PTR_ERR(task);
7981         if (sync == false)
7982                 goto out;
7983         status = nfs4_wait_for_completion_rpc_task(task);
7984         if (status != 0)
7985                 goto out;
7986         status = task->tk_status;
7987         trace_nfs4_layoutcommit(data->args.inode, status);
7988 out:
7989         dprintk("%s: status %d\n", __func__, status);
7990         rpc_put_task(task);
7991         return status;
7992 }
7993
7994 /**
7995  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7996  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7997  */
7998 static int
7999 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8000                     struct nfs_fsinfo *info,
8001                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8002 {
8003         struct nfs41_secinfo_no_name_args args = {
8004                 .style = SECINFO_STYLE_CURRENT_FH,
8005         };
8006         struct nfs4_secinfo_res res = {
8007                 .flavors = flavors,
8008         };
8009         struct rpc_message msg = {
8010                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8011                 .rpc_argp = &args,
8012                 .rpc_resp = &res,
8013         };
8014         struct rpc_clnt *clnt = server->client;
8015         struct rpc_cred *cred = NULL;
8016         int status;
8017
8018         if (use_integrity) {
8019                 clnt = server->nfs_client->cl_rpcclient;
8020                 cred = nfs4_get_clid_cred(server->nfs_client);
8021                 msg.rpc_cred = cred;
8022         }
8023
8024         dprintk("--> %s\n", __func__);
8025         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8026                                 &res.seq_res, 0);
8027         dprintk("<-- %s status=%d\n", __func__, status);
8028
8029         if (cred)
8030                 put_rpccred(cred);
8031
8032         return status;
8033 }
8034
8035 static int
8036 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8037                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8038 {
8039         struct nfs4_exception exception = { };
8040         int err;
8041         do {
8042                 /* first try using integrity protection */
8043                 err = -NFS4ERR_WRONGSEC;
8044
8045                 /* try to use integrity protection with machine cred */
8046                 if (_nfs4_is_integrity_protected(server->nfs_client))
8047                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8048                                                           flavors, true);
8049
8050                 /*
8051                  * if unable to use integrity protection, or SECINFO with
8052                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8053                  * disallowed by spec, but exists in deployed servers) use
8054                  * the current filesystem's rpc_client and the user cred.
8055                  */
8056                 if (err == -NFS4ERR_WRONGSEC)
8057                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8058                                                           flavors, false);
8059
8060                 switch (err) {
8061                 case 0:
8062                 case -NFS4ERR_WRONGSEC:
8063                 case -ENOTSUPP:
8064                         goto out;
8065                 default:
8066                         err = nfs4_handle_exception(server, err, &exception);
8067                 }
8068         } while (exception.retry);
8069 out:
8070         return err;
8071 }
8072
8073 static int
8074 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8075                     struct nfs_fsinfo *info)
8076 {
8077         int err;
8078         struct page *page;
8079         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8080         struct nfs4_secinfo_flavors *flavors;
8081         struct nfs4_secinfo4 *secinfo;
8082         int i;
8083
8084         page = alloc_page(GFP_KERNEL);
8085         if (!page) {
8086                 err = -ENOMEM;
8087                 goto out;
8088         }
8089
8090         flavors = page_address(page);
8091         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8092
8093         /*
8094          * Fall back on "guess and check" method if
8095          * the server doesn't support SECINFO_NO_NAME
8096          */
8097         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8098                 err = nfs4_find_root_sec(server, fhandle, info);
8099                 goto out_freepage;
8100         }
8101         if (err)
8102                 goto out_freepage;
8103
8104         for (i = 0; i < flavors->num_flavors; i++) {
8105                 secinfo = &flavors->flavors[i];
8106
8107                 switch (secinfo->flavor) {
8108                 case RPC_AUTH_NULL:
8109                 case RPC_AUTH_UNIX:
8110                 case RPC_AUTH_GSS:
8111                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8112                                         &secinfo->flavor_info);
8113                         break;
8114                 default:
8115                         flavor = RPC_AUTH_MAXFLAVOR;
8116                         break;
8117                 }
8118
8119                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8120                         flavor = RPC_AUTH_MAXFLAVOR;
8121
8122                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8123                         err = nfs4_lookup_root_sec(server, fhandle,
8124                                                    info, flavor);
8125                         if (!err)
8126                                 break;
