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         if (NFS_SERVER(dir)->nfs_client->cl_minorversion < 2)
 109                 return NULL;
 110
 111         err = security_dentry_init_security(dentry, sattr->ia_mode,
 112                                 &dentry->d_name, (void **)&label->label, &label->len);
 113         if (err == 0)
 114                 return label;
 115
 116         return NULL;
 117 }
 118 static inline void
 119 nfs4_label_release_security(struct nfs4_label *label)
 120 {
 121         if (label)
 122                 security_release_secctx(label->label, label->len);
 123 }
 124 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 125 {
 126         if (label)
 127                 return server->attr_bitmask;
 128
 129         return server->attr_bitmask_nl;
 130 }
 131 #else
 132 static inline struct nfs4_label *
 133 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 134         struct iattr *sattr, struct nfs4_label *l)
 135 { return NULL; }
 136 static inline void
 137 nfs4_label_release_security(struct nfs4_label *label)
 138 { return; }
 139 static inline u32 *
 140 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 141 { return server->attr_bitmask; }
 142 #endif
 143
 144 /* Prevent leaks of NFSv4 errors into userland */
 145 static int nfs4_map_errors(int err)
 146 {
 147         if (err >= -1000)
 148                 return err;
 149         switch (err) {
 150         case -NFS4ERR_RESOURCE:
 151         case -NFS4ERR_LAYOUTTRYLATER:
 152         case -NFS4ERR_RECALLCONFLICT:
 153                 return -EREMOTEIO;
 154         case -NFS4ERR_WRONGSEC:
 155                 return -EPERM;
 156         case -NFS4ERR_BADOWNER:
 157         case -NFS4ERR_BADNAME:
 158                 return -EINVAL;
 159         case -NFS4ERR_SHARE_DENIED:
 160                 return -EACCES;
 161         case -NFS4ERR_MINOR_VERS_MISMATCH:
 162                 return -EPROTONOSUPPORT;
 163         case -NFS4ERR_ACCESS:
 164                 return -EACCES;
 165         case -NFS4ERR_FILE_OPEN:
 166                 return -EBUSY;
 167         default:
 168                 dprintk("%s could not handle NFSv4 error %d\n",
 169                                 __func__, -err);
 170                 break;
 171         }
 172         return -EIO;
 173 }
 174
 175 /*
 176  * This is our standard bitmap for GETATTR requests.
 177  */
 178 const u32 nfs4_fattr_bitmap[3] = {
 179         FATTR4_WORD0_TYPE
 180         | FATTR4_WORD0_CHANGE
 181         | FATTR4_WORD0_SIZE
 182         | FATTR4_WORD0_FSID
 183         | FATTR4_WORD0_FILEID,
 184         FATTR4_WORD1_MODE
 185         | FATTR4_WORD1_NUMLINKS
 186         | FATTR4_WORD1_OWNER
 187         | FATTR4_WORD1_OWNER_GROUP
 188         | FATTR4_WORD1_RAWDEV
 189         | FATTR4_WORD1_SPACE_USED
 190         | FATTR4_WORD1_TIME_ACCESS
 191         | FATTR4_WORD1_TIME_METADATA
 192         | FATTR4_WORD1_TIME_MODIFY,
 193 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 194         FATTR4_WORD2_SECURITY_LABEL
 195 #endif
 196 };
 197
 198 static const u32 nfs4_pnfs_open_bitmap[3] = {
 199         FATTR4_WORD0_TYPE
 200         | FATTR4_WORD0_CHANGE
 201         | FATTR4_WORD0_SIZE
 202         | FATTR4_WORD0_FSID
 203         | FATTR4_WORD0_FILEID,
 204         FATTR4_WORD1_MODE
 205         | FATTR4_WORD1_NUMLINKS
 206         | FATTR4_WORD1_OWNER
 207         | FATTR4_WORD1_OWNER_GROUP
 208         | FATTR4_WORD1_RAWDEV
 209         | FATTR4_WORD1_SPACE_USED
 210         | FATTR4_WORD1_TIME_ACCESS
 211         | FATTR4_WORD1_TIME_METADATA
 212         | FATTR4_WORD1_TIME_MODIFY,
 213         FATTR4_WORD2_MDSTHRESHOLD
 214 };
 215
 216 static const u32 nfs4_open_noattr_bitmap[3] = {
 217         FATTR4_WORD0_TYPE
 218         | FATTR4_WORD0_CHANGE
 219         | FATTR4_WORD0_FILEID,
 220 };
 221
 222 const u32 nfs4_statfs_bitmap[3] = {
 223         FATTR4_WORD0_FILES_AVAIL
 224         | FATTR4_WORD0_FILES_FREE
 225         | FATTR4_WORD0_FILES_TOTAL,
 226         FATTR4_WORD1_SPACE_AVAIL
 227         | FATTR4_WORD1_SPACE_FREE
 228         | FATTR4_WORD1_SPACE_TOTAL
 229 };
 230
 231 const u32 nfs4_pathconf_bitmap[3] = {
 232         FATTR4_WORD0_MAXLINK
 233         | FATTR4_WORD0_MAXNAME,
 234         0
 235 };
 236
 237 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 238                         | FATTR4_WORD0_MAXREAD
 239                         | FATTR4_WORD0_MAXWRITE
 240                         | FATTR4_WORD0_LEASE_TIME,
 241                         FATTR4_WORD1_TIME_DELTA
 242                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 243                         FATTR4_WORD2_LAYOUT_BLKSIZE
 244 };
 245
 246 const u32 nfs4_fs_locations_bitmap[3] = {
 247         FATTR4_WORD0_TYPE
 248         | FATTR4_WORD0_CHANGE
 249         | FATTR4_WORD0_SIZE
 250         | FATTR4_WORD0_FSID
 251         | FATTR4_WORD0_FILEID
 252         | FATTR4_WORD0_FS_LOCATIONS,
 253         FATTR4_WORD1_MODE
 254         | FATTR4_WORD1_NUMLINKS
 255         | FATTR4_WORD1_OWNER
 256         | FATTR4_WORD1_OWNER_GROUP
 257         | FATTR4_WORD1_RAWDEV
 258         | FATTR4_WORD1_SPACE_USED
 259         | FATTR4_WORD1_TIME_ACCESS
 260         | FATTR4_WORD1_TIME_METADATA
 261         | FATTR4_WORD1_TIME_MODIFY
 262         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 263 };
 264
 265 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 266                 struct nfs4_readdir_arg *readdir)
 267 {
 268         __be32 *start, *p;
 269
 270         if (cookie > 2) {
 271                 readdir->cookie = cookie;
 272                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 273                 return;
 274         }
 275
 276         readdir->cookie = 0;
 277         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 278         if (cookie == 2)
 279                 return;
 280
 281         /*
 282          * NFSv4 servers do not return entries for '.' and '..'
 283          * Therefore, we fake these entries here.  We let '.'
 284          * have cookie 0 and '..' have cookie 1.  Note that
 285          * when talking to the server, we always send cookie 0
 286          * instead of 1 or 2.
 287          */
 288         start = p = kmap_atomic(*readdir->pages);
 289
 290         if (cookie == 0) {
 291                 *p++ = xdr_one;                                  /* next */
 292                 *p++ = xdr_zero;                   /* cookie, first word */
 293                 *p++ = xdr_one;                   /* cookie, second word */
 294                 *p++ = xdr_one;                             /* entry len */
 295                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 296                 p++;
 297                 *p++ = xdr_one;                         /* bitmap length */
 298                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 299                 *p++ = htonl(8);              /* attribute buffer length */
 300                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 301         }
 302
 303         *p++ = xdr_one;                                  /* next */
 304         *p++ = xdr_zero;                   /* cookie, first word */
 305         *p++ = xdr_two;                   /* cookie, second word */
 306         *p++ = xdr_two;                             /* entry len */
 307         memcpy(p, "..\0\0", 4);                         /* entry */
 308         p++;
 309         *p++ = xdr_one;                         /* bitmap length */
 310         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 311         *p++ = htonl(8);              /* attribute buffer length */
 312         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 313
 314         readdir->pgbase = (char *)p - (char *)start;
 315         readdir->count -= readdir->pgbase;
 316         kunmap_atomic(start);
 317 }
 318
 319 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 320 {
 321         int res = 0;
 322
 323         might_sleep();
 324
 325         if (*timeout <= 0)
 326                 *timeout = NFS4_POLL_RETRY_MIN;
 327         if (*timeout > NFS4_POLL_RETRY_MAX)
 328                 *timeout = NFS4_POLL_RETRY_MAX;
 329         freezable_schedule_timeout_killable_unsafe(*timeout);
 330         if (fatal_signal_pending(current))
 331                 res = -ERESTARTSYS;
 332         *timeout <<= 1;
 333         return res;
 334 }
 335
 336 /* This is the error handling routine for processes that are allowed
 337  * to sleep.
 338  */
 339 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 340 {
 341         struct nfs_client *clp = server->nfs_client;
 342         struct nfs4_state *state = exception->state;
 343         struct inode *inode = exception->inode;
 344         int ret = errorcode;
 345
 346         exception->retry = 0;
 347         switch(errorcode) {
 348                 case 0:
 349                         return 0;
 350                 case -NFS4ERR_OPENMODE:
 351                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 352                                 nfs4_inode_return_delegation(inode);
 353                                 exception->retry = 1;
 354                                 return 0;
 355                         }
 356                         if (state == NULL)
 357                                 break;
 358                         ret = nfs4_schedule_stateid_recovery(server, state);
 359                         if (ret < 0)
 360                                 break;
 361                         goto wait_on_recovery;
 362                 case -NFS4ERR_DELEG_REVOKED:
 363                 case -NFS4ERR_ADMIN_REVOKED:
 364                 case -NFS4ERR_BAD_STATEID:
 365                         if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
 366                                 nfs_remove_bad_delegation(inode);
 367                                 exception->retry = 1;
 368                                 break;
 369                         }
 370                         if (state == NULL)
 371                                 break;
 372                         ret = nfs4_schedule_stateid_recovery(server, state);
 373                         if (ret < 0)
 374                                 break;
 375                         goto wait_on_recovery;
 376                 case -NFS4ERR_EXPIRED:
 377                         if (state != NULL) {
 378                                 ret = nfs4_schedule_stateid_recovery(server, state);
 379                                 if (ret < 0)
 380                                         break;
 381                         }
 382                 case -NFS4ERR_STALE_STATEID:
 383                 case -NFS4ERR_STALE_CLIENTID:
 384                         nfs4_schedule_lease_recovery(clp);
 385                         goto wait_on_recovery;
 386 #if defined(CONFIG_NFS_V4_1)
 387                 case -NFS4ERR_BADSESSION:
 388                 case -NFS4ERR_BADSLOT:
 389                 case -NFS4ERR_BAD_HIGH_SLOT:
 390                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 391                 case -NFS4ERR_DEADSESSION:
 392                 case -NFS4ERR_SEQ_FALSE_RETRY:
 393                 case -NFS4ERR_SEQ_MISORDERED:
 394                         dprintk("%s ERROR: %d Reset session\n", __func__,
 395                                 errorcode);
 396                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 397                         goto wait_on_recovery;
 398 #endif /* defined(CONFIG_NFS_V4_1) */
 399                 case -NFS4ERR_FILE_OPEN:
 400                         if (exception->timeout > HZ) {
 401                                 /* We have retried a decent amount, time to
 402                                  * fail
 403                                  */
 404                                 ret = -EBUSY;
 405                                 break;
 406                         }
 407                 case -NFS4ERR_GRACE:
 408                 case -NFS4ERR_DELAY:
 409                         ret = nfs4_delay(server->client, &exception->timeout);
 410                         if (ret != 0)
 411                                 break;
 412                 case -NFS4ERR_RETRY_UNCACHED_REP:
 413                 case -NFS4ERR_OLD_STATEID:
 414                         exception->retry = 1;
 415                         break;
 416                 case -NFS4ERR_BADOWNER:
 417                         /* The following works around a Linux server bug! */
 418                 case -NFS4ERR_BADNAME:
 419                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 420                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 421                                 exception->retry = 1;
 422                                 printk(KERN_WARNING "NFS: v4 server %s "
 423                                                 "does not accept raw "
 424                                                 "uid/gids. "
 425                                                 "Reenabling the idmapper.\n",
 426                                                 server->nfs_client->cl_hostname);
 427                         }
 428         }
 429         /* We failed to handle the error */
 430         return nfs4_map_errors(ret);
 431 wait_on_recovery:
 432         ret = nfs4_wait_clnt_recover(clp);
 433         if (ret == 0)
 434                 exception->retry = 1;
 435         return ret;
 436 }
 437
 438
 439 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 440 {
 441         spin_lock(&clp->cl_lock);
 442         if (time_before(clp->cl_last_renewal,timestamp))
 443                 clp->cl_last_renewal = timestamp;
 444         spin_unlock(&clp->cl_lock);
 445 }
 446
 447 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 448 {
 449         do_renew_lease(server->nfs_client, timestamp);
 450 }
 451
 452 #if defined(CONFIG_NFS_V4_1)
 453
 454 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 455 {
 456         struct nfs4_session *session;
 457         struct nfs4_slot_table *tbl;
 458         bool send_new_highest_used_slotid = false;
 459
 460         if (!res->sr_slot) {
 461                 /* just wake up the next guy waiting since
 462                  * we may have not consumed a slot after all */
 463                 dprintk("%s: No slot\n", __func__);
 464                 return;
 465         }
 466         tbl = res->sr_slot->table;
 467         session = tbl->session;
 468
 469         spin_lock(&tbl->slot_tbl_lock);
 470         /* Be nice to the server: try to ensure that the last transmitted
 471          * value for highest_user_slotid <= target_highest_slotid
 472          */
 473         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 474                 send_new_highest_used_slotid = true;
 475
 476         if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
 477                 send_new_highest_used_slotid = false;
 478                 goto out_unlock;
 479         }
 480         nfs4_free_slot(tbl, res->sr_slot);
 481
 482         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 483                 send_new_highest_used_slotid = false;
 484 out_unlock:
 485         spin_unlock(&tbl->slot_tbl_lock);
 486         res->sr_slot = NULL;
 487         if (send_new_highest_used_slotid)
 488                 nfs41_server_notify_highest_slotid_update(session->clp);
 489 }
 490
 491 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 492 {
 493         struct nfs4_session *session;
 494         struct nfs4_slot *slot;
 495         struct nfs_client *clp;
 496         bool interrupted = false;
 497         int ret = 1;
 498
 499         /* don't increment the sequence number if the task wasn't sent */
 500         if (!RPC_WAS_SENT(task))
 501                 goto out;
 502
 503         slot = res->sr_slot;
 504         session = slot->table->session;
 505
 506         if (slot->interrupted) {
 507                 slot->interrupted = 0;
 508                 interrupted = true;
 509         }
 510
 511         /* Check the SEQUENCE operation status */
 512         switch (res->sr_status) {
 513         case 0:
 514                 /* Update the slot's sequence and clientid lease timer */
 515                 ++slot->seq_nr;
 516                 clp = session->clp;
 517                 do_renew_lease(clp, res->sr_timestamp);
 518                 /* Check sequence flags */
 519                 if (res->sr_status_flags != 0)
 520                         nfs4_schedule_lease_recovery(clp);
 521                 nfs41_update_target_slotid(slot->table, slot, res);
 522                 break;
 523         case 1:
 524                 /*
 525                  * sr_status remains 1 if an RPC level error occurred.
 526                  * The server may or may not have processed the sequence
 527                  * operation..
 528                  * Mark the slot as having hosted an interrupted RPC call.
 529                  */
 530                 slot->interrupted = 1;
 531                 goto out;
 532         case -NFS4ERR_DELAY:
 533                 /* The server detected a resend of the RPC call and
 534                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 535                  * of RFC5661.
 536                  */
 537                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 538                         __func__,
 539                         slot->slot_nr,
 540                         slot->seq_nr);
 541                 goto out_retry;
 542         case -NFS4ERR_BADSLOT:
 543                 /*
 544                  * The slot id we used was probably retired. Try again
 545                  * using a different slot id.
 546                  */
 547                 goto retry_nowait;
 548         case -NFS4ERR_SEQ_MISORDERED:
 549                 /*
 550                  * Was the last operation on this sequence interrupted?
 551                  * If so, retry after bumping the sequence number.
 552                  */
 553                 if (interrupted) {
 554                         ++slot->seq_nr;
 555                         goto retry_nowait;
 556                 }
 557                 /*
 558                  * Could this slot have been previously retired?
 559                  * If so, then the server may be expecting seq_nr = 1!
 560                  */
 561                 if (slot->seq_nr != 1) {
 562                         slot->seq_nr = 1;
 563                         goto retry_nowait;
 564                 }
 565                 break;
 566         case -NFS4ERR_SEQ_FALSE_RETRY:
 567                 ++slot->seq_nr;
 568                 goto retry_nowait;
 569         default:
 570                 /* Just update the slot sequence no. */
 571                 ++slot->seq_nr;
 572         }
 573 out:
 574         /* The session may be reset by one of the error handlers. */
 575         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 576         nfs41_sequence_free_slot(res);
 577         return ret;
 578 retry_nowait:
 579         if (rpc_restart_call_prepare(task)) {
 580                 task->tk_status = 0;
 581                 ret = 0;
 582         }
 583         goto out;
 584 out_retry:
 585         if (!rpc_restart_call(task))
 586                 goto out;
 587         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 588         return 0;
 589 }
 590
 591 static int nfs4_sequence_done(struct rpc_task *task,
 592                                struct nfs4_sequence_res *res)
 593 {
 594         if (res->sr_slot == NULL)
 595                 return 1;
 596         return nfs41_sequence_done(task, res);
 597 }
 598
 599 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 600                 struct nfs4_sequence_res *res, int cache_reply)
 601 {
 602         args->sa_slot = NULL;
 603         args->sa_cache_this = 0;
 604         args->sa_privileged = 0;
 605         if (cache_reply)
 606                 args->sa_cache_this = 1;
 607         res->sr_slot = NULL;
 608 }
 609
 610 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 611 {
 612         args->sa_privileged = 1;
 613 }
 614
 615 int nfs41_setup_sequence(struct nfs4_session *session,
 616                                 struct nfs4_sequence_args *args,
 617                                 struct nfs4_sequence_res *res,
 618                                 struct rpc_task *task)
 619 {
 620         struct nfs4_slot *slot;
 621         struct nfs4_slot_table *tbl;
 622
 623         dprintk("--> %s\n", __func__);
 624         /* slot already allocated? */
 625         if (res->sr_slot != NULL)
 626                 goto out_success;
 627
 628         tbl = &session->fc_slot_table;
 629
 630         task->tk_timeout = 0;
 631
 632         spin_lock(&tbl->slot_tbl_lock);
 633         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 634             !args->sa_privileged) {
 635                 /* The state manager will wait until the slot table is empty */
 636                 dprintk("%s session is draining\n", __func__);
 637                 goto out_sleep;
 638         }
 639
 640         slot = nfs4_alloc_slot(tbl);
 641         if (IS_ERR(slot)) {
 642                 /* If out of memory, try again in 1/4 second */
 643                 if (slot == ERR_PTR(-ENOMEM))
 644                         task->tk_timeout = HZ >> 2;
 645                 dprintk("<-- %s: no free slots\n", __func__);
 646                 goto out_sleep;
 647         }
 648         spin_unlock(&tbl->slot_tbl_lock);
 649
 650         args->sa_slot = slot;
 651
 652         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
 653                         slot->slot_nr, slot->seq_nr);
 654
 655         res->sr_slot = slot;
 656         res->sr_timestamp = jiffies;
 657         res->sr_status_flags = 0;
 658         /*
 659          * sr_status is only set in decode_sequence, and so will remain
 660          * set to 1 if an rpc level failure occurs.
 661          */
 662         res->sr_status = 1;
 663 out_success:
 664         rpc_call_start(task);
 665         return 0;
 666 out_sleep:
 667         /* Privileged tasks are queued with top priority */
 668         if (args->sa_privileged)
 669                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 670                                 NULL, RPC_PRIORITY_PRIVILEGED);
 671         else
 672                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 673         spin_unlock(&tbl->slot_tbl_lock);
 674         return -EAGAIN;
 675 }
 676 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 677
 678 int nfs4_setup_sequence(const struct nfs_server *server,
 679                         struct nfs4_sequence_args *args,
 680                         struct nfs4_sequence_res *res,
 681                         struct rpc_task *task)
 682 {
 683         struct nfs4_session *session = nfs4_get_session(server);
 684         int ret = 0;
 685
 686         if (session == NULL) {
 687                 rpc_call_start(task);
 688                 goto out;
 689         }
 690
 691         dprintk("--> %s clp %p session %p sr_slot %d\n",
 692                 __func__, session->clp, session, res->sr_slot ?
 693                         res->sr_slot->slot_nr : -1);
 694
 695         ret = nfs41_setup_sequence(session, args, res, task);
 696 out:
 697         dprintk("<-- %s status=%d\n", __func__, ret);
 698         return ret;
 699 }
 700
 701 struct nfs41_call_sync_data {
 702         const struct nfs_server *seq_server;
 703         struct nfs4_sequence_args *seq_args;
 704         struct nfs4_sequence_res *seq_res;
 705 };
 706
 707 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 708 {
 709         struct nfs41_call_sync_data *data = calldata;
 710         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 711
 712         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 713
 714         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 715 }
 716
 717 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 718 {
 719         struct nfs41_call_sync_data *data = calldata;
 720
 721         nfs41_sequence_done(task, data->seq_res);
 722 }
 723
 724 static const struct rpc_call_ops nfs41_call_sync_ops = {
 725         .rpc_call_prepare = nfs41_call_sync_prepare,
 726         .rpc_call_done = nfs41_call_sync_done,
 727 };
 728
 729 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 730                                    struct nfs_server *server,
 731                                    struct rpc_message *msg,
 732                                    struct nfs4_sequence_args *args,
 733                                    struct nfs4_sequence_res *res)
 734 {
 735         int ret;
 736         struct rpc_task *task;
 737         struct nfs41_call_sync_data data = {
 738                 .seq_server = server,
 739                 .seq_args = args,
 740                 .seq_res = res,
 741         };
 742         struct rpc_task_setup task_setup = {
 743                 .rpc_client = clnt,
 744                 .rpc_message = msg,
 745                 .callback_ops = &nfs41_call_sync_ops,
 746                 .callback_data = &data
 747         };
 748
 749         task = rpc_run_task(&task_setup);
 750         if (IS_ERR(task))
 751                 ret = PTR_ERR(task);
 752         else {
 753                 ret = task->tk_status;
 754                 rpc_put_task(task);
 755         }
 756         return ret;
 757 }
 758
 759 #else
 760 static
 761 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 762                 struct nfs4_sequence_res *res, int cache_reply)
 763 {
 764 }
 765
 766 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 767 {
 768 }
 769
 770
 771 static int nfs4_sequence_done(struct rpc_task *task,
 772                                struct nfs4_sequence_res *res)
 773 {
 774         return 1;
 775 }
 776 #endif /* CONFIG_NFS_V4_1 */
 777
 778 static
 779 int _nfs4_call_sync(struct rpc_clnt *clnt,
 780                     struct nfs_server *server,
 781                     struct rpc_message *msg,
 782                     struct nfs4_sequence_args *args,
 783                     struct nfs4_sequence_res *res)
 784 {
 785         return rpc_call_sync(clnt, msg, 0);
 786 }
 787
 788 static
 789 int nfs4_call_sync(struct rpc_clnt *clnt,
 790                    struct nfs_server *server,
 791                    struct rpc_message *msg,
 792                    struct nfs4_sequence_args *args,
 793                    struct nfs4_sequence_res *res,
 794                    int cache_reply)
 795 {
 796         nfs41_init_sequence(args, res, cache_reply);
 797         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 798                                                 args, res);
 799 }
 800
 801 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 802 {
 803         struct nfs_inode *nfsi = NFS_I(dir);
 804
 805         spin_lock(&dir->i_lock);
 806         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 807         if (!cinfo->atomic || cinfo->before != dir->i_version)
 808                 nfs_force_lookup_revalidate(dir);
 809         dir->i_version = cinfo->after;
 810         nfs_fscache_invalidate(dir);
 811         spin_unlock(&dir->i_lock);
 812 }
 813
 814 struct nfs4_opendata {
 815         struct kref kref;
 816         struct nfs_openargs o_arg;
 817         struct nfs_openres o_res;
 818         struct nfs_open_confirmargs c_arg;
 819         struct nfs_open_confirmres c_res;
 820         struct nfs4_string owner_name;
 821         struct nfs4_string group_name;
 822         struct nfs_fattr f_attr;
 823         struct nfs4_label *f_label;
 824         struct dentry *dir;
 825         struct dentry *dentry;
 826         struct nfs4_state_owner *owner;
 827         struct nfs4_state *state;
 828         struct iattr attrs;
 829         unsigned long timestamp;
 830         unsigned int rpc_done : 1;
 831         unsigned int is_recover : 1;
 832         int rpc_status;
 833         int cancelled;
 834 };
 835
 836 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 837                 int err, struct nfs4_exception *exception)
 838 {
 839         if (err != -EINVAL)
 840                 return false;
 841         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 842                 return false;
 843         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 844         exception->retry = 1;
 845         return true;
 846 }
 847
 848 static enum open_claim_type4
 849 nfs4_map_atomic_open_claim(struct nfs_server *server,
 850                 enum open_claim_type4 claim)
 851 {
 852         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 853                 return claim;
 854         switch (claim) {
 855         default:
 856                 return claim;
 857         case NFS4_OPEN_CLAIM_FH:
 858                 return NFS4_OPEN_CLAIM_NULL;
 859         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 860                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 861         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 862                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 863         }
 864 }
 865
 866 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 867 {
 868         p->o_res.f_attr = &p->f_attr;
 869         p->o_res.f_label = p->f_label;
 870         p->o_res.seqid = p->o_arg.seqid;
 871         p->c_res.seqid = p->c_arg.seqid;
 872         p->o_res.server = p->o_arg.server;
 873         p->o_res.access_request = p->o_arg.access;
 874         nfs_fattr_init(&p->f_attr);
 875         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 876 }
 877
 878 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 879                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 880                 const struct iattr *attrs,
 881                 struct nfs4_label *label,
 882                 enum open_claim_type4 claim,
 883                 gfp_t gfp_mask)
 884 {
 885         struct dentry *parent = dget_parent(dentry);
 886         struct inode *dir = parent->d_inode;
 887         struct nfs_server *server = NFS_SERVER(dir);
 888         struct nfs4_opendata *p;
 889
 890         p = kzalloc(sizeof(*p), gfp_mask);
 891         if (p == NULL)
 892                 goto err;
 893
 894         p->f_label = nfs4_label_alloc(server, gfp_mask);
 895         if (IS_ERR(p->f_label))
 896                 goto err_free_p;
 897
 898         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 899         if (p->o_arg.seqid == NULL)
 900                 goto err_free_label;
 901         nfs_sb_active(dentry->d_sb);
 902         p->dentry = dget(dentry);
 903         p->dir = parent;
 904         p->owner = sp;
 905         atomic_inc(&sp->so_count);
 906         p->o_arg.open_flags = flags;
 907         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 908         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 909          * will return permission denied for all bits until close */
 910         if (!(flags & O_EXCL)) {
 911                 /* ask server to check for all possible rights as results
 912                  * are cached */
 913                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 914                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 915         }
 916         p->o_arg.clientid = server->nfs_client->cl_clientid;
 917         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 918         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 919         p->o_arg.name = &dentry->d_name;
 920         p->o_arg.server = server;
 921         p->o_arg.bitmask = nfs4_bitmask(server, label);
 922         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 923         p->o_arg.label = label;
 924         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
 925         switch (p->o_arg.claim) {
 926         case NFS4_OPEN_CLAIM_NULL:
 927         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 928         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 929                 p->o_arg.fh = NFS_FH(dir);
 930                 break;
 931         case NFS4_OPEN_CLAIM_PREVIOUS:
 932         case NFS4_OPEN_CLAIM_FH:
 933         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 934         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 935                 p->o_arg.fh = NFS_FH(dentry->d_inode);
 936         }
 937         if (attrs != NULL && attrs->ia_valid != 0) {
 938                 __u32 verf[2];
 939
 940                 p->o_arg.u.attrs = &p->attrs;
 941                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 942
 943                 verf[0] = jiffies;
 944                 verf[1] = current->pid;
 945                 memcpy(p->o_arg.u.verifier.data, verf,
 946                                 sizeof(p->o_arg.u.verifier.data));
 947         }
 948         p->c_arg.fh = &p->o_res.fh;
 949         p->c_arg.stateid = &p->o_res.stateid;
 950         p->c_arg.seqid = p->o_arg.seqid;
 951         nfs4_init_opendata_res(p);
 952         kref_init(&p->kref);
 953         return p;
 954
 955 err_free_label:
 956         nfs4_label_free(p->f_label);
 957 err_free_p:
 958         kfree(p);
 959 err:
 960         dput(parent);
 961         return NULL;
 962 }
 963
 964 static void nfs4_opendata_free(struct kref *kref)
 965 {
 966         struct nfs4_opendata *p = container_of(kref,
 967                         struct nfs4_opendata, kref);
 968         struct super_block *sb = p->dentry->d_sb;
 969
 970         nfs_free_seqid(p->o_arg.seqid);
 971         if (p->state != NULL)
 972                 nfs4_put_open_state(p->state);
 973         nfs4_put_state_owner(p->owner);
 974
 975         nfs4_label_free(p->f_label);
 976
 977         dput(p->dir);
 978         dput(p->dentry);
 979         nfs_sb_deactive(sb);
 980         nfs_fattr_free_names(&p->f_attr);
 981         kfree(p);
 982 }
 983
 984 static void nfs4_opendata_put(struct nfs4_opendata *p)
 985 {
 986         if (p != NULL)
 987                 kref_put(&p->kref, nfs4_opendata_free);
 988 }
 989
 990 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 991 {
 992         int ret;
 993
 994         ret = rpc_wait_for_completion_task(task);
 995         return ret;
 996 }
 997
 998 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 999 {
1000         int ret = 0;
1001
1002         if (open_mode & (O_EXCL|O_TRUNC))
1003                 goto out;
1004         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1005                 case FMODE_READ:
1006                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1007                                 && state->n_rdonly != 0;
1008                         break;
1009                 case FMODE_WRITE:
1010                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1011                                 && state->n_wronly != 0;
1012                         break;
1013                 case FMODE_READ|FMODE_WRITE:
1014                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1015                                 && state->n_rdwr != 0;
1016         }
1017 out:
1018         return ret;
1019 }
1020
1021 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1022 {
1023         if (delegation == NULL)
1024                 return 0;
1025         if ((delegation->type & fmode) != fmode)
1026                 return 0;
1027         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1028                 return 0;
1029         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1030                 return 0;
1031         nfs_mark_delegation_referenced(delegation);
1032         return 1;
1033 }
1034
1035 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1036 {
1037         switch (fmode) {
1038                 case FMODE_WRITE:
1039                         state->n_wronly++;
1040                         break;
1041                 case FMODE_READ:
1042                         state->n_rdonly++;
1043                         break;
1044                 case FMODE_READ|FMODE_WRITE:
1045                         state->n_rdwr++;
1046         }
1047         nfs4_state_set_mode_locked(state, state->state | fmode);
1048 }
1049
1050 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1051 {
1052         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1053                 nfs4_stateid_copy(&state->stateid, stateid);
1054         nfs4_stateid_copy(&state->open_stateid, stateid);
1055         set_bit(NFS_OPEN_STATE, &state->flags);
1056         switch (fmode) {
1057                 case FMODE_READ:
1058                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1059                         break;
1060                 case FMODE_WRITE:
1061                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1062                         break;
1063                 case FMODE_READ|FMODE_WRITE:
1064                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1065         }
1066 }
1067
1068 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1069 {
1070         write_seqlock(&state->seqlock);
1071         nfs_set_open_stateid_locked(state, stateid, fmode);
1072         write_sequnlock(&state->seqlock);
1073 }
1074
1075 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1076 {
1077         /*
1078          * Protect the call to nfs4_state_set_mode_locked and
1079          * serialise the stateid update
1080          */
1081         write_seqlock(&state->seqlock);
1082         if (deleg_stateid != NULL) {
1083                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1084                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1085         }
1086         if (open_stateid != NULL)
1087                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1088         write_sequnlock(&state->seqlock);
1089         spin_lock(&state->owner->so_lock);
1090         update_open_stateflags(state, fmode);
1091         spin_unlock(&state->owner->so_lock);
1092 }
1093
1094 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1095 {
1096         struct nfs_inode *nfsi = NFS_I(state->inode);
1097         struct nfs_delegation *deleg_cur;
1098         int ret = 0;
1099
1100         fmode &= (FMODE_READ|FMODE_WRITE);
1101
1102         rcu_read_lock();
1103         deleg_cur = rcu_dereference(nfsi->delegation);
1104         if (deleg_cur == NULL)
1105                 goto no_delegation;
1106
1107         spin_lock(&deleg_cur->lock);
1108         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1109            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1110             (deleg_cur->type & fmode) != fmode)
1111                 goto no_delegation_unlock;
1112
1113         if (delegation == NULL)
1114                 delegation = &deleg_cur->stateid;
1115         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1116                 goto no_delegation_unlock;
1117
1118         nfs_mark_delegation_referenced(deleg_cur);
1119         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1120         ret = 1;
1121 no_delegation_unlock:
1122         spin_unlock(&deleg_cur->lock);
1123 no_delegation:
1124         rcu_read_unlock();
1125
1126         if (!ret && open_stateid != NULL) {
1127                 __update_open_stateid(state, open_stateid, NULL, fmode);
1128                 ret = 1;
1129         }
1130
1131         return ret;
1132 }
1133
1134
1135 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1136 {
1137         struct nfs_delegation *delegation;
1138
1139         rcu_read_lock();
1140         delegation = rcu_dereference(NFS_I(inode)->delegation);
1141         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1142                 rcu_read_unlock();
1143                 return;
1144         }
1145         rcu_read_unlock();
1146         nfs4_inode_return_delegation(inode);
1147 }
1148
1149 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1150 {
1151         struct nfs4_state *state = opendata->state;
1152         struct nfs_inode *nfsi = NFS_I(state->inode);
1153         struct nfs_delegation *delegation;
1154         int open_mode = opendata->o_arg.open_flags;
1155         fmode_t fmode = opendata->o_arg.fmode;
1156         nfs4_stateid stateid;
1157         int ret = -EAGAIN;
1158
1159         for (;;) {
1160                 if (can_open_cached(state, fmode, open_mode)) {
1161                         spin_lock(&state->owner->so_lock);
1162                         if (can_open_cached(state, fmode, open_mode)) {
1163                                 update_open_stateflags(state, fmode);
1164                                 spin_unlock(&state->owner->so_lock);
1165                                 goto out_return_state;
1166                         }
1167                         spin_unlock(&state->owner->so_lock);
1168                 }
1169                 rcu_read_lock();
1170                 delegation = rcu_dereference(nfsi->delegation);
1171                 if (!can_open_delegated(delegation, fmode)) {
1172                         rcu_read_unlock();
1173                         break;
1174                 }
1175                 /* Save the delegation */
1176                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1177                 rcu_read_unlock();
1178                 nfs_release_seqid(opendata->o_arg.seqid);
1179                 if (!opendata->is_recover) {
1180                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1181                         if (ret != 0)
1182                                 goto out;
1183                 }
1184                 ret = -EAGAIN;
1185
1186                 /* Try to update the stateid using the delegation */
1187                 if (update_open_stateid(state, NULL, &stateid, fmode))
1188                         goto out_return_state;
1189         }
1190 out:
1191         return ERR_PTR(ret);
1192 out_return_state:
1193         atomic_inc(&state->count);
1194         return state;
1195 }
1196
1197 static void
1198 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1199 {
1200         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1201         struct nfs_delegation *delegation;
1202         int delegation_flags = 0;
1203
1204         rcu_read_lock();
1205         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1206         if (delegation)
1207                 delegation_flags = delegation->flags;
1208         rcu_read_unlock();
1209         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1210                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1211                                    "returning a delegation for "
1212                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1213                                    clp->cl_hostname);
1214         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1215                 nfs_inode_set_delegation(state->inode,
1216                                          data->owner->so_cred,
1217                                          &data->o_res);
1218         else
1219                 nfs_inode_reclaim_delegation(state->inode,
1220                                              data->owner->so_cred,
1221                                              &data->o_res);
1222 }
1223
1224 /*
1225  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1226  * and update the nfs4_state.
