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 *, long *);
  81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state, struct nfs4_label *ilabel,
  87                             struct nfs4_label *olabel);
  88 #ifdef CONFIG_NFS_V4_1
  89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  90                 struct rpc_cred *);
  91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  92                 struct rpc_cred *);
  93 #endif
  94
  95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  96 static inline struct nfs4_label *
  97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  98         struct iattr *sattr, struct nfs4_label *label)
  99 {
 100         int err;
 101
 102         if (label == NULL)
 103                 return NULL;
 104
 105         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 106                 return NULL;
 107
 108         err = security_dentry_init_security(dentry, sattr->ia_mode,
 109                                 &dentry->d_name, (void **)&label->label, &label->len);
 110         if (err == 0)
 111                 return label;
 112
 113         return NULL;
 114 }
 115 static inline void
 116 nfs4_label_release_security(struct nfs4_label *label)
 117 {
 118         if (label)
 119                 security_release_secctx(label->label, label->len);
 120 }
 121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 122 {
 123         if (label)
 124                 return server->attr_bitmask;
 125
 126         return server->attr_bitmask_nl;
 127 }
 128 #else
 129 static inline struct nfs4_label *
 130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 131         struct iattr *sattr, struct nfs4_label *l)
 132 { return NULL; }
 133 static inline void
 134 nfs4_label_release_security(struct nfs4_label *label)
 135 { return; }
 136 static inline u32 *
 137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 { return server->attr_bitmask; }
 139 #endif
 140
 141 /* Prevent leaks of NFSv4 errors into userland */
 142 static int nfs4_map_errors(int err)
 143 {
 144         if (err >= -1000)
 145                 return err;
 146         switch (err) {
 147         case -NFS4ERR_RESOURCE:
 148         case -NFS4ERR_LAYOUTTRYLATER:
 149         case -NFS4ERR_RECALLCONFLICT:
 150                 return -EREMOTEIO;
 151         case -NFS4ERR_WRONGSEC:
 152         case -NFS4ERR_WRONG_CRED:
 153                 return -EPERM;
 154         case -NFS4ERR_BADOWNER:
 155         case -NFS4ERR_BADNAME:
 156                 return -EINVAL;
 157         case -NFS4ERR_SHARE_DENIED:
 158                 return -EACCES;
 159         case -NFS4ERR_MINOR_VERS_MISMATCH:
 160                 return -EPROTONOSUPPORT;
 161         case -NFS4ERR_FILE_OPEN:
 162                 return -EBUSY;
 163         default:
 164                 dprintk("%s could not handle NFSv4 error %d\n",
 165                                 __func__, -err);
 166                 break;
 167         }
 168         return -EIO;
 169 }
 170
 171 /*
 172  * This is our standard bitmap for GETATTR requests.
 173  */
 174 const u32 nfs4_fattr_bitmap[3] = {
 175         FATTR4_WORD0_TYPE
 176         | FATTR4_WORD0_CHANGE
 177         | FATTR4_WORD0_SIZE
 178         | FATTR4_WORD0_FSID
 179         | FATTR4_WORD0_FILEID,
 180         FATTR4_WORD1_MODE
 181         | FATTR4_WORD1_NUMLINKS
 182         | FATTR4_WORD1_OWNER
 183         | FATTR4_WORD1_OWNER_GROUP
 184         | FATTR4_WORD1_RAWDEV
 185         | FATTR4_WORD1_SPACE_USED
 186         | FATTR4_WORD1_TIME_ACCESS
 187         | FATTR4_WORD1_TIME_METADATA
 188         | FATTR4_WORD1_TIME_MODIFY,
 189 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 190         FATTR4_WORD2_SECURITY_LABEL
 191 #endif
 192 };
 193
 194 static const u32 nfs4_pnfs_open_bitmap[3] = {
 195         FATTR4_WORD0_TYPE
 196         | FATTR4_WORD0_CHANGE
 197         | FATTR4_WORD0_SIZE
 198         | FATTR4_WORD0_FSID
 199         | FATTR4_WORD0_FILEID,
 200         FATTR4_WORD1_MODE
 201         | FATTR4_WORD1_NUMLINKS
 202         | FATTR4_WORD1_OWNER
 203         | FATTR4_WORD1_OWNER_GROUP
 204         | FATTR4_WORD1_RAWDEV
 205         | FATTR4_WORD1_SPACE_USED
 206         | FATTR4_WORD1_TIME_ACCESS
 207         | FATTR4_WORD1_TIME_METADATA
 208         | FATTR4_WORD1_TIME_MODIFY,
 209         FATTR4_WORD2_MDSTHRESHOLD
 210 };
 211
 212 static const u32 nfs4_open_noattr_bitmap[3] = {
 213         FATTR4_WORD0_TYPE
 214         | FATTR4_WORD0_CHANGE
 215         | FATTR4_WORD0_FILEID,
 216 };
 217
 218 const u32 nfs4_statfs_bitmap[3] = {
 219         FATTR4_WORD0_FILES_AVAIL
 220         | FATTR4_WORD0_FILES_FREE
 221         | FATTR4_WORD0_FILES_TOTAL,
 222         FATTR4_WORD1_SPACE_AVAIL
 223         | FATTR4_WORD1_SPACE_FREE
 224         | FATTR4_WORD1_SPACE_TOTAL
 225 };
 226
 227 const u32 nfs4_pathconf_bitmap[3] = {
 228         FATTR4_WORD0_MAXLINK
 229         | FATTR4_WORD0_MAXNAME,
 230         0
 231 };
 232
 233 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 234                         | FATTR4_WORD0_MAXREAD
 235                         | FATTR4_WORD0_MAXWRITE
 236                         | FATTR4_WORD0_LEASE_TIME,
 237                         FATTR4_WORD1_TIME_DELTA
 238                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 239                         FATTR4_WORD2_LAYOUT_BLKSIZE
 240 };
 241
 242 const u32 nfs4_fs_locations_bitmap[3] = {
 243         FATTR4_WORD0_TYPE
 244         | FATTR4_WORD0_CHANGE
 245         | FATTR4_WORD0_SIZE
 246         | FATTR4_WORD0_FSID
 247         | FATTR4_WORD0_FILEID
 248         | FATTR4_WORD0_FS_LOCATIONS,
 249         FATTR4_WORD1_MODE
 250         | FATTR4_WORD1_NUMLINKS
 251         | FATTR4_WORD1_OWNER
 252         | FATTR4_WORD1_OWNER_GROUP
 253         | FATTR4_WORD1_RAWDEV
 254         | FATTR4_WORD1_SPACE_USED
 255         | FATTR4_WORD1_TIME_ACCESS
 256         | FATTR4_WORD1_TIME_METADATA
 257         | FATTR4_WORD1_TIME_MODIFY
 258         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 259 };
 260
 261 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 262                 struct nfs4_readdir_arg *readdir)
 263 {
 264         __be32 *start, *p;
 265
 266         if (cookie > 2) {
 267                 readdir->cookie = cookie;
 268                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 269                 return;
 270         }
 271
 272         readdir->cookie = 0;
 273         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 274         if (cookie == 2)
 275                 return;
 276
 277         /*
 278          * NFSv4 servers do not return entries for '.' and '..'
 279          * Therefore, we fake these entries here.  We let '.'
 280          * have cookie 0 and '..' have cookie 1.  Note that
 281          * when talking to the server, we always send cookie 0
 282          * instead of 1 or 2.
 283          */
 284         start = p = kmap_atomic(*readdir->pages);
 285
 286         if (cookie == 0) {
 287                 *p++ = xdr_one;                                  /* next */
 288                 *p++ = xdr_zero;                   /* cookie, first word */
 289                 *p++ = xdr_one;                   /* cookie, second word */
 290                 *p++ = xdr_one;                             /* entry len */
 291                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 292                 p++;
 293                 *p++ = xdr_one;                         /* bitmap length */
 294                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 295                 *p++ = htonl(8);              /* attribute buffer length */
 296                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 297         }
 298
 299         *p++ = xdr_one;                                  /* next */
 300         *p++ = xdr_zero;                   /* cookie, first word */
 301         *p++ = xdr_two;                   /* cookie, second word */
 302         *p++ = xdr_two;                             /* entry len */
 303         memcpy(p, "..\0\0", 4);                         /* entry */
 304         p++;
 305         *p++ = xdr_one;                         /* bitmap length */
 306         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 307         *p++ = htonl(8);              /* attribute buffer length */
 308         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 309
 310         readdir->pgbase = (char *)p - (char *)start;
 311         readdir->count -= readdir->pgbase;
 312         kunmap_atomic(start);
 313 }
 314
 315 static long nfs4_update_delay(long *timeout)
 316 {
 317         long ret;
 318         if (!timeout)
 319                 return NFS4_POLL_RETRY_MAX;
 320         if (*timeout <= 0)
 321                 *timeout = NFS4_POLL_RETRY_MIN;
 322         if (*timeout > NFS4_POLL_RETRY_MAX)
 323                 *timeout = NFS4_POLL_RETRY_MAX;
 324         ret = *timeout;
 325         *timeout <<= 1;
 326         return ret;
 327 }
 328
 329 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 330 {
 331         int res = 0;
 332
 333         might_sleep();
 334
 335         freezable_schedule_timeout_killable_unsafe(
 336                 nfs4_update_delay(timeout));
 337         if (fatal_signal_pending(current))
 338                 res = -ERESTARTSYS;
 339         return res;
 340 }
 341
 342 /* This is the error handling routine for processes that are allowed
 343  * to sleep.
 344  */
 345 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 346 {
 347         struct nfs_client *clp = server->nfs_client;
 348         struct nfs4_state *state = exception->state;
 349         struct inode *inode = exception->inode;
 350         int ret = errorcode;
 351
 352         exception->retry = 0;
 353         switch(errorcode) {
 354                 case 0:
 355                         return 0;
 356                 case -NFS4ERR_OPENMODE:
 357                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 358                                 nfs4_inode_return_delegation(inode);
 359                                 exception->retry = 1;
 360                                 return 0;
 361                         }
 362                         if (state == NULL)
 363                                 break;
 364                         ret = nfs4_schedule_stateid_recovery(server, state);
 365                         if (ret < 0)
 366                                 break;
 367                         goto wait_on_recovery;
 368                 case -NFS4ERR_DELEG_REVOKED:
 369                 case -NFS4ERR_ADMIN_REVOKED:
 370                 case -NFS4ERR_BAD_STATEID:
 371                         if (state == NULL)
 372                                 break;
 373                         ret = nfs4_schedule_stateid_recovery(server, state);
 374                         if (ret < 0)
 375                                 break;
 376                         goto wait_on_recovery;
 377                 case -NFS4ERR_EXPIRED:
 378                         if (state != NULL) {
 379                                 ret = nfs4_schedule_stateid_recovery(server, state);
 380                                 if (ret < 0)
 381                                         break;
 382                         }
 383                 case -NFS4ERR_STALE_STATEID:
 384                 case -NFS4ERR_STALE_CLIENTID:
 385                         nfs4_schedule_lease_recovery(clp);
 386                         goto wait_on_recovery;
 387                 case -NFS4ERR_MOVED:
 388                         ret = nfs4_schedule_migration_recovery(server);
 389                         if (ret < 0)
 390                                 break;
 391                         goto wait_on_recovery;
 392                 case -NFS4ERR_LEASE_MOVED:
 393                         nfs4_schedule_lease_moved_recovery(clp);
 394                         goto wait_on_recovery;
 395 #if defined(CONFIG_NFS_V4_1)
 396                 case -NFS4ERR_BADSESSION:
 397                 case -NFS4ERR_BADSLOT:
 398                 case -NFS4ERR_BAD_HIGH_SLOT:
 399                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 400                 case -NFS4ERR_DEADSESSION:
 401                 case -NFS4ERR_SEQ_FALSE_RETRY:
 402                 case -NFS4ERR_SEQ_MISORDERED:
 403                         dprintk("%s ERROR: %d Reset session\n", __func__,
 404                                 errorcode);
 405                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 406                         goto wait_on_recovery;
 407 #endif /* defined(CONFIG_NFS_V4_1) */
 408                 case -NFS4ERR_FILE_OPEN:
 409                         if (exception->timeout > HZ) {
 410                                 /* We have retried a decent amount, time to
 411                                  * fail
 412                                  */
 413                                 ret = -EBUSY;
 414                                 break;
 415                         }
 416                 case -NFS4ERR_GRACE:
 417                 case -NFS4ERR_DELAY:
 418                         ret = nfs4_delay(server->client, &exception->timeout);
 419                         if (ret != 0)
 420                                 break;
 421                 case -NFS4ERR_RETRY_UNCACHED_REP:
 422                 case -NFS4ERR_OLD_STATEID:
 423                         exception->retry = 1;
 424                         break;
 425                 case -NFS4ERR_BADOWNER:
 426                         /* The following works around a Linux server bug! */
 427                 case -NFS4ERR_BADNAME:
 428                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 429                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 430                                 exception->retry = 1;
 431                                 printk(KERN_WARNING "NFS: v4 server %s "
 432                                                 "does not accept raw "
 433                                                 "uid/gids. "
 434                                                 "Reenabling the idmapper.\n",
 435                                                 server->nfs_client->cl_hostname);
 436                         }
 437         }
 438         /* We failed to handle the error */
 439         return nfs4_map_errors(ret);
 440 wait_on_recovery:
 441         ret = nfs4_wait_clnt_recover(clp);
 442         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 443                 return -EIO;
 444         if (ret == 0)
 445                 exception->retry = 1;
 446         return ret;
 447 }
 448
 449 /*
 450  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 451  * or 'false' otherwise.