8127                 }
8128         }
8129
8130         if (flavor == RPC_AUTH_MAXFLAVOR)
8131                 err = -EPERM;
8132
8133 out_freepage:
8134         put_page(page);
8135         if (err == -EACCES)
8136                 return -EPERM;
8137 out:
8138         return err;
8139 }
8140
8141 static int _nfs41_test_stateid(struct nfs_server *server,
8142                 nfs4_stateid *stateid,
8143                 struct rpc_cred *cred)
8144 {
8145         int status;
8146         struct nfs41_test_stateid_args args = {
8147                 .stateid = stateid,
8148         };
8149         struct nfs41_test_stateid_res res;
8150         struct rpc_message msg = {
8151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8152                 .rpc_argp = &args,
8153                 .rpc_resp = &res,
8154                 .rpc_cred = cred,
8155         };
8156         struct rpc_clnt *rpc_client = server->client;
8157
8158         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8159                 &rpc_client, &msg);
8160
8161         dprintk("NFS call  test_stateid %p\n", stateid);
8162         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8163         nfs4_set_sequence_privileged(&args.seq_args);
8164         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8165                         &args.seq_args, &res.seq_res);
8166         if (status != NFS_OK) {
8167                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8168                 return status;
8169         }
8170         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8171         return -res.status;
8172 }
8173
8174 /**
8175  * nfs41_test_stateid - perform a TEST_STATEID operation
8176  *
8177  * @server: server / transport on which to perform the operation
8178  * @stateid: state ID to test
8179  * @cred: credential
8180  *
8181  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8182  * Otherwise a negative NFS4ERR value is returned if the operation
8183  * failed or the state ID is not currently valid.
8184  */
8185 static int nfs41_test_stateid(struct nfs_server *server,
8186                 nfs4_stateid *stateid,
8187                 struct rpc_cred *cred)
8188 {
8189         struct nfs4_exception exception = { };
8190         int err;
8191         do {
8192                 err = _nfs41_test_stateid(server, stateid, cred);
8193                 if (err != -NFS4ERR_DELAY)
8194                         break;
8195                 nfs4_handle_exception(server, err, &exception);
8196         } while (exception.retry);
8197         return err;
8198 }
8199
8200 struct nfs_free_stateid_data {
8201         struct nfs_server *server;
8202         struct nfs41_free_stateid_args args;
8203         struct nfs41_free_stateid_res res;
8204 };
8205
8206 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8207 {
8208         struct nfs_free_stateid_data *data = calldata;
8209         nfs41_setup_sequence(nfs4_get_session(data->server),
8210                         &data->args.seq_args,
8211                         &data->res.seq_res,
8212                         task);
8213 }
8214
8215 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8216 {
8217         struct nfs_free_stateid_data *data = calldata;
8218
8219         nfs41_sequence_done(task, &data->res.seq_res);
8220
8221         switch (task->tk_status) {
8222         case -NFS4ERR_DELAY:
8223                 if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
8224                         rpc_restart_call_prepare(task);
8225         }
8226 }
8227
8228 static void nfs41_free_stateid_release(void *calldata)
8229 {
8230         kfree(calldata);
8231 }
8232
8233 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8234         .rpc_call_prepare = nfs41_free_stateid_prepare,
8235         .rpc_call_done = nfs41_free_stateid_done,
8236         .rpc_release = nfs41_free_stateid_release,
8237 };
8238
8239 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8240                 nfs4_stateid *stateid,
8241                 struct rpc_cred *cred,
8242                 bool privileged)
8243 {
8244         struct rpc_message msg = {
8245                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8246                 .rpc_cred = cred,
8247         };
8248         struct rpc_task_setup task_setup = {
8249                 .rpc_client = server->client,
8250                 .rpc_message = &msg,
8251                 .callback_ops = &nfs41_free_stateid_ops,
8252                 .flags = RPC_TASK_ASYNC,
8253         };
8254         struct nfs_free_stateid_data *data;
8255
8256         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8257                 &task_setup.rpc_client, &msg);
8258
8259         dprintk("NFS call  free_stateid %p\n", stateid);
8260         data = kmalloc(sizeof(*data), GFP_NOFS);
8261         if (!data)
8262                 return ERR_PTR(-ENOMEM);
8263         data->server = server;
8264         nfs4_stateid_copy(&data->args.stateid, stateid);
8265
8266         task_setup.callback_data = data;
8267
8268         msg.rpc_argp = &data->args;
8269         msg.rpc_resp = &data->res;
8270         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8271         if (privileged)
8272                 nfs4_set_sequence_privileged(&data->args.seq_args);
8273
8274         return rpc_run_task(&task_setup);
8275 }
8276
8277 /**
8278  * nfs41_free_stateid - perform a FREE_STATEID operation
8279  *
8280  * @server: server / transport on which to perform the operation
8281  * @stateid: state ID to release
8282  * @cred: credential
8283  *
8284  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8285  * negative NFS4ERR value is returned.