1227  */
1228 static struct nfs4_state *
1229 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1230 {
1231         struct inode *inode = data->state->inode;
1232         struct nfs4_state *state = data->state;
1233         int ret;
1234
1235         if (!data->rpc_done) {
1236                 ret = data->rpc_status;
1237                 goto err;
1238         }
1239
1240         ret = -ESTALE;
1241         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1242             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1243             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1244                 goto err;
1245
1246         ret = -ENOMEM;
1247         state = nfs4_get_open_state(inode, data->owner);
1248         if (state == NULL)
1249                 goto err;
1250
1251         ret = nfs_refresh_inode(inode, &data->f_attr);
1252         if (ret)
1253                 goto err;
1254
1255         nfs_setsecurity(inode, &data->f_attr, data->f_label);
1256
1257         if (data->o_res.delegation_type != 0)
1258                 nfs4_opendata_check_deleg(data, state);
1259         update_open_stateid(state, &data->o_res.stateid, NULL,
1260                             data->o_arg.fmode);
1261
1262         return state;
1263 err:
1264         return ERR_PTR(ret);
1265
1266 }
1267
1268 static struct nfs4_state *
1269 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1270 {
1271         struct inode *inode;
1272         struct nfs4_state *state = NULL;
1273         int ret;
1274
1275         if (!data->rpc_done) {
1276                 state = nfs4_try_open_cached(data);
1277                 goto out;
1278         }
1279
1280         ret = -EAGAIN;
1281         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1282                 goto err;
1283         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1284         ret = PTR_ERR(inode);
1285         if (IS_ERR(inode))
1286                 goto err;
1287         ret = -ENOMEM;
1288         state = nfs4_get_open_state(inode, data->owner);
1289         if (state == NULL)
1290                 goto err_put_inode;
1291         if (data->o_res.delegation_type != 0)
1292                 nfs4_opendata_check_deleg(data, state);
1293         update_open_stateid(state, &data->o_res.stateid, NULL,
1294                         data->o_arg.fmode);
1295         iput(inode);
1296 out:
1297         nfs_release_seqid(data->o_arg.seqid);
1298         return state;
1299 err_put_inode:
1300         iput(inode);
1301 err:
1302         return ERR_PTR(ret);
1303 }
1304
1305 static struct nfs4_state *
1306 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1307 {
1308         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1309                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1310         return _nfs4_opendata_to_nfs4_state(data);
1311 }
1312
1313 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1314 {
1315         struct nfs_inode *nfsi = NFS_I(state->inode);
1316         struct nfs_open_context *ctx;
1317
1318         spin_lock(&state->inode->i_lock);
1319         list_for_each_entry(ctx, &nfsi->open_files, list) {
1320                 if (ctx->state != state)
1321                         continue;
1322                 get_nfs_open_context(ctx);
1323                 spin_unlock(&state->inode->i_lock);
1324                 return ctx;
1325         }
1326         spin_unlock(&state->inode->i_lock);
1327         return ERR_PTR(-ENOENT);
1328 }
1329
1330 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1331                 struct nfs4_state *state, enum open_claim_type4 claim)
1332 {
1333         struct nfs4_opendata *opendata;
1334
1335         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1336                         NULL, NULL, claim, GFP_NOFS);
1337         if (opendata == NULL)
1338                 return ERR_PTR(-ENOMEM);
1339         opendata->state = state;
1340         atomic_inc(&state->count);
1341         return opendata;
1342 }
1343
1344 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1345 {
1346         struct nfs4_state *newstate;
1347         int ret;
1348
1349         opendata->o_arg.open_flags = 0;
1350         opendata->o_arg.fmode = fmode;
1351         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1352         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1353         nfs4_init_opendata_res(opendata);
1354         ret = _nfs4_recover_proc_open(opendata);
1355         if (ret != 0)
1356                 return ret;
1357         newstate = nfs4_opendata_to_nfs4_state(opendata);
1358         if (IS_ERR(newstate))
1359                 return PTR_ERR(newstate);
1360         nfs4_close_state(newstate, fmode);
1361         *res = newstate;
1362         return 0;
1363 }
1364
1365 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1366 {
1367         struct nfs4_state *newstate;
1368         int ret;
1369
1370         /* memory barrier prior to reading state->n_* */
1371         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1372         clear_bit(NFS_OPEN_STATE, &state->flags);
1373         smp_rmb();
1374         if (state->n_rdwr != 0) {
1375                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1376                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1377                 if (ret != 0)
1378                         return ret;
1379                 if (newstate != state)
1380                         return -ESTALE;
1381         }
1382         if (state->n_wronly != 0) {
1383                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1384                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1385                 if (ret != 0)
1386                         return ret;
1387                 if (newstate != state)
1388                         return -ESTALE;
1389         }
1390         if (state->n_rdonly != 0) {
1391                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1392                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1393                 if (ret != 0)
1394                         return ret;
1395                 if (newstate != state)
1396                         return -ESTALE;
1397         }
1398         /*
1399          * We may have performed cached opens for all three recoveries.
1400          * Check if we need to update the current stateid.
1401          */
1402         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1403             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1404                 write_seqlock(&state->seqlock);
1405                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1406                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1407                 write_sequnlock(&state->seqlock);
1408         }
1409         return 0;
1410 }
1411
1412 /*
1413  * OPEN_RECLAIM:
1414  *      reclaim state on the server after a reboot.
1415  */
1416 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1417 {
1418         struct nfs_delegation *delegation;
1419         struct nfs4_opendata *opendata;
1420         fmode_t delegation_type = 0;
1421         int status;
1422
1423         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1424                         NFS4_OPEN_CLAIM_PREVIOUS);
1425         if (IS_ERR(opendata))
1426                 return PTR_ERR(opendata);
1427         rcu_read_lock();
1428         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1429         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1430                 delegation_type = delegation->type;
1431         rcu_read_unlock();
1432         opendata->o_arg.u.delegation_type = delegation_type;
1433         status = nfs4_open_recover(opendata, state);
1434         nfs4_opendata_put(opendata);
1435         return status;
1436 }
1437
1438 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1439 {
1440         struct nfs_server *server = NFS_SERVER(state->inode);
1441         struct nfs4_exception exception = { };
1442         int err;
1443         do {
1444                 err = _nfs4_do_open_reclaim(ctx, state);
1445                 trace_nfs4_open_reclaim(ctx, 0, err);
1446                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1447                         continue;
1448                 if (err != -NFS4ERR_DELAY)
1449                         break;
1450                 nfs4_handle_exception(server, err, &exception);
1451         } while (exception.retry);
1452         return err;
1453 }
1454
1455 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1456 {
1457         struct nfs_open_context *ctx;
1458         int ret;
1459
1460         ctx = nfs4_state_find_open_context(state);
1461         if (IS_ERR(ctx))
1462                 return -EAGAIN;
1463         ret = nfs4_do_open_reclaim(ctx, state);
1464         put_nfs_open_context(ctx);
1465         return ret;
1466 }
1467
1468 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1469 {
1470         switch (err) {
1471                 default:
1472                         printk(KERN_ERR "NFS: %s: unhandled error "
1473                                         "%d.\n", __func__, err);
1474                 case 0:
1475                 case -ENOENT:
1476                 case -ESTALE:
1477                         break;
1478                 case -NFS4ERR_BADSESSION:
1479                 case -NFS4ERR_BADSLOT:
1480                 case -NFS4ERR_BAD_HIGH_SLOT:
1481                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1482                 case -NFS4ERR_DEADSESSION:
1483                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1484                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1485                         return -EAGAIN;
1486                 case -NFS4ERR_STALE_CLIENTID:
1487                 case -NFS4ERR_STALE_STATEID:
1488                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1489                 case -NFS4ERR_EXPIRED:
1490                         /* Don't recall a delegation if it was lost */
1491                         nfs4_schedule_lease_recovery(server->nfs_client);
1492                         return -EAGAIN;
1493                 case -NFS4ERR_DELEG_REVOKED:
1494                 case -NFS4ERR_ADMIN_REVOKED:
1495                 case -NFS4ERR_BAD_STATEID:
1496                 case -NFS4ERR_OPENMODE:
1497                         nfs_inode_find_state_and_recover(state->inode,
1498                                         stateid);
1499                         nfs4_schedule_stateid_recovery(server, state);
1500                         return 0;
1501                 case -NFS4ERR_DELAY:
1502                 case -NFS4ERR_GRACE:
1503                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1504                         ssleep(1);
1505                         return -EAGAIN;
1506                 case -ENOMEM:
1507                 case -NFS4ERR_DENIED:
1508                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1509                         return 0;
1510         }
1511         return err;
1512 }
1513
1514 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1515 {
1516         struct nfs_server *server = NFS_SERVER(state->inode);
1517         struct nfs4_opendata *opendata;
1518         int err;
1519
1520         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1521                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1522         if (IS_ERR(opendata))
1523                 return PTR_ERR(opendata);
1524         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1525         err = nfs4_open_recover(opendata, state);
1526         nfs4_opendata_put(opendata);
1527         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1528 }
1529
1530 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1531 {
1532         struct nfs4_opendata *data = calldata;
1533
1534         data->rpc_status = task->tk_status;
1535         if (data->rpc_status == 0) {
1536                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1537                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1538                 renew_lease(data->o_res.server, data->timestamp);
1539                 data->rpc_done = 1;
1540         }
1541 }
1542
1543 static void nfs4_open_confirm_release(void *calldata)
1544 {
1545         struct nfs4_opendata *data = calldata;
1546         struct nfs4_state *state = NULL;
1547
1548         /* If this request hasn't been cancelled, do nothing */
1549         if (data->cancelled == 0)
1550                 goto out_free;
1551         /* In case of error, no cleanup! */
1552         if (!data->rpc_done)
1553                 goto out_free;
1554         state = nfs4_opendata_to_nfs4_state(data);
1555         if (!IS_ERR(state))
1556                 nfs4_close_state(state, data->o_arg.fmode);
1557 out_free:
1558         nfs4_opendata_put(data);
1559 }
1560
1561 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1562         .rpc_call_done = nfs4_open_confirm_done,
1563         .rpc_release = nfs4_open_confirm_release,
1564 };
1565
1566 /*
1567  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1568  */
1569 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1570 {
1571         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1572         struct rpc_task *task;
1573         struct  rpc_message msg = {
1574                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1575                 .rpc_argp = &data->c_arg,
1576                 .rpc_resp = &data->c_res,
1577                 .rpc_cred = data->owner->so_cred,
1578         };
1579         struct rpc_task_setup task_setup_data = {
1580                 .rpc_client = server->client,
1581                 .rpc_message = &msg,
1582                 .callback_ops = &nfs4_open_confirm_ops,
1583                 .callback_data = data,
1584                 .workqueue = nfsiod_workqueue,
1585                 .flags = RPC_TASK_ASYNC,
1586         };
1587         int status;
1588
1589         kref_get(&data->kref);
1590         data->rpc_done = 0;
1591         data->rpc_status = 0;
1592         data->timestamp = jiffies;
1593         task = rpc_run_task(&task_setup_data);
1594         if (IS_ERR(task))
1595                 return PTR_ERR(task);
1596         status = nfs4_wait_for_completion_rpc_task(task);
1597         if (status != 0) {
1598                 data->cancelled = 1;
1599                 smp_wmb();
1600         } else
1601                 status = data->rpc_status;
1602         rpc_put_task(task);
1603         return status;
1604 }
1605
1606 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1607 {
1608         struct nfs4_opendata *data = calldata;
1609         struct nfs4_state_owner *sp = data->owner;
1610         struct nfs_client *clp = sp->so_server->nfs_client;
1611
1612         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1613                 goto out_wait;
1614         /*
1615          * Check if we still need to send an OPEN call, or if we can use
1616          * a delegation instead.
1617          */
1618         if (data->state != NULL) {
1619                 struct nfs_delegation *delegation;
1620
1621                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1622                         goto out_no_action;
1623                 rcu_read_lock();
1624                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1625                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1626                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1627                     can_open_delegated(delegation, data->o_arg.fmode))
1628                         goto unlock_no_action;
1629                 rcu_read_unlock();
1630         }
1631         /* Update client id. */
1632         data->o_arg.clientid = clp->cl_clientid;
1633         switch (data->o_arg.claim) {
1634         case NFS4_OPEN_CLAIM_PREVIOUS:
1635         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1636         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1637                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1638         case NFS4_OPEN_CLAIM_FH:
1639                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1640                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1641         }
1642         data->timestamp = jiffies;
1643         if (nfs4_setup_sequence(data->o_arg.server,
1644                                 &data->o_arg.seq_args,
1645                                 &data->o_res.seq_res,
1646                                 task) != 0)
1647                 nfs_release_seqid(data->o_arg.seqid);
1648
1649         /* Set the create mode (note dependency on the session type) */
1650         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1651         if (data->o_arg.open_flags & O_EXCL) {
1652                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1653                 if (nfs4_has_persistent_session(clp))
1654                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1655                 else if (clp->cl_mvops->minor_version > 0)
1656                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1657         }
1658         return;
1659 unlock_no_action:
1660         rcu_read_unlock();
1661 out_no_action:
1662         task->tk_action = NULL;
1663 out_wait:
1664         nfs4_sequence_done(task, &data->o_res.seq_res);
1665 }
1666
1667 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1668 {
1669         struct nfs4_opendata *data = calldata;
1670
1671         data->rpc_status = task->tk_status;
1672
1673         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1674                 return;
1675
1676         if (task->tk_status == 0) {
1677                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1678                         switch (data->o_res.f_attr->mode & S_IFMT) {
1679                         case S_IFREG:
1680                                 break;
1681                         case S_IFLNK:
1682                                 data->rpc_status = -ELOOP;
1683                                 break;
1684                         case S_IFDIR:
1685                                 data->rpc_status = -EISDIR;
1686                                 break;
1687                         default:
1688                                 data->rpc_status = -ENOTDIR;
1689                         }
1690                 }
1691                 renew_lease(data->o_res.server, data->timestamp);
1692                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1693                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1694         }
1695         data->rpc_done = 1;
1696 }
1697
1698 static void nfs4_open_release(void *calldata)
1699 {
1700         struct nfs4_opendata *data = calldata;
1701         struct nfs4_state *state = NULL;
1702
1703         /* If this request hasn't been cancelled, do nothing */
1704         if (data->cancelled == 0)
1705                 goto out_free;
1706         /* In case of error, no cleanup! */
1707         if (data->rpc_status != 0 || !data->rpc_done)
1708                 goto out_free;
1709         /* In case we need an open_confirm, no cleanup! */
1710         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1711                 goto out_free;
1712         state = nfs4_opendata_to_nfs4_state(data);
1713         if (!IS_ERR(state))
1714                 nfs4_close_state(state, data->o_arg.fmode);
1715 out_free:
1716         nfs4_opendata_put(data);
1717 }
1718
1719 static const struct rpc_call_ops nfs4_open_ops = {
1720         .rpc_call_prepare = nfs4_open_prepare,
1721         .rpc_call_done = nfs4_open_done,
1722         .rpc_release = nfs4_open_release,
1723 };
1724
1725 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1726 {
1727         struct inode *dir = data->dir->d_inode;
1728         struct nfs_server *server = NFS_SERVER(dir);
1729         struct nfs_openargs *o_arg = &data->o_arg;
1730         struct nfs_openres *o_res = &data->o_res;
1731         struct rpc_task *task;
1732         struct rpc_message msg = {
1733                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1734                 .rpc_argp = o_arg,
1735                 .rpc_resp = o_res,
1736                 .rpc_cred = data->owner->so_cred,
1737         };
1738         struct rpc_task_setup task_setup_data = {
1739                 .rpc_client = server->client,
1740                 .rpc_message = &msg,
1741                 .callback_ops = &nfs4_open_ops,
1742                 .callback_data = data,
1743                 .workqueue = nfsiod_workqueue,
1744                 .flags = RPC_TASK_ASYNC,
1745         };
1746         int status;
1747
1748         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1749         kref_get(&data->kref);
1750         data->rpc_done = 0;
1751         data->rpc_status = 0;
1752         data->cancelled = 0;
1753         data->is_recover = 0;
1754         if (isrecover) {
1755                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1756                 data->is_recover = 1;
1757         }
1758         task = rpc_run_task(&task_setup_data);
1759         if (IS_ERR(task))
1760                 return PTR_ERR(task);
1761         status = nfs4_wait_for_completion_rpc_task(task);
1762         if (status != 0) {
1763                 data->cancelled = 1;
1764                 smp_wmb();
1765         } else
1766                 status = data->rpc_status;
1767         rpc_put_task(task);
1768
1769         return status;
1770 }
1771
1772 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1773 {
1774         struct inode *dir = data->dir->d_inode;
1775         struct nfs_openres *o_res = &data->o_res;
1776         int status;
1777
1778         status = nfs4_run_open_task(data, 1);
1779         if (status != 0 || !data->rpc_done)
1780                 return status;
1781
1782         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1783
1784         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1785                 status = _nfs4_proc_open_confirm(data);
1786                 if (status != 0)
1787                         return status;
1788         }
1789
1790         return status;
1791 }
1792
1793 static int nfs4_opendata_access(struct rpc_cred *cred,
1794                                 struct nfs4_opendata *opendata,
1795                                 struct nfs4_state *state, fmode_t fmode,
1796                                 int openflags)
1797 {
1798         struct nfs_access_entry cache;
1799         u32 mask;
1800
1801         /* access call failed or for some reason the server doesn't
1802          * support any access modes -- defer access call until later */
1803         if (opendata->o_res.access_supported == 0)
1804                 return 0;
1805
1806         mask = 0;
1807         /* don't check MAY_WRITE - a newly created file may not have
1808          * write mode bits, but POSIX allows the creating process to write.
1809          * use openflags to check for exec, because fmode won't
1810          * always have FMODE_EXEC set when file open for exec. */
1811         if (openflags & __FMODE_EXEC) {
1812                 /* ONLY check for exec rights */
1813                 mask = MAY_EXEC;
1814         } else if (fmode & FMODE_READ)
1815                 mask = MAY_READ;
1816
1817         cache.cred = cred;
1818         cache.jiffies = jiffies;
1819         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1820         nfs_access_add_cache(state->inode, &cache);
1821
1822         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1823                 return 0;
1824
1825         /* even though OPEN succeeded, access is denied. Close the file */
1826         nfs4_close_state(state, fmode);
1827         return -EACCES;
1828 }
1829
1830 /*
1831  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1832  */
1833 static int _nfs4_proc_open(struct nfs4_opendata *data)
1834 {
1835         struct inode *dir = data->dir->d_inode;
1836         struct nfs_server *server = NFS_SERVER(dir);
1837         struct nfs_openargs *o_arg = &data->o_arg;
1838         struct nfs_openres *o_res = &data->o_res;
1839         int status;
1840
1841         status = nfs4_run_open_task(data, 0);
1842         if (!data->rpc_done)
1843                 return status;
1844         if (status != 0) {
1845                 if (status == -NFS4ERR_BADNAME &&
1846                                 !(o_arg->open_flags & O_CREAT))
1847                         return -ENOENT;
1848                 return status;
1849         }
1850
1851         nfs_fattr_map_and_free_names(server, &data->f_attr);
1852
1853         if (o_arg->open_flags & O_CREAT)
1854                 update_changeattr(dir, &o_res->cinfo);
1855         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1856                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1857         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1858                 status = _nfs4_proc_open_confirm(data);
1859                 if (status != 0)
1860                         return status;
1861         }
1862         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1863                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
1864         return 0;
1865 }
1866
1867 static int nfs4_recover_expired_lease(struct nfs_server *server)
1868 {
1869         return nfs4_client_recover_expired_lease(server->nfs_client);
1870 }
1871
1872 /*
1873  * OPEN_EXPIRED:
1874  *      reclaim state on the server after a network partition.