 452  */
 453 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 454 {
 455         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 456
 457         if (flavor == RPC_AUTH_GSS_KRB5I ||
 458             flavor == RPC_AUTH_GSS_KRB5P)
 459                 return true;
 460
 461         return false;
 462 }
 463
 464 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 465 {
 466         spin_lock(&clp->cl_lock);
 467         if (time_before(clp->cl_last_renewal,timestamp))
 468                 clp->cl_last_renewal = timestamp;
 469         spin_unlock(&clp->cl_lock);
 470 }
 471
 472 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 473 {
 474         do_renew_lease(server->nfs_client, timestamp);
 475 }
 476
 477 struct nfs4_call_sync_data {
 478         const struct nfs_server *seq_server;
 479         struct nfs4_sequence_args *seq_args;
 480         struct nfs4_sequence_res *seq_res;
 481 };
 482
 483 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
 484                                struct nfs4_sequence_res *res, int cache_reply)
 485 {
 486         args->sa_slot = NULL;
 487         args->sa_cache_this = cache_reply;
 488         args->sa_privileged = 0;
 489
 490         res->sr_slot = NULL;
 491 }
 492
 493 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 494 {
 495         args->sa_privileged = 1;
 496 }
 497
 498 static int nfs40_setup_sequence(const struct nfs_server *server,
 499                                 struct nfs4_sequence_args *args,
 500                                 struct nfs4_sequence_res *res,
 501                                 struct rpc_task *task)
 502 {
 503         struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
 504         struct nfs4_slot *slot;
 505
 506         /* slot already allocated? */
 507         if (res->sr_slot != NULL)
 508                 goto out_start;
 509
 510         spin_lock(&tbl->slot_tbl_lock);
 511         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 512                 goto out_sleep;
 513
 514         slot = nfs4_alloc_slot(tbl);
 515         if (IS_ERR(slot)) {
 516                 if (slot == ERR_PTR(-ENOMEM))
 517                         task->tk_timeout = HZ >> 2;
 518                 goto out_sleep;
 519         }
 520         spin_unlock(&tbl->slot_tbl_lock);
 521
 522         args->sa_slot = slot;
 523         res->sr_slot = slot;
 524
 525 out_start:
 526         rpc_call_start(task);
 527         return 0;
 528
 529 out_sleep:
 530         if (args->sa_privileged)
 531                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 532                                 NULL, RPC_PRIORITY_PRIVILEGED);
 533         else
 534                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 535         spin_unlock(&tbl->slot_tbl_lock);
 536         return -EAGAIN;
 537 }
 538
 539 static int nfs40_sequence_done(struct rpc_task *task,
 540                                struct nfs4_sequence_res *res)
 541 {
 542         struct nfs4_slot *slot = res->sr_slot;
 543         struct nfs4_slot_table *tbl;
 544
 545         if (slot == NULL)
 546                 goto out;
 547
 548         tbl = slot->table;
 549         spin_lock(&tbl->slot_tbl_lock);
 550         if (!nfs41_wake_and_assign_slot(tbl, slot))
 551                 nfs4_free_slot(tbl, slot);
 552         spin_unlock(&tbl->slot_tbl_lock);
 553
 554         res->sr_slot = NULL;
 555 out:
 556         return 1;
 557 }
 558
 559 #if defined(CONFIG_NFS_V4_1)
 560
 561 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 562 {
 563         struct nfs4_session *session;
 564         struct nfs4_slot_table *tbl;
 565         struct nfs4_slot *slot = res->sr_slot;
 566         bool send_new_highest_used_slotid = false;
 567
 568         tbl = slot->table;
 569         session = tbl->session;
 570
 571         spin_lock(&tbl->slot_tbl_lock);
 572         /* Be nice to the server: try to ensure that the last transmitted
 573          * value for highest_user_slotid <= target_highest_slotid
 574          */
 575         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 576                 send_new_highest_used_slotid = true;
 577
 578         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 579                 send_new_highest_used_slotid = false;
 580                 goto out_unlock;
 581         }
 582         nfs4_free_slot(tbl, slot);
 583
 584         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 585                 send_new_highest_used_slotid = false;
 586 out_unlock:
 587         spin_unlock(&tbl->slot_tbl_lock);
 588         res->sr_slot = NULL;
 589         if (send_new_highest_used_slotid)
 590                 nfs41_server_notify_highest_slotid_update(session->clp);
 591 }
 592
 593 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 594 {
 595         struct nfs4_session *session;
 596         struct nfs4_slot *slot = res->sr_slot;
 597         struct nfs_client *clp;
 598         bool interrupted = false;
 599         int ret = 1;
 600
 601         if (slot == NULL)
 602                 goto out_noaction;
 603         /* don't increment the sequence number if the task wasn't sent */
 604         if (!RPC_WAS_SENT(task))
 605                 goto out;
 606
 607         session = slot->table->session;
 608
 609         if (slot->interrupted) {
 610                 slot->interrupted = 0;
 611                 interrupted = true;
 612         }
 613
 614         trace_nfs4_sequence_done(session, res);
 615         /* Check the SEQUENCE operation status */
 616         switch (res->sr_status) {
 617         case 0:
 618                 /* Update the slot's sequence and clientid lease timer */
 619                 ++slot->seq_nr;
 620                 clp = session->clp;
 621                 do_renew_lease(clp, res->sr_timestamp);
 622                 /* Check sequence flags */
 623                 if (res->sr_status_flags != 0)
 624                         nfs4_schedule_lease_recovery(clp);
 625                 nfs41_update_target_slotid(slot->table, slot, res);
 626                 break;
 627         case 1:
 628                 /*
 629                  * sr_status remains 1 if an RPC level error occurred.
 630                  * The server may or may not have processed the sequence
 631                  * operation..
 632                  * Mark the slot as having hosted an interrupted RPC call.
 633                  */
 634                 slot->interrupted = 1;
 635                 goto out;
 636         case -NFS4ERR_DELAY:
 637                 /* The server detected a resend of the RPC call and
 638                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 639                  * of RFC5661.
 640                  */
 641                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 642                         __func__,
 643                         slot->slot_nr,
 644                         slot->seq_nr);
 645                 goto out_retry;
 646         case -NFS4ERR_BADSLOT:
 647                 /*
 648                  * The slot id we used was probably retired. Try again
 649                  * using a different slot id.
 650                  */
 651                 goto retry_nowait;
 652         case -NFS4ERR_SEQ_MISORDERED:
 653                 /*
 654                  * Was the last operation on this sequence interrupted?
 655                  * If so, retry after bumping the sequence number.
 656                  */
 657                 if (interrupted) {
 658                         ++slot->seq_nr;
 659                         goto retry_nowait;
 660                 }
 661                 /*
 662                  * Could this slot have been previously retired?
 663                  * If so, then the server may be expecting seq_nr = 1!
 664                  */
 665                 if (slot->seq_nr != 1) {
 666                         slot->seq_nr = 1;
 667                         goto retry_nowait;
 668                 }
 669                 break;
 670         case -NFS4ERR_SEQ_FALSE_RETRY:
 671                 ++slot->seq_nr;
 672                 goto retry_nowait;
 673         default:
 674                 /* Just update the slot sequence no. */
 675                 ++slot->seq_nr;
 676         }
 677 out:
 678         /* The session may be reset by one of the error handlers. */
 679         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 680         nfs41_sequence_free_slot(res);
 681 out_noaction:
 682         return ret;
 683 retry_nowait:
 684         if (rpc_restart_call_prepare(task)) {
 685                 task->tk_status = 0;
 686                 ret = 0;
 687         }
 688         goto out;
 689 out_retry:
 690         if (!rpc_restart_call(task))
 691                 goto out;
 692         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 693         return 0;
 694 }
 695 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 696
 697 static int nfs4_sequence_done(struct rpc_task *task,
 698                                struct nfs4_sequence_res *res)
 699 {
 700         if (res->sr_slot == NULL)
 701                 return 1;
 702         if (!res->sr_slot->table->session)
 703                 return nfs40_sequence_done(task, res);
 704         return nfs41_sequence_done(task, res);
 705 }
 706
 707 int nfs41_setup_sequence(struct nfs4_session *session,
 708                                 struct nfs4_sequence_args *args,
 709                                 struct nfs4_sequence_res *res,
 710                                 struct rpc_task *task)
 711 {
 712         struct nfs4_slot *slot;
 713         struct nfs4_slot_table *tbl;
 714
 715         dprintk("--> %s\n", __func__);
 716         /* slot already allocated? */
 717         if (res->sr_slot != NULL)
 718                 goto out_success;
 719
 720         tbl = &session->fc_slot_table;
 721
 722         task->tk_timeout = 0;
 723
 724         spin_lock(&tbl->slot_tbl_lock);
 725         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 726             !args->sa_privileged) {
 727                 /* The state manager will wait until the slot table is empty */
 728                 dprintk("%s session is draining\n", __func__);
 729                 goto out_sleep;
 730         }
 731
 732         slot = nfs4_alloc_slot(tbl);
 733         if (IS_ERR(slot)) {
 734                 /* If out of memory, try again in 1/4 second */
 735                 if (slot == ERR_PTR(-ENOMEM))
 736                         task->tk_timeout = HZ >> 2;
 737                 dprintk("<-- %s: no free slots\n", __func__);
 738                 goto out_sleep;
 739         }
 740         spin_unlock(&tbl->slot_tbl_lock);
 741
 742         args->sa_slot = slot;
 743
 744         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 745                         slot->slot_nr, slot->seq_nr);
 746
 747         res->sr_slot = slot;
 748         res->sr_timestamp = jiffies;
 749         res->sr_status_flags = 0;
 750         /*
 751          * sr_status is only set in decode_sequence, and so will remain
 752          * set to 1 if an rpc level failure occurs.