8286  */
8287 static int nfs41_free_stateid(struct nfs_server *server,
8288                 nfs4_stateid *stateid,
8289                 struct rpc_cred *cred)
8290 {
8291         struct rpc_task *task;
8292         int ret;
8293
8294         task = _nfs41_free_stateid(server, stateid, cred, true);
8295         if (IS_ERR(task))
8296                 return PTR_ERR(task);
8297         ret = rpc_wait_for_completion_task(task);
8298         if (!ret)
8299                 ret = task->tk_status;
8300         rpc_put_task(task);
8301         return ret;
8302 }
8303
8304 static void
8305 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8306 {
8307         struct rpc_task *task;
8308         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8309
8310         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8311         nfs4_free_lock_state(server, lsp);
8312         if (IS_ERR(task))
8313                 return;
8314         rpc_put_task(task);
8315 }
8316
8317 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8318                 const nfs4_stateid *s2)
8319 {
8320         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8321                 return false;
8322
8323         if (s1->seqid == s2->seqid)
8324                 return true;
8325         if (s1->seqid == 0 || s2->seqid == 0)
8326                 return true;
8327
8328         return false;
8329 }
8330
8331 #endif /* CONFIG_NFS_V4_1 */
8332
8333 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8334                 const nfs4_stateid *s2)
8335 {
8336         return nfs4_stateid_match(s1, s2);
8337 }
8338
8339
8340 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8341         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8342         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8343         .recover_open   = nfs4_open_reclaim,
8344         .recover_lock   = nfs4_lock_reclaim,
8345         .establish_clid = nfs4_init_clientid,
8346         .detect_trunking = nfs40_discover_server_trunking,
8347 };
8348
8349 #if defined(CONFIG_NFS_V4_1)
8350 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8351         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8352         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8353         .recover_open   = nfs4_open_reclaim,
8354         .recover_lock   = nfs4_lock_reclaim,
8355         .establish_clid = nfs41_init_clientid,
8356         .reclaim_complete = nfs41_proc_reclaim_complete,
8357         .detect_trunking = nfs41_discover_server_trunking,
8358 };
8359 #endif /* CONFIG_NFS_V4_1 */
8360
8361 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8362         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8363         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8364         .recover_open   = nfs4_open_expired,
8365         .recover_lock   = nfs4_lock_expired,
8366         .establish_clid = nfs4_init_clientid,
8367 };
8368
8369 #if defined(CONFIG_NFS_V4_1)
8370 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8371         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8372         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8373         .recover_open   = nfs41_open_expired,
8374         .recover_lock   = nfs41_lock_expired,
8375         .establish_clid = nfs41_init_clientid,
8376 };
8377 #endif /* CONFIG_NFS_V4_1 */
8378
8379 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8380         .sched_state_renewal = nfs4_proc_async_renew,
8381         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8382         .renew_lease = nfs4_proc_renew,
8383 };
8384
8385 #if defined(CONFIG_NFS_V4_1)
8386 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8387         .sched_state_renewal = nfs41_proc_async_sequence,
8388         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8389         .renew_lease = nfs4_proc_sequence,
8390 };
8391 #endif
8392
8393 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8394         .get_locations = _nfs40_proc_get_locations,