1875  *      Assumes caller holds the appropriate lock
1876  */
1877 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1878 {
1879         struct nfs4_opendata *opendata;
1880         int ret;
1881
1882         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1883                         NFS4_OPEN_CLAIM_FH);
1884         if (IS_ERR(opendata))
1885                 return PTR_ERR(opendata);
1886         ret = nfs4_open_recover(opendata, state);
1887         if (ret == -ESTALE)
1888                 d_drop(ctx->dentry);
1889         nfs4_opendata_put(opendata);
1890         return ret;
1891 }
1892
1893 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1894 {
1895         struct nfs_server *server = NFS_SERVER(state->inode);
1896         struct nfs4_exception exception = { };
1897         int err;
1898
1899         do {
1900                 err = _nfs4_open_expired(ctx, state);
1901                 trace_nfs4_open_expired(ctx, 0, err);
1902                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1903                         continue;
1904                 switch (err) {
1905                 default:
1906                         goto out;
1907                 case -NFS4ERR_GRACE:
1908                 case -NFS4ERR_DELAY:
1909                         nfs4_handle_exception(server, err, &exception);
1910                         err = 0;
1911                 }
1912         } while (exception.retry);
1913 out:
1914         return err;
1915 }
1916
1917 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1918 {
1919         struct nfs_open_context *ctx;
1920         int ret;
1921
1922         ctx = nfs4_state_find_open_context(state);
1923         if (IS_ERR(ctx))
1924                 return -EAGAIN;
1925         ret = nfs4_do_open_expired(ctx, state);
1926         put_nfs_open_context(ctx);
1927         return ret;
1928 }
1929
1930 #if defined(CONFIG_NFS_V4_1)
1931 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1932 {
1933         struct nfs_server *server = NFS_SERVER(state->inode);
1934         nfs4_stateid *stateid = &state->stateid;
1935         struct nfs_delegation *delegation;
1936         struct rpc_cred *cred = NULL;
1937         int status = -NFS4ERR_BAD_STATEID;
1938
1939         /* If a state reset has been done, test_stateid is unneeded */
1940         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1941                 return;
1942
1943         /* Get the delegation credential for use by test/free_stateid */
1944         rcu_read_lock();
1945         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1946         if (delegation != NULL &&
1947             nfs4_stateid_match(&delegation->stateid, stateid)) {
1948                 cred = get_rpccred(delegation->cred);
1949                 rcu_read_unlock();
1950                 status = nfs41_test_stateid(server, stateid, cred);
1951         } else
1952                 rcu_read_unlock();
1953
1954         if (status != NFS_OK) {
1955                 /* Free the stateid unless the server explicitly
1956                  * informs us the stateid is unrecognized. */
1957                 if (status != -NFS4ERR_BAD_STATEID)
1958                         nfs41_free_stateid(server, stateid, cred);
1959                 nfs_remove_bad_delegation(state->inode);
1960
1961                 write_seqlock(&state->seqlock);
1962                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1963                 write_sequnlock(&state->seqlock);
1964                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1965         }
1966
1967         if (cred != NULL)
1968                 put_rpccred(cred);
1969 }
1970
1971 /**
1972  * nfs41_check_open_stateid - possibly free an open stateid
1973  *
1974  * @state: NFSv4 state for an inode
1975  *
1976  * Returns NFS_OK if recovery for this stateid is now finished.
1977  * Otherwise a negative NFS4ERR value is returned.
1978  */
1979 static int nfs41_check_open_stateid(struct nfs4_state *state)
1980 {
1981         struct nfs_server *server = NFS_SERVER(state->inode);
1982         nfs4_stateid *stateid = &state->open_stateid;
1983         struct rpc_cred *cred = state->owner->so_cred;
1984         int status;
1985
1986         /* If a state reset has been done, test_stateid is unneeded */
1987         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1988             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1989             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1990                 return -NFS4ERR_BAD_STATEID;
1991
1992         status = nfs41_test_stateid(server, stateid, cred);
1993         if (status != NFS_OK) {
1994                 /* Free the stateid unless the server explicitly
1995                  * informs us the stateid is unrecognized. */
1996                 if (status != -NFS4ERR_BAD_STATEID)
1997                         nfs41_free_stateid(server, stateid, cred);
1998
1999                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2000                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2001                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2002                 clear_bit(NFS_OPEN_STATE, &state->flags);
2003         }
2004         return status;
2005 }
2006
2007 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2008 {
2009         int status;
2010
2011         nfs41_clear_delegation_stateid(state);
2012         status = nfs41_check_open_stateid(state);
2013         if (status != NFS_OK)
2014                 status = nfs4_open_expired(sp, state);
2015         return status;
2016 }
2017 #endif
2018
2019 /*
2020  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2021  * fields corresponding to attributes that were used to store the verifier.
2022  * Make sure we clobber those fields in the later setattr call
2023  */
2024 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2025 {
2026         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2027             !(sattr->ia_valid & ATTR_ATIME_SET))
2028                 sattr->ia_valid |= ATTR_ATIME;
2029
2030         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2031             !(sattr->ia_valid & ATTR_MTIME_SET))
2032                 sattr->ia_valid |= ATTR_MTIME;
2033 }
2034
2035 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2036                 fmode_t fmode,
2037                 int flags,
2038                 struct nfs_open_context *ctx)
2039 {
2040         struct nfs4_state_owner *sp = opendata->owner;
2041         struct nfs_server *server = sp->so_server;
2042         struct dentry *dentry;
2043         struct nfs4_state *state;
2044         unsigned int seq;
2045         int ret;
2046
2047         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2048
2049         ret = _nfs4_proc_open(opendata);
2050         if (ret != 0)
2051                 goto out;
2052
2053         state = nfs4_opendata_to_nfs4_state(opendata);
2054         ret = PTR_ERR(state);
2055         if (IS_ERR(state))
2056                 goto out;
2057         if (server->caps & NFS_CAP_POSIX_LOCK)
2058                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2059
2060         dentry = opendata->dentry;
2061         if (dentry->d_inode == NULL) {
2062                 /* FIXME: Is this d_drop() ever needed? */
2063                 d_drop(dentry);
2064                 dentry = d_add_unique(dentry, igrab(state->inode));
2065                 if (dentry == NULL) {
2066                         dentry = opendata->dentry;
2067                 } else if (dentry != ctx->dentry) {
2068                         dput(ctx->dentry);
2069                         ctx->dentry = dget(dentry);
2070                 }
2071                 nfs_set_verifier(dentry,
2072                                 nfs_save_change_attribute(opendata->dir->d_inode));
2073         }
2074
2075         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2076         if (ret != 0)
2077                 goto out;
2078
2079         ctx->state = state;
2080         if (dentry->d_inode == state->inode) {
2081                 nfs_inode_attach_open_context(ctx);
2082                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2083                         nfs4_schedule_stateid_recovery(server, state);
2084         }
2085 out:
2086         return ret;
2087 }
2088
2089 /*
2090  * Returns a referenced nfs4_state
2091  */
2092 static int _nfs4_do_open(struct inode *dir,
2093                         struct nfs_open_context *ctx,
2094                         int flags,
2095                         struct iattr *sattr,
2096                         struct nfs4_label *label)
2097 {
2098         struct nfs4_state_owner  *sp;
2099         struct nfs4_state     *state = NULL;
2100         struct nfs_server       *server = NFS_SERVER(dir);
2101         struct nfs4_opendata *opendata;
2102         struct dentry *dentry = ctx->dentry;
2103         struct rpc_cred *cred = ctx->cred;
2104         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2105         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2106         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2107         struct nfs4_label *olabel = NULL;
2108         int status;
2109
2110         /* Protect against reboot recovery conflicts */
2111         status = -ENOMEM;
2112         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2113         if (sp == NULL) {
2114                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2115                 goto out_err;
2116         }
2117         status = nfs4_recover_expired_lease(server);
2118         if (status != 0)
2119                 goto err_put_state_owner;
2120         if (dentry->d_inode != NULL)
2121                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2122         status = -ENOMEM;
2123         if (dentry->d_inode)
2124                 claim = NFS4_OPEN_CLAIM_FH;
2125         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2126                         label, claim, GFP_KERNEL);
2127         if (opendata == NULL)
2128                 goto err_put_state_owner;
2129
2130         if (label) {
2131                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2132                 if (IS_ERR(olabel)) {
2133                         status = PTR_ERR(olabel);
2134                         goto err_opendata_put;
2135                 }
2136         }
2137
2138         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2139                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2140                 if (!opendata->f_attr.mdsthreshold)
2141                         goto err_free_label;
2142                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2143         }
2144         if (dentry->d_inode != NULL)
2145                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2146
2147         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2148         if (status != 0)
2149                 goto err_free_label;
2150         state = ctx->state;
2151
2152         if ((opendata->o_arg.open_flags & O_EXCL) &&
2153             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2154                 nfs4_exclusive_attrset(opendata, sattr);
2155
2156                 nfs_fattr_init(opendata->o_res.f_attr);
2157                 status = nfs4_do_setattr(state->inode, cred,
2158                                 opendata->o_res.f_attr, sattr,
2159                                 state, label, olabel);
2160                 if (status == 0) {
2161                         nfs_setattr_update_inode(state->inode, sattr);
2162                         nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2163                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2164                 }
2165         }
2166
2167         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2168                 *ctx_th = opendata->f_attr.mdsthreshold;
2169         else
2170                 kfree(opendata->f_attr.mdsthreshold);
2171         opendata->f_attr.mdsthreshold = NULL;
2172
2173         nfs4_label_free(olabel);
2174
2175         nfs4_opendata_put(opendata);
2176         nfs4_put_state_owner(sp);
2177         return 0;
2178 err_free_label:
2179         nfs4_label_free(olabel);
2180 err_opendata_put:
2181         kfree(opendata->f_attr.mdsthreshold);
2182         nfs4_opendata_put(opendata);
2183 err_put_state_owner:
2184         nfs4_put_state_owner(sp);
2185 out_err:
2186         return status;
2187 }
2188
2189
2190 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2191                                         struct nfs_open_context *ctx,
2192                                         int flags,
2193                                         struct iattr *sattr,
2194                                         struct nfs4_label *label)
2195 {
2196         struct nfs_server *server = NFS_SERVER(dir);
2197         struct nfs4_exception exception = { };
2198         struct nfs4_state *res;
2199         int status;
2200
2201         do {
2202                 status = _nfs4_do_open(dir, ctx, flags, sattr, label);
2203                 res = ctx->state;
2204                 trace_nfs4_open_file(ctx, flags, status);
2205                 if (status == 0)
2206                         break;
2207                 /* NOTE: BAD_SEQID means the server and client disagree about the
2208                  * book-keeping w.r.t. state-changing operations
2209                  * (OPEN/CLOSE/LOCK/LOCKU...)
2210                  * It is actually a sign of a bug on the client or on the server.
2211                  *
2212                  * If we receive a BAD_SEQID error in the particular case of
2213                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2214                  * have unhashed the old state_owner for us, and that we can
2215                  * therefore safely retry using a new one. We should still warn
2216                  * the user though...
2217                  */
2218                 if (status == -NFS4ERR_BAD_SEQID) {
2219                         pr_warn_ratelimited("NFS: v4 server %s "
2220                                         " returned a bad sequence-id error!\n",
2221                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2222                         exception.retry = 1;
2223                         continue;
2224                 }
2225                 /*
2226                  * BAD_STATEID on OPEN means that the server cancelled our
2227                  * state before it received the OPEN_CONFIRM.
2228                  * Recover by retrying the request as per the discussion
2229                  * on Page 181 of RFC3530.
2230                  */
2231                 if (status == -NFS4ERR_BAD_STATEID) {
2232                         exception.retry = 1;
2233                         continue;
2234                 }
2235                 if (status == -EAGAIN) {
2236                         /* We must have found a delegation */
2237                         exception.retry = 1;
2238                         continue;
2239                 }
2240                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2241                         continue;
2242                 res = ERR_PTR(nfs4_handle_exception(server,
2243                                         status, &exception));
2244         } while (exception.retry);
2245         return res;
2246 }
2247
2248 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2249                             struct nfs_fattr *fattr, struct iattr *sattr,
2250                             struct nfs4_state *state, struct nfs4_label *ilabel,
2251                             struct nfs4_label *olabel)
2252 {
2253         struct nfs_server *server = NFS_SERVER(inode);
2254         struct nfs_setattrargs  arg = {
2255                 .fh             = NFS_FH(inode),
2256                 .iap            = sattr,
2257                 .server         = server,
2258                 .bitmask = server->attr_bitmask,
2259                 .label          = ilabel,
2260         };
2261         struct nfs_setattrres  res = {
2262                 .fattr          = fattr,
2263                 .label          = olabel,
2264                 .server         = server,
2265         };
2266         struct rpc_message msg = {
2267                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2268                 .rpc_argp       = &arg,
2269                 .rpc_resp       = &res,
2270                 .rpc_cred       = cred,
2271         };
2272         unsigned long timestamp = jiffies;
2273         fmode_t fmode;
2274         bool truncate;
2275         int status;
2276
2277         arg.bitmask = nfs4_bitmask(server, ilabel);
2278         if (ilabel)
2279                 arg.bitmask = nfs4_bitmask(server, olabel);
2280
2281         nfs_fattr_init(fattr);
2282
2283         /* Servers should only apply open mode checks for file size changes */
2284         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2285         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2286
2287         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2288                 /* Use that stateid */
2289         } else if (truncate && state != NULL && nfs4_valid_open_stateid(state)) {
2290                 struct nfs_lockowner lockowner = {
2291                         .l_owner = current->files,
2292                         .l_pid = current->tgid,
2293                 };
2294                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2295                                 &lockowner);
2296         } else
2297                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2298
2299         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2300         if (status == 0 && state != NULL)
2301                 renew_lease(server, timestamp);
2302         return status;
2303 }
2304
2305 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2306                            struct nfs_fattr *fattr, struct iattr *sattr,
2307                            struct nfs4_state *state, struct nfs4_label *ilabel,
2308                            struct nfs4_label *olabel)
2309 {
2310         struct nfs_server *server = NFS_SERVER(inode);
2311         struct nfs4_exception exception = {
2312                 .state = state,
2313                 .inode = inode,
2314         };
2315         int err;
2316         do {
2317                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2318                 switch (err) {
2319                 case -NFS4ERR_OPENMODE:
2320                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2321                                 pr_warn_once("NFSv4: server %s is incorrectly "
2322                                                 "applying open mode checks to "
2323                                                 "a SETATTR that is not "
2324                                                 "changing file size.\n",
2325                                                 server->nfs_client->cl_hostname);
2326                         }
2327                         if (state && !(state->state & FMODE_WRITE)) {
2328                                 err = -EBADF;
2329                                 if (sattr->ia_valid & ATTR_OPEN)
2330                                         err = -EACCES;
2331                                 goto out;
2332                         }
2333                 }
2334                 err = nfs4_handle_exception(server, err, &exception);
2335         } while (exception.retry);
2336 out:
2337         return err;
2338 }
2339
2340 struct nfs4_closedata {
2341         struct inode *inode;
2342         struct nfs4_state *state;
2343         struct nfs_closeargs arg;
2344         struct nfs_closeres res;
2345         struct nfs_fattr fattr;
2346         unsigned long timestamp;
2347         bool roc;
2348         u32 roc_barrier;
2349 };
2350
2351 static void nfs4_free_closedata(void *data)
2352 {
2353         struct nfs4_closedata *calldata = data;
2354         struct nfs4_state_owner *sp = calldata->state->owner;
2355         struct super_block *sb = calldata->state->inode->i_sb;
2356
2357         if (calldata->roc)
2358                 pnfs_roc_release(calldata->state->inode);
2359         nfs4_put_open_state(calldata->state);
2360         nfs_free_seqid(calldata->arg.seqid);
2361         nfs4_put_state_owner(sp);
2362         nfs_sb_deactive(sb);
2363         kfree(calldata);
2364 }
2365
2366 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2367                 fmode_t fmode)
2368 {
2369         spin_lock(&state->owner->so_lock);
2370         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2371         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
2372         case FMODE_WRITE:
2373                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2374                 break;
2375         case FMODE_READ:
2376                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2377                 break;
2378         case 0:
2379                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2380                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2381                 clear_bit(NFS_OPEN_STATE, &state->flags);
2382         }
2383         spin_unlock(&state->owner->so_lock);
2384 }
2385
2386 static void nfs4_close_done(struct rpc_task *task, void *data)
2387 {
2388         struct nfs4_closedata *calldata = data;
2389         struct nfs4_state *state = calldata->state;
2390         struct nfs_server *server = NFS_SERVER(calldata->inode);
2391
2392         dprintk("%s: begin!\n", __func__);
2393         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2394                 return;
2395         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2396         /* hmm. we are done with the inode, and in the process of freeing
2397          * the state_owner. we keep this around to process errors
2398          */
2399         switch (task->tk_status) {
2400                 case 0:
2401                         if (calldata->roc)
2402                                 pnfs_roc_set_barrier(state->inode,
2403                                                      calldata->roc_barrier);
2404                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2405                         renew_lease(server, calldata->timestamp);
2406                         nfs4_close_clear_stateid_flags(state,
2407                                         calldata->arg.fmode);
2408                         break;
2409                 case -NFS4ERR_STALE_STATEID:
2410                 case -NFS4ERR_OLD_STATEID:
2411                 case -NFS4ERR_BAD_STATEID:
2412                 case -NFS4ERR_EXPIRED:
2413                         if (calldata->arg.fmode == 0)
2414                                 break;
2415                 default:
2416                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2417                                 rpc_restart_call_prepare(task);
2418         }
2419         nfs_release_seqid(calldata->arg.seqid);
2420         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2421         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2422 }
2423
2424 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2425 {
2426         struct nfs4_closedata *calldata = data;
2427         struct nfs4_state *state = calldata->state;
2428         struct inode *inode = calldata->inode;
2429         int call_close = 0;
2430
2431         dprintk("%s: begin!\n", __func__);
2432         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2433                 goto out_wait;
2434
2435         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2436         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2437         spin_lock(&state->owner->so_lock);
2438         /* Calculate the change in open mode */
2439         if (state->n_rdwr == 0) {
2440                 if (state->n_rdonly == 0) {
2441                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2442                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2443                         calldata->arg.fmode &= ~FMODE_READ;
2444                 }
2445                 if (state->n_wronly == 0) {
2446                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2447                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2448                         calldata->arg.fmode &= ~FMODE_WRITE;
2449                 }
2450         }
2451         if (!nfs4_valid_open_stateid(state))
2452                 call_close = 0;
2453         spin_unlock(&state->owner->so_lock);
2454
2455         if (!call_close) {
2456                 /* Note: exit _without_ calling nfs4_close_done */
2457                 goto out_no_action;
2458         }
2459
2460         if (calldata->arg.fmode == 0) {
2461                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2462                 if (calldata->roc &&
2463                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2464                         nfs_release_seqid(calldata->arg.seqid);
2465                         goto out_wait;
2466                     }
2467         }
2468
2469         nfs_fattr_init(calldata->res.fattr);
2470         calldata->timestamp = jiffies;
2471         if (nfs4_setup_sequence(NFS_SERVER(inode),
2472                                 &calldata->arg.seq_args,
2473                                 &calldata->res.seq_res,
2474                                 task) != 0)
2475                 nfs_release_seqid(calldata->arg.seqid);
2476         dprintk("%s: done!\n", __func__);
2477         return;
2478 out_no_action:
2479         task->tk_action = NULL;
2480 out_wait:
2481         nfs4_sequence_done(task, &calldata->res.seq_res);
2482 }
2483
2484 static const struct rpc_call_ops nfs4_close_ops = {
2485         .rpc_call_prepare = nfs4_close_prepare,
2486         .rpc_call_done = nfs4_close_done,
2487         .rpc_release = nfs4_free_closedata,
2488 };
2489
2490 /*
2491  * It is possible for data to be read/written from a mem-mapped file
2492  * after the sys_close call (which hits the vfs layer as a flush).
2493  * This means that we can't safely call nfsv4 close on a file until
2494  * the inode is cleared. This in turn means that we are not good
2495  * NFSv4 citizens - we do not indicate to the server to update the file's
2496  * share state even when we are done with one of the three share
2497  * stateid's in the inode.
2498  *
2499  * NOTE: Caller must be holding the sp->so_owner semaphore!
2500  */
2501 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2502 {
2503         struct nfs_server *server = NFS_SERVER(state->inode);
2504         struct nfs4_closedata *calldata;
2505         struct nfs4_state_owner *sp = state->owner;
2506         struct rpc_task *task;
2507         struct rpc_message msg = {
2508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2509                 .rpc_cred = state->owner->so_cred,
2510         };
2511         struct rpc_task_setup task_setup_data = {
2512                 .rpc_client = server->client,
2513                 .rpc_message = &msg,
2514                 .callback_ops = &nfs4_close_ops,
2515                 .workqueue = nfsiod_workqueue,
2516                 .flags = RPC_TASK_ASYNC,
2517         };
2518         int status = -ENOMEM;
2519
2520         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2521         if (calldata == NULL)
2522                 goto out;
2523         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2524         calldata->inode = state->inode;
2525         calldata->state = state;
2526         calldata->arg.fh = NFS_FH(state->inode);
2527         calldata->arg.stateid = &state->open_stateid;
2528         /* Serialization for the sequence id */
2529         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2530         if (calldata->arg.seqid == NULL)
2531                 goto out_free_calldata;
2532         calldata->arg.fmode = 0;
2533         calldata->arg.bitmask = server->cache_consistency_bitmask;
2534         calldata->res.fattr = &calldata->fattr;
2535         calldata->res.seqid = calldata->arg.seqid;
2536         calldata->res.server = server;
2537         calldata->roc = pnfs_roc(state->inode);
2538         nfs_sb_active(calldata->inode->i_sb);
2539
2540         msg.rpc_argp = &calldata->arg;
2541         msg.rpc_resp = &calldata->res;
2542         task_setup_data.callback_data = calldata;
2543         task = rpc_run_task(&task_setup_data);
2544         if (IS_ERR(task))
2545                 return PTR_ERR(task);
2546         status = 0;
2547         if (wait)
2548                 status = rpc_wait_for_completion_task(task);
2549         rpc_put_task(task);
2550         return status;
2551 out_free_calldata:
2552         kfree(calldata);
2553 out:
2554         nfs4_put_open_state(state);
2555         nfs4_put_state_owner(sp);
2556         return status;
2557 }
2558
2559 static struct inode *
2560 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2561 {
2562         struct nfs4_state *state;
2563         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2564
2565         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2566
2567         /* Protect against concurrent sillydeletes */
2568         state = nfs4_do_open(dir, ctx, open_flags, attr, label);
2569
2570         nfs4_label_release_security(label);
2571
2572         if (IS_ERR(state))
2573                 return ERR_CAST(state);
2574         return state->inode;
2575 }
2576
2577 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2578 {
2579         if (ctx->state == NULL)
2580                 return;
2581         if (is_sync)
2582                 nfs4_close_sync(ctx->state, ctx->mode);
2583         else
2584                 nfs4_close_state(ctx->state, ctx->mode);
2585 }
2586
2587 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2588 {
2589         struct nfs4_server_caps_arg args = {
2590                 .fhandle = fhandle,
2591         };
2592         struct nfs4_server_caps_res res = {};
2593         struct rpc_message msg = {
2594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2595                 .rpc_argp = &args,
2596                 .rpc_resp = &res,
2597         };
2598         int status;
2599
2600         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2601         if (status == 0) {
2602                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2603                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2604                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2605                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2606                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2607                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2608                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2609                         server->caps |= NFS_CAP_ACLS;
2610                 if (res.has_links != 0)
2611                         server->caps |= NFS_CAP_HARDLINKS;
2612                 if (res.has_symlinks != 0)
2613                         server->caps |= NFS_CAP_SYMLINKS;
2614                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2615                         server->caps |= NFS_CAP_FILEID;
2616                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2617                         server->caps |= NFS_CAP_MODE;
2618                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2619                         server->caps |= NFS_CAP_NLINK;
2620                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2621                         server->caps |= NFS_CAP_OWNER;
2622                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2623                         server->caps |= NFS_CAP_OWNER_GROUP;
2624                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2625                         server->caps |= NFS_CAP_ATIME;
2626                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2627                         server->caps |= NFS_CAP_CTIME;
2628                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2629                         server->caps |= NFS_CAP_MTIME;
2630 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2631                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2632                         server->caps |= NFS_CAP_SECURITY_LABEL;
2633 #endif
2634                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2635                                 sizeof(server->attr_bitmask));
2636
2637                 if (server->caps & NFS_CAP_SECURITY_LABEL) {
2638                         server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2639                         res.attr_bitmask[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2640                 }
2641                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2642                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2643                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2644                 server->acl_bitmask = res.acl_bitmask;
2645                 server->fh_expire_type = res.fh_expire_type;
2646         }
2647
2648         return status;
2649 }
2650
2651 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2652 {
2653         struct nfs4_exception exception = { };
2654         int err;
2655         do {
2656                 err = nfs4_handle_exception(server,
2657                                 _nfs4_server_capabilities(server, fhandle),
2658                                 &exception);
2659         } while (exception.retry);
2660         return err;
2661 }
2662
2663 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2664                 struct nfs_fsinfo *info)
2665 {
2666         u32 bitmask[3];
2667         struct nfs4_lookup_root_arg args = {
2668                 .bitmask = bitmask,
2669         };
2670         struct nfs4_lookup_res res = {
2671                 .server = server,
2672                 .fattr = info->fattr,
2673                 .fh = fhandle,
2674         };
2675         struct rpc_message msg = {
2676                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2677                 .rpc_argp = &args,
2678                 .rpc_resp = &res,
2679         };
2680
2681         bitmask[0] = nfs4_fattr_bitmap[0];
2682         bitmask[1] = nfs4_fattr_bitmap[1];
2683         /*
2684          * Process the label in the upcoming getfattr
2685          */
2686         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2687
2688         nfs_fattr_init(info->fattr);
2689         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2690 }
2691
2692 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2693                 struct nfs_fsinfo *info)
2694 {
2695         struct nfs4_exception exception = { };
2696         int err;
2697         do {
2698                 err = _nfs4_lookup_root(server, fhandle, info);
2699                 switch (err) {
2700                 case 0:
2701                 case -NFS4ERR_WRONGSEC:
2702                         goto out;
2703                 default:
2704                         err = nfs4_handle_exception(server, err, &exception);
2705                 }
2706         } while (exception.retry);
2707 out:
2708         return err;
2709 }
2710
2711 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2712                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2713 {
2714         struct rpc_auth *auth;
2715         int ret;
2716
2717         auth = rpcauth_create(flavor, server->client);
2718         if (IS_ERR(auth)) {
2719                 ret = -EACCES;
2720                 goto out;
2721         }
2722         ret = nfs4_lookup_root(server, fhandle, info);
2723 out:
2724         return ret;
2725 }
2726
2727 /*
2728  * Retry pseudoroot lookup with various security flavors.  We do this when:
2729  *
2730  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2731  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2732  *
2733  * Returns zero on success, or a negative NFS4ERR value, or a
2734  * negative errno value.
2735  */
2736 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2737                               struct nfs_fsinfo *info)
2738 {
2739         /* Per 3530bis 15.33.5 */
2740         static const rpc_authflavor_t flav_array[] = {
2741                 RPC_AUTH_GSS_KRB5P,
2742                 RPC_AUTH_GSS_KRB5I,
2743                 RPC_AUTH_GSS_KRB5,
2744                 RPC_AUTH_UNIX,                  /* courtesy */
2745                 RPC_AUTH_NULL,
2746         };
2747         int status = -EPERM;
2748         size_t i;
2749
2750         for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2751                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2752                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2753                         continue;
2754                 break;
2755         }
2756
2757         /*
2758          * -EACCESS could mean that the user doesn't have correct permissions
2759          * to access the mount.  It could also mean that we tried to mount
2760          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2761          * existing mount programs don't handle -EACCES very well so it should
2762          * be mapped to -EPERM instead.
2763          */
2764         if (status == -EACCES)
2765                 status = -EPERM;
2766         return status;
2767 }
2768
2769 static int nfs4_do_find_root_sec(struct nfs_server *server,
2770                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2771 {
2772         int mv = server->nfs_client->cl_minorversion;
2773         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2774 }
2775
2776 /**
2777  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2778  * @server: initialized nfs_server handle
2779  * @fhandle: we fill in the pseudo-fs root file handle
2780  * @info: we fill in an FSINFO struct
2781  *
2782  * Returns zero on success, or a negative errno.
2783  */
2784 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2785                          struct nfs_fsinfo *info)
2786 {
2787         int status;
2788
2789         status = nfs4_lookup_root(server, fhandle, info);
2790         if ((status == -NFS4ERR_WRONGSEC) &&
2791             !(server->flags & NFS_MOUNT_SECFLAVOUR))
2792                 status = nfs4_do_find_root_sec(server, fhandle, info);
2793
2794         if (status == 0)
2795                 status = nfs4_server_capabilities(server, fhandle);
2796         if (status == 0)
2797                 status = nfs4_do_fsinfo(server, fhandle, info);
2798
2799         return nfs4_map_errors(status);
2800 }
2801
2802 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2803                               struct nfs_fsinfo *info)
2804 {
2805         int error;
2806         struct nfs_fattr *fattr = info->fattr;
2807         struct nfs4_label *label = NULL;
2808
2809         error = nfs4_server_capabilities(server, mntfh);
2810         if (error < 0) {
2811                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2812                 return error;
2813         }
2814
2815         label = nfs4_label_alloc(server, GFP_KERNEL);
2816         if (IS_ERR(label))
2817                 return PTR_ERR(label);
2818
2819         error = nfs4_proc_getattr(server, mntfh, fattr, label);
2820         if (error < 0) {
2821                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2822                 goto err_free_label;
2823         }
2824
2825         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2826             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2827                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2828
2829 err_free_label:
2830         nfs4_label_free(label);
2831
2832         return error;
2833 }
2834
2835 /*
2836  * Get locations and (maybe) other attributes of a referral.