 753          */
 754         res->sr_status = 1;
 755         trace_nfs4_setup_sequence(session, args);
 756 out_success:
 757         rpc_call_start(task);
 758         return 0;
 759 out_sleep:
 760         /* Privileged tasks are queued with top priority */
 761         if (args->sa_privileged)
 762                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 763                                 NULL, RPC_PRIORITY_PRIVILEGED);
 764         else
 765                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 766         spin_unlock(&tbl->slot_tbl_lock);
 767         return -EAGAIN;
 768 }
 769 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 770
 771 static int nfs4_setup_sequence(const struct nfs_server *server,
 772                                struct nfs4_sequence_args *args,
 773                                struct nfs4_sequence_res *res,
 774                                struct rpc_task *task)
 775 {
 776         struct nfs4_session *session = nfs4_get_session(server);
 777         int ret = 0;
 778
 779         if (!session)
 780                 return nfs40_setup_sequence(server, args, res, task);
 781
 782         dprintk("--> %s clp %p session %p sr_slot %u\n",
 783                 __func__, session->clp, session, res->sr_slot ?
 784                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 785
 786         ret = nfs41_setup_sequence(session, args, res, task);
 787
 788         dprintk("<-- %s status=%d\n", __func__, ret);
 789         return ret;
 790 }
 791
 792 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 793 {
 794         struct nfs4_call_sync_data *data = calldata;
 795         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 796
 797         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 798
 799         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 800 }
 801
 802 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 803 {
 804         struct nfs4_call_sync_data *data = calldata;
 805
 806         nfs41_sequence_done(task, data->seq_res);
 807 }
 808
 809 static const struct rpc_call_ops nfs41_call_sync_ops = {
 810         .rpc_call_prepare = nfs41_call_sync_prepare,
 811         .rpc_call_done = nfs41_call_sync_done,
 812 };
 813
 814 #else   /* !CONFIG_NFS_V4_1 */
 815
 816 static int nfs4_setup_sequence(const struct nfs_server *server,
 817                                struct nfs4_sequence_args *args,
 818                                struct nfs4_sequence_res *res,
 819                                struct rpc_task *task)
 820 {
 821         return nfs40_setup_sequence(server, args, res, task);
 822 }
 823
 824 static int nfs4_sequence_done(struct rpc_task *task,
 825                                struct nfs4_sequence_res *res)
 826 {
 827         return nfs40_sequence_done(task, res);
 828 }
 829
 830 #endif  /* !CONFIG_NFS_V4_1 */
 831
 832 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 833 {
 834         struct nfs4_call_sync_data *data = calldata;
 835         nfs4_setup_sequence(data->seq_server,
 836                                 data->seq_args, data->seq_res, task);
 837 }
 838
 839 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 840 {
 841         struct nfs4_call_sync_data *data = calldata;
 842         nfs4_sequence_done(task, data->seq_res);
 843 }
 844
 845 static const struct rpc_call_ops nfs40_call_sync_ops = {
 846         .rpc_call_prepare = nfs40_call_sync_prepare,
 847         .rpc_call_done = nfs40_call_sync_done,
 848 };
 849
 850 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 851                                    struct nfs_server *server,
 852                                    struct rpc_message *msg,
 853                                    struct nfs4_sequence_args *args,
 854                                    struct nfs4_sequence_res *res)
 855 {
 856         int ret;
 857         struct rpc_task *task;
 858         struct nfs_client *clp = server->nfs_client;
 859         struct nfs4_call_sync_data data = {
 860                 .seq_server = server,
 861                 .seq_args = args,
 862                 .seq_res = res,
 863         };
 864         struct rpc_task_setup task_setup = {
 865                 .rpc_client = clnt,
 866                 .rpc_message = msg,
 867                 .callback_ops = clp->cl_mvops->call_sync_ops,
 868                 .callback_data = &data
 869         };
 870
 871         task = rpc_run_task(&task_setup);
 872         if (IS_ERR(task))
 873                 ret = PTR_ERR(task);
 874         else {
 875                 ret = task->tk_status;
 876                 rpc_put_task(task);
 877         }
 878         return ret;
 879 }
 880
 881 int nfs4_call_sync(struct rpc_clnt *clnt,
 882                    struct nfs_server *server,
 883                    struct rpc_message *msg,
 884                    struct nfs4_sequence_args *args,
 885                    struct nfs4_sequence_res *res,
 886                    int cache_reply)
 887 {
 888         nfs4_init_sequence(args, res, cache_reply);
 889         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 890 }
 891
 892 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 893 {
 894         struct nfs_inode *nfsi = NFS_I(dir);
 895
 896         spin_lock(&dir->i_lock);
 897         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 898         if (!cinfo->atomic || cinfo->before != dir->i_version)
 899                 nfs_force_lookup_revalidate(dir);
 900         dir->i_version = cinfo->after;
 901         nfs_fscache_invalidate(dir);
 902         spin_unlock(&dir->i_lock);
 903 }
 904
 905 struct nfs4_opendata {
 906         struct kref kref;
 907         struct nfs_openargs o_arg;
 908         struct nfs_openres o_res;
 909         struct nfs_open_confirmargs c_arg;
 910         struct nfs_open_confirmres c_res;
 911         struct nfs4_string owner_name;
 912         struct nfs4_string group_name;
 913         struct nfs_fattr f_attr;
 914         struct nfs4_label *f_label;
 915         struct dentry *dir;
 916         struct dentry *dentry;
 917         struct nfs4_state_owner *owner;
 918         struct nfs4_state *state;
 919         struct iattr attrs;
 920         unsigned long timestamp;
 921         unsigned int rpc_done : 1;
 922         unsigned int file_created : 1;
 923         unsigned int is_recover : 1;
 924         int rpc_status;
 925         int cancelled;
 926 };
 927
 928 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 929                 int err, struct nfs4_exception *exception)
 930 {
 931         if (err != -EINVAL)
 932                 return false;
 933         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 934                 return false;
 935         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 936         exception->retry = 1;
 937         return true;
 938 }
 939
 940 static enum open_claim_type4
 941 nfs4_map_atomic_open_claim(struct nfs_server *server,
 942                 enum open_claim_type4 claim)
 943 {
 944         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 945                 return claim;
 946         switch (claim) {
 947         default:
 948                 return claim;
 949         case NFS4_OPEN_CLAIM_FH:
 950                 return NFS4_OPEN_CLAIM_NULL;
 951         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 952                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 953         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 954                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 955         }
 956 }
 957
 958 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 959 {
 960         p->o_res.f_attr = &p->f_attr;
 961         p->o_res.f_label = p->f_label;
 962         p->o_res.seqid = p->o_arg.seqid;
 963         p->c_res.seqid = p->c_arg.seqid;
 964         p->o_res.server = p->o_arg.server;
 965         p->o_res.access_request = p->o_arg.access;
 966         nfs_fattr_init(&p->f_attr);
 967         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 968 }
 969
 970 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 971                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 972                 const struct iattr *attrs,
 973                 struct nfs4_label *label,
 974                 enum open_claim_type4 claim,
 975                 gfp_t gfp_mask)
 976 {
 977         struct dentry *parent = dget_parent(dentry);
 978         struct inode *dir = parent->d_inode;
 979         struct nfs_server *server = NFS_SERVER(dir);
 980         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
 981         struct nfs4_opendata *p;
 982
 983         p = kzalloc(sizeof(*p), gfp_mask);
 984         if (p == NULL)
 985                 goto err;
 986
 987         p->f_label = nfs4_label_alloc(server, gfp_mask);
 988         if (IS_ERR(p->f_label))
 989                 goto err_free_p;
 990
 991         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
 992         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
 993         if (IS_ERR(p->o_arg.seqid))
 994                 goto err_free_label;
 995         nfs_sb_active(dentry->d_sb);
 996         p->dentry = dget(dentry);
 997         p->dir = parent;
 998         p->owner = sp;
 999         atomic_inc(&sp->so_count);
1000         p->o_arg.open_flags = flags;
1001         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1002         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1003          * will return permission denied for all bits until close */
1004         if (!(flags & O_EXCL)) {
1005                 /* ask server to check for all possible rights as results
1006                  * are cached */
1007                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1008                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1009         }
1010         p->o_arg.clientid = server->nfs_client->cl_clientid;
1011         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1012         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1013         p->o_arg.name = &dentry->d_name;
1014         p->o_arg.server = server;
1015         p->o_arg.bitmask = nfs4_bitmask(server, label);
1016         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1017         p->o_arg.label = label;
1018         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1019         switch (p->o_arg.claim) {
1020         case NFS4_OPEN_CLAIM_NULL:
1021         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1022         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1023                 p->o_arg.fh = NFS_FH(dir);
1024                 break;
1025         case NFS4_OPEN_CLAIM_PREVIOUS:
1026         case NFS4_OPEN_CLAIM_FH:
1027         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1028         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1029                 p->o_arg.fh = NFS_FH(dentry->d_inode);
1030         }
1031         if (attrs != NULL && attrs->ia_valid != 0) {
1032                 __u32 verf[2];
1033
1034                 p->o_arg.u.attrs = &p->attrs;
1035                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1036
1037                 verf[0] = jiffies;
1038                 verf[1] = current->pid;
1039                 memcpy(p->o_arg.u.verifier.data, verf,
1040                                 sizeof(p->o_arg.u.verifier.data));
1041         }
1042         p->c_arg.fh = &p->o_res.fh;
1043         p->c_arg.stateid = &p->o_res.stateid;
1044         p->c_arg.seqid = p->o_arg.seqid;
1045         nfs4_init_opendata_res(p);
1046         kref_init(&p->kref);
1047         return p;
1048
1049 err_free_label:
1050         nfs4_label_free(p->f_label);
1051 err_free_p:
1052         kfree(p);
1053 err:
1054         dput(parent);
1055         return NULL;
1056 }
1057
1058 static void nfs4_opendata_free(struct kref *kref)
1059 {
1060         struct nfs4_opendata *p = container_of(kref,
1061                         struct nfs4_opendata, kref);
1062         struct super_block *sb = p->dentry->d_sb;
1063
1064         nfs_free_seqid(p->o_arg.seqid);
1065         if (p->state != NULL)
1066                 nfs4_put_open_state(p->state);
1067         nfs4_put_state_owner(p->owner);
1068
1069         nfs4_label_free(p->f_label);
1070
1071         dput(p->dir);
1072         dput(p->dentry);
1073         nfs_sb_deactive(sb);
1074         nfs_fattr_free_names(&p->f_attr);
1075         kfree(p->f_attr.mdsthreshold);
1076         kfree(p);
1077 }
1078
1079 static void nfs4_opendata_put(struct nfs4_opendata *p)
1080 {
1081         if (p != NULL)
1082                 kref_put(&p->kref, nfs4_opendata_free);
1083 }
1084
1085 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1086 {
1087         int ret;
1088
1089         ret = rpc_wait_for_completion_task(task);
1090         return ret;
1091 }
1092
1093 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1094 {
1095         int ret = 0;
1096
1097         if (open_mode & (O_EXCL|O_TRUNC))
1098                 goto out;
1099         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1100                 case FMODE_READ:
1101                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1102                                 && state->n_rdonly != 0;
1103                         break;
1104                 case FMODE_WRITE:
1105                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1106                                 && state->n_wronly != 0;
1107                         break;
1108                 case FMODE_READ|FMODE_WRITE:
1109                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1110                                 && state->n_rdwr != 0;
1111         }
1112 out:
1113         return ret;
1114 }
1115
1116 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1117 {
1118         if (delegation == NULL)
1119                 return 0;
1120         if ((delegation->type & fmode) != fmode)
1121                 return 0;
1122         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1123                 return 0;
1124         nfs_mark_delegation_referenced(delegation);
1125         return 1;
1126 }
1127
1128 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1129 {
1130         switch (fmode) {
1131                 case FMODE_WRITE:
1132                         state->n_wronly++;
1133                         break;
1134                 case FMODE_READ:
1135                         state->n_rdonly++;
1136                         break;
1137                 case FMODE_READ|FMODE_WRITE:
1138                         state->n_rdwr++;
1139         }
1140         nfs4_state_set_mode_locked(state, state->state | fmode);
1141 }
1142
1143 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1144 {
1145         struct nfs_client *clp = state->owner->so_server->nfs_client;
1146         bool need_recover = false;
1147
1148         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1149                 need_recover = true;
1150         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1151                 need_recover = true;
1152         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1153                 need_recover = true;
1154         if (need_recover)
1155                 nfs4_state_mark_reclaim_nograce(clp, state);
1156 }
1157
1158 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1159                 nfs4_stateid *stateid)
1160 {
1161         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1162                 return true;
1163         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1164                 nfs_test_and_clear_all_open_stateid(state);
1165                 return true;
1166         }
1167         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1168                 return true;