8395         .fsid_present = _nfs40_proc_fsid_present,
8396 };
8397
8398 #if defined(CONFIG_NFS_V4_1)
8399 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8400         .get_locations = _nfs41_proc_get_locations,
8401         .fsid_present = _nfs41_proc_fsid_present,
8402 };
8403 #endif  /* CONFIG_NFS_V4_1 */
8404
8405 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8406         .minor_version = 0,
8407         .init_caps = NFS_CAP_READDIRPLUS
8408                 | NFS_CAP_ATOMIC_OPEN
8409                 | NFS_CAP_CHANGE_ATTR
8410                 | NFS_CAP_POSIX_LOCK,
8411         .init_client = nfs40_init_client,
8412         .shutdown_client = nfs40_shutdown_client,
8413         .match_stateid = nfs4_match_stateid,
8414         .find_root_sec = nfs4_find_root_sec,
8415         .free_lock_state = nfs4_release_lockowner,
8416         .call_sync_ops = &nfs40_call_sync_ops,
8417         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8418         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8419         .state_renewal_ops = &nfs40_state_renewal_ops,
8420         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8421 };
8422
8423 #if defined(CONFIG_NFS_V4_1)
8424 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8425         .minor_version = 1,
8426         .init_caps = NFS_CAP_READDIRPLUS
8427                 | NFS_CAP_ATOMIC_OPEN
8428                 | NFS_CAP_CHANGE_ATTR
8429                 | NFS_CAP_POSIX_LOCK
8430                 | NFS_CAP_STATEID_NFSV41
8431                 | NFS_CAP_ATOMIC_OPEN_V1,
8432         .init_client = nfs41_init_client,
8433         .shutdown_client = nfs41_shutdown_client,
8434         .match_stateid = nfs41_match_stateid,
8435         .find_root_sec = nfs41_find_root_sec,
8436         .free_lock_state = nfs41_free_lock_state,
8437         .call_sync_ops = &nfs41_call_sync_ops,
8438         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8439         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8440         .state_renewal_ops = &nfs41_state_renewal_ops,
8441         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8442 };
8443 #endif
8444
8445 #if defined(CONFIG_NFS_V4_2)
8446 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8447         .minor_version = 2,
8448         .init_caps = NFS_CAP_READDIRPLUS
8449                 | NFS_CAP_ATOMIC_OPEN
8450                 | NFS_CAP_CHANGE_ATTR
8451                 | NFS_CAP_POSIX_LOCK
8452                 | NFS_CAP_STATEID_NFSV41
8453                 | NFS_CAP_ATOMIC_OPEN_V1,
8454         .init_client = nfs41_init_client,
8455         .shutdown_client = nfs41_shutdown_client,
8456         .match_stateid = nfs41_match_stateid,
8457         .find_root_sec = nfs41_find_root_sec,
8458         .free_lock_state = nfs41_free_lock_state,
8459         .call_sync_ops = &nfs41_call_sync_ops,
8460         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8461         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8462         .state_renewal_ops = &nfs41_state_renewal_ops,
8463 };
8464 #endif
8465
8466 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8467         [0] = &nfs_v4_0_minor_ops,
8468 #if defined(CONFIG_NFS_V4_1)
8469         [1] = &nfs_v4_1_minor_ops,
8470 #endif
8471 #if defined(CONFIG_NFS_V4_2)
8472         [2] = &nfs_v4_2_minor_ops,
8473 #endif
8474 };
8475
8476 static const struct inode_operations nfs4_dir_inode_operations = {
8477         .create         = nfs_create,
8478         .lookup         = nfs_lookup,
8479         .atomic_open    = nfs_atomic_open,
8480         .link           = nfs_link,
8481         .unlink         = nfs_unlink,
8482         .symlink        = nfs_symlink,
8483         .mkdir          = nfs_mkdir,
8484         .rmdir          = nfs_rmdir,
8485         .mknod          = nfs_mknod,
8486         .rename         = nfs_rename,
8487         .permission     = nfs_permission,
8488         .getattr        = nfs_getattr,
8489         .setattr        = nfs_setattr,