2837  * Note that we'll actually follow the referral later when
2838  * we detect fsid mismatch in inode revalidation
2839  */
2840 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2841                              const struct qstr *name, struct nfs_fattr *fattr,
2842                              struct nfs_fh *fhandle)
2843 {
2844         int status = -ENOMEM;
2845         struct page *page = NULL;
2846         struct nfs4_fs_locations *locations = NULL;
2847
2848         page = alloc_page(GFP_KERNEL);
2849         if (page == NULL)
2850                 goto out;
2851         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2852         if (locations == NULL)
2853                 goto out;
2854
2855         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2856         if (status != 0)
2857                 goto out;
2858         /* Make sure server returned a different fsid for the referral */
2859         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2860                 dprintk("%s: server did not return a different fsid for"
2861                         " a referral at %s\n", __func__, name->name);
2862                 status = -EIO;
2863                 goto out;
2864         }
2865         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2866         nfs_fixup_referral_attributes(&locations->fattr);
2867
2868         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2869         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2870         memset(fhandle, 0, sizeof(struct nfs_fh));
2871 out:
2872         if (page)
2873                 __free_page(page);
2874         kfree(locations);
2875         return status;
2876 }
2877
2878 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
2879                                 struct nfs_fattr *fattr, struct nfs4_label *label)
2880 {
2881         struct nfs4_getattr_arg args = {
2882                 .fh = fhandle,
2883                 .bitmask = server->attr_bitmask,
2884         };
2885         struct nfs4_getattr_res res = {
2886                 .fattr = fattr,
2887                 .label = label,
2888                 .server = server,
2889         };
2890         struct rpc_message msg = {
2891                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2892                 .rpc_argp = &args,
2893                 .rpc_resp = &res,
2894         };
2895
2896         args.bitmask = nfs4_bitmask(server, label);
2897
2898         nfs_fattr_init(fattr);
2899         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2900 }
2901
2902 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
2903                                 struct nfs_fattr *fattr, struct nfs4_label *label)
2904 {
2905         struct nfs4_exception exception = { };
2906         int err;
2907         do {
2908                 err = nfs4_handle_exception(server,
2909                                 _nfs4_proc_getattr(server, fhandle, fattr, label),
2910                                 &exception);
2911         } while (exception.retry);
2912         return err;
2913 }
2914
2915 /*
2916  * The file is not closed if it is opened due to the a request to change
2917  * the size of the file. The open call will not be needed once the
2918  * VFS layer lookup-intents are implemented.
2919  *
2920  * Close is called when the inode is destroyed.
2921  * If we haven't opened the file for O_WRONLY, we
2922  * need to in the size_change case to obtain a stateid.
2923  *
2924  * Got race?
2925  * Because OPEN is always done by name in nfsv4, it is
2926  * possible that we opened a different file by the same
2927  * name.  We can recognize this race condition, but we
2928  * can't do anything about it besides returning an error.
2929  *
2930  * This will be fixed with VFS changes (lookup-intent).
2931  */
2932 static int
2933 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2934                   struct iattr *sattr)
2935 {
2936         struct inode *inode = dentry->d_inode;
2937         struct rpc_cred *cred = NULL;
2938         struct nfs4_state *state = NULL;
2939         struct nfs4_label *label = NULL;
2940         int status;
2941
2942         if (pnfs_ld_layoutret_on_setattr(inode))
2943                 pnfs_commit_and_return_layout(inode);
2944
2945         nfs_fattr_init(fattr);
2946
2947         /* Deal with open(O_TRUNC) */
2948         if (sattr->ia_valid & ATTR_OPEN)
2949                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
2950
2951         /* Optimization: if the end result is no change, don't RPC */
2952         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
2953                 return 0;
2954
2955         /* Search for an existing open(O_WRITE) file */
2956         if (sattr->ia_valid & ATTR_FILE) {
2957                 struct nfs_open_context *ctx;
2958
2959                 ctx = nfs_file_open_context(sattr->ia_file);
2960                 if (ctx) {
2961                         cred = ctx->cred;
2962                         state = ctx->state;
2963                 }
2964         }
2965
2966         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
2967         if (IS_ERR(label))
2968                 return PTR_ERR(label);
2969
2970         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
2971         if (status == 0) {
2972                 nfs_setattr_update_inode(inode, sattr);
2973                 nfs_setsecurity(inode, fattr, label);
2974         }
2975         nfs4_label_free(label);
2976         return status;
2977 }
2978
2979 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2980                 const struct qstr *name, struct nfs_fh *fhandle,
2981                 struct nfs_fattr *fattr, struct nfs4_label *label)
2982 {
2983         struct nfs_server *server = NFS_SERVER(dir);
2984         int                    status;
2985         struct nfs4_lookup_arg args = {
2986                 .bitmask = server->attr_bitmask,
2987                 .dir_fh = NFS_FH(dir),
2988                 .name = name,
2989         };
2990         struct nfs4_lookup_res res = {
2991                 .server = server,
2992                 .fattr = fattr,
2993                 .label = label,
2994                 .fh = fhandle,
2995         };
2996         struct rpc_message msg = {
2997                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2998                 .rpc_argp = &args,
2999                 .rpc_resp = &res,
3000         };
3001
3002         args.bitmask = nfs4_bitmask(server, label);
3003
3004         nfs_fattr_init(fattr);
3005
3006         dprintk("NFS call  lookup %s\n", name->name);
3007         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3008         dprintk("NFS reply lookup: %d\n", status);
3009         return status;
3010 }
3011
3012 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3013 {
3014         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3015                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3016         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3017         fattr->nlink = 2;
3018 }
3019
3020 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3021                                    struct qstr *name, struct nfs_fh *fhandle,
3022                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3023 {
3024         struct nfs4_exception exception = { };
3025         struct rpc_clnt *client = *clnt;
3026         int err;
3027         do {
3028                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3029                 switch (err) {
3030                 case -NFS4ERR_BADNAME:
3031                         err = -ENOENT;
3032                         goto out;
3033                 case -NFS4ERR_MOVED:
3034                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3035                         goto out;
3036                 case -NFS4ERR_WRONGSEC:
3037                         err = -EPERM;
3038                         if (client != *clnt)
3039                                 goto out;
3040
3041                         client = nfs4_create_sec_client(client, dir, name);
3042                         if (IS_ERR(client))
3043                                 return PTR_ERR(client);
3044
3045                         exception.retry = 1;
3046                         break;
3047                 default:
3048                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3049                 }
3050         } while (exception.retry);
3051
3052 out:
3053         if (err == 0)
3054                 *clnt = client;
3055         else if (client != *clnt)
3056                 rpc_shutdown_client(client);
3057
3058         return err;
3059 }
3060
3061 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3062                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3063                             struct nfs4_label *label)
3064 {
3065         int status;
3066         struct rpc_clnt *client = NFS_CLIENT(dir);
3067
3068         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3069         if (client != NFS_CLIENT(dir)) {
3070                 rpc_shutdown_client(client);
3071                 nfs_fixup_secinfo_attributes(fattr);
3072         }
3073         return status;
3074 }
3075
3076 struct rpc_clnt *
3077 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3078                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3079 {
3080         int status;
3081         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
3082
3083         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3084         if (status < 0) {
3085                 rpc_shutdown_client(client);
3086                 return ERR_PTR(status);
3087         }
3088         return client;
3089 }
3090
3091 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3092 {
3093         struct nfs_server *server = NFS_SERVER(inode);
3094         struct nfs4_accessargs args = {
3095                 .fh = NFS_FH(inode),
3096                 .bitmask = server->cache_consistency_bitmask,
3097         };
3098         struct nfs4_accessres res = {
3099                 .server = server,
3100         };
3101         struct rpc_message msg = {
3102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3103                 .rpc_argp = &args,
3104                 .rpc_resp = &res,
3105                 .rpc_cred = entry->cred,
3106         };
3107         int mode = entry->mask;
3108         int status = 0;
3109
3110         /*
3111          * Determine which access bits we want to ask for...
3112          */
3113         if (mode & MAY_READ)
3114                 args.access |= NFS4_ACCESS_READ;
3115         if (S_ISDIR(inode->i_mode)) {
3116                 if (mode & MAY_WRITE)
3117                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3118                 if (mode & MAY_EXEC)
3119                         args.access |= NFS4_ACCESS_LOOKUP;
3120         } else {
3121                 if (mode & MAY_WRITE)
3122                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3123                 if (mode & MAY_EXEC)
3124                         args.access |= NFS4_ACCESS_EXECUTE;
3125         }
3126
3127         res.fattr = nfs_alloc_fattr();
3128         if (res.fattr == NULL)
3129                 return -ENOMEM;
3130
3131         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3132         if (!status) {
3133                 nfs_access_set_mask(entry, res.access);
3134                 nfs_refresh_inode(inode, res.fattr);
3135         }
3136         nfs_free_fattr(res.fattr);
3137         return status;
3138 }
3139
3140 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3141 {
3142         struct nfs4_exception exception = { };
3143         int err;
3144         do {
3145                 err = nfs4_handle_exception(NFS_SERVER(inode),
3146                                 _nfs4_proc_access(inode, entry),
3147                                 &exception);
3148         } while (exception.retry);
3149         return err;
3150 }
3151
3152 /*
3153  * TODO: For the time being, we don't try to get any attributes
3154  * along with any of the zero-copy operations READ, READDIR,
3155  * READLINK, WRITE.
3156  *
3157  * In the case of the first three, we want to put the GETATTR
3158  * after the read-type operation -- this is because it is hard
3159  * to predict the length of a GETATTR response in v4, and thus
3160  * align the READ data correctly.  This means that the GETATTR
3161  * may end up partially falling into the page cache, and we should
3162  * shift it into the 'tail' of the xdr_buf before processing.
3163  * To do this efficiently, we need to know the total length
3164  * of data received, which doesn't seem to be available outside
3165  * of the RPC layer.
3166  *
3167  * In the case of WRITE, we also want to put the GETATTR after
3168  * the operation -- in this case because we want to make sure
3169  * we get the post-operation mtime and size.
3170  *
3171  * Both of these changes to the XDR layer would in fact be quite
3172  * minor, but I decided to leave them for a subsequent patch.
3173  */
3174 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3175                 unsigned int pgbase, unsigned int pglen)
3176 {
3177         struct nfs4_readlink args = {
3178                 .fh       = NFS_FH(inode),
3179                 .pgbase   = pgbase,
3180                 .pglen    = pglen,
3181                 .pages    = &page,
3182         };
3183         struct nfs4_readlink_res res;
3184         struct rpc_message msg = {
3185                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3186                 .rpc_argp = &args,
3187                 .rpc_resp = &res,
3188         };
3189
3190         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3191 }
3192
3193 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3194                 unsigned int pgbase, unsigned int pglen)
3195 {
3196         struct nfs4_exception exception = { };
3197         int err;
3198         do {
3199                 err = nfs4_handle_exception(NFS_SERVER(inode),
3200                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
3201                                 &exception);
3202         } while (exception.retry);
3203         return err;
3204 }
3205
3206 /*
3207  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3208  */
3209 static int
3210 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3211                  int flags)
3212 {
3213         struct nfs4_label l, *ilabel = NULL;
3214         struct nfs_open_context *ctx;
3215         struct nfs4_state *state;
3216         int status = 0;
3217
3218         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3219         if (IS_ERR(ctx))
3220                 return PTR_ERR(ctx);
3221
3222         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3223
3224         sattr->ia_mode &= ~current_umask();
3225         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel);
3226         if (IS_ERR(state)) {
3227                 status = PTR_ERR(state);
3228                 goto out;
3229         }
3230 out:
3231         nfs4_label_release_security(ilabel);
3232         put_nfs_open_context(ctx);
3233         return status;
3234 }
3235
3236 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3237 {
3238         struct nfs_server *server = NFS_SERVER(dir);
3239         struct nfs_removeargs args = {
3240                 .fh = NFS_FH(dir),
3241                 .name = *name,
3242         };
3243         struct nfs_removeres res = {
3244                 .server = server,
3245         };
3246         struct rpc_message msg = {
3247                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3248                 .rpc_argp = &args,
3249                 .rpc_resp = &res,
3250         };
3251         int status;
3252
3253         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3254         if (status == 0)
3255                 update_changeattr(dir, &res.cinfo);
3256         return status;
3257 }
3258
3259 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3260 {
3261         struct nfs4_exception exception = { };
3262         int err;
3263         do {
3264                 err = nfs4_handle_exception(NFS_SERVER(dir),
3265                                 _nfs4_proc_remove(dir, name),
3266                                 &exception);
3267         } while (exception.retry);
3268         return err;
3269 }
3270
3271 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3272 {
3273         struct nfs_server *server = NFS_SERVER(dir);
3274         struct nfs_removeargs *args = msg->rpc_argp;
3275         struct nfs_removeres *res = msg->rpc_resp;
3276
3277         res->server = server;
3278         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3279         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3280
3281         nfs_fattr_init(res->dir_attr);
3282 }
3283
3284 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3285 {
3286         nfs4_setup_sequence(NFS_SERVER(data->dir),
3287                         &data->args.seq_args,
3288                         &data->res.seq_res,
3289                         task);
3290 }
3291
3292 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3293 {
3294         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3295
3296         if (!nfs4_sequence_done(task, &res->seq_res))
3297                 return 0;
3298         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3299                 return 0;
3300         update_changeattr(dir, &res->cinfo);
3301         return 1;
3302 }
3303
3304 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3305 {
3306         struct nfs_server *server = NFS_SERVER(dir);
3307         struct nfs_renameargs *arg = msg->rpc_argp;
3308         struct nfs_renameres *res = msg->rpc_resp;
3309
3310         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3311         res->server = server;
3312         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3313 }
3314
3315 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3316 {
3317         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3318                         &data->args.seq_args,
3319                         &data->res.seq_res,
3320                         task);
3321 }
3322
3323 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3324                                  struct inode *new_dir)
3325 {
3326         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3327
3328         if (!nfs4_sequence_done(task, &res->seq_res))
3329                 return 0;
3330         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3331                 return 0;
3332
3333         update_changeattr(old_dir, &res->old_cinfo);
3334         update_changeattr(new_dir, &res->new_cinfo);
3335         return 1;
3336 }
3337
3338 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3339                 struct inode *new_dir, struct qstr *new_name)
3340 {
3341         struct nfs_server *server = NFS_SERVER(old_dir);
3342         struct nfs_renameargs arg = {
3343                 .old_dir = NFS_FH(old_dir),
3344                 .new_dir = NFS_FH(new_dir),
3345                 .old_name = old_name,
3346                 .new_name = new_name,
3347         };
3348         struct nfs_renameres res = {
3349                 .server = server,
3350         };
3351         struct rpc_message msg = {
3352                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3353                 .rpc_argp = &arg,
3354                 .rpc_resp = &res,
3355         };
3356         int status = -ENOMEM;
3357
3358         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3359         if (!status) {
3360                 update_changeattr(old_dir, &res.old_cinfo);
3361                 update_changeattr(new_dir, &res.new_cinfo);
3362         }
3363         return status;
3364 }
3365
3366 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3367                 struct inode *new_dir, struct qstr *new_name)
3368 {
3369         struct nfs4_exception exception = { };
3370         int err;
3371         do {
3372                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3373                                 _nfs4_proc_rename(old_dir, old_name,
3374                                         new_dir, new_name),
3375                                 &exception);
3376         } while (exception.retry);
3377         return err;
3378 }
3379
3380 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3381 {
3382         struct nfs_server *server = NFS_SERVER(inode);
3383         struct nfs4_link_arg arg = {
3384                 .fh     = NFS_FH(inode),
3385                 .dir_fh = NFS_FH(dir),
3386                 .name   = name,
3387                 .bitmask = server->attr_bitmask,
3388         };
3389         struct nfs4_link_res res = {
3390                 .server = server,
3391                 .label = NULL,
3392         };
3393         struct rpc_message msg = {
3394                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3395                 .rpc_argp = &arg,
3396                 .rpc_resp = &res,
3397         };
3398         int status = -ENOMEM;
3399
3400         res.fattr = nfs_alloc_fattr();
3401         if (res.fattr == NULL)
3402                 goto out;
3403
3404         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3405         if (IS_ERR(res.label)) {
3406                 status = PTR_ERR(res.label);
3407                 goto out;
3408         }
3409         arg.bitmask = nfs4_bitmask(server, res.label);
3410
3411         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3412         if (!status) {
3413                 update_changeattr(dir, &res.cinfo);
3414                 status = nfs_post_op_update_inode(inode, res.fattr);
3415                 if (!status)
3416                         nfs_setsecurity(inode, res.fattr, res.label);
3417         }
3418
3419
3420         nfs4_label_free(res.label);
3421
3422 out:
3423         nfs_free_fattr(res.fattr);
3424         return status;
3425 }
3426
3427 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3428 {
3429         struct nfs4_exception exception = { };
3430         int err;
3431         do {
3432                 err = nfs4_handle_exception(NFS_SERVER(inode),
3433                                 _nfs4_proc_link(inode, dir, name),
3434                                 &exception);
3435         } while (exception.retry);
3436         return err;
3437 }
3438
3439 struct nfs4_createdata {
3440         struct rpc_message msg;
3441         struct nfs4_create_arg arg;
3442         struct nfs4_create_res res;
3443         struct nfs_fh fh;
3444         struct nfs_fattr fattr;
3445         struct nfs4_label *label;
3446 };
3447
3448 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3449                 struct qstr *name, struct iattr *sattr, u32 ftype)
3450 {
3451         struct nfs4_createdata *data;
3452
3453         data = kzalloc(sizeof(*data), GFP_KERNEL);
3454         if (data != NULL) {
3455                 struct nfs_server *server = NFS_SERVER(dir);
3456
3457                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3458                 if (IS_ERR(data->label))
3459                         goto out_free;
3460
3461                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3462                 data->msg.rpc_argp = &data->arg;
3463                 data->msg.rpc_resp = &data->res;
3464                 data->arg.dir_fh = NFS_FH(dir);
3465                 data->arg.server = server;
3466                 data->arg.name = name;
3467                 data->arg.attrs = sattr;
3468                 data->arg.ftype = ftype;
3469                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3470                 data->res.server = server;
3471                 data->res.fh = &data->fh;
3472                 data->res.fattr = &data->fattr;
3473                 data->res.label = data->label;
3474                 nfs_fattr_init(data->res.fattr);
3475         }
3476         return data;
3477 out_free:
3478         kfree(data);
3479         return NULL;
3480 }
3481
3482 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3483 {
3484         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3485                                     &data->arg.seq_args, &data->res.seq_res, 1);
3486         if (status == 0) {
3487                 update_changeattr(dir, &data->res.dir_cinfo);
3488                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3489         }
3490         return status;
3491 }
3492
3493 static void nfs4_free_createdata(struct nfs4_createdata *data)
3494 {
3495         nfs4_label_free(data->label);
3496         kfree(data);
3497 }
3498
3499 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3500                 struct page *page, unsigned int len, struct iattr *sattr,
3501                 struct nfs4_label *label)
3502 {
3503         struct nfs4_createdata *data;
3504         int status = -ENAMETOOLONG;
3505
3506         if (len > NFS4_MAXPATHLEN)
3507                 goto out;
3508
3509         status = -ENOMEM;
3510         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3511         if (data == NULL)
3512                 goto out;
3513
3514         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3515         data->arg.u.symlink.pages = &page;
3516         data->arg.u.symlink.len = len;
3517         data->arg.label = label;
3518
3519         status = nfs4_do_create(dir, dentry, data);
3520
3521         nfs4_free_createdata(data);
3522 out:
3523         return status;
3524 }
3525
3526 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3527                 struct page *page, unsigned int len, struct iattr *sattr)
3528 {
3529         struct nfs4_exception exception = { };
3530         struct nfs4_label l, *label = NULL;
3531         int err;
3532
3533         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3534
3535         do {
3536                 err = nfs4_handle_exception(NFS_SERVER(dir),
3537                                 _nfs4_proc_symlink(dir, dentry, page,
3538                                                         len, sattr, label),
3539                                 &exception);
3540         } while (exception.retry);
3541
3542         nfs4_label_release_security(label);
3543         return err;
3544 }
3545
3546 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3547                 struct iattr *sattr, struct nfs4_label *label)
3548 {
3549         struct nfs4_createdata *data;
3550         int status = -ENOMEM;
3551
3552         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3553         if (data == NULL)
3554                 goto out;
3555
3556         data->arg.label = label;
3557         status = nfs4_do_create(dir, dentry, data);
3558
3559         nfs4_free_createdata(data);
3560 out:
3561         return status;
3562 }
3563
3564 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3565                 struct iattr *sattr)
3566 {
3567         struct nfs4_exception exception = { };
3568         struct nfs4_label l, *label = NULL;
3569         int err;
3570
3571         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3572
3573         sattr->ia_mode &= ~current_umask();
3574         do {
3575                 err = nfs4_handle_exception(NFS_SERVER(dir),
3576                                 _nfs4_proc_mkdir(dir, dentry, sattr, label),
3577                                 &exception);
3578         } while (exception.retry);
3579         nfs4_label_release_security(label);
3580
3581         return err;
3582 }
3583
3584 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3585                 u64 cookie, struct page **pages, unsigned int count, int plus)
3586 {
3587         struct inode            *dir = dentry->d_inode;
3588         struct nfs4_readdir_arg args = {
3589                 .fh = NFS_FH(dir),
3590                 .pages = pages,
3591                 .pgbase = 0,
3592                 .count = count,
3593                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3594                 .plus = plus,
3595         };
3596         struct nfs4_readdir_res res;
3597         struct rpc_message msg = {
3598                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3599                 .rpc_argp = &args,
3600                 .rpc_resp = &res,
3601                 .rpc_cred = cred,
3602         };
3603         int                     status;
3604
3605         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3606                         dentry->d_parent->d_name.name,
3607                         dentry->d_name.name,
3608                         (unsigned long long)cookie);
3609         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3610         res.pgbase = args.pgbase;
3611         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3612         if (status >= 0) {
3613                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3614                 status += args.pgbase;
3615         }
3616
3617         nfs_invalidate_atime(dir);
3618
3619         dprintk("%s: returns %d\n", __func__, status);
3620         return status;
3621 }
3622
3623 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3624                 u64 cookie, struct page **pages, unsigned int count, int plus)
3625 {
3626         struct nfs4_exception exception = { };
3627         int err;
3628         do {
3629                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3630                                 _nfs4_proc_readdir(dentry, cred, cookie,
3631                                         pages, count, plus),
3632                                 &exception);
3633         } while (exception.retry);
3634         return err;
3635 }
3636
3637 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3638                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3639 {
3640         struct nfs4_createdata *data;
3641         int mode = sattr->ia_mode;
3642         int status = -ENOMEM;
3643
3644         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3645         if (data == NULL)
3646                 goto out;
3647
3648         if (S_ISFIFO(mode))
3649                 data->arg.ftype = NF4FIFO;
3650         else if (S_ISBLK(mode)) {
3651                 data->arg.ftype = NF4BLK;
3652                 data->arg.u.device.specdata1 = MAJOR(rdev);
3653                 data->arg.u.device.specdata2 = MINOR(rdev);
3654         }
3655         else if (S_ISCHR(mode)) {
3656                 data->arg.ftype = NF4CHR;
3657                 data->arg.u.device.specdata1 = MAJOR(rdev);
3658                 data->arg.u.device.specdata2 = MINOR(rdev);
3659         } else if (!S_ISSOCK(mode)) {
3660                 status = -EINVAL;
3661                 goto out_free;
3662         }
3663
3664         data->arg.label = label;
3665         status = nfs4_do_create(dir, dentry, data);
3666 out_free:
3667         nfs4_free_createdata(data);
3668 out:
3669         return status;
3670 }
3671
3672 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3673                 struct iattr *sattr, dev_t rdev)
3674 {
3675         struct nfs4_exception exception = { };
3676         struct nfs4_label l, *label = NULL;
3677         int err;
3678
3679         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3680
3681         sattr->ia_mode &= ~current_umask();
3682         do {
3683                 err = nfs4_handle_exception(NFS_SERVER(dir),
3684                                 _nfs4_proc_mknod(dir, dentry, sattr, label, rdev),
3685                                 &exception);
3686         } while (exception.retry);
3687
3688         nfs4_label_release_security(label);
3689
3690         return err;
3691 }
3692
3693 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3694                  struct nfs_fsstat *fsstat)
3695 {
3696         struct nfs4_statfs_arg args = {
3697                 .fh = fhandle,
3698                 .bitmask = server->attr_bitmask,
3699         };
3700         struct nfs4_statfs_res res = {
3701                 .fsstat = fsstat,
3702         };
3703         struct rpc_message msg = {
3704                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3705                 .rpc_argp = &args,
3706                 .rpc_resp = &res,
3707         };
3708
3709         nfs_fattr_init(fsstat->fattr);
3710         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3711 }
3712
3713 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3714 {
3715         struct nfs4_exception exception = { };
3716         int err;
3717         do {
3718                 err = nfs4_handle_exception(server,
3719                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3720                                 &exception);
3721         } while (exception.retry);
3722         return err;
3723 }
3724
3725 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3726                 struct nfs_fsinfo *fsinfo)
3727 {
3728         struct nfs4_fsinfo_arg args = {
3729                 .fh = fhandle,
3730                 .bitmask = server->attr_bitmask,
3731         };
3732         struct nfs4_fsinfo_res res = {
3733                 .fsinfo = fsinfo,
3734         };
3735         struct rpc_message msg = {
3736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3737                 .rpc_argp = &args,
3738                 .rpc_resp = &res,
3739         };
3740
3741         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3742 }
3743
3744 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3745 {
3746         struct nfs4_exception exception = { };
3747         unsigned long now = jiffies;
3748         int err;
3749
3750         do {
3751                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3752                 if (err == 0) {
3753                         struct nfs_client *clp = server->nfs_client;
3754
3755                         spin_lock(&clp->cl_lock);
3756                         clp->cl_lease_time = fsinfo->lease_time * HZ;
3757                         clp->cl_last_renewal = now;
3758                         spin_unlock(&clp->cl_lock);
3759                         break;
3760                 }
3761                 err = nfs4_handle_exception(server, err, &exception);
3762         } while (exception.retry);
3763         return err;
3764 }
3765
3766 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3767 {
3768         int error;
3769
3770         nfs_fattr_init(fsinfo->fattr);
3771         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3772         if (error == 0) {
3773                 /* block layout checks this! */
3774                 server->pnfs_blksize = fsinfo->blksize;
3775                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3776         }
3777
3778         return error;
3779 }
3780
3781 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3782                 struct nfs_pathconf *pathconf)
3783 {
3784         struct nfs4_pathconf_arg args = {
3785                 .fh = fhandle,
3786                 .bitmask = server->attr_bitmask,
3787         };
3788         struct nfs4_pathconf_res res = {
3789                 .pathconf = pathconf,
3790         };
3791         struct rpc_message msg = {
3792                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3793                 .rpc_argp = &args,
3794                 .rpc_resp = &res,
3795         };
3796
3797         /* None of the pathconf attributes are mandatory to implement */
3798         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3799                 memset(pathconf, 0, sizeof(*pathconf));
3800                 return 0;
3801         }
3802
3803         nfs_fattr_init(pathconf->fattr);
3804         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3805 }
3806
3807 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3808                 struct nfs_pathconf *pathconf)
3809 {
3810         struct nfs4_exception exception = { };
3811         int err;
3812
3813         do {
3814                 err = nfs4_handle_exception(server,
3815                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3816                                 &exception);
3817         } while (exception.retry);
3818         return err;
3819 }
3820
3821 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3822                 const struct nfs_open_context *ctx,
3823                 const struct nfs_lock_context *l_ctx,
3824                 fmode_t fmode)
3825 {
3826         const struct nfs_lockowner *lockowner = NULL;
3827
3828         if (l_ctx != NULL)
3829                 lockowner = &l_ctx->lockowner;
3830         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3831 }
3832 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3833
3834 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3835                 const struct nfs_open_context *ctx,
3836                 const struct nfs_lock_context *l_ctx,
3837                 fmode_t fmode)
3838 {
3839         nfs4_stateid current_stateid;
3840
3841         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode))
3842                 return false;
3843         return nfs4_stateid_match(stateid, &current_stateid);
3844 }
3845
3846 static bool nfs4_error_stateid_expired(int err)
3847 {
3848         switch (err) {
3849         case -NFS4ERR_DELEG_REVOKED:
3850         case -NFS4ERR_ADMIN_REVOKED:
3851         case -NFS4ERR_BAD_STATEID:
3852         case -NFS4ERR_STALE_STATEID:
3853         case -NFS4ERR_OLD_STATEID:
3854         case -NFS4ERR_OPENMODE:
3855         case -NFS4ERR_EXPIRED:
3856                 return true;
3857         }
3858         return false;
3859 }
3860
3861 void __nfs4_read_done_cb(struct nfs_read_data *data)
3862 {
3863         nfs_invalidate_atime(data->header->inode);
3864 }
3865
3866 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3867 {
3868         struct nfs_server *server = NFS_SERVER(data->header->inode);
3869
3870         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3871                 rpc_restart_call_prepare(task);
3872                 return -EAGAIN;
3873         }
3874
3875         __nfs4_read_done_cb(data);
3876         if (task->tk_status > 0)
3877                 renew_lease(server, data->timestamp);
3878         return 0;
3879 }
3880
3881 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3882                 struct nfs_readargs *args)
3883 {
3884
3885         if (!nfs4_error_stateid_expired(task->tk_status) ||
3886                 nfs4_stateid_is_current(&args->stateid,
3887                                 args->context,
3888                                 args->lock_context,
3889                                 FMODE_READ))
3890                 return false;
3891         rpc_restart_call_prepare(task);
3892         return true;
3893 }
3894
3895 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3896 {
3897
3898         dprintk("--> %s\n", __func__);
3899
3900         if (!nfs4_sequence_done(task, &data->res.seq_res))
3901                 return -EAGAIN;
3902         if (nfs4_read_stateid_changed(task, &data->args))
3903                 return -EAGAIN;
3904         return data->read_done_cb ? data->read_done_cb(task, data) :
3905                                     nfs4_read_done_cb(task, data);
3906 }
3907
3908 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3909 {
3910         data->timestamp   = jiffies;
3911         data->read_done_cb = nfs4_read_done_cb;
3912         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3913         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3914 }
3915
3916 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3917 {
3918         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3919                         &data->args.seq_args,
3920                         &data->res.seq_res,
3921                         task))
3922                 return;
3923         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3924                         data->args.lock_context, FMODE_READ);
3925 }
3926
3927 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3928 {
3929         struct inode *inode = data->header->inode;
3930
3931         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3932                 rpc_restart_call_prepare(task);
3933                 return -EAGAIN;
3934         }
3935         if (task->tk_status >= 0) {
3936                 renew_lease(NFS_SERVER(inode), data->timestamp);
3937                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3938         }
3939         return 0;
3940 }
3941
3942 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3943                 struct nfs_writeargs *args)
3944 {
3945
3946         if (!nfs4_error_stateid_expired(task->tk_status) ||
3947                 nfs4_stateid_is_current(&args->stateid,
3948                                 args->context,
3949                                 args->lock_context,
3950                                 FMODE_WRITE))
3951                 return false;
3952         rpc_restart_call_prepare(task);
3953         return true;
3954 }
3955
3956 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3957 {
3958         if (!nfs4_sequence_done(task, &data->res.seq_res))
3959                 return -EAGAIN;
3960         if (nfs4_write_stateid_changed(task, &data->args))
3961                 return -EAGAIN;
3962         return data->write_done_cb ? data->write_done_cb(task, data) :
3963                 nfs4_write_done_cb(task, data);
3964 }
3965
3966 static
3967 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3968 {
3969         const struct nfs_pgio_header *hdr = data->header;
3970
3971         /* Don't request attributes for pNFS or O_DIRECT writes */
3972         if (data->ds_clp != NULL || hdr->dreq != NULL)
3973                 return false;
3974         /* Otherwise, request attributes if and only if we don't hold
3975          * a delegation
3976          */
3977         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3978 }
3979
3980 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3981 {
3982         struct nfs_server *server = NFS_SERVER(data->header->inode);
3983
3984         if (!nfs4_write_need_cache_consistency_data(data)) {
3985                 data->args.bitmask = NULL;
3986                 data->res.fattr = NULL;
3987         } else
3988                 data->args.bitmask = server->cache_consistency_bitmask;
3989
3990         if (!data->write_done_cb)
3991                 data->write_done_cb = nfs4_write_done_cb;
3992         data->res.server = server;
3993         data->timestamp   = jiffies;
3994
3995         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3996         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3997 }
3998
3999 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
4000 {
4001         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
4002                         &data->args.seq_args,
4003                         &data->res.seq_res,
4004                         task))
4005                 return;
4006         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
4007                         data->args.lock_context, FMODE_WRITE);
4008 }
4009
4010 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4011 {
4012         nfs4_setup_sequence(NFS_SERVER(data->inode),
4013                         &data->args.seq_args,
4014                         &data->res.seq_res,
4015                         task);
4016 }
4017
4018 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4019 {
4020         struct inode *inode = data->inode;
4021
4022         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4023                 rpc_restart_call_prepare(task);
4024                 return -EAGAIN;
4025         }
4026         return 0;
4027 }
4028
4029 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4030 {
4031         if (!nfs4_sequence_done(task, &data->res.seq_res))
4032                 return -EAGAIN;
4033         return data->commit_done_cb(task, data);
4034 }
4035
4036 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4037 {
4038         struct nfs_server *server = NFS_SERVER(data->inode);
4039
4040         if (data->commit_done_cb == NULL)
4041                 data->commit_done_cb = nfs4_commit_done_cb;
4042         data->res.server = server;
4043         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4044         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4045 }
4046
4047 struct nfs4_renewdata {
4048         struct nfs_client       *client;
4049         unsigned long           timestamp;
4050 };
4051
4052 /*
4053  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4054  * standalone procedure for queueing an asynchronous RENEW.