1169         return false;
1170 }
1171
1172 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1173 {
1174         if (state->n_wronly)
1175                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1176         if (state->n_rdonly)
1177                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1178         if (state->n_rdwr)
1179                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1180 }
1181
1182 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1183                 nfs4_stateid *stateid, fmode_t fmode)
1184 {
1185         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1186         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1187         case FMODE_WRITE:
1188                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1189                 break;
1190         case FMODE_READ:
1191                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1192                 break;
1193         case 0:
1194                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1195                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1196                 clear_bit(NFS_OPEN_STATE, &state->flags);
1197         }
1198         if (stateid == NULL)
1199                 return;
1200         /* Handle races with OPEN */
1201         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1202             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1203                 nfs_resync_open_stateid_locked(state);
1204                 return;
1205         }
1206         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1207                 nfs4_stateid_copy(&state->stateid, stateid);
1208         nfs4_stateid_copy(&state->open_stateid, stateid);
1209 }
1210
1211 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1212 {
1213         write_seqlock(&state->seqlock);
1214         nfs_clear_open_stateid_locked(state, stateid, fmode);
1215         write_sequnlock(&state->seqlock);
1216         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1217                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1218 }
1219
1220 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1221 {
1222         switch (fmode) {
1223                 case FMODE_READ:
1224                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1225                         break;
1226                 case FMODE_WRITE:
1227                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1228                         break;
1229                 case FMODE_READ|FMODE_WRITE:
1230                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1231         }
1232         if (!nfs_need_update_open_stateid(state, stateid))
1233                 return;
1234         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1235                 nfs4_stateid_copy(&state->stateid, stateid);
1236         nfs4_stateid_copy(&state->open_stateid, stateid);
1237 }
1238
1239 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1240 {
1241         /*
1242          * Protect the call to nfs4_state_set_mode_locked and
1243          * serialise the stateid update
1244          */
1245         write_seqlock(&state->seqlock);
1246         if (deleg_stateid != NULL) {
1247                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1248                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1249         }
1250         if (open_stateid != NULL)
1251                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1252         write_sequnlock(&state->seqlock);
1253         spin_lock(&state->owner->so_lock);
1254         update_open_stateflags(state, fmode);
1255         spin_unlock(&state->owner->so_lock);
1256 }
1257
1258 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1259 {
1260         struct nfs_inode *nfsi = NFS_I(state->inode);
1261         struct nfs_delegation *deleg_cur;
1262         int ret = 0;
1263
1264         fmode &= (FMODE_READ|FMODE_WRITE);
1265
1266         rcu_read_lock();
1267         deleg_cur = rcu_dereference(nfsi->delegation);
1268         if (deleg_cur == NULL)
1269                 goto no_delegation;
1270
1271         spin_lock(&deleg_cur->lock);
1272         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1273            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1274             (deleg_cur->type & fmode) != fmode)
1275                 goto no_delegation_unlock;
1276
1277         if (delegation == NULL)
1278                 delegation = &deleg_cur->stateid;
1279         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1280                 goto no_delegation_unlock;
1281
1282         nfs_mark_delegation_referenced(deleg_cur);
1283         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1284         ret = 1;
1285 no_delegation_unlock:
1286         spin_unlock(&deleg_cur->lock);
1287 no_delegation:
1288         rcu_read_unlock();
1289
1290         if (!ret && open_stateid != NULL) {
1291                 __update_open_stateid(state, open_stateid, NULL, fmode);
1292                 ret = 1;
1293         }
1294         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1295                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1296
1297         return ret;
1298 }
1299
1300 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1301                 const nfs4_stateid *stateid)
1302 {
1303         struct nfs4_state *state = lsp->ls_state;
1304         bool ret = false;
1305
1306         spin_lock(&state->state_lock);
1307         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1308                 goto out_noupdate;
1309         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1310                 goto out_noupdate;
1311         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1312         ret = true;
1313 out_noupdate:
1314         spin_unlock(&state->state_lock);
1315         return ret;
1316 }
1317
1318 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1319 {
1320         struct nfs_delegation *delegation;
1321
1322         rcu_read_lock();
1323         delegation = rcu_dereference(NFS_I(inode)->delegation);
1324         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1325                 rcu_read_unlock();
1326                 return;
1327         }
1328         rcu_read_unlock();
1329         nfs4_inode_return_delegation(inode);
1330 }
1331
1332 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1333 {
1334         struct nfs4_state *state = opendata->state;
1335         struct nfs_inode *nfsi = NFS_I(state->inode);
1336         struct nfs_delegation *delegation;
1337         int open_mode = opendata->o_arg.open_flags;
1338         fmode_t fmode = opendata->o_arg.fmode;
1339         nfs4_stateid stateid;
1340         int ret = -EAGAIN;
1341
1342         for (;;) {
1343                 spin_lock(&state->owner->so_lock);
1344                 if (can_open_cached(state, fmode, open_mode)) {
1345                         update_open_stateflags(state, fmode);
1346                         spin_unlock(&state->owner->so_lock);
1347                         goto out_return_state;
1348                 }
1349                 spin_unlock(&state->owner->so_lock);
1350                 rcu_read_lock();
1351                 delegation = rcu_dereference(nfsi->delegation);
1352                 if (!can_open_delegated(delegation, fmode)) {
1353                         rcu_read_unlock();
1354                         break;
1355                 }
1356                 /* Save the delegation */
1357                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1358                 rcu_read_unlock();
1359                 nfs_release_seqid(opendata->o_arg.seqid);
1360                 if (!opendata->is_recover) {
1361                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1362                         if (ret != 0)
1363                                 goto out;
1364                 }
1365                 ret = -EAGAIN;
1366
1367                 /* Try to update the stateid using the delegation */
1368                 if (update_open_stateid(state, NULL, &stateid, fmode))
1369                         goto out_return_state;
1370         }
1371 out:
1372         return ERR_PTR(ret);
1373 out_return_state:
1374         atomic_inc(&state->count);
1375         return state;
1376 }
1377
1378 static void
1379 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1380 {
1381         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1382         struct nfs_delegation *delegation;
1383         int delegation_flags = 0;
1384
1385         rcu_read_lock();
1386         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1387         if (delegation)
1388                 delegation_flags = delegation->flags;
1389         rcu_read_unlock();
1390         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1391                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1392                                    "returning a delegation for "
1393                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1394                                    clp->cl_hostname);
1395         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1396                 nfs_inode_set_delegation(state->inode,
1397                                          data->owner->so_cred,
1398                                          &data->o_res);
1399         else
1400                 nfs_inode_reclaim_delegation(state->inode,
1401                                              data->owner->so_cred,
1402                                              &data->o_res);
1403 }
1404
1405 /*
1406  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1407  * and update the nfs4_state.
1408  */
1409 static struct nfs4_state *
1410 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1411 {
1412         struct inode *inode = data->state->inode;
1413         struct nfs4_state *state = data->state;
1414         int ret;
1415
1416         if (!data->rpc_done) {
1417                 if (data->rpc_status) {
1418                         ret = data->rpc_status;
1419                         goto err;
1420                 }
1421                 /* cached opens have already been processed */
1422                 goto update;
1423         }
1424
1425         ret = nfs_refresh_inode(inode, &data->f_attr);
1426         if (ret)
1427                 goto err;
1428
1429         if (data->o_res.delegation_type != 0)
1430                 nfs4_opendata_check_deleg(data, state);
1431 update:
1432         update_open_stateid(state, &data->o_res.stateid, NULL,
1433                             data->o_arg.fmode);
1434         atomic_inc(&state->count);
1435
1436         return state;
1437 err:
1438         return ERR_PTR(ret);
1439
1440 }
1441
1442 static struct nfs4_state *
1443 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1444 {
1445         struct inode *inode;
1446         struct nfs4_state *state = NULL;
1447         int ret;
1448
1449         if (!data->rpc_done) {
1450                 state = nfs4_try_open_cached(data);
1451                 goto out;
1452         }
1453
1454         ret = -EAGAIN;
1455         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1456                 goto err;
1457         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1458         ret = PTR_ERR(inode);
1459         if (IS_ERR(inode))
1460                 goto err;
1461         ret = -ENOMEM;
1462         state = nfs4_get_open_state(inode, data->owner);
1463         if (state == NULL)
1464                 goto err_put_inode;
1465         if (data->o_res.delegation_type != 0)
1466                 nfs4_opendata_check_deleg(data, state);
1467         update_open_stateid(state, &data->o_res.stateid, NULL,
1468                         data->o_arg.fmode);
1469         iput(inode);
1470 out:
1471         nfs_release_seqid(data->o_arg.seqid);
1472         return state;
1473 err_put_inode:
1474         iput(inode);
1475 err:
1476         return ERR_PTR(ret);
1477 }
1478
1479 static struct nfs4_state *
1480 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1481 {
1482         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1483                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1484         return _nfs4_opendata_to_nfs4_state(data);
1485 }
1486
1487 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1488 {
1489         struct nfs_inode *nfsi = NFS_I(state->inode);
1490         struct nfs_open_context *ctx;
1491
1492         spin_lock(&state->inode->i_lock);
1493         list_for_each_entry(ctx, &nfsi->open_files, list) {
1494                 if (ctx->state != state)
1495                         continue;
1496                 get_nfs_open_context(ctx);
1497                 spin_unlock(&state->inode->i_lock);
1498                 return ctx;
1499         }
1500         spin_unlock(&state->inode->i_lock);
1501         return ERR_PTR(-ENOENT);
1502 }
1503
1504 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1505                 struct nfs4_state *state, enum open_claim_type4 claim)
1506 {
1507         struct nfs4_opendata *opendata;
1508
1509         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1510                         NULL, NULL, claim, GFP_NOFS);
1511         if (opendata == NULL)
1512                 return ERR_PTR(-ENOMEM);
1513         opendata->state = state;
1514         atomic_inc(&state->count);
1515         return opendata;
1516 }
1517
1518 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1519 {
1520         struct nfs4_state *newstate;
1521         int ret;
1522
1523         opendata->o_arg.open_flags = 0;
1524         opendata->o_arg.fmode = fmode;
1525         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1526         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1527         nfs4_init_opendata_res(opendata);
1528         ret = _nfs4_recover_proc_open(opendata);
1529         if (ret != 0)
1530                 return ret;
1531         newstate = nfs4_opendata_to_nfs4_state(opendata);
1532         if (IS_ERR(newstate))
1533                 return PTR_ERR(newstate);
1534         nfs4_close_state(newstate, fmode);
1535         *res = newstate;
1536         return 0;
1537 }
1538
1539 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1540 {
1541         struct nfs4_state *newstate;
1542         int ret;
1543
1544         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1545         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1546         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1547         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1548         /* memory barrier prior to reading state->n_* */
1549         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1550         clear_bit(NFS_OPEN_STATE, &state->flags);
1551         smp_rmb();
1552         if (state->n_rdwr != 0) {
1553                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1554                 if (ret != 0)
1555                         return ret;
1556                 if (newstate != state)
1557                         return -ESTALE;
1558         }
1559         if (state->n_wronly != 0) {
1560                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1561                 if (ret != 0)
1562                         return ret;
1563                 if (newstate != state)
1564                         return -ESTALE;
1565         }
1566         if (state->n_rdonly != 0) {
1567                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1568                 if (ret != 0)
1569                         return ret;
1570                 if (newstate != state)
1571                         return -ESTALE;
1572         }
1573         /*
1574          * We may have performed cached opens for all three recoveries.