8490         .getxattr       = generic_getxattr,
8491         .setxattr       = generic_setxattr,
8492         .listxattr      = generic_listxattr,
8493         .removexattr    = generic_removexattr,
8494 };
8495
8496 static const struct inode_operations nfs4_file_inode_operations = {
8497         .permission     = nfs_permission,
8498         .getattr        = nfs_getattr,
8499         .setattr        = nfs_setattr,
8500         .getxattr       = generic_getxattr,
8501         .setxattr       = generic_setxattr,
8502         .listxattr      = generic_listxattr,
8503         .removexattr    = generic_removexattr,
8504 };
8505
8506 const struct nfs_rpc_ops nfs_v4_clientops = {
8507         .version        = 4,                    /* protocol version */
8508         .dentry_ops     = &nfs4_dentry_operations,
8509         .dir_inode_ops  = &nfs4_dir_inode_operations,
8510         .file_inode_ops = &nfs4_file_inode_operations,
8511         .file_ops       = &nfs4_file_operations,
8512         .getroot        = nfs4_proc_get_root,
8513         .submount       = nfs4_submount,
8514         .try_mount      = nfs4_try_mount,
8515         .getattr        = nfs4_proc_getattr,
8516         .setattr        = nfs4_proc_setattr,
8517         .lookup         = nfs4_proc_lookup,
8518         .access         = nfs4_proc_access,
8519         .readlink       = nfs4_proc_readlink,
8520         .create         = nfs4_proc_create,
8521         .remove         = nfs4_proc_remove,
8522         .unlink_setup   = nfs4_proc_unlink_setup,
8523         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8524         .unlink_done    = nfs4_proc_unlink_done,
8525         .rename_setup   = nfs4_proc_rename_setup,
8526         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8527         .rename_done    = nfs4_proc_rename_done,
8528         .link           = nfs4_proc_link,
8529         .symlink        = nfs4_proc_symlink,
8530         .mkdir          = nfs4_proc_mkdir,
8531         .rmdir          = nfs4_proc_remove,
8532         .readdir        = nfs4_proc_readdir,
8533         .mknod          = nfs4_proc_mknod,
8534         .statfs         = nfs4_proc_statfs,
8535         .fsinfo         = nfs4_proc_fsinfo,
8536         .pathconf       = nfs4_proc_pathconf,
8537         .set_capabilities = nfs4_server_capabilities,
8538         .decode_dirent  = nfs4_decode_dirent,
8539         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8540         .read_setup     = nfs4_proc_read_setup,
8541         .read_done      = nfs4_read_done,
8542         .write_setup    = nfs4_proc_write_setup,
8543         .write_done     = nfs4_write_done,
8544         .commit_setup   = nfs4_proc_commit_setup,
8545         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8546         .commit_done    = nfs4_commit_done,
8547         .lock           = nfs4_proc_lock,
8548         .clear_acl_cache = nfs4_zap_acl_attr,
8549         .close_context  = nfs4_close_context,
8550         .open_context   = nfs4_atomic_open,
8551         .have_delegation = nfs4_have_delegation,
8552         .return_delegation = nfs4_inode_return_delegation,
8553         .alloc_client   = nfs4_alloc_client,
8554         .init_client    = nfs4_init_client,
8555         .free_client    = nfs4_free_client,
8556         .create_server  = nfs4_create_server,
8557         .clone_server   = nfs_clone_server,
8558 };
8559
8560 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8561         .prefix = XATTR_NAME_NFSV4_ACL,
8562         .list   = nfs4_xattr_list_nfs4_acl,
8563         .get    = nfs4_xattr_get_nfs4_acl,
8564         .set    = nfs4_xattr_set_nfs4_acl,
8565 };
8566
8567 const struct xattr_handler *nfs4_xattr_handlers[] = {
8568         &nfs4_xattr_nfs4_acl_handler,
8569 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8570         &nfs4_xattr_nfs4_label_handler,
8571 #endif
8572         NULL
8573 };
8574
8575 /*
8576  * Local variables:
8577  *  c-basic-offset: 8
8578  * End:
8579  */