4055  */
4056 static void nfs4_renew_release(void *calldata)
4057 {
4058         struct nfs4_renewdata *data = calldata;
4059         struct nfs_client *clp = data->client;
4060
4061         if (atomic_read(&clp->cl_count) > 1)
4062                 nfs4_schedule_state_renewal(clp);
4063         nfs_put_client(clp);
4064         kfree(data);
4065 }
4066
4067 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4068 {
4069         struct nfs4_renewdata *data = calldata;
4070         struct nfs_client *clp = data->client;
4071         unsigned long timestamp = data->timestamp;
4072
4073         trace_nfs4_renew_async(clp, task->tk_status);
4074         if (task->tk_status < 0) {
4075                 /* Unless we're shutting down, schedule state recovery! */
4076                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4077                         return;
4078                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4079                         nfs4_schedule_lease_recovery(clp);
4080                         return;
4081                 }
4082                 nfs4_schedule_path_down_recovery(clp);
4083         }
4084         do_renew_lease(clp, timestamp);
4085 }
4086
4087 static const struct rpc_call_ops nfs4_renew_ops = {
4088         .rpc_call_done = nfs4_renew_done,
4089         .rpc_release = nfs4_renew_release,
4090 };
4091
4092 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4093 {
4094         struct rpc_message msg = {
4095                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4096                 .rpc_argp       = clp,
4097                 .rpc_cred       = cred,
4098         };
4099         struct nfs4_renewdata *data;
4100
4101         if (renew_flags == 0)
4102                 return 0;
4103         if (!atomic_inc_not_zero(&clp->cl_count))
4104                 return -EIO;
4105         data = kmalloc(sizeof(*data), GFP_NOFS);
4106         if (data == NULL)
4107                 return -ENOMEM;
4108         data->client = clp;
4109         data->timestamp = jiffies;
4110         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4111                         &nfs4_renew_ops, data);
4112 }
4113
4114 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4115 {
4116         struct rpc_message msg = {
4117                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4118                 .rpc_argp       = clp,
4119                 .rpc_cred       = cred,
4120         };
4121         unsigned long now = jiffies;
4122         int status;
4123
4124         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4125         if (status < 0)
4126                 return status;
4127         do_renew_lease(clp, now);
4128         return 0;
4129 }
4130
4131 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4132 {
4133         return (server->caps & NFS_CAP_ACLS)
4134                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
4135                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
4136 }
4137
4138 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4139  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4140  * the stack.
4141  */
4142 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4143
4144 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4145                 struct page **pages, unsigned int *pgbase)
4146 {
4147         struct page *newpage, **spages;
4148         int rc = 0;
4149         size_t len;
4150         spages = pages;
4151
4152         do {
4153                 len = min_t(size_t, PAGE_SIZE, buflen);
4154                 newpage = alloc_page(GFP_KERNEL);
4155
4156                 if (newpage == NULL)
4157                         goto unwind;
4158                 memcpy(page_address(newpage), buf, len);
4159                 buf += len;
4160                 buflen -= len;
4161                 *pages++ = newpage;
4162                 rc++;
4163         } while (buflen != 0);
4164
4165         return rc;
4166
4167 unwind:
4168         for(; rc > 0; rc--)
4169                 __free_page(spages[rc-1]);
4170         return -ENOMEM;
4171 }
4172
4173 struct nfs4_cached_acl {
4174         int cached;
4175         size_t len;
4176         char data[0];
4177 };
4178
4179 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4180 {
4181         struct nfs_inode *nfsi = NFS_I(inode);
4182
4183         spin_lock(&inode->i_lock);
4184         kfree(nfsi->nfs4_acl);
4185         nfsi->nfs4_acl = acl;
4186         spin_unlock(&inode->i_lock);
4187 }
4188
4189 static void nfs4_zap_acl_attr(struct inode *inode)
4190 {
4191         nfs4_set_cached_acl(inode, NULL);
4192 }
4193
4194 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4195 {
4196         struct nfs_inode *nfsi = NFS_I(inode);
4197         struct nfs4_cached_acl *acl;
4198         int ret = -ENOENT;
4199
4200         spin_lock(&inode->i_lock);
4201         acl = nfsi->nfs4_acl;
4202         if (acl == NULL)
4203                 goto out;
4204         if (buf == NULL) /* user is just asking for length */
4205                 goto out_len;
4206         if (acl->cached == 0)
4207                 goto out;
4208         ret = -ERANGE; /* see getxattr(2) man page */
4209         if (acl->len > buflen)
4210                 goto out;
4211         memcpy(buf, acl->data, acl->len);
4212 out_len:
4213         ret = acl->len;
4214 out:
4215         spin_unlock(&inode->i_lock);
4216         return ret;
4217 }
4218
4219 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4220 {
4221         struct nfs4_cached_acl *acl;
4222         size_t buflen = sizeof(*acl) + acl_len;
4223
4224         if (buflen <= PAGE_SIZE) {
4225                 acl = kmalloc(buflen, GFP_KERNEL);
4226                 if (acl == NULL)
4227                         goto out;
4228                 acl->cached = 1;
4229                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4230         } else {
4231                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4232                 if (acl == NULL)
4233                         goto out;
4234                 acl->cached = 0;
4235         }
4236         acl->len = acl_len;
4237 out:
4238         nfs4_set_cached_acl(inode, acl);
4239 }
4240
4241 /*
4242  * The getxattr API returns the required buffer length when called with a
4243  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4244  * the required buf.  On a NULL buf, we send a page of data to the server
4245  * guessing that the ACL request can be serviced by a page. If so, we cache
4246  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4247  * the cache. If not so, we throw away the page, and cache the required
4248  * length. The next getxattr call will then produce another round trip to
4249  * the server, this time with the input buf of the required size.
4250  */
4251 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4252 {
4253         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4254         struct nfs_getaclargs args = {
4255                 .fh = NFS_FH(inode),
4256                 .acl_pages = pages,
4257                 .acl_len = buflen,
4258         };
4259         struct nfs_getaclres res = {
4260                 .acl_len = buflen,
4261         };
4262         struct rpc_message msg = {
4263                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4264                 .rpc_argp = &args,
4265                 .rpc_resp = &res,
4266         };
4267         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4268         int ret = -ENOMEM, i;
4269
4270         /* As long as we're doing a round trip to the server anyway,
4271          * let's be prepared for a page of acl data. */
4272         if (npages == 0)
4273                 npages = 1;
4274         if (npages > ARRAY_SIZE(pages))
4275                 return -ERANGE;
4276
4277         for (i = 0; i < npages; i++) {
4278                 pages[i] = alloc_page(GFP_KERNEL);
4279                 if (!pages[i])
4280                         goto out_free;
4281         }
4282
4283         /* for decoding across pages */
4284         res.acl_scratch = alloc_page(GFP_KERNEL);
4285         if (!res.acl_scratch)
4286                 goto out_free;
4287
4288         args.acl_len = npages * PAGE_SIZE;
4289         args.acl_pgbase = 0;
4290
4291         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4292                 __func__, buf, buflen, npages, args.acl_len);
4293         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4294                              &msg, &args.seq_args, &res.seq_res, 0);
4295         if (ret)
4296                 goto out_free;
4297
4298         /* Handle the case where the passed-in buffer is too short */
4299         if (res.acl_flags & NFS4_ACL_TRUNC) {
4300                 /* Did the user only issue a request for the acl length? */
4301                 if (buf == NULL)
4302                         goto out_ok;
4303                 ret = -ERANGE;
4304                 goto out_free;
4305         }
4306         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4307         if (buf) {
4308                 if (res.acl_len > buflen) {
4309                         ret = -ERANGE;
4310                         goto out_free;
4311                 }
4312                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4313         }
4314 out_ok:
4315         ret = res.acl_len;
4316 out_free:
4317         for (i = 0; i < npages; i++)
4318                 if (pages[i])
4319                         __free_page(pages[i]);
4320         if (res.acl_scratch)
4321                 __free_page(res.acl_scratch);
4322         return ret;
4323 }
4324
4325 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4326 {
4327         struct nfs4_exception exception = { };
4328         ssize_t ret;
4329         do {
4330                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4331                 if (ret >= 0)
4332                         break;
4333                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4334         } while (exception.retry);
4335         return ret;
4336 }
4337
4338 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4339 {
4340         struct nfs_server *server = NFS_SERVER(inode);
4341         int ret;
4342
4343         if (!nfs4_server_supports_acls(server))
4344                 return -EOPNOTSUPP;
4345         ret = nfs_revalidate_inode(server, inode);
4346         if (ret < 0)
4347                 return ret;
4348         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4349                 nfs_zap_acl_cache(inode);
4350         ret = nfs4_read_cached_acl(inode, buf, buflen);
4351         if (ret != -ENOENT)
4352                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4353                  * but no cached acl data, just the acl length */
4354                 return ret;
4355         return nfs4_get_acl_uncached(inode, buf, buflen);
4356 }
4357
4358 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4359 {
4360         struct nfs_server *server = NFS_SERVER(inode);
4361         struct page *pages[NFS4ACL_MAXPAGES];
4362         struct nfs_setaclargs arg = {
4363                 .fh             = NFS_FH(inode),
4364                 .acl_pages      = pages,
4365                 .acl_len        = buflen,
4366         };
4367         struct nfs_setaclres res;
4368         struct rpc_message msg = {
4369                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4370                 .rpc_argp       = &arg,
4371                 .rpc_resp       = &res,
4372         };
4373         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4374         int ret, i;
4375
4376         if (!nfs4_server_supports_acls(server))
4377                 return -EOPNOTSUPP;
4378         if (npages > ARRAY_SIZE(pages))
4379                 return -ERANGE;
4380         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4381         if (i < 0)
4382                 return i;
4383         nfs4_inode_return_delegation(inode);
4384         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4385
4386         /*
4387          * Free each page after tx, so the only ref left is
4388          * held by the network stack
4389          */
4390         for (; i > 0; i--)
4391                 put_page(pages[i-1]);
4392
4393         /*
4394          * Acl update can result in inode attribute update.
4395          * so mark the attribute cache invalid.
4396          */
4397         spin_lock(&inode->i_lock);
4398         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4399         spin_unlock(&inode->i_lock);
4400         nfs_access_zap_cache(inode);
4401         nfs_zap_acl_cache(inode);
4402         return ret;
4403 }
4404
4405 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4406 {
4407         struct nfs4_exception exception = { };
4408         int err;
4409         do {
4410                 err = nfs4_handle_exception(NFS_SERVER(inode),
4411                                 __nfs4_proc_set_acl(inode, buf, buflen),
4412                                 &exception);
4413         } while (exception.retry);
4414         return err;
4415 }
4416
4417 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4418 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4419                                         size_t buflen)
4420 {
4421         struct nfs_server *server = NFS_SERVER(inode);
4422         struct nfs_fattr fattr;
4423         struct nfs4_label label = {0, 0, buflen, buf};
4424
4425         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4426         struct nfs4_getattr_arg args = {
4427                 .fh             = NFS_FH(inode),
4428                 .bitmask        = bitmask,
4429         };
4430         struct nfs4_getattr_res res = {
4431                 .fattr          = &fattr,
4432                 .label          = &label,
4433                 .server         = server,
4434         };
4435         struct rpc_message msg = {
4436                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4437                 .rpc_argp       = &args,
4438                 .rpc_resp       = &res,
4439         };
4440         int ret;
4441
4442         nfs_fattr_init(&fattr);
4443
4444         ret = rpc_call_sync(server->client, &msg, 0);
4445         if (ret)
4446                 return ret;
4447         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4448                 return -ENOENT;
4449         if (buflen < label.len)
4450                 return -ERANGE;
4451         return 0;
4452 }
4453
4454 static int nfs4_get_security_label(struct inode *inode, void *buf,
4455                                         size_t buflen)
4456 {
4457         struct nfs4_exception exception = { };
4458         int err;
4459
4460         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4461                 return -EOPNOTSUPP;
4462
4463         do {
4464                 err = nfs4_handle_exception(NFS_SERVER(inode),
4465                                 _nfs4_get_security_label(inode, buf, buflen),
4466                                 &exception);
4467         } while (exception.retry);
4468         return err;
4469 }
4470
4471 static int _nfs4_do_set_security_label(struct inode *inode,
4472                 struct nfs4_label *ilabel,
4473                 struct nfs_fattr *fattr,
4474                 struct nfs4_label *olabel)
4475 {
4476
4477         struct iattr sattr = {0};
4478         struct nfs_server *server = NFS_SERVER(inode);
4479         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4480         struct nfs_setattrargs args = {
4481                 .fh             = NFS_FH(inode),
4482                 .iap            = &sattr,
4483                 .server         = server,
4484                 .bitmask        = bitmask,
4485                 .label          = ilabel,
4486         };
4487         struct nfs_setattrres res = {
4488                 .fattr          = fattr,
4489                 .label          = olabel,
4490                 .server         = server,
4491         };
4492         struct rpc_message msg = {
4493                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4494                 .rpc_argp       = &args,
4495                 .rpc_resp       = &res,
4496         };
4497         int status;
4498
4499         nfs4_stateid_copy(&args.stateid, &zero_stateid);
4500
4501         status = rpc_call_sync(server->client, &msg, 0);
4502         if (status)
4503                 dprintk("%s failed: %d\n", __func__, status);
4504
4505         return status;
4506 }
4507
4508 static int nfs4_do_set_security_label(struct inode *inode,
4509                 struct nfs4_label *ilabel,
4510                 struct nfs_fattr *fattr,
4511                 struct nfs4_label *olabel)
4512 {
4513         struct nfs4_exception exception = { };
4514         int err;
4515
4516         do {
4517                 err = nfs4_handle_exception(NFS_SERVER(inode),
4518                                 _nfs4_do_set_security_label(inode, ilabel,
4519                                 fattr, olabel),
4520                                 &exception);
4521         } while (exception.retry);
4522         return err;
4523 }
4524
4525 static int
4526 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4527 {
4528         struct nfs4_label ilabel, *olabel = NULL;
4529         struct nfs_fattr fattr;
4530         struct rpc_cred *cred;
4531         struct inode *inode = dentry->d_inode;
4532         int status;
4533
4534         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4535                 return -EOPNOTSUPP;
4536
4537         nfs_fattr_init(&fattr);
4538
4539         ilabel.pi = 0;
4540         ilabel.lfs = 0;
4541         ilabel.label = (char *)buf;
4542         ilabel.len = buflen;
4543
4544         cred = rpc_lookup_cred();
4545         if (IS_ERR(cred))
4546                 return PTR_ERR(cred);
4547
4548         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4549         if (IS_ERR(olabel)) {
4550                 status = -PTR_ERR(olabel);
4551                 goto out;
4552         }
4553
4554         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4555         if (status == 0)
4556                 nfs_setsecurity(inode, &fattr, olabel);
4557
4558         nfs4_label_free(olabel);
4559 out:
4560         put_rpccred(cred);
4561         return status;
4562 }
4563 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4564
4565
4566 static int
4567 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4568 {
4569         struct nfs_client *clp = server->nfs_client;
4570
4571         if (task->tk_status >= 0)
4572                 return 0;
4573         switch(task->tk_status) {
4574                 case -NFS4ERR_DELEG_REVOKED:
4575                 case -NFS4ERR_ADMIN_REVOKED:
4576                 case -NFS4ERR_BAD_STATEID:
4577                         if (state == NULL)
4578                                 break;
4579                         nfs_remove_bad_delegation(state->inode);
4580                 case -NFS4ERR_OPENMODE:
4581                         if (state == NULL)
4582                                 break;
4583                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4584                                 goto stateid_invalid;
4585                         goto wait_on_recovery;
4586                 case -NFS4ERR_EXPIRED:
4587                         if (state != NULL) {
4588                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4589                                         goto stateid_invalid;
4590                         }
4591                 case -NFS4ERR_STALE_STATEID:
4592                 case -NFS4ERR_STALE_CLIENTID:
4593                         nfs4_schedule_lease_recovery(clp);
4594                         goto wait_on_recovery;
4595 #if defined(CONFIG_NFS_V4_1)
4596                 case -NFS4ERR_BADSESSION:
4597                 case -NFS4ERR_BADSLOT:
4598                 case -NFS4ERR_BAD_HIGH_SLOT:
4599                 case -NFS4ERR_DEADSESSION:
4600                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4601                 case -NFS4ERR_SEQ_FALSE_RETRY:
4602                 case -NFS4ERR_SEQ_MISORDERED:
4603                         dprintk("%s ERROR %d, Reset session\n", __func__,
4604                                 task->tk_status);
4605                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4606                         task->tk_status = 0;
4607                         return -EAGAIN;
4608 #endif /* CONFIG_NFS_V4_1 */
4609                 case -NFS4ERR_DELAY:
4610                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4611                 case -NFS4ERR_GRACE:
4612                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4613                         task->tk_status = 0;
4614                         return -EAGAIN;
4615                 case -NFS4ERR_RETRY_UNCACHED_REP:
4616                 case -NFS4ERR_OLD_STATEID:
4617                         task->tk_status = 0;
4618                         return -EAGAIN;
4619         }
4620         task->tk_status = nfs4_map_errors(task->tk_status);
4621         return 0;
4622 stateid_invalid:
4623         task->tk_status = -EIO;
4624         return 0;
4625 wait_on_recovery:
4626         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4627         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4628                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4629         task->tk_status = 0;
4630         return -EAGAIN;
4631 }
4632
4633 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4634                                     nfs4_verifier *bootverf)
4635 {
4636         __be32 verf[2];
4637
4638         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4639                 /* An impossible timestamp guarantees this value
4640                  * will never match a generated boot time. */
4641                 verf[0] = 0;
4642                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4643         } else {
4644                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4645                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4646                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4647         }
4648         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4649 }
4650
4651 static unsigned int
4652 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4653                                    char *buf, size_t len)
4654 {
4655         unsigned int result;
4656
4657         rcu_read_lock();
4658         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4659                                 clp->cl_ipaddr,
4660                                 rpc_peeraddr2str(clp->cl_rpcclient,
4661                                                         RPC_DISPLAY_ADDR),
4662                                 rpc_peeraddr2str(clp->cl_rpcclient,
4663                                                         RPC_DISPLAY_PROTO));
4664         rcu_read_unlock();
4665         return result;
4666 }
4667
4668 static unsigned int
4669 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4670                                 char *buf, size_t len)
4671 {
4672         const char *nodename = clp->cl_rpcclient->cl_nodename;
4673
4674         if (nfs4_client_id_uniquifier[0] != '\0')
4675                 return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4676                                 clp->rpc_ops->version,
4677                                 clp->cl_minorversion,
4678                                 nfs4_client_id_uniquifier,
4679                                 nodename);
4680         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4681                                 clp->rpc_ops->version, clp->cl_minorversion,
4682                                 nodename);
4683 }
4684
4685 /**
4686  * nfs4_proc_setclientid - Negotiate client ID
4687  * @clp: state data structure
4688  * @program: RPC program for NFSv4 callback service
4689  * @port: IP port number for NFS4 callback service
4690  * @cred: RPC credential to use for this call
4691  * @res: where to place the result
4692  *
4693  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4694  */
4695 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4696                 unsigned short port, struct rpc_cred *cred,
4697                 struct nfs4_setclientid_res *res)
4698 {
4699         nfs4_verifier sc_verifier;
4700         struct nfs4_setclientid setclientid = {
4701                 .sc_verifier = &sc_verifier,
4702                 .sc_prog = program,
4703                 .sc_cb_ident = clp->cl_cb_ident,
4704         };
4705         struct rpc_message msg = {
4706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4707                 .rpc_argp = &setclientid,
4708                 .rpc_resp = res,
4709                 .rpc_cred = cred,
4710         };
4711         int status;
4712
4713         /* nfs_client_id4 */
4714         nfs4_init_boot_verifier(clp, &sc_verifier);
4715         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4716                 setclientid.sc_name_len =
4717                                 nfs4_init_uniform_client_string(clp,
4718                                                 setclientid.sc_name,
4719                                                 sizeof(setclientid.sc_name));
4720         else
4721                 setclientid.sc_name_len =
4722                                 nfs4_init_nonuniform_client_string(clp,
4723                                                 setclientid.sc_name,
4724                                                 sizeof(setclientid.sc_name));
4725         /* cb_client4 */
4726         rcu_read_lock();
4727         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4728                                 sizeof(setclientid.sc_netid), "%s",
4729                                 rpc_peeraddr2str(clp->cl_rpcclient,
4730                                                         RPC_DISPLAY_NETID));
4731         rcu_read_unlock();
4732         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4733                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4734                                 clp->cl_ipaddr, port >> 8, port & 255);
4735
4736         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4737                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4738                 setclientid.sc_name_len, setclientid.sc_name);
4739         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4740         trace_nfs4_setclientid(clp, status);
4741         dprintk("NFS reply setclientid: %d\n", status);
4742         return status;
4743 }
4744
4745 /**
4746  * nfs4_proc_setclientid_confirm - Confirm client ID
4747  * @clp: state data structure
4748  * @res: result of a previous SETCLIENTID
4749  * @cred: RPC credential to use for this call
4750  *
4751  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4752  */
4753 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4754                 struct nfs4_setclientid_res *arg,
4755                 struct rpc_cred *cred)
4756 {
4757         struct rpc_message msg = {
4758                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4759                 .rpc_argp = arg,
4760                 .rpc_cred = cred,
4761         };
4762         int status;
4763
4764         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4765                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4766                 clp->cl_clientid);
4767         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4768         trace_nfs4_setclientid_confirm(clp, status);
4769         dprintk("NFS reply setclientid_confirm: %d\n", status);
4770         return status;
4771 }
4772
4773 struct nfs4_delegreturndata {
4774         struct nfs4_delegreturnargs args;
4775         struct nfs4_delegreturnres res;
4776         struct nfs_fh fh;
4777         nfs4_stateid stateid;
4778         unsigned long timestamp;
4779         struct nfs_fattr fattr;
4780         int rpc_status;
4781 };
4782
4783 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4784 {
4785         struct nfs4_delegreturndata *data = calldata;
4786
4787         if (!nfs4_sequence_done(task, &data->res.seq_res))
4788                 return;
4789
4790         switch (task->tk_status) {
4791         case -NFS4ERR_STALE_STATEID:
4792         case -NFS4ERR_EXPIRED:
4793         case 0:
4794                 renew_lease(data->res.server, data->timestamp);
4795                 break;
4796         default:
4797                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4798                                 -EAGAIN) {
4799                         rpc_restart_call_prepare(task);
4800                         return;
4801                 }
4802         }
4803         data->rpc_status = task->tk_status;
4804 }
4805
4806 static void nfs4_delegreturn_release(void *calldata)
4807 {
4808         kfree(calldata);
4809 }
4810
4811 #if defined(CONFIG_NFS_V4_1)
4812 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4813 {
4814         struct nfs4_delegreturndata *d_data;
4815
4816         d_data = (struct nfs4_delegreturndata *)data;
4817
4818         nfs4_setup_sequence(d_data->res.server,
4819                         &d_data->args.seq_args,
4820                         &d_data->res.seq_res,
4821                         task);
4822 }
4823 #endif /* CONFIG_NFS_V4_1 */
4824
4825 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4826 #if defined(CONFIG_NFS_V4_1)
4827         .rpc_call_prepare = nfs4_delegreturn_prepare,
4828 #endif /* CONFIG_NFS_V4_1 */
4829         .rpc_call_done = nfs4_delegreturn_done,
4830         .rpc_release = nfs4_delegreturn_release,
4831 };
4832
4833 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4834 {
4835         struct nfs4_delegreturndata *data;
4836         struct nfs_server *server = NFS_SERVER(inode);
4837         struct rpc_task *task;
4838         struct rpc_message msg = {
4839                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4840                 .rpc_cred = cred,
4841         };
4842         struct rpc_task_setup task_setup_data = {
4843                 .rpc_client = server->client,
4844                 .rpc_message = &msg,
4845                 .callback_ops = &nfs4_delegreturn_ops,
4846                 .flags = RPC_TASK_ASYNC,
4847         };
4848         int status = 0;
4849
4850         data = kzalloc(sizeof(*data), GFP_NOFS);
4851         if (data == NULL)
4852                 return -ENOMEM;
4853         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4854         data->args.fhandle = &data->fh;
4855         data->args.stateid = &data->stateid;
4856         data->args.bitmask = server->cache_consistency_bitmask;
4857         nfs_copy_fh(&data->fh, NFS_FH(inode));
4858         nfs4_stateid_copy(&data->stateid, stateid);
4859         data->res.fattr = &data->fattr;
4860         data->res.server = server;
4861         nfs_fattr_init(data->res.fattr);
4862         data->timestamp = jiffies;
4863         data->rpc_status = 0;
4864
4865         task_setup_data.callback_data = data;
4866         msg.rpc_argp = &data->args;
4867         msg.rpc_resp = &data->res;
4868         task = rpc_run_task(&task_setup_data);
4869         if (IS_ERR(task))
4870                 return PTR_ERR(task);
4871         if (!issync)
4872                 goto out;
4873         status = nfs4_wait_for_completion_rpc_task(task);
4874         if (status != 0)
4875                 goto out;
4876         status = data->rpc_status;
4877         if (status == 0)
4878                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4879         else
4880                 nfs_refresh_inode(inode, &data->fattr);
4881 out:
4882         rpc_put_task(task);
4883         return status;
4884 }
4885
4886 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4887 {
4888         struct nfs_server *server = NFS_SERVER(inode);
4889         struct nfs4_exception exception = { };
4890         int err;
4891         do {
4892                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4893                 switch (err) {
4894                         case -NFS4ERR_STALE_STATEID:
4895                         case -NFS4ERR_EXPIRED:
4896                         case 0:
4897                                 return 0;
4898                 }
4899                 err = nfs4_handle_exception(server, err, &exception);
4900         } while (exception.retry);
4901         return err;
4902 }
4903
4904 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4905 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4906
4907 /*
4908  * sleep, with exponential backoff, and retry the LOCK operation.