1575          * Check if we need to update the current stateid.
1576          */
1577         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1578             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1579                 write_seqlock(&state->seqlock);
1580                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1581                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1582                 write_sequnlock(&state->seqlock);
1583         }
1584         return 0;
1585 }
1586
1587 /*
1588  * OPEN_RECLAIM:
1589  *      reclaim state on the server after a reboot.
1590  */
1591 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1592 {
1593         struct nfs_delegation *delegation;
1594         struct nfs4_opendata *opendata;
1595         fmode_t delegation_type = 0;
1596         int status;
1597
1598         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1599                         NFS4_OPEN_CLAIM_PREVIOUS);
1600         if (IS_ERR(opendata))
1601                 return PTR_ERR(opendata);
1602         rcu_read_lock();
1603         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1604         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1605                 delegation_type = delegation->type;
1606         rcu_read_unlock();
1607         opendata->o_arg.u.delegation_type = delegation_type;
1608         status = nfs4_open_recover(opendata, state);
1609         nfs4_opendata_put(opendata);
1610         return status;
1611 }
1612
1613 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1614 {
1615         struct nfs_server *server = NFS_SERVER(state->inode);
1616         struct nfs4_exception exception = { };
1617         int err;
1618         do {
1619                 err = _nfs4_do_open_reclaim(ctx, state);
1620                 trace_nfs4_open_reclaim(ctx, 0, err);
1621                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1622                         continue;
1623                 if (err != -NFS4ERR_DELAY)
1624                         break;
1625                 nfs4_handle_exception(server, err, &exception);
1626         } while (exception.retry);
1627         return err;
1628 }
1629
1630 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1631 {
1632         struct nfs_open_context *ctx;
1633         int ret;
1634
1635         ctx = nfs4_state_find_open_context(state);
1636         if (IS_ERR(ctx))
1637                 return -EAGAIN;
1638         ret = nfs4_do_open_reclaim(ctx, state);
1639         put_nfs_open_context(ctx);
1640         return ret;
1641 }
1642
1643 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1644 {
1645         switch (err) {
1646                 default:
1647                         printk(KERN_ERR "NFS: %s: unhandled error "
1648                                         "%d.\n", __func__, err);
1649                 case 0:
1650                 case -ENOENT:
1651                 case -ESTALE:
1652                         break;
1653                 case -NFS4ERR_BADSESSION:
1654                 case -NFS4ERR_BADSLOT:
1655                 case -NFS4ERR_BAD_HIGH_SLOT:
1656                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1657                 case -NFS4ERR_DEADSESSION:
1658                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1659                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1660                         return -EAGAIN;
1661                 case -NFS4ERR_STALE_CLIENTID:
1662                 case -NFS4ERR_STALE_STATEID:
1663                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1664                 case -NFS4ERR_EXPIRED:
1665                         /* Don't recall a delegation if it was lost */
1666                         nfs4_schedule_lease_recovery(server->nfs_client);
1667                         return -EAGAIN;
1668                 case -NFS4ERR_MOVED:
1669                         nfs4_schedule_migration_recovery(server);
1670                         return -EAGAIN;
1671                 case -NFS4ERR_LEASE_MOVED:
1672                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1673                         return -EAGAIN;
1674                 case -NFS4ERR_DELEG_REVOKED:
1675                 case -NFS4ERR_ADMIN_REVOKED:
1676                 case -NFS4ERR_BAD_STATEID:
1677                 case -NFS4ERR_OPENMODE:
1678                         nfs_inode_find_state_and_recover(state->inode,
1679                                         stateid);
1680                         nfs4_schedule_stateid_recovery(server, state);
1681                         return -EAGAIN;
1682                 case -NFS4ERR_DELAY:
1683                 case -NFS4ERR_GRACE:
1684                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1685                         ssleep(1);
1686                         return -EAGAIN;
1687                 case -ENOMEM:
1688                 case -NFS4ERR_DENIED:
1689                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1690                         return 0;
1691         }
1692         return err;
1693 }
1694
1695 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1696 {
1697         struct nfs_server *server = NFS_SERVER(state->inode);
1698         struct nfs4_opendata *opendata;
1699         int err;
1700
1701         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1702                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1703         if (IS_ERR(opendata))
1704                 return PTR_ERR(opendata);
1705         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1706         err = nfs4_open_recover(opendata, state);
1707         nfs4_opendata_put(opendata);
1708         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1709 }
1710
1711 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1712 {
1713         struct nfs4_opendata *data = calldata;
1714
1715         nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1716                                 &data->c_res.seq_res, task);
1717 }
1718
1719 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1720 {
1721         struct nfs4_opendata *data = calldata;
1722
1723         nfs40_sequence_done(task, &data->c_res.seq_res);
1724
1725         data->rpc_status = task->tk_status;
1726         if (data->rpc_status == 0) {
1727                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1728                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1729                 renew_lease(data->o_res.server, data->timestamp);
1730                 data->rpc_done = 1;
1731         }
1732 }
1733
1734 static void nfs4_open_confirm_release(void *calldata)
1735 {
1736         struct nfs4_opendata *data = calldata;
1737         struct nfs4_state *state = NULL;
1738
1739         /* If this request hasn't been cancelled, do nothing */
1740         if (data->cancelled == 0)
1741                 goto out_free;
1742         /* In case of error, no cleanup! */
1743         if (!data->rpc_done)
1744                 goto out_free;
1745         state = nfs4_opendata_to_nfs4_state(data);
1746         if (!IS_ERR(state))
1747                 nfs4_close_state(state, data->o_arg.fmode);
1748 out_free:
1749         nfs4_opendata_put(data);
1750 }
1751
1752 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1753         .rpc_call_prepare = nfs4_open_confirm_prepare,
1754         .rpc_call_done = nfs4_open_confirm_done,
1755         .rpc_release = nfs4_open_confirm_release,
1756 };
1757
1758 /*
1759  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1760  */
1761 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1762 {
1763         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1764         struct rpc_task *task;
1765         struct  rpc_message msg = {
1766                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1767                 .rpc_argp = &data->c_arg,
1768                 .rpc_resp = &data->c_res,
1769                 .rpc_cred = data->owner->so_cred,
1770         };
1771         struct rpc_task_setup task_setup_data = {
1772                 .rpc_client = server->client,
1773                 .rpc_message = &msg,
1774                 .callback_ops = &nfs4_open_confirm_ops,
1775                 .callback_data = data,
1776                 .workqueue = nfsiod_workqueue,
1777                 .flags = RPC_TASK_ASYNC,
1778         };
1779         int status;
1780
1781         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1782         kref_get(&data->kref);
1783         data->rpc_done = 0;
1784         data->rpc_status = 0;
1785         data->timestamp = jiffies;
1786         task = rpc_run_task(&task_setup_data);
1787         if (IS_ERR(task))
1788                 return PTR_ERR(task);
1789         status = nfs4_wait_for_completion_rpc_task(task);
1790         if (status != 0) {
1791                 data->cancelled = 1;
1792                 smp_wmb();
1793         } else
1794                 status = data->rpc_status;
1795         rpc_put_task(task);
1796         return status;
1797 }
1798
1799 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1800 {
1801         struct nfs4_opendata *data = calldata;
1802         struct nfs4_state_owner *sp = data->owner;
1803         struct nfs_client *clp = sp->so_server->nfs_client;
1804
1805         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1806                 goto out_wait;
1807         /*
1808          * Check if we still need to send an OPEN call, or if we can use
1809          * a delegation instead.
1810          */
1811         if (data->state != NULL) {
1812                 struct nfs_delegation *delegation;
1813
1814                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1815                         goto out_no_action;
1816                 rcu_read_lock();
1817                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1818                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1819                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1820                     can_open_delegated(delegation, data->o_arg.fmode))
1821                         goto unlock_no_action;
1822                 rcu_read_unlock();
1823         }
1824         /* Update client id. */
1825         data->o_arg.clientid = clp->cl_clientid;
1826         switch (data->o_arg.claim) {
1827         case NFS4_OPEN_CLAIM_PREVIOUS:
1828         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1829         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1830                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1831         case NFS4_OPEN_CLAIM_FH:
1832                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1833                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1834         }
1835         data->timestamp = jiffies;
1836         if (nfs4_setup_sequence(data->o_arg.server,
1837                                 &data->o_arg.seq_args,
1838                                 &data->o_res.seq_res,
1839                                 task) != 0)
1840                 nfs_release_seqid(data->o_arg.seqid);
1841
1842         /* Set the create mode (note dependency on the session type) */
1843         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1844         if (data->o_arg.open_flags & O_EXCL) {
1845                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1846                 if (nfs4_has_persistent_session(clp))
1847                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1848                 else if (clp->cl_mvops->minor_version > 0)
1849                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1850         }
1851         return;
1852 unlock_no_action:
1853         rcu_read_unlock();
1854 out_no_action:
1855         task->tk_action = NULL;
1856 out_wait:
1857         nfs4_sequence_done(task, &data->o_res.seq_res);
1858 }
1859
1860 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1861 {
1862         struct nfs4_opendata *data = calldata;
1863
1864         data->rpc_status = task->tk_status;
1865
1866         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1867                 return;
1868
1869         if (task->tk_status == 0) {
1870                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1871                         switch (data->o_res.f_attr->mode & S_IFMT) {
1872                         case S_IFREG:
1873                                 break;
1874                         case S_IFLNK:
1875                                 data->rpc_status = -ELOOP;
1876                                 break;
1877                         case S_IFDIR:
1878                                 data->rpc_status = -EISDIR;
1879                                 break;
1880                         default:
1881                                 data->rpc_status = -ENOTDIR;
1882                         }
1883                 }
1884                 renew_lease(data->o_res.server, data->timestamp);
1885                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1886                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1887         }
1888         data->rpc_done = 1;
1889 }
1890
1891 static void nfs4_open_release(void *calldata)
1892 {
1893         struct nfs4_opendata *data = calldata;
1894         struct nfs4_state *state = NULL;
1895
1896         /* If this request hasn't been cancelled, do nothing */
1897         if (data->cancelled == 0)
1898                 goto out_free;
1899         /* In case of error, no cleanup! */
1900         if (data->rpc_status != 0 || !data->rpc_done)
1901                 goto out_free;
1902         /* In case we need an open_confirm, no cleanup! */
1903         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1904                 goto out_free;
1905         state = nfs4_opendata_to_nfs4_state(data);
1906         if (!IS_ERR(state))
1907                 nfs4_close_state(state, data->o_arg.fmode);
1908 out_free:
1909         nfs4_opendata_put(data);
1910 }
1911
1912 static const struct rpc_call_ops nfs4_open_ops = {
1913         .rpc_call_prepare = nfs4_open_prepare,
1914         .rpc_call_done = nfs4_open_done,
1915         .rpc_release = nfs4_open_release,
1916 };
1917
1918 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1919 {
1920         struct inode *dir = data->dir->d_inode;
1921         struct nfs_server *server = NFS_SERVER(dir);
1922         struct nfs_openargs *o_arg = &data->o_arg;
1923         struct nfs_openres *o_res = &data->o_res;
1924         struct rpc_task *task;
1925         struct rpc_message msg = {
1926                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1927                 .rpc_argp = o_arg,
1928                 .rpc_resp = o_res,
1929                 .rpc_cred = data->owner->so_cred,
1930         };
1931         struct rpc_task_setup task_setup_data = {
1932                 .rpc_client = server->client,
1933                 .rpc_message = &msg,
1934                 .callback_ops = &nfs4_open_ops,
1935                 .callback_data = data,
1936                 .workqueue = nfsiod_workqueue,
1937                 .flags = RPC_TASK_ASYNC,
1938         };
1939         int status;
1940
1941         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1942         kref_get(&data->kref);
1943         data->rpc_done = 0;
1944         data->rpc_status = 0;
1945         data->cancelled = 0;
1946         data->is_recover = 0;
1947         if (isrecover) {
1948                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1949                 data->is_recover = 1;
1950         }
1951         task = rpc_run_task(&task_setup_data);
1952         if (IS_ERR(task))
1953                 return PTR_ERR(task);
1954         status = nfs4_wait_for_completion_rpc_task(task);
1955         if (status != 0) {
1956                 data->cancelled = 1;
1957                 smp_wmb();
1958         } else
1959                 status = data->rpc_status;
1960         rpc_put_task(task);
1961
1962         return status;
1963 }
1964
1965 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1966 {
1967         struct inode *dir = data->dir->d_inode;
1968         struct nfs_openres *o_res = &data->o_res;
1969         int status;
1970
1971         status = nfs4_run_open_task(data, 1);
1972         if (status != 0 || !data->rpc_done)
1973                 return status;
1974
1975         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1976
1977         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1978                 status = _nfs4_proc_open_confirm(data);
1979                 if (status != 0)
1980                         return status;
1981         }
1982
1983         return status;
1984 }
1985
1986 /*
1987  * Additional permission checks in order to distinguish between an
1988  * open for read, and an open for execute. This works around the
1989  * fact that NFSv4 OPEN treats read and execute permissions as being
1990  * the same.