4909  */
4910 static unsigned long
4911 nfs4_set_lock_task_retry(unsigned long timeout)
4912 {
4913         freezable_schedule_timeout_killable_unsafe(timeout);
4914         timeout <<= 1;
4915         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4916                 return NFS4_LOCK_MAXTIMEOUT;
4917         return timeout;
4918 }
4919
4920 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4921 {
4922         struct inode *inode = state->inode;
4923         struct nfs_server *server = NFS_SERVER(inode);
4924         struct nfs_client *clp = server->nfs_client;
4925         struct nfs_lockt_args arg = {
4926                 .fh = NFS_FH(inode),
4927                 .fl = request,
4928         };
4929         struct nfs_lockt_res res = {
4930                 .denied = request,
4931         };
4932         struct rpc_message msg = {
4933                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4934                 .rpc_argp       = &arg,
4935                 .rpc_resp       = &res,
4936                 .rpc_cred       = state->owner->so_cred,
4937         };
4938         struct nfs4_lock_state *lsp;
4939         int status;
4940
4941         arg.lock_owner.clientid = clp->cl_clientid;
4942         status = nfs4_set_lock_state(state, request);
4943         if (status != 0)
4944                 goto out;
4945         lsp = request->fl_u.nfs4_fl.owner;
4946         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4947         arg.lock_owner.s_dev = server->s_dev;
4948         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4949         switch (status) {
4950                 case 0:
4951                         request->fl_type = F_UNLCK;
4952                         break;
4953                 case -NFS4ERR_DENIED:
4954                         status = 0;
4955         }
4956         request->fl_ops->fl_release_private(request);
4957 out:
4958         return status;
4959 }
4960
4961 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4962 {
4963         struct nfs4_exception exception = { };
4964         int err;
4965
4966         do {
4967                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4968                                 _nfs4_proc_getlk(state, cmd, request),
4969                                 &exception);
4970         } while (exception.retry);
4971         return err;
4972 }
4973
4974 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4975 {
4976         int res = 0;
4977         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4978                 case FL_POSIX:
4979                         res = posix_lock_file_wait(file, fl);
4980                         break;
4981                 case FL_FLOCK:
4982                         res = flock_lock_file_wait(file, fl);
4983                         break;
4984                 default:
4985                         BUG();
4986         }
4987         return res;
4988 }
4989
4990 struct nfs4_unlockdata {
4991         struct nfs_locku_args arg;
4992         struct nfs_locku_res res;
4993         struct nfs4_lock_state *lsp;
4994         struct nfs_open_context *ctx;
4995         struct file_lock fl;
4996         const struct nfs_server *server;
4997         unsigned long timestamp;
4998 };
4999
5000 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5001                 struct nfs_open_context *ctx,
5002                 struct nfs4_lock_state *lsp,
5003                 struct nfs_seqid *seqid)
5004 {
5005         struct nfs4_unlockdata *p;
5006         struct inode *inode = lsp->ls_state->inode;
5007
5008         p = kzalloc(sizeof(*p), GFP_NOFS);
5009         if (p == NULL)
5010                 return NULL;
5011         p->arg.fh = NFS_FH(inode);
5012         p->arg.fl = &p->fl;
5013         p->arg.seqid = seqid;
5014         p->res.seqid = seqid;
5015         p->arg.stateid = &lsp->ls_stateid;
5016         p->lsp = lsp;
5017         atomic_inc(&lsp->ls_count);
5018         /* Ensure we don't close file until we're done freeing locks! */
5019         p->ctx = get_nfs_open_context(ctx);
5020         memcpy(&p->fl, fl, sizeof(p->fl));
5021         p->server = NFS_SERVER(inode);
5022         return p;
5023 }
5024
5025 static void nfs4_locku_release_calldata(void *data)
5026 {
5027         struct nfs4_unlockdata *calldata = data;
5028         nfs_free_seqid(calldata->arg.seqid);
5029         nfs4_put_lock_state(calldata->lsp);
5030         put_nfs_open_context(calldata->ctx);
5031         kfree(calldata);
5032 }
5033
5034 static void nfs4_locku_done(struct rpc_task *task, void *data)
5035 {
5036         struct nfs4_unlockdata *calldata = data;
5037
5038         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5039                 return;
5040         switch (task->tk_status) {
5041                 case 0:
5042                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5043                                         &calldata->res.stateid);
5044                         renew_lease(calldata->server, calldata->timestamp);
5045                         break;
5046                 case -NFS4ERR_BAD_STATEID:
5047                 case -NFS4ERR_OLD_STATEID:
5048                 case -NFS4ERR_STALE_STATEID:
5049                 case -NFS4ERR_EXPIRED:
5050                         break;
5051                 default:
5052                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5053                                 rpc_restart_call_prepare(task);
5054         }
5055         nfs_release_seqid(calldata->arg.seqid);
5056 }
5057
5058 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5059 {
5060         struct nfs4_unlockdata *calldata = data;
5061
5062         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5063                 goto out_wait;
5064         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5065                 /* Note: exit _without_ running nfs4_locku_done */
5066                 goto out_no_action;
5067         }
5068         calldata->timestamp = jiffies;
5069         if (nfs4_setup_sequence(calldata->server,
5070                                 &calldata->arg.seq_args,
5071                                 &calldata->res.seq_res,
5072                                 task) != 0)
5073                 nfs_release_seqid(calldata->arg.seqid);
5074         return;
5075 out_no_action:
5076         task->tk_action = NULL;
5077 out_wait:
5078         nfs4_sequence_done(task, &calldata->res.seq_res);
5079 }
5080
5081 static const struct rpc_call_ops nfs4_locku_ops = {
5082         .rpc_call_prepare = nfs4_locku_prepare,
5083         .rpc_call_done = nfs4_locku_done,
5084         .rpc_release = nfs4_locku_release_calldata,
5085 };
5086
5087 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5088                 struct nfs_open_context *ctx,
5089                 struct nfs4_lock_state *lsp,
5090                 struct nfs_seqid *seqid)
5091 {
5092         struct nfs4_unlockdata *data;
5093         struct rpc_message msg = {
5094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5095                 .rpc_cred = ctx->cred,
5096         };
5097         struct rpc_task_setup task_setup_data = {
5098                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5099                 .rpc_message = &msg,
5100                 .callback_ops = &nfs4_locku_ops,
5101                 .workqueue = nfsiod_workqueue,
5102                 .flags = RPC_TASK_ASYNC,
5103         };
5104
5105         /* Ensure this is an unlock - when canceling a lock, the
5106          * canceled lock is passed in, and it won't be an unlock.
5107          */
5108         fl->fl_type = F_UNLCK;
5109
5110         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5111         if (data == NULL) {
5112                 nfs_free_seqid(seqid);
5113                 return ERR_PTR(-ENOMEM);
5114         }
5115
5116         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5117         msg.rpc_argp = &data->arg;
5118         msg.rpc_resp = &data->res;
5119         task_setup_data.callback_data = data;
5120         return rpc_run_task(&task_setup_data);
5121 }
5122
5123 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5124 {
5125         struct inode *inode = state->inode;
5126         struct nfs4_state_owner *sp = state->owner;
5127         struct nfs_inode *nfsi = NFS_I(inode);
5128         struct nfs_seqid *seqid;
5129         struct nfs4_lock_state *lsp;
5130         struct rpc_task *task;
5131         int status = 0;
5132         unsigned char fl_flags = request->fl_flags;
5133
5134         status = nfs4_set_lock_state(state, request);
5135         /* Unlock _before_ we do the RPC call */
5136         request->fl_flags |= FL_EXISTS;
5137         /* Exclude nfs_delegation_claim_locks() */
5138         mutex_lock(&sp->so_delegreturn_mutex);
5139         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5140         down_read(&nfsi->rwsem);
5141         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5142                 up_read(&nfsi->rwsem);
5143                 mutex_unlock(&sp->so_delegreturn_mutex);
5144                 goto out;
5145         }
5146         up_read(&nfsi->rwsem);
5147         mutex_unlock(&sp->so_delegreturn_mutex);
5148         if (status != 0)
5149                 goto out;
5150         /* Is this a delegated lock? */
5151         lsp = request->fl_u.nfs4_fl.owner;
5152         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5153                 goto out;
5154         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5155         status = -ENOMEM;
5156         if (seqid == NULL)
5157                 goto out;
5158         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5159         status = PTR_ERR(task);
5160         if (IS_ERR(task))
5161                 goto out;
5162         status = nfs4_wait_for_completion_rpc_task(task);
5163         rpc_put_task(task);
5164 out:
5165         request->fl_flags = fl_flags;
5166         return status;
5167 }
5168
5169 struct nfs4_lockdata {
5170         struct nfs_lock_args arg;
5171         struct nfs_lock_res res;
5172         struct nfs4_lock_state *lsp;
5173         struct nfs_open_context *ctx;
5174         struct file_lock fl;
5175         unsigned long timestamp;
5176         int rpc_status;
5177         int cancelled;
5178         struct nfs_server *server;
5179 };
5180
5181 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5182                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5183                 gfp_t gfp_mask)
5184 {
5185         struct nfs4_lockdata *p;
5186         struct inode *inode = lsp->ls_state->inode;
5187         struct nfs_server *server = NFS_SERVER(inode);
5188
5189         p = kzalloc(sizeof(*p), gfp_mask);
5190         if (p == NULL)
5191                 return NULL;
5192
5193         p->arg.fh = NFS_FH(inode);
5194         p->arg.fl = &p->fl;
5195         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5196         if (p->arg.open_seqid == NULL)
5197                 goto out_free;
5198         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5199         if (p->arg.lock_seqid == NULL)
5200                 goto out_free_seqid;
5201         p->arg.lock_stateid = &lsp->ls_stateid;
5202         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5203         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5204         p->arg.lock_owner.s_dev = server->s_dev;
5205         p->res.lock_seqid = p->arg.lock_seqid;
5206         p->lsp = lsp;
5207         p->server = server;
5208         atomic_inc(&lsp->ls_count);
5209         p->ctx = get_nfs_open_context(ctx);
5210         memcpy(&p->fl, fl, sizeof(p->fl));
5211         return p;
5212 out_free_seqid:
5213         nfs_free_seqid(p->arg.open_seqid);
5214 out_free:
5215         kfree(p);
5216         return NULL;
5217 }
5218
5219 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5220 {
5221         struct nfs4_lockdata *data = calldata;
5222         struct nfs4_state *state = data->lsp->ls_state;
5223
5224         dprintk("%s: begin!\n", __func__);
5225         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5226                 goto out_wait;
5227         /* Do we need to do an open_to_lock_owner? */
5228         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5229                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5230                         goto out_release_lock_seqid;
5231                 }
5232                 data->arg.open_stateid = &state->open_stateid;
5233                 data->arg.new_lock_owner = 1;
5234                 data->res.open_seqid = data->arg.open_seqid;
5235         } else
5236                 data->arg.new_lock_owner = 0;
5237         if (!nfs4_valid_open_stateid(state)) {
5238                 data->rpc_status = -EBADF;
5239                 task->tk_action = NULL;
5240                 goto out_release_open_seqid;
5241         }
5242         data->timestamp = jiffies;
5243         if (nfs4_setup_sequence(data->server,
5244                                 &data->arg.seq_args,
5245                                 &data->res.seq_res,
5246                                 task) == 0)
5247                 return;
5248 out_release_open_seqid:
5249         nfs_release_seqid(data->arg.open_seqid);
5250 out_release_lock_seqid:
5251         nfs_release_seqid(data->arg.lock_seqid);
5252 out_wait:
5253         nfs4_sequence_done(task, &data->res.seq_res);
5254         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5255 }
5256
5257 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5258 {
5259         struct nfs4_lockdata *data = calldata;
5260
5261         dprintk("%s: begin!\n", __func__);
5262
5263         if (!nfs4_sequence_done(task, &data->res.seq_res))
5264                 return;
5265
5266         data->rpc_status = task->tk_status;
5267         if (data->arg.new_lock_owner != 0) {
5268                 if (data->rpc_status == 0)
5269                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5270                 else
5271                         goto out;
5272         }
5273         if (data->rpc_status == 0) {
5274                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5275                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5276                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5277         }
5278 out:
5279         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5280 }
5281
5282 static void nfs4_lock_release(void *calldata)
5283 {
5284         struct nfs4_lockdata *data = calldata;
5285
5286         dprintk("%s: begin!\n", __func__);
5287         nfs_free_seqid(data->arg.open_seqid);
5288         if (data->cancelled != 0) {
5289                 struct rpc_task *task;
5290                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5291                                 data->arg.lock_seqid);
5292                 if (!IS_ERR(task))
5293                         rpc_put_task_async(task);
5294                 dprintk("%s: cancelling lock!\n", __func__);
5295         } else
5296                 nfs_free_seqid(data->arg.lock_seqid);
5297         nfs4_put_lock_state(data->lsp);
5298         put_nfs_open_context(data->ctx);
5299         kfree(data);
5300         dprintk("%s: done!\n", __func__);
5301 }
5302
5303 static const struct rpc_call_ops nfs4_lock_ops = {
5304         .rpc_call_prepare = nfs4_lock_prepare,
5305         .rpc_call_done = nfs4_lock_done,
5306         .rpc_release = nfs4_lock_release,
5307 };
5308
5309 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5310 {
5311         switch (error) {
5312         case -NFS4ERR_ADMIN_REVOKED:
5313         case -NFS4ERR_BAD_STATEID:
5314                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5315                 if (new_lock_owner != 0 ||
5316                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5317                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5318                 break;
5319         case -NFS4ERR_STALE_STATEID:
5320                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5321         case -NFS4ERR_EXPIRED:
5322                 nfs4_schedule_lease_recovery(server->nfs_client);
5323         };
5324 }
5325
5326 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5327 {
5328         struct nfs4_lockdata *data;
5329         struct rpc_task *task;
5330         struct rpc_message msg = {
5331                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5332                 .rpc_cred = state->owner->so_cred,
5333         };
5334         struct rpc_task_setup task_setup_data = {
5335                 .rpc_client = NFS_CLIENT(state->inode),
5336                 .rpc_message = &msg,
5337                 .callback_ops = &nfs4_lock_ops,
5338                 .workqueue = nfsiod_workqueue,
5339                 .flags = RPC_TASK_ASYNC,
5340         };
5341         int ret;
5342
5343         dprintk("%s: begin!\n", __func__);
5344         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5345                         fl->fl_u.nfs4_fl.owner,
5346                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5347         if (data == NULL)
5348                 return -ENOMEM;
5349         if (IS_SETLKW(cmd))
5350                 data->arg.block = 1;
5351         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5352         msg.rpc_argp = &data->arg;
5353         msg.rpc_resp = &data->res;
5354         task_setup_data.callback_data = data;
5355         if (recovery_type > NFS_LOCK_NEW) {
5356                 if (recovery_type == NFS_LOCK_RECLAIM)
5357                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5358                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5359         }
5360         task = rpc_run_task(&task_setup_data);
5361         if (IS_ERR(task))
5362                 return PTR_ERR(task);
5363         ret = nfs4_wait_for_completion_rpc_task(task);
5364         if (ret == 0) {
5365                 ret = data->rpc_status;
5366                 if (ret)
5367                         nfs4_handle_setlk_error(data->server, data->lsp,
5368                                         data->arg.new_lock_owner, ret);
5369         } else
5370                 data->cancelled = 1;
5371         rpc_put_task(task);
5372         dprintk("%s: done, ret = %d!\n", __func__, ret);
5373         return ret;
5374 }
5375
5376 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5377 {
5378         struct nfs_server *server = NFS_SERVER(state->inode);
5379         struct nfs4_exception exception = {
5380                 .inode = state->inode,
5381         };
5382         int err;
5383
5384         do {
5385                 /* Cache the lock if possible... */
5386                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5387                         return 0;
5388                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5389                 if (err != -NFS4ERR_DELAY)
5390                         break;
5391                 nfs4_handle_exception(server, err, &exception);
5392         } while (exception.retry);
5393         return err;
5394 }
5395
5396 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5397 {
5398         struct nfs_server *server = NFS_SERVER(state->inode);
5399         struct nfs4_exception exception = {
5400                 .inode = state->inode,
5401         };
5402         int err;
5403
5404         err = nfs4_set_lock_state(state, request);
5405         if (err != 0)
5406                 return err;
5407         do {
5408                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5409                         return 0;
5410                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5411                 switch (err) {
5412                 default:
5413                         goto out;
5414                 case -NFS4ERR_GRACE:
5415                 case -NFS4ERR_DELAY:
5416                         nfs4_handle_exception(server, err, &exception);
5417                         err = 0;
5418                 }
5419         } while (exception.retry);
5420 out:
5421         return err;
5422 }
5423
5424 #if defined(CONFIG_NFS_V4_1)
5425 /**
5426  * nfs41_check_expired_locks - possibly free a lock stateid
5427  *
5428  * @state: NFSv4 state for an inode
5429  *
5430  * Returns NFS_OK if recovery for this stateid is now finished.
5431  * Otherwise a negative NFS4ERR value is returned.
5432  */
5433 static int nfs41_check_expired_locks(struct nfs4_state *state)
5434 {
5435         int status, ret = -NFS4ERR_BAD_STATEID;
5436         struct nfs4_lock_state *lsp;
5437         struct nfs_server *server = NFS_SERVER(state->inode);
5438
5439         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5440                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5441                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5442
5443                         status = nfs41_test_stateid(server,
5444                                         &lsp->ls_stateid,
5445                                         cred);
5446                         if (status != NFS_OK) {
5447                                 /* Free the stateid unless the server
5448                                  * informs us the stateid is unrecognized. */
5449                                 if (status != -NFS4ERR_BAD_STATEID)
5450                                         nfs41_free_stateid(server,
5451                                                         &lsp->ls_stateid,
5452                                                         cred);
5453                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5454                                 ret = status;
5455                         }
5456                 }
5457         };
5458
5459         return ret;
5460 }
5461
5462 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5463 {
5464         int status = NFS_OK;
5465
5466         if (test_bit(LK_STATE_IN_USE, &state->flags))
5467                 status = nfs41_check_expired_locks(state);
5468         if (status != NFS_OK)
5469                 status = nfs4_lock_expired(state, request);
5470         return status;
5471 }
5472 #endif
5473
5474 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5475 {
5476         struct nfs4_state_owner *sp = state->owner;
5477         struct nfs_inode *nfsi = NFS_I(state->inode);
5478         unsigned char fl_flags = request->fl_flags;
5479         unsigned int seq;
5480         int status = -ENOLCK;
5481
5482         if ((fl_flags & FL_POSIX) &&
5483                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5484                 goto out;
5485         /* Is this a delegated open? */
5486         status = nfs4_set_lock_state(state, request);
5487         if (status != 0)
5488                 goto out;
5489         request->fl_flags |= FL_ACCESS;
5490         status = do_vfs_lock(request->fl_file, request);
5491         if (status < 0)
5492                 goto out;
5493         down_read(&nfsi->rwsem);
5494         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5495                 /* Yes: cache locks! */
5496                 /* ...but avoid races with delegation recall... */
5497                 request->fl_flags = fl_flags & ~FL_SLEEP;
5498                 status = do_vfs_lock(request->fl_file, request);
5499                 goto out_unlock;
5500         }
5501         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5502         up_read(&nfsi->rwsem);
5503         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5504         if (status != 0)
5505                 goto out;
5506         down_read(&nfsi->rwsem);
5507         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5508                 status = -NFS4ERR_DELAY;
5509                 goto out_unlock;
5510         }
5511         /* Note: we always want to sleep here! */
5512         request->fl_flags = fl_flags | FL_SLEEP;
5513         if (do_vfs_lock(request->fl_file, request) < 0)
5514                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5515                         "manager!\n", __func__);
5516 out_unlock:
5517         up_read(&nfsi->rwsem);
5518 out:
5519         request->fl_flags = fl_flags;
5520         return status;
5521 }
5522
5523 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5524 {
5525         struct nfs4_exception exception = {
5526                 .state = state,
5527                 .inode = state->inode,
5528         };
5529         int err;
5530
5531         do {
5532                 err = _nfs4_proc_setlk(state, cmd, request);
5533                 if (err == -NFS4ERR_DENIED)
5534                         err = -EAGAIN;
5535                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5536                                 err, &exception);
5537         } while (exception.retry);
5538         return err;
5539 }
5540
5541 static int
5542 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5543 {
5544         struct nfs_open_context *ctx;
5545         struct nfs4_state *state;
5546         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5547         int status;
5548
5549         /* verify open state */
5550         ctx = nfs_file_open_context(filp);
5551         state = ctx->state;
5552
5553         if (request->fl_start < 0 || request->fl_end < 0)
5554                 return -EINVAL;
5555
5556         if (IS_GETLK(cmd)) {
5557                 if (state != NULL)
5558                         return nfs4_proc_getlk(state, F_GETLK, request);
5559                 return 0;
5560         }
5561
5562         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5563                 return -EINVAL;
5564
5565         if (request->fl_type == F_UNLCK) {
5566                 if (state != NULL)
5567                         return nfs4_proc_unlck(state, cmd, request);
5568                 return 0;
5569         }
5570
5571         if (state == NULL)
5572                 return -ENOLCK;
5573         /*
5574          * Don't rely on the VFS having checked the file open mode,
5575          * since it won't do this for flock() locks.
5576          */
5577         switch (request->fl_type) {
5578         case F_RDLCK:
5579                 if (!(filp->f_mode & FMODE_READ))
5580                         return -EBADF;
5581                 break;
5582         case F_WRLCK:
5583                 if (!(filp->f_mode & FMODE_WRITE))
5584                         return -EBADF;
5585         }
5586
5587         do {
5588                 status = nfs4_proc_setlk(state, cmd, request);
5589                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5590                         break;
5591                 timeout = nfs4_set_lock_task_retry(timeout);
5592                 status = -ERESTARTSYS;
5593                 if (signalled())
5594                         break;
5595         } while(status < 0);
5596         return status;
5597 }
5598
5599 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5600 {
5601         struct nfs_server *server = NFS_SERVER(state->inode);
5602         int err;
5603
5604         err = nfs4_set_lock_state(state, fl);
5605         if (err != 0)
5606                 return err;
5607         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5608         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5609 }
5610
5611 struct nfs_release_lockowner_data {
5612         struct nfs4_lock_state *lsp;
5613         struct nfs_server *server;
5614         struct nfs_release_lockowner_args args;
5615 };
5616
5617 static void nfs4_release_lockowner_release(void *calldata)
5618 {
5619         struct nfs_release_lockowner_data *data = calldata;
5620         nfs4_free_lock_state(data->server, data->lsp);
5621         kfree(calldata);
5622 }
5623
5624 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5625         .rpc_release = nfs4_release_lockowner_release,
5626 };
5627
5628 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5629 {
5630         struct nfs_release_lockowner_data *data;
5631         struct rpc_message msg = {
5632                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5633         };
5634
5635         if (server->nfs_client->cl_mvops->minor_version != 0)
5636                 return -EINVAL;
5637         data = kmalloc(sizeof(*data), GFP_NOFS);
5638         if (!data)
5639                 return -ENOMEM;
5640         data->lsp = lsp;
5641         data->server = server;
5642         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5643         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5644         data->args.lock_owner.s_dev = server->s_dev;
5645         msg.rpc_argp = &data->args;
5646         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5647         return 0;
5648 }
5649
5650 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5651
5652 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5653                                    const void *buf, size_t buflen,
5654                                    int flags, int type)
5655 {
5656         if (strcmp(key, "") != 0)
5657                 return -EINVAL;
5658
5659         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5660 }
5661
5662 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5663                                    void *buf, size_t buflen, int type)
5664 {
5665         if (strcmp(key, "") != 0)
5666                 return -EINVAL;
5667
5668         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5669 }
5670
5671 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5672                                        size_t list_len, const char *name,
5673                                        size_t name_len, int type)
5674 {
5675         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5676
5677         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5678                 return 0;
5679
5680         if (list && len <= list_len)
5681                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5682         return len;
5683 }
5684
5685 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5686 static inline int nfs4_server_supports_labels(struct nfs_server *server)
5687 {
5688         return server->caps & NFS_CAP_SECURITY_LABEL;
5689 }
5690
5691 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
5692                                    const void *buf, size_t buflen,
5693                                    int flags, int type)
5694 {
5695         if (security_ismaclabel(key))
5696                 return nfs4_set_security_label(dentry, buf, buflen);
5697
5698         return -EOPNOTSUPP;
5699 }
5700
5701 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
5702                                    void *buf, size_t buflen, int type)
5703 {
5704         if (security_ismaclabel(key))
5705                 return nfs4_get_security_label(dentry->d_inode, buf, buflen);
5706         return -EOPNOTSUPP;
5707 }
5708
5709 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
5710                                        size_t list_len, const char *name,
5711                                        size_t name_len, int type)
5712 {
5713         size_t len = 0;
5714
5715         if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
5716                 len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
5717                 if (list && len <= list_len)
5718                         security_inode_listsecurity(dentry->d_inode, list, len);
5719         }
5720         return len;
5721 }
5722
5723 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
5724         .prefix = XATTR_SECURITY_PREFIX,
5725         .list   = nfs4_xattr_list_nfs4_label,
5726         .get    = nfs4_xattr_get_nfs4_label,
5727         .set    = nfs4_xattr_set_nfs4_label,
5728 };
5729 #endif
5730
5731
5732 /*
5733  * nfs_fhget will use either the mounted_on_fileid or the fileid
5734  */
5735 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5736 {
5737         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5738                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5739               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5740               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5741                 return;
5742
5743         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5744                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5745         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5746         fattr->nlink = 2;
5747 }
5748
5749 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5750                                    const struct qstr *name,
5751                                    struct nfs4_fs_locations *fs_locations,
5752                                    struct page *page)
5753 {
5754         struct nfs_server *server = NFS_SERVER(dir);
5755         u32 bitmask[3] = {
5756                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5757         };
5758         struct nfs4_fs_locations_arg args = {
5759                 .dir_fh = NFS_FH(dir),
5760                 .name = name,
5761                 .page = page,
5762                 .bitmask = bitmask,
5763         };
5764         struct nfs4_fs_locations_res res = {
5765                 .fs_locations = fs_locations,
5766         };
5767         struct rpc_message msg = {
5768                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5769                 .rpc_argp = &args,
5770                 .rpc_resp = &res,
5771         };
5772         int status;
5773
5774         dprintk("%s: start\n", __func__);
5775
5776         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5777          * is not supported */
5778         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5779                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5780         else
5781                 bitmask[0] |= FATTR4_WORD0_FILEID;
5782
5783         nfs_fattr_init(&fs_locations->fattr);
5784         fs_locations->server = server;
5785         fs_locations->nlocations = 0;
5786         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5787         dprintk("%s: returned status = %d\n", __func__, status);
5788         return status;
5789 }
5790
5791 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5792                            const struct qstr *name,
5793                            struct nfs4_fs_locations *fs_locations,
5794                            struct page *page)
5795 {
5796         struct nfs4_exception exception = { };
5797         int err;
5798         do {
5799                 err = nfs4_handle_exception(NFS_SERVER(dir),
5800                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5801                                 &exception);
5802         } while (exception.retry);
5803         return err;
5804 }
5805
5806 /**
5807  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
5808  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
5809  */
5810 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5811 {
5812         int status;
5813         struct nfs4_secinfo_arg args = {
5814                 .dir_fh = NFS_FH(dir),
5815                 .name   = name,
5816         };
5817         struct nfs4_secinfo_res res = {
5818                 .flavors     = flavors,
5819         };
5820         struct rpc_message msg = {
5821                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5822                 .rpc_argp = &args,
5823                 .rpc_resp = &res,
5824         };
5825         struct rpc_clnt *clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
5826
5827         dprintk("NFS call  secinfo %s\n", name->name);
5828         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5829         dprintk("NFS reply  secinfo: %d\n", status);
5830         return status;
5831 }
5832
5833 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5834                       struct nfs4_secinfo_flavors *flavors)
5835 {
5836         struct nfs4_exception exception = { };
5837         int err;
5838         do {
5839                 err = nfs4_handle_exception(NFS_SERVER(dir),
5840                                 _nfs4_proc_secinfo(dir, name, flavors),
5841                                 &exception);
5842         } while (exception.retry);
5843         return err;
5844 }
5845
5846 #ifdef CONFIG_NFS_V4_1
5847 /*
5848  * Check the exchange flags returned by the server for invalid flags, having
5849  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5850  * DS flags set.