1991  * Note that in the non-execute case, we want to turn off permission
1992  * checking if we just created a new file (POSIX open() semantics).
1993  */
1994 static int nfs4_opendata_access(struct rpc_cred *cred,
1995                                 struct nfs4_opendata *opendata,
1996                                 struct nfs4_state *state, fmode_t fmode,
1997                                 int openflags)
1998 {
1999         struct nfs_access_entry cache;
2000         u32 mask;
2001
2002         /* access call failed or for some reason the server doesn't
2003          * support any access modes -- defer access call until later */
2004         if (opendata->o_res.access_supported == 0)
2005                 return 0;
2006
2007         mask = 0;
2008         /*
2009          * Use openflags to check for exec, because fmode won't
2010          * always have FMODE_EXEC set when file open for exec.
2011          */
2012         if (openflags & __FMODE_EXEC) {
2013                 /* ONLY check for exec rights */
2014                 mask = MAY_EXEC;
2015         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2016                 mask = MAY_READ;
2017
2018         cache.cred = cred;
2019         cache.jiffies = jiffies;
2020         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2021         nfs_access_add_cache(state->inode, &cache);
2022
2023         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2024                 return 0;
2025
2026         /* even though OPEN succeeded, access is denied. Close the file */
2027         nfs4_close_state(state, fmode);
2028         return -EACCES;
2029 }
2030
2031 /*
2032  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2033  */
2034 static int _nfs4_proc_open(struct nfs4_opendata *data)
2035 {
2036         struct inode *dir = data->dir->d_inode;
2037         struct nfs_server *server = NFS_SERVER(dir);
2038         struct nfs_openargs *o_arg = &data->o_arg;
2039         struct nfs_openres *o_res = &data->o_res;
2040         int status;
2041
2042         status = nfs4_run_open_task(data, 0);
2043         if (!data->rpc_done)
2044                 return status;
2045         if (status != 0) {
2046                 if (status == -NFS4ERR_BADNAME &&
2047                                 !(o_arg->open_flags & O_CREAT))
2048                         return -ENOENT;
2049                 return status;
2050         }
2051
2052         nfs_fattr_map_and_free_names(server, &data->f_attr);
2053
2054         if (o_arg->open_flags & O_CREAT) {
2055                 update_changeattr(dir, &o_res->cinfo);
2056                 if (o_arg->open_flags & O_EXCL)
2057                         data->file_created = 1;
2058                 else if (o_res->cinfo.before != o_res->cinfo.after)
2059                         data->file_created = 1;
2060         }
2061         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2062                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2063         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2064                 status = _nfs4_proc_open_confirm(data);
2065                 if (status != 0)
2066                         return status;
2067         }
2068         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2069                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2070         return 0;
2071 }
2072
2073 static int nfs4_recover_expired_lease(struct nfs_server *server)
2074 {
2075         return nfs4_client_recover_expired_lease(server->nfs_client);
2076 }
2077
2078 /*
2079  * OPEN_EXPIRED:
2080  *      reclaim state on the server after a network partition.
2081  *      Assumes caller holds the appropriate lock
2082  */
2083 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2084 {
2085         struct nfs4_opendata *opendata;
2086         int ret;
2087
2088         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2089                         NFS4_OPEN_CLAIM_FH);
2090         if (IS_ERR(opendata))
2091                 return PTR_ERR(opendata);
2092         ret = nfs4_open_recover(opendata, state);
2093         if (ret == -ESTALE)
2094                 d_drop(ctx->dentry);
2095         nfs4_opendata_put(opendata);
2096         return ret;
2097 }
2098
2099 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2100 {
2101         struct nfs_server *server = NFS_SERVER(state->inode);
2102         struct nfs4_exception exception = { };
2103         int err;
2104
2105         do {
2106                 err = _nfs4_open_expired(ctx, state);
2107                 trace_nfs4_open_expired(ctx, 0, err);
2108                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2109                         continue;
2110                 switch (err) {
2111                 default:
2112                         goto out;
2113                 case -NFS4ERR_GRACE:
2114                 case -NFS4ERR_DELAY:
2115                         nfs4_handle_exception(server, err, &exception);
2116                         err = 0;
2117                 }
2118         } while (exception.retry);
2119 out:
2120         return err;
2121 }
2122
2123 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2124 {
2125         struct nfs_open_context *ctx;
2126         int ret;
2127
2128         ctx = nfs4_state_find_open_context(state);
2129         if (IS_ERR(ctx))
2130                 return -EAGAIN;
2131         ret = nfs4_do_open_expired(ctx, state);
2132         put_nfs_open_context(ctx);
2133         return ret;
2134 }
2135
2136 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2137 {
2138         nfs_remove_bad_delegation(state->inode);
2139         write_seqlock(&state->seqlock);
2140         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2141         write_sequnlock(&state->seqlock);
2142         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2143 }
2144
2145 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2146 {
2147         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2148                 nfs_finish_clear_delegation_stateid(state);
2149 }
2150
2151 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2152 {
2153         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2154         nfs40_clear_delegation_stateid(state);
2155         return nfs4_open_expired(sp, state);
2156 }
2157
2158 #if defined(CONFIG_NFS_V4_1)
2159 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2160 {
2161         struct nfs_server *server = NFS_SERVER(state->inode);
2162         nfs4_stateid stateid;
2163         struct nfs_delegation *delegation;
2164         struct rpc_cred *cred;
2165         int status;
2166
2167         /* Get the delegation credential for use by test/free_stateid */
2168         rcu_read_lock();
2169         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2170         if (delegation == NULL) {
2171                 rcu_read_unlock();
2172                 return;
2173         }
2174
2175         nfs4_stateid_copy(&stateid, &delegation->stateid);
2176         cred = get_rpccred(delegation->cred);
2177         rcu_read_unlock();
2178         status = nfs41_test_stateid(server, &stateid, cred);
2179         trace_nfs4_test_delegation_stateid(state, NULL, status);
2180
2181         if (status != NFS_OK) {
2182                 /* Free the stateid unless the server explicitly
2183                  * informs us the stateid is unrecognized. */
2184                 if (status != -NFS4ERR_BAD_STATEID)
2185                         nfs41_free_stateid(server, &stateid, cred);
2186                 nfs_finish_clear_delegation_stateid(state);
2187         }
2188
2189         put_rpccred(cred);
2190 }
2191
2192 /**
2193  * nfs41_check_open_stateid - possibly free an open stateid
2194  *
2195  * @state: NFSv4 state for an inode
2196  *
2197  * Returns NFS_OK if recovery for this stateid is now finished.
2198  * Otherwise a negative NFS4ERR value is returned.
2199  */
2200 static int nfs41_check_open_stateid(struct nfs4_state *state)
2201 {
2202         struct nfs_server *server = NFS_SERVER(state->inode);
2203         nfs4_stateid *stateid = &state->open_stateid;
2204         struct rpc_cred *cred = state->owner->so_cred;
2205         int status;
2206
2207         /* If a state reset has been done, test_stateid is unneeded */
2208         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2209             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2210             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2211                 return -NFS4ERR_BAD_STATEID;
2212
2213         status = nfs41_test_stateid(server, stateid, cred);
2214         trace_nfs4_test_open_stateid(state, NULL, status);
2215         if (status != NFS_OK) {
2216                 /* Free the stateid unless the server explicitly
2217                  * informs us the stateid is unrecognized. */
2218                 if (status != -NFS4ERR_BAD_STATEID)
2219                         nfs41_free_stateid(server, stateid, cred);
2220
2221                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2222                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2223                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2224                 clear_bit(NFS_OPEN_STATE, &state->flags);
2225         }
2226         return status;
2227 }
2228
2229 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2230 {
2231         int status;
2232
2233         nfs41_check_delegation_stateid(state);
2234         status = nfs41_check_open_stateid(state);
2235         if (status != NFS_OK)
2236                 status = nfs4_open_expired(sp, state);
2237         return status;
2238 }
2239 #endif
2240
2241 /*
2242  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2243  * fields corresponding to attributes that were used to store the verifier.