5851  */
5852 static int nfs4_check_cl_exchange_flags(u32 flags)
5853 {
5854         if (flags & ~EXCHGID4_FLAG_MASK_R)
5855                 goto out_inval;
5856         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5857             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5858                 goto out_inval;
5859         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5860                 goto out_inval;
5861         return NFS_OK;
5862 out_inval:
5863         return -NFS4ERR_INVAL;
5864 }
5865
5866 static bool
5867 nfs41_same_server_scope(struct nfs41_server_scope *a,
5868                         struct nfs41_server_scope *b)
5869 {
5870         if (a->server_scope_sz == b->server_scope_sz &&
5871             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5872                 return true;
5873
5874         return false;
5875 }
5876
5877 /*
5878  * nfs4_proc_bind_conn_to_session()
5879  *
5880  * The 4.1 client currently uses the same TCP connection for the
5881  * fore and backchannel.
5882  */
5883 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5884 {
5885         int status;
5886         struct nfs41_bind_conn_to_session_res res;
5887         struct rpc_message msg = {
5888                 .rpc_proc =
5889                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5890                 .rpc_argp = clp,
5891                 .rpc_resp = &res,
5892                 .rpc_cred = cred,
5893         };
5894
5895         dprintk("--> %s\n", __func__);
5896
5897         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5898         if (unlikely(res.session == NULL)) {
5899                 status = -ENOMEM;
5900                 goto out;
5901         }
5902
5903         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5904         trace_nfs4_bind_conn_to_session(clp, status);
5905         if (status == 0) {
5906                 if (memcmp(res.session->sess_id.data,
5907                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5908                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5909                         status = -EIO;
5910                         goto out_session;
5911                 }
5912                 if (res.dir != NFS4_CDFS4_BOTH) {
5913                         dprintk("NFS: %s: Unexpected direction from server\n",
5914                                 __func__);
5915                         status = -EIO;
5916                         goto out_session;
5917                 }
5918                 if (res.use_conn_in_rdma_mode) {
5919                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5920                                 __func__);
5921                         status = -EIO;
5922                         goto out_session;
5923                 }
5924         }
5925 out_session:
5926         kfree(res.session);
5927 out:
5928         dprintk("<-- %s status= %d\n", __func__, status);
5929         return status;
5930 }
5931
5932 /*
5933  * nfs4_proc_exchange_id()
5934  *
5935  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5936  *
5937  * Since the clientid has expired, all compounds using sessions
5938  * associated with the stale clientid will be returning
5939  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5940  * be in some phase of session reset.
5941  */
5942 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5943 {
5944         nfs4_verifier verifier;
5945         struct nfs41_exchange_id_args args = {
5946                 .verifier = &verifier,
5947                 .client = clp,
5948                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
5949                         EXCHGID4_FLAG_BIND_PRINC_STATEID,
5950         };
5951         struct nfs41_exchange_id_res res = {
5952                 0
5953         };
5954         int status;
5955         struct rpc_message msg = {
5956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5957                 .rpc_argp = &args,
5958                 .rpc_resp = &res,
5959                 .rpc_cred = cred,
5960         };
5961
5962         nfs4_init_boot_verifier(clp, &verifier);
5963         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5964                                                         sizeof(args.id));
5965         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5966                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5967                 args.id_len, args.id);
5968
5969         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5970                                         GFP_NOFS);
5971         if (unlikely(res.server_owner == NULL)) {
5972                 status = -ENOMEM;
5973                 goto out;
5974         }
5975
5976         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5977                                         GFP_NOFS);
5978         if (unlikely(res.server_scope == NULL)) {
5979                 status = -ENOMEM;
5980                 goto out_server_owner;
5981         }
5982
5983         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5984         if (unlikely(res.impl_id == NULL)) {
5985                 status = -ENOMEM;
5986                 goto out_server_scope;
5987         }
5988
5989         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5990         trace_nfs4_exchange_id(clp, status);
5991         if (status == 0)
5992                 status = nfs4_check_cl_exchange_flags(res.flags);
5993
5994         if (status == 0) {
5995                 clp->cl_clientid = res.clientid;
5996                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5997                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5998                         clp->cl_seqid = res.seqid;
5999
6000                 kfree(clp->cl_serverowner);
6001                 clp->cl_serverowner = res.server_owner;
6002                 res.server_owner = NULL;
6003
6004                 /* use the most recent implementation id */
6005                 kfree(clp->cl_implid);
6006                 clp->cl_implid = res.impl_id;
6007
6008                 if (clp->cl_serverscope != NULL &&
6009                     !nfs41_same_server_scope(clp->cl_serverscope,
6010                                              res.server_scope)) {
6011                         dprintk("%s: server_scope mismatch detected\n",
6012                                 __func__);
6013                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6014                         kfree(clp->cl_serverscope);
6015                         clp->cl_serverscope = NULL;
6016                 }
6017
6018                 if (clp->cl_serverscope == NULL) {
6019                         clp->cl_serverscope = res.server_scope;
6020                         goto out;
6021                 }
6022         } else
6023                 kfree(res.impl_id);
6024
6025 out_server_owner:
6026         kfree(res.server_owner);
6027 out_server_scope:
6028         kfree(res.server_scope);
6029 out:
6030         if (clp->cl_implid != NULL)
6031                 dprintk("NFS reply exchange_id: Server Implementation ID: "
6032                         "domain: %s, name: %s, date: %llu,%u\n",
6033                         clp->cl_implid->domain, clp->cl_implid->name,
6034                         clp->cl_implid->date.seconds,
6035                         clp->cl_implid->date.nseconds);
6036         dprintk("NFS reply exchange_id: %d\n", status);
6037         return status;
6038 }
6039
6040 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6041                 struct rpc_cred *cred)
6042 {
6043         struct rpc_message msg = {
6044                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6045                 .rpc_argp = clp,
6046                 .rpc_cred = cred,
6047         };
6048         int status;
6049
6050         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6051         trace_nfs4_destroy_clientid(clp, status);
6052         if (status)
6053                 dprintk("NFS: Got error %d from the server %s on "
6054                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
6055         return status;
6056 }
6057
6058 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6059                 struct rpc_cred *cred)
6060 {
6061         unsigned int loop;
6062         int ret;
6063
6064         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6065                 ret = _nfs4_proc_destroy_clientid(clp, cred);
6066                 switch (ret) {
6067                 case -NFS4ERR_DELAY:
6068                 case -NFS4ERR_CLIENTID_BUSY:
6069                         ssleep(1);
6070                         break;
6071                 default:
6072                         return ret;
6073                 }
6074         }
6075         return 0;
6076 }
6077
6078 int nfs4_destroy_clientid(struct nfs_client *clp)
6079 {
6080         struct rpc_cred *cred;
6081         int ret = 0;
6082
6083         if (clp->cl_mvops->minor_version < 1)
6084                 goto out;
6085         if (clp->cl_exchange_flags == 0)
6086                 goto out;
6087         if (clp->cl_preserve_clid)
6088                 goto out;
6089         cred = nfs4_get_clid_cred(clp);
6090         ret = nfs4_proc_destroy_clientid(clp, cred);
6091         if (cred)
6092                 put_rpccred(cred);
6093         switch (ret) {
6094         case 0:
6095         case -NFS4ERR_STALE_CLIENTID:
6096                 clp->cl_exchange_flags = 0;
6097         }
6098 out:
6099         return ret;
6100 }
6101
6102 struct nfs4_get_lease_time_data {
6103         struct nfs4_get_lease_time_args *args;
6104         struct nfs4_get_lease_time_res *res;
6105         struct nfs_client *clp;
6106 };
6107
6108 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6109                                         void *calldata)
6110 {
6111         struct nfs4_get_lease_time_data *data =
6112                         (struct nfs4_get_lease_time_data *)calldata;
6113
6114         dprintk("--> %s\n", __func__);
6115         /* just setup sequence, do not trigger session recovery
6116            since we're invoked within one */
6117         nfs41_setup_sequence(data->clp->cl_session,
6118                         &data->args->la_seq_args,
6119                         &data->res->lr_seq_res,
6120                         task);
6121         dprintk("<-- %s\n", __func__);
6122 }
6123
6124 /*
6125  * Called from nfs4_state_manager thread for session setup, so don't recover
6126  * from sequence operation or clientid errors.
6127  */
6128 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6129 {
6130         struct nfs4_get_lease_time_data *data =
6131                         (struct nfs4_get_lease_time_data *)calldata;
6132
6133         dprintk("--> %s\n", __func__);
6134         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6135                 return;
6136         switch (task->tk_status) {
6137         case -NFS4ERR_DELAY:
6138         case -NFS4ERR_GRACE:
6139                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6140                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
6141                 task->tk_status = 0;
6142                 /* fall through */
6143         case -NFS4ERR_RETRY_UNCACHED_REP:
6144                 rpc_restart_call_prepare(task);
6145                 return;
6146         }
6147         dprintk("<-- %s\n", __func__);
6148 }
6149
6150 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
6151         .rpc_call_prepare = nfs4_get_lease_time_prepare,
6152         .rpc_call_done = nfs4_get_lease_time_done,
6153 };
6154
6155 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
6156 {
6157         struct rpc_task *task;
6158         struct nfs4_get_lease_time_args args;
6159         struct nfs4_get_lease_time_res res = {
6160                 .lr_fsinfo = fsinfo,
6161         };
6162         struct nfs4_get_lease_time_data data = {
6163                 .args = &args,
6164                 .res = &res,
6165                 .clp = clp,
6166         };
6167         struct rpc_message msg = {
6168                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
6169                 .rpc_argp = &args,
6170                 .rpc_resp = &res,
6171         };
6172         struct rpc_task_setup task_setup = {
6173                 .rpc_client = clp->cl_rpcclient,
6174                 .rpc_message = &msg,
6175                 .callback_ops = &nfs4_get_lease_time_ops,
6176                 .callback_data = &data,
6177                 .flags = RPC_TASK_TIMEOUT,
6178         };
6179         int status;
6180
6181         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
6182         nfs4_set_sequence_privileged(&args.la_seq_args);
6183         dprintk("--> %s\n", __func__);
6184         task = rpc_run_task(&task_setup);
6185
6186         if (IS_ERR(task))
6187                 status = PTR_ERR(task);
6188         else {
6189                 status = task->tk_status;
6190                 rpc_put_task(task);
6191         }
6192         dprintk("<-- %s return %d\n", __func__, status);
6193
6194         return status;
6195 }
6196
6197 /*
6198  * Initialize the values to be used by the client in CREATE_SESSION
6199  * If nfs4_init_session set the fore channel request and response sizes,
6200  * use them.
6201  *
6202  * Set the back channel max_resp_sz_cached to zero to force the client to
6203  * always set csa_cachethis to FALSE because the current implementation
6204  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
6205  */
6206 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
6207 {
6208         unsigned int max_rqst_sz, max_resp_sz;
6209
6210         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
6211         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
6212
6213         /* Fore channel attributes */
6214         args->fc_attrs.max_rqst_sz = max_rqst_sz;
6215         args->fc_attrs.max_resp_sz = max_resp_sz;
6216         args->fc_attrs.max_ops = NFS4_MAX_OPS;
6217         args->fc_attrs.max_reqs = max_session_slots;
6218
6219         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
6220                 "max_ops=%u max_reqs=%u\n",
6221                 __func__,
6222                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
6223                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
6224
6225         /* Back channel attributes */
6226         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
6227         args->bc_attrs.max_resp_sz = PAGE_SIZE;
6228         args->bc_attrs.max_resp_sz_cached = 0;
6229         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
6230         args->bc_attrs.max_reqs = 1;
6231
6232         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
6233                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
6234                 __func__,
6235                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
6236                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
6237                 args->bc_attrs.max_reqs);
6238 }
6239
6240 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6241 {
6242         struct nfs4_channel_attrs *sent = &args->fc_attrs;
6243         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
6244
6245         if (rcvd->max_resp_sz > sent->max_resp_sz)
6246                 return -EINVAL;
6247         /*
6248          * Our requested max_ops is the minimum we need; we're not
6249          * prepared to break up compounds into smaller pieces than that.
6250          * So, no point even trying to continue if the server won't
6251          * cooperate:
6252          */
6253         if (rcvd->max_ops < sent->max_ops)
6254                 return -EINVAL;
6255         if (rcvd->max_reqs == 0)
6256                 return -EINVAL;
6257         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
6258                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
6259         return 0;
6260 }
6261
6262 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6263 {
6264         struct nfs4_channel_attrs *sent = &args->bc_attrs;
6265         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
6266
6267         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
6268                 return -EINVAL;
6269         if (rcvd->max_resp_sz < sent->max_resp_sz)
6270                 return -EINVAL;
6271         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
6272                 return -EINVAL;
6273         /* These would render the backchannel useless: */
6274         if (rcvd->max_ops != sent->max_ops)
6275                 return -EINVAL;
6276         if (rcvd->max_reqs != sent->max_reqs)
6277                 return -EINVAL;
6278         return 0;
6279 }
6280
6281 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
6282                                      struct nfs4_session *session)
6283 {
6284         int ret;
6285
6286         ret = nfs4_verify_fore_channel_attrs(args, session);
6287         if (ret)
6288                 return ret;
6289         return nfs4_verify_back_channel_attrs(args, session);
6290 }
6291
6292 static int _nfs4_proc_create_session(struct nfs_client *clp,
6293                 struct rpc_cred *cred)
6294 {
6295         struct nfs4_session *session = clp->cl_session;
6296         struct nfs41_create_session_args args = {
6297                 .client = clp,
6298                 .cb_program = NFS4_CALLBACK,
6299         };
6300         struct nfs41_create_session_res res = {
6301                 .client = clp,
6302         };
6303         struct rpc_message msg = {
6304                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
6305                 .rpc_argp = &args,
6306                 .rpc_resp = &res,
6307                 .rpc_cred = cred,
6308         };
6309         int status;
6310
6311         nfs4_init_channel_attrs(&args);
6312         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
6313
6314         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6315         trace_nfs4_create_session(clp, status);
6316
6317         if (!status) {
6318                 /* Verify the session's negotiated channel_attrs values */
6319                 status = nfs4_verify_channel_attrs(&args, session);
6320                 /* Increment the clientid slot sequence id */
6321                 clp->cl_seqid++;
6322         }
6323
6324         return status;
6325 }
6326
6327 /*
6328  * Issues a CREATE_SESSION operation to the server.
6329  * It is the responsibility of the caller to verify the session is
6330  * expired before calling this routine.
6331  */
6332 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
6333 {
6334         int status;
6335         unsigned *ptr;
6336         struct nfs4_session *session = clp->cl_session;
6337
6338         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
6339
6340         status = _nfs4_proc_create_session(clp, cred);
6341         if (status)
6342                 goto out;
6343
6344         /* Init or reset the session slot tables */
6345         status = nfs4_setup_session_slot_tables(session);
6346         dprintk("slot table setup returned %d\n", status);
6347         if (status)
6348                 goto out;
6349
6350         ptr = (unsigned *)&session->sess_id.data[0];
6351         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
6352                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
6353 out:
6354         dprintk("<-- %s\n", __func__);
6355         return status;
6356 }
6357
6358 /*
6359  * Issue the over-the-wire RPC DESTROY_SESSION.
6360  * The caller must serialize access to this routine.
6361  */
6362 int nfs4_proc_destroy_session(struct nfs4_session *session,
6363                 struct rpc_cred *cred)
6364 {
6365         struct rpc_message msg = {
6366                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
6367                 .rpc_argp = session,
6368                 .rpc_cred = cred,
6369         };
6370         int status = 0;
6371
6372         dprintk("--> nfs4_proc_destroy_session\n");
6373
6374         /* session is still being setup */
6375         if (session->clp->cl_cons_state != NFS_CS_READY)
6376                 return status;
6377
6378         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6379         trace_nfs4_destroy_session(session->clp, status);
6380
6381         if (status)
6382                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
6383                         "Session has been destroyed regardless...\n", status);
6384
6385         dprintk("<-- nfs4_proc_destroy_session\n");
6386         return status;
6387 }
6388
6389 /*
6390  * Renew the cl_session lease.
6391  */
6392 struct nfs4_sequence_data {
6393         struct nfs_client *clp;
6394         struct nfs4_sequence_args args;
6395         struct nfs4_sequence_res res;
6396 };
6397
6398 static void nfs41_sequence_release(void *data)
6399 {
6400         struct nfs4_sequence_data *calldata = data;
6401         struct nfs_client *clp = calldata->clp;
6402
6403         if (atomic_read(&clp->cl_count) > 1)
6404                 nfs4_schedule_state_renewal(clp);
6405         nfs_put_client(clp);
6406         kfree(calldata);
6407 }
6408
6409 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6410 {
6411         switch(task->tk_status) {
6412         case -NFS4ERR_DELAY:
6413                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6414                 return -EAGAIN;
6415         default:
6416                 nfs4_schedule_lease_recovery(clp);
6417         }
6418         return 0;
6419 }
6420
6421 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6422 {
6423         struct nfs4_sequence_data *calldata = data;
6424         struct nfs_client *clp = calldata->clp;
6425
6426         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6427                 return;
6428
6429         trace_nfs4_sequence(clp, task->tk_status);
6430         if (task->tk_status < 0) {
6431                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
6432                 if (atomic_read(&clp->cl_count) == 1)
6433                         goto out;
6434
6435                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6436                         rpc_restart_call_prepare(task);
6437                         return;
6438                 }
6439         }
6440         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6441 out:
6442         dprintk("<-- %s\n", __func__);
6443 }
6444
6445 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6446 {
6447         struct nfs4_sequence_data *calldata = data;
6448         struct nfs_client *clp = calldata->clp;
6449         struct nfs4_sequence_args *args;
6450         struct nfs4_sequence_res *res;
6451
6452         args = task->tk_msg.rpc_argp;
6453         res = task->tk_msg.rpc_resp;
6454
6455         nfs41_setup_sequence(clp->cl_session, args, res, task);
6456 }
6457
6458 static const struct rpc_call_ops nfs41_sequence_ops = {
6459         .rpc_call_done = nfs41_sequence_call_done,
6460         .rpc_call_prepare = nfs41_sequence_prepare,
6461         .rpc_release = nfs41_sequence_release,
6462 };
6463
6464 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
6465                 struct rpc_cred *cred,
6466                 bool is_privileged)
6467 {
6468         struct nfs4_sequence_data *calldata;
6469         struct rpc_message msg = {
6470                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6471                 .rpc_cred = cred,
6472         };
6473         struct rpc_task_setup task_setup_data = {
6474                 .rpc_client = clp->cl_rpcclient,
6475                 .rpc_message = &msg,
6476                 .callback_ops = &nfs41_sequence_ops,
6477                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6478         };
6479
6480         if (!atomic_inc_not_zero(&clp->cl_count))
6481                 return ERR_PTR(-EIO);
6482         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6483         if (calldata == NULL) {
6484                 nfs_put_client(clp);
6485                 return ERR_PTR(-ENOMEM);
6486         }
6487         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6488         if (is_privileged)
6489                 nfs4_set_sequence_privileged(&calldata->args);
6490         msg.rpc_argp = &calldata->args;
6491         msg.rpc_resp = &calldata->res;
6492         calldata->clp = clp;
6493         task_setup_data.callback_data = calldata;
6494
6495         return rpc_run_task(&task_setup_data);
6496 }
6497
6498 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6499 {
6500         struct rpc_task *task;
6501         int ret = 0;
6502
6503         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6504                 return 0;
6505         task = _nfs41_proc_sequence(clp, cred, false);
6506         if (IS_ERR(task))
6507                 ret = PTR_ERR(task);
6508         else
6509                 rpc_put_task_async(task);
6510         dprintk("<-- %s status=%d\n", __func__, ret);
6511         return ret;
6512 }
6513
6514 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6515 {
6516         struct rpc_task *task;
6517         int ret;
6518
6519         task = _nfs41_proc_sequence(clp, cred, true);
6520         if (IS_ERR(task)) {
6521                 ret = PTR_ERR(task);
6522                 goto out;
6523         }
6524         ret = rpc_wait_for_completion_task(task);
6525         if (!ret) {
6526                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6527
6528                 if (task->tk_status == 0)
6529                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6530                 ret = task->tk_status;
6531         }
6532         rpc_put_task(task);
6533 out:
6534         dprintk("<-- %s status=%d\n", __func__, ret);
6535         return ret;
6536 }
6537
6538 struct nfs4_reclaim_complete_data {
6539         struct nfs_client *clp;
6540         struct nfs41_reclaim_complete_args arg;
6541         struct nfs41_reclaim_complete_res res;
6542 };
6543
6544 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6545 {
6546         struct nfs4_reclaim_complete_data *calldata = data;
6547
6548         nfs41_setup_sequence(calldata->clp->cl_session,
6549                         &calldata->arg.seq_args,
6550                         &calldata->res.seq_res,
6551                         task);
6552 }
6553
6554 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6555 {
6556         switch(task->tk_status) {
6557         case 0:
6558         case -NFS4ERR_COMPLETE_ALREADY:
6559         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6560                 break;
6561         case -NFS4ERR_DELAY:
6562                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6563                 /* fall through */
6564         case -NFS4ERR_RETRY_UNCACHED_REP:
6565                 return -EAGAIN;
6566         default:
6567                 nfs4_schedule_lease_recovery(clp);
6568         }
6569         return 0;
6570 }
6571
6572 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6573 {
6574         struct nfs4_reclaim_complete_data *calldata = data;
6575         struct nfs_client *clp = calldata->clp;
6576         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6577
6578         dprintk("--> %s\n", __func__);
6579         if (!nfs41_sequence_done(task, res))
6580                 return;
6581
6582         trace_nfs4_reclaim_complete(clp, task->tk_status);
6583         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6584                 rpc_restart_call_prepare(task);
6585                 return;
6586         }
6587         dprintk("<-- %s\n", __func__);
6588 }
6589
6590 static void nfs4_free_reclaim_complete_data(void *data)
6591 {
6592         struct nfs4_reclaim_complete_data *calldata = data;
6593
6594         kfree(calldata);
6595 }
6596
6597 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6598         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6599         .rpc_call_done = nfs4_reclaim_complete_done,
6600         .rpc_release = nfs4_free_reclaim_complete_data,
6601 };
6602
6603 /*
6604  * Issue a global reclaim complete.
6605  */
6606 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
6607                 struct rpc_cred *cred)
6608 {
6609         struct nfs4_reclaim_complete_data *calldata;
6610         struct rpc_task *task;
6611         struct rpc_message msg = {
6612                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6613                 .rpc_cred = cred,
6614         };
6615         struct rpc_task_setup task_setup_data = {
6616                 .rpc_client = clp->cl_rpcclient,
6617                 .rpc_message = &msg,
6618                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6619                 .flags = RPC_TASK_ASYNC,
6620         };
6621         int status = -ENOMEM;
6622
6623         dprintk("--> %s\n", __func__);
6624         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6625         if (calldata == NULL)
6626                 goto out;
6627         calldata->clp = clp;
6628         calldata->arg.one_fs = 0;
6629
6630         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6631         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6632         msg.rpc_argp = &calldata->arg;
6633         msg.rpc_resp = &calldata->res;
6634         task_setup_data.callback_data = calldata;
6635         task = rpc_run_task(&task_setup_data);
6636         if (IS_ERR(task)) {
6637                 status = PTR_ERR(task);
6638                 goto out;
6639         }
6640         status = nfs4_wait_for_completion_rpc_task(task);
6641         if (status == 0)
6642                 status = task->tk_status;
6643         rpc_put_task(task);
6644         return 0;
6645 out:
6646         dprintk("<-- %s status=%d\n", __func__, status);
6647         return status;
6648 }
6649
6650 static void
6651 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6652 {
6653         struct nfs4_layoutget *lgp = calldata;
6654         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6655         struct nfs4_session *session = nfs4_get_session(server);
6656
6657         dprintk("--> %s\n", __func__);
6658         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6659          * right now covering the LAYOUTGET we are about to send.
6660          * However, that is not so catastrophic, and there seems
6661          * to be no way to prevent it completely.