2244  * Make sure we clobber those fields in the later setattr call
2245  */
2246 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2247 {
2248         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2249             !(sattr->ia_valid & ATTR_ATIME_SET))
2250                 sattr->ia_valid |= ATTR_ATIME;
2251
2252         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2253             !(sattr->ia_valid & ATTR_MTIME_SET))
2254                 sattr->ia_valid |= ATTR_MTIME;
2255 }
2256
2257 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2258                 fmode_t fmode,
2259                 int flags,
2260                 struct nfs_open_context *ctx)
2261 {
2262         struct nfs4_state_owner *sp = opendata->owner;
2263         struct nfs_server *server = sp->so_server;
2264         struct dentry *dentry;
2265         struct nfs4_state *state;
2266         unsigned int seq;
2267         int ret;
2268
2269         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2270
2271         ret = _nfs4_proc_open(opendata);
2272         if (ret != 0)
2273                 goto out;
2274
2275         state = nfs4_opendata_to_nfs4_state(opendata);
2276         ret = PTR_ERR(state);
2277         if (IS_ERR(state))
2278                 goto out;
2279         if (server->caps & NFS_CAP_POSIX_LOCK)
2280                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2281
2282         dentry = opendata->dentry;
2283         if (dentry->d_inode == NULL) {
2284                 /* FIXME: Is this d_drop() ever needed? */
2285                 d_drop(dentry);
2286                 dentry = d_add_unique(dentry, igrab(state->inode));
2287                 if (dentry == NULL) {
2288                         dentry = opendata->dentry;
2289                 } else if (dentry != ctx->dentry) {
2290                         dput(ctx->dentry);
2291                         ctx->dentry = dget(dentry);
2292                 }
2293                 nfs_set_verifier(dentry,
2294                                 nfs_save_change_attribute(opendata->dir->d_inode));
2295         }
2296
2297         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2298         if (ret != 0)
2299                 goto out;
2300
2301         ctx->state = state;
2302         if (dentry->d_inode == state->inode) {
2303                 nfs_inode_attach_open_context(ctx);
2304                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2305                         nfs4_schedule_stateid_recovery(server, state);
2306         }
2307 out:
2308         return ret;
2309 }
2310
2311 /*
2312  * Returns a referenced nfs4_state
2313  */
2314 static int _nfs4_do_open(struct inode *dir,
2315                         struct nfs_open_context *ctx,
2316                         int flags,
2317                         struct iattr *sattr,
2318                         struct nfs4_label *label,
2319                         int *opened)
2320 {
2321         struct nfs4_state_owner  *sp;
2322         struct nfs4_state     *state = NULL;
2323         struct nfs_server       *server = NFS_SERVER(dir);
2324         struct nfs4_opendata *opendata;
2325         struct dentry *dentry = ctx->dentry;
2326         struct rpc_cred *cred = ctx->cred;
2327         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2328         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2329         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2330         struct nfs4_label *olabel = NULL;
2331         int status;
2332
2333         /* Protect against reboot recovery conflicts */
2334         status = -ENOMEM;
2335         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2336         if (sp == NULL) {
2337                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2338                 goto out_err;
2339         }
2340         status = nfs4_recover_expired_lease(server);
2341         if (status != 0)
2342                 goto err_put_state_owner;
2343         if (dentry->d_inode != NULL)
2344                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2345         status = -ENOMEM;
2346         if (dentry->d_inode)
2347                 claim = NFS4_OPEN_CLAIM_FH;
2348         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2349                         label, claim, GFP_KERNEL);
2350         if (opendata == NULL)
2351                 goto err_put_state_owner;
2352
2353         if (label) {
2354                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2355                 if (IS_ERR(olabel)) {
2356                         status = PTR_ERR(olabel);
2357                         goto err_opendata_put;
2358                 }
2359         }
2360
2361         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2362                 if (!opendata->f_attr.mdsthreshold) {
2363                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2364                         if (!opendata->f_attr.mdsthreshold)
2365                                 goto err_free_label;
2366                 }
2367                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2368         }
2369         if (dentry->d_inode != NULL)
2370                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2371
2372         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2373         if (status != 0)
2374                 goto err_free_label;
2375         state = ctx->state;
2376
2377         if ((opendata->o_arg.open_flags & O_EXCL) &&
2378             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2379                 nfs4_exclusive_attrset(opendata, sattr);
2380
2381                 nfs_fattr_init(opendata->o_res.f_attr);
2382                 status = nfs4_do_setattr(state->inode, cred,
2383                                 opendata->o_res.f_attr, sattr,
2384                                 state, label, olabel);
2385                 if (status == 0) {
2386                         nfs_setattr_update_inode(state->inode, sattr);
2387                         nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2388                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2389                 }
2390         }
2391         if (opendata->file_created)
2392                 *opened |= FILE_CREATED;
2393
2394         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2395                 *ctx_th = opendata->f_attr.mdsthreshold;
2396                 opendata->f_attr.mdsthreshold = NULL;
2397         }
2398
2399         nfs4_label_free(olabel);
2400
2401         nfs4_opendata_put(opendata);
2402         nfs4_put_state_owner(sp);
2403         return 0;
2404 err_free_label:
2405         nfs4_label_free(olabel);
2406 err_opendata_put:
2407         nfs4_opendata_put(opendata);
2408 err_put_state_owner:
2409         nfs4_put_state_owner(sp);
2410 out_err:
2411         return status;
2412 }
2413
2414
2415 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2416                                         struct nfs_open_context *ctx,
2417                                         int flags,
2418                                         struct iattr *sattr,
2419                                         struct nfs4_label *label,
2420                                         int *opened)
2421 {
2422         struct nfs_server *server = NFS_SERVER(dir);
2423         struct nfs4_exception exception = { };
2424         struct nfs4_state *res;
2425         int status;
2426
2427         do {
2428                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2429                 res = ctx->state;
2430                 trace_nfs4_open_file(ctx, flags, status);
2431                 if (status == 0)
2432                         break;
2433                 /* NOTE: BAD_SEQID means the server and client disagree about the
2434                  * book-keeping w.r.t. state-changing operations
2435                  * (OPEN/CLOSE/LOCK/LOCKU...)
2436                  * It is actually a sign of a bug on the client or on the server.
2437                  *
2438                  * If we receive a BAD_SEQID error in the particular case of
2439                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2440                  * have unhashed the old state_owner for us, and that we can
2441                  * therefore safely retry using a new one. We should still warn
2442                  * the user though...
2443                  */
2444                 if (status == -NFS4ERR_BAD_SEQID) {
2445                         pr_warn_ratelimited("NFS: v4 server %s "
2446                                         " returned a bad sequence-id error!\n",
2447                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2448                         exception.retry = 1;
2449                         continue;
2450                 }
2451                 /*
2452                  * BAD_STATEID on OPEN means that the server cancelled our
2453                  * state before it received the OPEN_CONFIRM.
2454                  * Recover by retrying the request as per the discussion
2455                  * on Page 181 of RFC3530.
2456                  */
2457                 if (status == -NFS4ERR_BAD_STATEID) {
2458                         exception.retry = 1;
2459                         continue;
2460                 }
2461                 if (status == -EAGAIN) {
2462                         /* We must have found a delegation */
2463                         exception.retry = 1;
2464                         continue;
2465                 }
2466                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2467                         continue;
2468                 res = ERR_PTR(nfs4_handle_exception(server,
2469                                         status, &exception));
2470         } while (exception.retry);
2471         return res;
2472 }
2473
2474 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2475                             struct nfs_fattr *fattr, struct iattr *sattr,
2476                             struct nfs4_state *state, struct nfs4_label *ilabel,
2477                             struct nfs4_label *olabel)
2478 {
2479         struct nfs_server *server = NFS_SERVER(inode);
2480         struct nfs_setattrargs  arg = {
2481                 .fh             = NFS_FH(inode),
2482                 .iap            = sattr,
2483                 .server         = server,
2484                 .bitmask = server->attr_bitmask,
2485                 .label          = ilabel,
2486         };
2487         struct nfs_setattrres  res = {
2488                 .fattr          = fattr,
2489                 .label          = olabel,
2490                 .server         = server,
2491         };
2492         struct rpc_message msg = {
2493                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2494                 .rpc_argp       = &arg,
2495                 .rpc_resp       = &res,
2496                 .rpc_cred       = cred,
2497         };
2498         unsigned long timestamp = jiffies;
2499         fmode_t fmode;
2500         bool truncate;
2501         int status;
2502
2503         arg.bitmask = nfs4_bitmask(server, ilabel);
2504         if (ilabel)
2505                 arg.bitmask = nfs4_bitmask(server, olabel);
2506
2507         nfs_fattr_init(fattr);
2508
2509         /* Servers should only apply open mode checks for file size changes */
2510         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2511         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2512
2513         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2514                 /* Use that stateid */
2515         } else if (truncate && state != NULL) {
2516                 struct nfs_lockowner lockowner = {
2517                         .l_owner = current->files,
2518                         .l_pid = current->tgid,
2519                 };
2520                 if (!nfs4_valid_open_stateid(state))
2521                         return -EBADF;
2522                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2523                                 &lockowner) == -EIO)
2524                         return -EBADF;
2525         } else
2526                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2527
2528         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2529         if (status == 0 && state != NULL)
2530                 renew_lease(server, timestamp);
2531         return status;
2532 }
2533
2534 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2535                            struct nfs_fattr *fattr, struct iattr *sattr,
2536                            struct nfs4_state *state, struct nfs4_label *ilabel,
2537                            struct nfs4_label *olabel)
2538 {
2539         struct nfs_server *server = NFS_SERVER(inode);
2540         struct nfs4_exception exception = {
2541                 .state = state,
2542                 .inode = inode,
2543         };
2544         int err;
2545         do {
2546                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2547                 trace_nfs4_setattr(inode, err);
2548                 switch (err) {
2549                 case -NFS4ERR_OPENMODE:
2550                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2551                                 pr_warn_once("NFSv4: server %s is incorrectly "
2552                                                 "applying open mode checks to "
2553                                                 "a SETATTR that is not "
2554                                                 "changing file size.\n",
2555                                                 server->nfs_client->cl_hostname);
2556                         }
2557                         if (state && !(state->state & FMODE_WRITE)) {
2558                                 err = -EBADF;
2559                                 if (sattr->ia_valid & ATTR_OPEN)
2560                                         err = -EACCES;
2561                                 goto out;
2562                         }
2563                 }
2564                 err = nfs4_handle_exception(server, err, &exception);
2565         } while (exception.retry);
2566 out:
2567         return err;
2568 }
2569
2570 struct nfs4_closedata {
2571         struct inode *inode;
2572         struct nfs4_state *state;
2573         struct nfs_closeargs arg;
2574         struct nfs_closeres res;
2575         struct nfs_fattr fattr;
2576         unsigned long timestamp;
2577         bool roc;
2578         u32 roc_barrier;
2579 };
2580
2581 static void nfs4_free_closedata(void *data)
2582 {
2583         struct nfs4_closedata *calldata = data;
2584         struct nfs4_state_owner *sp = calldata->state->owner;
2585         struct super_block *sb = calldata->state->inode->i_sb;
2586
2587         if (calldata->roc)
2588                 pnfs_roc_release(calldata->state->inode);
2589         nfs4_put_open_state(calldata->state);
2590         nfs_free_seqid(calldata->arg.seqid);
2591         nfs4_put_state_owner(sp);
2592         nfs_sb_deactive(sb);
2593         kfree(calldata);
2594 }
2595
2596 static void nfs4_close_done(struct rpc_task *task, void *data)
2597 {
2598         struct nfs4_closedata *calldata = data;
2599         struct nfs4_state *state = calldata->state;
2600         struct nfs_server *server = NFS_SERVER(calldata->inode);
2601         nfs4_stateid *res_stateid = NULL;
2602
2603         dprintk("%s: begin!\n", __func__);
2604         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2605                 return;
2606         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2607         /* hmm. we are done with the inode, and in the process of freeing
2608          * the state_owner. we keep this around to process errors
2609          */
2610         switch (task->tk_status) {
2611                 case 0:
2612                         res_stateid = &calldata->res.stateid;
2613                         if (calldata->arg.fmode == 0 && calldata->roc)
2614                                 pnfs_roc_set_barrier(state->inode,
2615                                                      calldata->roc_barrier);
2616                         renew_lease(server, calldata->timestamp);
2617                         break;
2618                 case -NFS4ERR_ADMIN_REVOKED:
2619                 case -NFS4ERR_STALE_STATEID:
2620                 case -NFS4ERR_OLD_STATEID:
2621                 case -NFS4ERR_BAD_STATEID:
2622                 case -NFS4ERR_EXPIRED:
2623                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2624                                                 &state->stateid)) {
2625                                 rpc_restart_call_prepare(task);
2626                                 goto out_release;
2627                         }
2628                         if (calldata->arg.fmode == 0)
2629                                 break;