6662          */
6663         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6664                                 &lgp->res.seq_res, task))
6665                 return;
6666         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6667                                           NFS_I(lgp->args.inode)->layout,
6668                                           lgp->args.ctx->state)) {
6669                 rpc_exit(task, NFS4_OK);
6670         }
6671 }
6672
6673 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6674 {
6675         struct nfs4_layoutget *lgp = calldata;
6676         struct inode *inode = lgp->args.inode;
6677         struct nfs_server *server = NFS_SERVER(inode);
6678         struct pnfs_layout_hdr *lo;
6679         struct nfs4_state *state = NULL;
6680         unsigned long timeo, giveup;
6681
6682         dprintk("--> %s\n", __func__);
6683
6684         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6685                 goto out;
6686
6687         switch (task->tk_status) {
6688         case 0:
6689                 goto out;
6690         case -NFS4ERR_LAYOUTTRYLATER:
6691         case -NFS4ERR_RECALLCONFLICT:
6692                 timeo = rpc_get_timeout(task->tk_client);
6693                 giveup = lgp->args.timestamp + timeo;
6694                 if (time_after(giveup, jiffies))
6695                         task->tk_status = -NFS4ERR_DELAY;
6696                 break;
6697         case -NFS4ERR_EXPIRED:
6698         case -NFS4ERR_BAD_STATEID:
6699                 spin_lock(&inode->i_lock);
6700                 lo = NFS_I(inode)->layout;
6701                 if (!lo || list_empty(&lo->plh_segs)) {
6702                         spin_unlock(&inode->i_lock);
6703                         /* If the open stateid was bad, then recover it. */
6704                         state = lgp->args.ctx->state;
6705                 } else {
6706                         LIST_HEAD(head);
6707
6708                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6709                         spin_unlock(&inode->i_lock);
6710                         /* Mark the bad layout state as invalid, then
6711                          * retry using the open stateid. */
6712                         pnfs_free_lseg_list(&head);
6713                 }
6714         }
6715         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6716                 rpc_restart_call_prepare(task);
6717 out:
6718         dprintk("<-- %s\n", __func__);
6719 }
6720
6721 static size_t max_response_pages(struct nfs_server *server)
6722 {
6723         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6724         return nfs_page_array_len(0, max_resp_sz);
6725 }
6726
6727 static void nfs4_free_pages(struct page **pages, size_t size)
6728 {
6729         int i;
6730
6731         if (!pages)
6732                 return;
6733
6734         for (i = 0; i < size; i++) {
6735                 if (!pages[i])
6736                         break;
6737                 __free_page(pages[i]);
6738         }
6739         kfree(pages);
6740 }
6741
6742 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6743 {
6744         struct page **pages;
6745         int i;
6746
6747         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6748         if (!pages) {
6749                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6750                 return NULL;
6751         }
6752
6753         for (i = 0; i < size; i++) {
6754                 pages[i] = alloc_page(gfp_flags);
6755                 if (!pages[i]) {
6756                         dprintk("%s: failed to allocate page\n", __func__);
6757                         nfs4_free_pages(pages, size);
6758                         return NULL;
6759                 }
6760         }
6761
6762         return pages;
6763 }
6764
6765 static void nfs4_layoutget_release(void *calldata)
6766 {
6767         struct nfs4_layoutget *lgp = calldata;
6768         struct inode *inode = lgp->args.inode;
6769         struct nfs_server *server = NFS_SERVER(inode);
6770         size_t max_pages = max_response_pages(server);
6771
6772         dprintk("--> %s\n", __func__);
6773         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6774         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6775         put_nfs_open_context(lgp->args.ctx);
6776         kfree(calldata);
6777         dprintk("<-- %s\n", __func__);
6778 }
6779
6780 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6781         .rpc_call_prepare = nfs4_layoutget_prepare,
6782         .rpc_call_done = nfs4_layoutget_done,
6783         .rpc_release = nfs4_layoutget_release,
6784 };
6785
6786 struct pnfs_layout_segment *
6787 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6788 {
6789         struct inode *inode = lgp->args.inode;
6790         struct nfs_server *server = NFS_SERVER(inode);
6791         size_t max_pages = max_response_pages(server);
6792         struct rpc_task *task;
6793         struct rpc_message msg = {
6794                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6795                 .rpc_argp = &lgp->args,
6796                 .rpc_resp = &lgp->res,
6797                 .rpc_cred = lgp->cred,
6798         };
6799         struct rpc_task_setup task_setup_data = {
6800                 .rpc_client = server->client,
6801                 .rpc_message = &msg,
6802                 .callback_ops = &nfs4_layoutget_call_ops,
6803                 .callback_data = lgp,
6804                 .flags = RPC_TASK_ASYNC,
6805         };
6806         struct pnfs_layout_segment *lseg = NULL;
6807         int status = 0;
6808
6809         dprintk("--> %s\n", __func__);
6810
6811         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6812         if (!lgp->args.layout.pages) {
6813                 nfs4_layoutget_release(lgp);
6814                 return ERR_PTR(-ENOMEM);
6815         }
6816         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6817         lgp->args.timestamp = jiffies;
6818
6819         lgp->res.layoutp = &lgp->args.layout;
6820         lgp->res.seq_res.sr_slot = NULL;
6821         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6822
6823         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6824         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6825
6826         task = rpc_run_task(&task_setup_data);
6827         if (IS_ERR(task))
6828                 return ERR_CAST(task);
6829         status = nfs4_wait_for_completion_rpc_task(task);
6830         if (status == 0)
6831                 status = task->tk_status;
6832         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6833         if (status == 0 && lgp->res.layoutp->len)
6834                 lseg = pnfs_layout_process(lgp);
6835         rpc_put_task(task);
6836         dprintk("<-- %s status=%d\n", __func__, status);
6837         if (status)
6838                 return ERR_PTR(status);
6839         return lseg;
6840 }
6841
6842 static void
6843 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6844 {
6845         struct nfs4_layoutreturn *lrp = calldata;
6846
6847         dprintk("--> %s\n", __func__);
6848         nfs41_setup_sequence(lrp->clp->cl_session,
6849                         &lrp->args.seq_args,
6850                         &lrp->res.seq_res,
6851                         task);
6852 }
6853
6854 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6855 {
6856         struct nfs4_layoutreturn *lrp = calldata;
6857         struct nfs_server *server;
6858
6859         dprintk("--> %s\n", __func__);
6860
6861         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6862                 return;
6863
6864         server = NFS_SERVER(lrp->args.inode);
6865         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6866                 rpc_restart_call_prepare(task);
6867                 return;
6868         }
6869         dprintk("<-- %s\n", __func__);
6870 }
6871
6872 static void nfs4_layoutreturn_release(void *calldata)
6873 {
6874         struct nfs4_layoutreturn *lrp = calldata;
6875         struct pnfs_layout_hdr *lo = lrp->args.layout;
6876
6877         dprintk("--> %s\n", __func__);
6878         spin_lock(&lo->plh_inode->i_lock);
6879         if (lrp->res.lrs_present)
6880                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6881         lo->plh_block_lgets--;
6882         spin_unlock(&lo->plh_inode->i_lock);
6883         pnfs_put_layout_hdr(lrp->args.layout);
6884         kfree(calldata);
6885         dprintk("<-- %s\n", __func__);
6886 }
6887
6888 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6889         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6890         .rpc_call_done = nfs4_layoutreturn_done,
6891         .rpc_release = nfs4_layoutreturn_release,
6892 };
6893
6894 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6895 {
6896         struct rpc_task *task;
6897         struct rpc_message msg = {
6898                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6899                 .rpc_argp = &lrp->args,
6900                 .rpc_resp = &lrp->res,
6901                 .rpc_cred = lrp->cred,
6902         };
6903         struct rpc_task_setup task_setup_data = {
6904                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
6905                 .rpc_message = &msg,
6906                 .callback_ops = &nfs4_layoutreturn_call_ops,
6907                 .callback_data = lrp,
6908         };
6909         int status;
6910
6911         dprintk("--> %s\n", __func__);
6912         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6913         task = rpc_run_task(&task_setup_data);
6914         if (IS_ERR(task))
6915                 return PTR_ERR(task);
6916         status = task->tk_status;
6917         dprintk("<-- %s status=%d\n", __func__, status);
6918         rpc_put_task(task);
6919         return status;
6920 }
6921
6922 /*
6923  * Retrieve the list of Data Server devices from the MDS.
6924  */
6925 static int _nfs4_getdevicelist(struct nfs_server *server,
6926                                     const struct nfs_fh *fh,
6927                                     struct pnfs_devicelist *devlist)
6928 {
6929         struct nfs4_getdevicelist_args args = {
6930                 .fh = fh,
6931                 .layoutclass = server->pnfs_curr_ld->id,
6932         };
6933         struct nfs4_getdevicelist_res res = {
6934                 .devlist = devlist,
6935         };
6936         struct rpc_message msg = {
6937                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6938                 .rpc_argp = &args,
6939                 .rpc_resp = &res,
6940         };
6941         int status;
6942
6943         dprintk("--> %s\n", __func__);
6944         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6945                                 &res.seq_res, 0);
6946         dprintk("<-- %s status=%d\n", __func__, status);
6947         return status;
6948 }
6949
6950 int nfs4_proc_getdevicelist(struct nfs_server *server,
6951                             const struct nfs_fh *fh,
6952                             struct pnfs_devicelist *devlist)
6953 {
6954         struct nfs4_exception exception = { };
6955         int err;
6956
6957         do {
6958                 err = nfs4_handle_exception(server,
6959                                 _nfs4_getdevicelist(server, fh, devlist),
6960                                 &exception);
6961         } while (exception.retry);
6962
6963         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6964                 err, devlist->num_devs);
6965
6966         return err;
6967 }
6968 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6969
6970 static int
6971 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
6972                 struct pnfs_device *pdev,
6973                 struct rpc_cred *cred)
6974 {
6975         struct nfs4_getdeviceinfo_args args = {
6976                 .pdev = pdev,
6977         };
6978         struct nfs4_getdeviceinfo_res res = {
6979                 .pdev = pdev,
6980         };
6981         struct rpc_message msg = {
6982                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6983                 .rpc_argp = &args,
6984                 .rpc_resp = &res,
6985                 .rpc_cred = cred,
6986         };
6987         int status;
6988
6989         dprintk("--> %s\n", __func__);
6990         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6991         dprintk("<-- %s status=%d\n", __func__, status);
6992
6993         return status;
6994 }
6995
6996 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
6997                 struct pnfs_device *pdev,
6998                 struct rpc_cred *cred)
6999 {
7000         struct nfs4_exception exception = { };
7001         int err;
7002
7003         do {
7004                 err = nfs4_handle_exception(server,
7005                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
7006                                         &exception);
7007         } while (exception.retry);
7008         return err;
7009 }
7010 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7011
7012 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7013 {
7014         struct nfs4_layoutcommit_data *data = calldata;
7015         struct nfs_server *server = NFS_SERVER(data->args.inode);
7016         struct nfs4_session *session = nfs4_get_session(server);
7017
7018         nfs41_setup_sequence(session,
7019                         &data->args.seq_args,
7020                         &data->res.seq_res,
7021                         task);
7022 }
7023
7024 static void
7025 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7026 {
7027         struct nfs4_layoutcommit_data *data = calldata;
7028         struct nfs_server *server = NFS_SERVER(data->args.inode);
7029
7030         if (!nfs41_sequence_done(task, &data->res.seq_res))
7031                 return;
7032
7033         switch (task->tk_status) { /* Just ignore these failures */
7034         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7035         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7036         case -NFS4ERR_BADLAYOUT:     /* no layout */
7037         case -NFS4ERR_GRACE:        /* loca_recalim always false */
7038                 task->tk_status = 0;
7039                 break;
7040         case 0:
7041                 nfs_post_op_update_inode_force_wcc(data->args.inode,
7042                                                    data->res.fattr);
7043                 break;
7044         default:
7045                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7046                         rpc_restart_call_prepare(task);
7047                         return;
7048                 }
7049         }
7050 }
7051
7052 static void nfs4_layoutcommit_release(void *calldata)
7053 {
7054         struct nfs4_layoutcommit_data *data = calldata;
7055
7056         pnfs_cleanup_layoutcommit(data);
7057         put_rpccred(data->cred);
7058         kfree(data);
7059 }
7060
7061 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7062         .rpc_call_prepare = nfs4_layoutcommit_prepare,
7063         .rpc_call_done = nfs4_layoutcommit_done,
7064         .rpc_release = nfs4_layoutcommit_release,
7065 };
7066
7067 int
7068 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7069 {
7070         struct rpc_message msg = {
7071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7072                 .rpc_argp = &data->args,
7073                 .rpc_resp = &data->res,
7074                 .rpc_cred = data->cred,
7075         };
7076         struct rpc_task_setup task_setup_data = {
7077                 .task = &data->task,
7078                 .rpc_client = NFS_CLIENT(data->args.inode),
7079                 .rpc_message = &msg,
7080                 .callback_ops = &nfs4_layoutcommit_ops,
7081                 .callback_data = data,
7082                 .flags = RPC_TASK_ASYNC,
7083         };
7084         struct rpc_task *task;
7085         int status = 0;
7086
7087         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7088                 "lbw: %llu inode %lu\n",
7089                 data->task.tk_pid, sync,
7090                 data->args.lastbytewritten,
7091                 data->args.inode->i_ino);
7092
7093         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7094         task = rpc_run_task(&task_setup_data);
7095         if (IS_ERR(task))
7096                 return PTR_ERR(task);
7097         if (sync == false)
7098                 goto out;
7099         status = nfs4_wait_for_completion_rpc_task(task);
7100         if (status != 0)
7101                 goto out;
7102         status = task->tk_status;
7103 out:
7104         dprintk("%s: status %d\n", __func__, status);
7105         rpc_put_task(task);
7106         return status;
7107 }
7108
7109 /**
7110  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7111  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7112  */
7113 static int
7114 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7115                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
7116 {
7117         struct nfs41_secinfo_no_name_args args = {
7118                 .style = SECINFO_STYLE_CURRENT_FH,
7119         };
7120         struct nfs4_secinfo_res res = {
7121                 .flavors = flavors,
7122         };
7123         struct rpc_message msg = {
7124                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
7125                 .rpc_argp = &args,
7126                 .rpc_resp = &res,
7127         };
7128         return nfs4_call_sync(server->nfs_client->cl_rpcclient, server, &msg,
7129                                 &args.seq_args, &res.seq_res, 0);
7130 }
7131
7132 static int
7133 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7134                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
7135 {
7136         struct nfs4_exception exception = { };
7137         int err;
7138         do {
7139                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
7140                 switch (err) {
7141                 case 0:
7142                 case -NFS4ERR_WRONGSEC:
7143                 case -NFS4ERR_NOTSUPP:
7144                         goto out;
7145                 default:
7146                         err = nfs4_handle_exception(server, err, &exception);
7147                 }
7148         } while (exception.retry);
7149 out:
7150         return err;
7151 }
7152
7153 static int
7154 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
7155                     struct nfs_fsinfo *info)
7156 {
7157         int err;
7158         struct page *page;
7159         rpc_authflavor_t flavor;
7160         struct nfs4_secinfo_flavors *flavors;
7161
7162         page = alloc_page(GFP_KERNEL);
7163         if (!page) {
7164                 err = -ENOMEM;
7165                 goto out;
7166         }
7167
7168         flavors = page_address(page);
7169         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
7170
7171         /*
7172          * Fall back on "guess and check" method if
7173          * the server doesn't support SECINFO_NO_NAME
7174          */
7175         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
7176                 err = nfs4_find_root_sec(server, fhandle, info);
7177                 goto out_freepage;
7178         }
7179         if (err)
7180                 goto out_freepage;
7181
7182         flavor = nfs_find_best_sec(flavors);
7183         if (err == 0)
7184                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
7185
7186 out_freepage:
7187         put_page(page);
7188         if (err == -EACCES)
7189                 return -EPERM;
7190 out:
7191         return err;
7192 }
7193
7194 static int _nfs41_test_stateid(struct nfs_server *server,
7195                 nfs4_stateid *stateid,
7196                 struct rpc_cred *cred)
7197 {
7198         int status;
7199         struct nfs41_test_stateid_args args = {
7200                 .stateid = stateid,
7201         };
7202         struct nfs41_test_stateid_res res;
7203         struct rpc_message msg = {
7204                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
7205                 .rpc_argp = &args,
7206                 .rpc_resp = &res,
7207                 .rpc_cred = cred,
7208         };
7209
7210         dprintk("NFS call  test_stateid %p\n", stateid);
7211         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
7212         nfs4_set_sequence_privileged(&args.seq_args);
7213         status = nfs4_call_sync_sequence(server->client, server, &msg,
7214                         &args.seq_args, &res.seq_res);
7215         if (status != NFS_OK) {
7216                 dprintk("NFS reply test_stateid: failed, %d\n", status);
7217                 return status;
7218         }
7219         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
7220         return -res.status;
7221 }
7222
7223 /**
7224  * nfs41_test_stateid - perform a TEST_STATEID operation
7225  *
7226  * @server: server / transport on which to perform the operation
7227  * @stateid: state ID to test
7228  * @cred: credential
7229  *
7230  * Returns NFS_OK if the server recognizes that "stateid" is valid.
7231  * Otherwise a negative NFS4ERR value is returned if the operation
7232  * failed or the state ID is not currently valid.
7233  */
7234 static int nfs41_test_stateid(struct nfs_server *server,
7235                 nfs4_stateid *stateid,
7236                 struct rpc_cred *cred)
7237 {
7238         struct nfs4_exception exception = { };
7239         int err;
7240         do {
7241                 err = _nfs41_test_stateid(server, stateid, cred);
7242                 if (err != -NFS4ERR_DELAY)
7243                         break;
7244                 nfs4_handle_exception(server, err, &exception);
7245         } while (exception.retry);
7246         return err;
7247 }
7248
7249 struct nfs_free_stateid_data {
7250         struct nfs_server *server;
7251         struct nfs41_free_stateid_args args;
7252         struct nfs41_free_stateid_res res;
7253 };
7254
7255 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
7256 {
7257         struct nfs_free_stateid_data *data = calldata;
7258         nfs41_setup_sequence(nfs4_get_session(data->server),
7259                         &data->args.seq_args,
7260                         &data->res.seq_res,
7261                         task);
7262 }
7263
7264 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
7265 {
7266         struct nfs_free_stateid_data *data = calldata;
7267
7268         nfs41_sequence_done(task, &data->res.seq_res);
7269
7270         switch (task->tk_status) {
7271         case -NFS4ERR_DELAY:
7272                 if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
7273                         rpc_restart_call_prepare(task);
7274         }
7275 }
7276
7277 static void nfs41_free_stateid_release(void *calldata)
7278 {
7279         kfree(calldata);
7280 }
7281
7282 static const struct rpc_call_ops nfs41_free_stateid_ops = {
7283         .rpc_call_prepare = nfs41_free_stateid_prepare,
7284         .rpc_call_done = nfs41_free_stateid_done,
7285         .rpc_release = nfs41_free_stateid_release,
7286 };
7287
7288 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
7289                 nfs4_stateid *stateid,
7290                 struct rpc_cred *cred,
7291                 bool privileged)
7292 {
7293         struct rpc_message msg = {
7294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
7295                 .rpc_cred = cred,
7296         };
7297         struct rpc_task_setup task_setup = {
7298                 .rpc_client = server->client,
7299                 .rpc_message = &msg,
7300                 .callback_ops = &nfs41_free_stateid_ops,
7301                 .flags = RPC_TASK_ASYNC,
7302         };
7303         struct nfs_free_stateid_data *data;
7304
7305         dprintk("NFS call  free_stateid %p\n", stateid);
7306         data = kmalloc(sizeof(*data), GFP_NOFS);
7307         if (!data)
7308                 return ERR_PTR(-ENOMEM);
7309         data->server = server;
7310         nfs4_stateid_copy(&data->args.stateid, stateid);
7311
7312         task_setup.callback_data = data;
7313
7314         msg.rpc_argp = &data->args;
7315         msg.rpc_resp = &data->res;
7316         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
7317         if (privileged)
7318                 nfs4_set_sequence_privileged(&data->args.seq_args);
7319
7320         return rpc_run_task(&task_setup);
7321 }
7322
7323 /**
7324  * nfs41_free_stateid - perform a FREE_STATEID operation
7325  *
7326  * @server: server / transport on which to perform the operation
7327  * @stateid: state ID to release
7328  * @cred: credential
7329  *
7330  * Returns NFS_OK if the server freed "stateid".  Otherwise a
7331  * negative NFS4ERR value is returned.
7332  */
7333 static int nfs41_free_stateid(struct nfs_server *server,
7334                 nfs4_stateid *stateid,
7335                 struct rpc_cred *cred)
7336 {
7337         struct rpc_task *task;
7338         int ret;
7339
7340         task = _nfs41_free_stateid(server, stateid, cred, true);
7341         if (IS_ERR(task))
7342                 return PTR_ERR(task);
7343         ret = rpc_wait_for_completion_task(task);
7344         if (!ret)
7345                 ret = task->tk_status;
7346         rpc_put_task(task);
7347         return ret;
7348 }
7349
7350 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
7351 {
7352         struct rpc_task *task;
7353         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
7354
7355         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
7356         nfs4_free_lock_state(server, lsp);
7357         if (IS_ERR(task))
7358                 return PTR_ERR(task);
7359         rpc_put_task(task);
7360         return 0;
7361 }
7362
7363 static bool nfs41_match_stateid(const nfs4_stateid *s1,
7364                 const nfs4_stateid *s2)
7365 {
7366         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
7367                 return false;
7368
7369         if (s1->seqid == s2->seqid)
7370                 return true;
7371         if (s1->seqid == 0 || s2->seqid == 0)
7372                 return true;
7373
7374         return false;
7375 }
7376
7377 #endif /* CONFIG_NFS_V4_1 */
7378
7379 static bool nfs4_match_stateid(const nfs4_stateid *s1,
7380                 const nfs4_stateid *s2)
7381 {
7382         return nfs4_stateid_match(s1, s2);
7383 }
7384
7385
7386 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
7387         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7388         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7389         .recover_open   = nfs4_open_reclaim,
7390         .recover_lock   = nfs4_lock_reclaim,
7391         .establish_clid = nfs4_init_clientid,
7392         .detect_trunking = nfs40_discover_server_trunking,
7393 };
7394
7395 #if defined(CONFIG_NFS_V4_1)
7396 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
7397         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7398         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7399         .recover_open   = nfs4_open_reclaim,
7400         .recover_lock   = nfs4_lock_reclaim,
7401         .establish_clid = nfs41_init_clientid,
7402         .reclaim_complete = nfs41_proc_reclaim_complete,
7403         .detect_trunking = nfs41_discover_server_trunking,
7404 };
7405 #endif /* CONFIG_NFS_V4_1 */
7406
7407 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
7408         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7409         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7410         .recover_open   = nfs4_open_expired,
7411         .recover_lock   = nfs4_lock_expired,
7412         .establish_clid = nfs4_init_clientid,
7413 };
7414
7415 #if defined(CONFIG_NFS_V4_1)
7416 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
7417         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7418         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7419         .recover_open   = nfs41_open_expired,
7420         .recover_lock   = nfs41_lock_expired,
7421         .establish_clid = nfs41_init_clientid,
7422 };
7423 #endif /* CONFIG_NFS_V4_1 */
7424
7425 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
7426         .sched_state_renewal = nfs4_proc_async_renew,
7427         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
7428         .renew_lease = nfs4_proc_renew,
7429 };
7430
7431 #if defined(CONFIG_NFS_V4_1)
7432 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7433         .sched_state_renewal = nfs41_proc_async_sequence,
7434         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7435         .renew_lease = nfs4_proc_sequence,
7436 };
7437 #endif
7438
7439 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7440         .minor_version = 0,
7441         .init_caps = NFS_CAP_READDIRPLUS
7442                 | NFS_CAP_ATOMIC_OPEN
7443                 | NFS_CAP_CHANGE_ATTR
7444                 | NFS_CAP_POSIX_LOCK,
7445         .call_sync = _nfs4_call_sync,
7446         .match_stateid = nfs4_match_stateid,
7447         .find_root_sec = nfs4_find_root_sec,
7448         .free_lock_state = nfs4_release_lockowner,
7449         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7450         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7451         .state_renewal_ops = &nfs40_state_renewal_ops,
7452 };
7453
7454 #if defined(CONFIG_NFS_V4_1)
7455 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7456         .minor_version = 1,
7457         .init_caps = NFS_CAP_READDIRPLUS
7458                 | NFS_CAP_ATOMIC_OPEN
7459                 | NFS_CAP_CHANGE_ATTR
7460                 | NFS_CAP_POSIX_LOCK
7461                 | NFS_CAP_STATEID_NFSV41
7462                 | NFS_CAP_ATOMIC_OPEN_V1,
7463         .call_sync = nfs4_call_sync_sequence,
7464         .match_stateid = nfs41_match_stateid,
7465         .find_root_sec = nfs41_find_root_sec,
7466         .free_lock_state = nfs41_free_lock_state,
7467         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7468         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7469         .state_renewal_ops = &nfs41_state_renewal_ops,
7470 };
7471 #endif
7472
7473 #if defined(CONFIG_NFS_V4_2)
7474 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
7475         .minor_version = 2,
7476         .init_caps = NFS_CAP_READDIRPLUS
7477                 | NFS_CAP_ATOMIC_OPEN
7478                 | NFS_CAP_CHANGE_ATTR
7479                 | NFS_CAP_POSIX_LOCK
7480                 | NFS_CAP_STATEID_NFSV41
7481                 | NFS_CAP_ATOMIC_OPEN_V1,
7482         .call_sync = nfs4_call_sync_sequence,
7483         .match_stateid = nfs41_match_stateid,
7484         .find_root_sec = nfs41_find_root_sec,
7485         .free_lock_state = nfs41_free_lock_state,
7486         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7487         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7488         .state_renewal_ops = &nfs41_state_renewal_ops,
7489 };
7490 #endif
7491
7492 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7493         [0] = &nfs_v4_0_minor_ops,
7494 #if defined(CONFIG_NFS_V4_1)
7495         [1] = &nfs_v4_1_minor_ops,
7496 #endif
7497 #if defined(CONFIG_NFS_V4_2)
7498         [2] = &nfs_v4_2_minor_ops,
7499 #endif
7500 };
7501
7502 static const struct inode_operations nfs4_dir_inode_operations = {
7503         .create         = nfs_create,
7504         .lookup         = nfs_lookup,
7505         .atomic_open    = nfs_atomic_open,
7506         .link           = nfs_link,
7507         .unlink         = nfs_unlink,
7508         .symlink        = nfs_symlink,
7509         .mkdir          = nfs_mkdir,
7510         .rmdir          = nfs_rmdir,
7511         .mknod          = nfs_mknod,
7512         .rename         = nfs_rename,
7513         .permission     = nfs_permission,
7514         .getattr        = nfs_getattr,
7515         .setattr        = nfs_setattr,
7516         .getxattr       = generic_getxattr,
7517         .setxattr       = generic_setxattr,
7518         .listxattr      = generic_listxattr,
7519         .removexattr    = generic_removexattr,
7520 };
7521
7522 static const struct inode_operations nfs4_file_inode_operations = {
7523         .permission     = nfs_permission,
7524         .getattr        = nfs_getattr,
7525         .setattr        = nfs_setattr,
7526         .getxattr       = generic_getxattr,
7527         .setxattr       = generic_setxattr,
7528         .listxattr      = generic_listxattr,
7529         .removexattr    = generic_removexattr,
7530 };
7531
7532 const struct nfs_rpc_ops nfs_v4_clientops = {
7533         .version        = 4,                    /* protocol version */
7534         .dentry_ops     = &nfs4_dentry_operations,
7535         .dir_inode_ops  = &nfs4_dir_inode_operations,
7536         .file_inode_ops = &nfs4_file_inode_operations,
7537         .file_ops       = &nfs4_file_operations,
7538         .getroot        = nfs4_proc_get_root,
7539         .submount       = nfs4_submount,
7540         .try_mount      = nfs4_try_mount,
7541         .getattr        = nfs4_proc_getattr,
7542         .setattr        = nfs4_proc_setattr,
7543         .lookup         = nfs4_proc_lookup,
7544         .access         = nfs4_proc_access,
7545         .readlink       = nfs4_proc_readlink,
7546         .create         = nfs4_proc_create,
7547         .remove         = nfs4_proc_remove,
7548         .unlink_setup   = nfs4_proc_unlink_setup,
7549         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7550         .unlink_done    = nfs4_proc_unlink_done,
7551         .rename         = nfs4_proc_rename,
7552         .rename_setup   = nfs4_proc_rename_setup,
7553         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7554         .rename_done    = nfs4_proc_rename_done,
7555         .link           = nfs4_proc_link,
7556         .symlink        = nfs4_proc_symlink,
7557         .mkdir          = nfs4_proc_mkdir,
7558         .rmdir          = nfs4_proc_remove,
7559         .readdir        = nfs4_proc_readdir,
7560         .mknod          = nfs4_proc_mknod,
7561         .statfs         = nfs4_proc_statfs,
7562         .fsinfo         = nfs4_proc_fsinfo,
7563         .pathconf       = nfs4_proc_pathconf,
7564         .set_capabilities = nfs4_server_capabilities,
7565         .decode_dirent  = nfs4_decode_dirent,
7566         .read_setup     = nfs4_proc_read_setup,
7567         .read_pageio_init = pnfs_pageio_init_read,
7568         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7569         .read_done      = nfs4_read_done,
7570         .write_setup    = nfs4_proc_write_setup,
7571         .write_pageio_init = pnfs_pageio_init_write,
7572         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7573         .write_done     = nfs4_write_done,
7574         .commit_setup   = nfs4_proc_commit_setup,
7575         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7576         .commit_done    = nfs4_commit_done,
7577         .lock           = nfs4_proc_lock,
7578         .clear_acl_cache = nfs4_zap_acl_attr,
7579         .close_context  = nfs4_close_context,
7580         .open_context   = nfs4_atomic_open,
7581         .have_delegation = nfs4_have_delegation,
7582         .return_delegation = nfs4_inode_return_delegation,
7583         .alloc_client   = nfs4_alloc_client,
7584         .init_client    = nfs4_init_client,
7585         .free_client    = nfs4_free_client,
7586         .create_server  = nfs4_create_server,
7587         .clone_server   = nfs_clone_server,
7588 };
7589
7590 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7591         .prefix = XATTR_NAME_NFSV4_ACL,
7592         .list   = nfs4_xattr_list_nfs4_acl,
7593         .get    = nfs4_xattr_get_nfs4_acl,
7594         .set    = nfs4_xattr_set_nfs4_acl,
7595 };
7596
7597 const struct xattr_handler *nfs4_xattr_handlers[] = {
7598         &nfs4_xattr_nfs4_acl_handler,
7599 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7600         &nfs4_xattr_nfs4_label_handler,
7601 #endif
7602         NULL
7603 };
7604
7605 /*
7606  * Local variables:
7607  *  c-basic-offset: 8
7608  * End:
7609  */