2630                 default:
2631                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2632                                 rpc_restart_call_prepare(task);
2633                                 goto out_release;
2634                         }
2635         }
2636         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2637 out_release:
2638         nfs_release_seqid(calldata->arg.seqid);
2639         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2640         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2641 }
2642
2643 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2644 {
2645         struct nfs4_closedata *calldata = data;
2646         struct nfs4_state *state = calldata->state;
2647         struct inode *inode = calldata->inode;
2648         bool is_rdonly, is_wronly, is_rdwr;
2649         int call_close = 0;
2650
2651         dprintk("%s: begin!\n", __func__);
2652         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2653                 goto out_wait;
2654
2655         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2656         spin_lock(&state->owner->so_lock);
2657         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2658         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2659         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2660         nfs4_stateid_copy(&calldata->arg.stateid, &state->stateid);
2661         /* Calculate the change in open mode */
2662         calldata->arg.fmode = 0;
2663         if (state->n_rdwr == 0) {
2664                 if (state->n_rdonly == 0)
2665                         call_close |= is_rdonly;
2666                 else if (is_rdonly)
2667                         calldata->arg.fmode |= FMODE_READ;
2668                 if (state->n_wronly == 0)
2669                         call_close |= is_wronly;
2670                 else if (is_wronly)
2671                         calldata->arg.fmode |= FMODE_WRITE;
2672         } else if (is_rdwr)
2673                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2674
2675         if (calldata->arg.fmode == 0)
2676                 call_close |= is_rdwr;
2677
2678         if (!nfs4_valid_open_stateid(state))
2679                 call_close = 0;
2680         spin_unlock(&state->owner->so_lock);
2681
2682         if (!call_close) {
2683                 /* Note: exit _without_ calling nfs4_close_done */
2684                 goto out_no_action;
2685         }
2686
2687         if (calldata->arg.fmode == 0) {
2688                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2689                 if (calldata->roc &&
2690                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2691                         nfs_release_seqid(calldata->arg.seqid);
2692                         goto out_wait;
2693                     }
2694         }
2695
2696         nfs_fattr_init(calldata->res.fattr);
2697         calldata->timestamp = jiffies;
2698         if (nfs4_setup_sequence(NFS_SERVER(inode),
2699                                 &calldata->arg.seq_args,
2700                                 &calldata->res.seq_res,
2701                                 task) != 0)
2702                 nfs_release_seqid(calldata->arg.seqid);
2703         dprintk("%s: done!\n", __func__);
2704         return;
2705 out_no_action:
2706         task->tk_action = NULL;
2707 out_wait:
2708         nfs4_sequence_done(task, &calldata->res.seq_res);
2709 }
2710
2711 static const struct rpc_call_ops nfs4_close_ops = {
2712         .rpc_call_prepare = nfs4_close_prepare,
2713         .rpc_call_done = nfs4_close_done,
2714         .rpc_release = nfs4_free_closedata,
2715 };
2716
2717 static bool nfs4_state_has_opener(struct nfs4_state *state)
2718 {
2719         /* first check existing openers */
2720         if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0 &&
2721             state->n_rdonly != 0)
2722                 return true;
2723
2724         if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0 &&
2725             state->n_wronly != 0)
2726                 return true;
2727
2728         if (test_bit(NFS_O_RDWR_STATE, &state->flags) != 0 &&
2729             state->n_rdwr != 0)
2730                 return true;
2731
2732         return false;
2733 }
2734
2735 static bool nfs4_roc(struct inode *inode)
2736 {
2737         struct nfs_inode *nfsi = NFS_I(inode);
2738         struct nfs_open_context *ctx;
2739         struct nfs4_state *state;
2740
2741         spin_lock(&inode->i_lock);
2742         list_for_each_entry(ctx, &nfsi->open_files, list) {
2743                 state = ctx->state;
2744                 if (state == NULL)
2745                         continue;
2746                 if (nfs4_state_has_opener(state)) {
2747                         spin_unlock(&inode->i_lock);
2748                         return false;
2749                 }
2750         }
2751         spin_unlock(&inode->i_lock);
2752
2753         if (nfs4_check_delegation(inode, FMODE_READ))
2754                 return false;
2755
2756         return pnfs_roc(inode);
2757 }
2758
2759 /*
2760  * It is possible for data to be read/written from a mem-mapped file
2761  * after the sys_close call (which hits the vfs layer as a flush).
2762  * This means that we can't safely call nfsv4 close on a file until
2763  * the inode is cleared. This in turn means that we are not good
2764  * NFSv4 citizens - we do not indicate to the server to update the file's
2765  * share state even when we are done with one of the three share
2766  * stateid's in the inode.
2767  *
2768  * NOTE: Caller must be holding the sp->so_owner semaphore!
2769  */
2770 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2771 {
2772         struct nfs_server *server = NFS_SERVER(state->inode);
2773         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2774         struct nfs4_closedata *calldata;
2775         struct nfs4_state_owner *sp = state->owner;
2776         struct rpc_task *task;
2777         struct rpc_message msg = {
2778                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2779                 .rpc_cred = state->owner->so_cred,
2780         };
2781         struct rpc_task_setup task_setup_data = {
2782                 .rpc_client = server->client,
2783                 .rpc_message = &msg,
2784                 .callback_ops = &nfs4_close_ops,
2785                 .workqueue = nfsiod_workqueue,
2786                 .flags = RPC_TASK_ASYNC,
2787         };
2788         int status = -ENOMEM;
2789
2790         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2791                 &task_setup_data.rpc_client, &msg);
2792
2793         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2794         if (calldata == NULL)
2795                 goto out;
2796         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2797         calldata->inode = state->inode;
2798         calldata->state = state;
2799         calldata->arg.fh = NFS_FH(state->inode);
2800         /* Serialization for the sequence id */
2801         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2802         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2803         if (IS_ERR(calldata->arg.seqid))
2804                 goto out_free_calldata;
2805         calldata->arg.fmode = 0;
2806         calldata->arg.bitmask = server->cache_consistency_bitmask;
2807         calldata->res.fattr = &calldata->fattr;
2808         calldata->res.seqid = calldata->arg.seqid;
2809         calldata->res.server = server;
2810         calldata->roc = nfs4_roc(state->inode);
2811         nfs_sb_active(calldata->inode->i_sb);
2812
2813         msg.rpc_argp = &calldata->arg;
2814         msg.rpc_resp = &calldata->res;
2815         task_setup_data.callback_data = calldata;
2816         task = rpc_run_task(&task_setup_data);
2817         if (IS_ERR(task))
2818                 return PTR_ERR(task);
2819         status = 0;
2820         if (wait)
2821                 status = rpc_wait_for_completion_task(task);
2822         rpc_put_task(task);
2823         return status;
2824 out_free_calldata:
2825         kfree(calldata);
2826 out:
2827         nfs4_put_open_state(state);
2828         nfs4_put_state_owner(sp);
2829         return status;
2830 }
2831
2832 static struct inode *
2833 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2834                 int open_flags, struct iattr *attr, int *opened)
2835 {
2836         struct nfs4_state *state;
2837         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2838
2839         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2840
2841         /* Protect against concurrent sillydeletes */
2842         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2843
2844         nfs4_label_release_security(label);
2845
2846         if (IS_ERR(state))
2847                 return ERR_CAST(state);
2848         return state->inode;
2849 }
2850
2851 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2852 {
2853         if (ctx->state == NULL)
2854                 return;
2855         if (is_sync)
2856                 nfs4_close_sync(ctx->state, ctx->mode);
2857         else
2858                 nfs4_close_state(ctx->state, ctx->mode);
2859 }
2860
2861 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2862 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2863 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2864
2865 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2866 {
2867         struct nfs4_server_caps_arg args = {
2868                 .fhandle = fhandle,
2869         };
2870         struct nfs4_server_caps_res res = {};
2871         struct rpc_message msg = {
2872                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2873                 .rpc_argp = &args,
2874                 .rpc_resp = &res,
2875         };
2876         int status;
2877
2878         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2879         if (status == 0) {
2880                 /* Sanity check the server answers */
2881                 switch (server->nfs_client->cl_minorversion) {
2882                 case 0:
2883                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2884                         res.attr_bitmask[2] = 0;
2885                         break;
2886                 case 1:
2887                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2888                         break;
2889                 case 2:
2890                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2891                 }
2892                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2893                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2894                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2895                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2896                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2897                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2898                                 NFS_CAP_SECURITY_LABEL);
2899                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2900                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2901                         server->caps |= NFS_CAP_ACLS;
2902                 if (res.has_links != 0)
2903                         server->caps |= NFS_CAP_HARDLINKS;
2904                 if (res.has_symlinks != 0)
2905                         server->caps |= NFS_CAP_SYMLINKS;
2906                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2907                         server->caps |= NFS_CAP_FILEID;
2908                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2909                         server->caps |= NFS_CAP_MODE;
2910                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2911                         server->caps |= NFS_CAP_NLINK;
2912                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2913                         server->caps |= NFS_CAP_OWNER;
2914                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2915                         server->caps |= NFS_CAP_OWNER_GROUP;
2916                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2917                         server->caps |= NFS_CAP_ATIME;
2918                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2919                         server->caps |= NFS_CAP_CTIME;
2920                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2921                         server->caps |= NFS_CAP_MTIME;
2922 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2923                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2924                         server->caps |= NFS_CAP_SECURITY_LABEL;
2925 #endif
2926                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2927                                 sizeof(server->attr_bitmask));
2928                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2929
2930                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2931                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2932                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2933                 server->cache_consistency_bitmask[2] = 0;
2934                 server->acl_bitmask = res.acl_bitmask;
2935                 server->fh_expire_type = res.fh_expire_type;
2936         }
2937
2938         return status;
2939 }
2940
2941 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2942 {
2943         struct nfs4_exception exception = { };
2944         int err;
2945         do {
2946                 err = nfs4_handle_exception(server,
2947                                 _nfs4_server_capabilities(server, fhandle),
2948                                 &exception);
2949         } while (exception.retry);
2950         return err;
2951 }
2952
2953 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2954                 struct nfs_fsinfo *info)
2955 {
2956         u32 bitmask[3];
2957         struct nfs4_lookup_root_arg args = {
2958                 .bitmask = bitmask,
2959         };
2960         struct nfs4_lookup_res res = {
2961                 .server = server,
2962                 .fattr = info->fattr,
2963                 .fh = fhandle,
2964         };
2965         struct rpc_message msg = {
2966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2967                 .rpc_argp = &args,
2968                 .rpc_resp = &res,
2969         };
2970
2971         bitmask[0] = nfs4_fattr_bitmap[0];
2972         bitmask[1] = nfs4_fattr_bitmap[1];
2973         /*
2974          * Process the label in the upcoming getfattr
2975          */
2976         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2977
2978         nfs_fattr_init(info->fattr);
2979         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2980 }
2981
2982 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2983                 struct nfs_fsinfo *info)
2984 {
2985         struct nfs4_exception exception = { };
2986         int err;
2987         do {
2988                 err = _nfs4_lookup_root(server, fhandle, info);
2989                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2990                 switch (err) {
2991                 case 0:
2992                 case -NFS4ERR_WRONGSEC:
2993                         goto out;
2994                 default:
2995                         err = nfs4_handle_exception(server, err, &exception);
2996                 }
2997         } while (exception.retry);
2998 out:
2999         return err;
3000 }
3001
3002 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3003                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3004 {
3005         struct rpc_auth_create_args auth_args = {
3006                 .pseudoflavor = flavor,
3007         };
3008         struct rpc_auth *auth;
3009         int ret;
3010
3011         auth = rpcauth_create(&auth_args, server->client);
3012         if (IS_ERR(auth)) {
3013                 ret = -EACCES;
3014                 goto out;
3015         }
3016         ret = nfs4_lookup_root(server, fhandle, info);
3017 out:
3018         return ret;
3019 }
3020
3021 /*
3022  * Retry pseudoroot lookup with various security flavors.  We do this when:
3023  *
3024  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3025  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3026  *
3027  * Returns zero on success, or a negative NFS4ERR value, or a
3028  * negative errno value.
3029  */
3030 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3031                               struct nfs_fsinfo *info)
3032 {
3033         /* Per 3530bis 15.33.5 */
3034         static const rpc_authflavor_t flav_array[] = {
3035                 RPC_AUTH_GSS_KRB5P,
3036                 RPC_AUTH_GSS_KRB5I,
3037                 RPC_AUTH_GSS_KRB5,
3038                 RPC_AUTH_UNIX,                  /* courtesy */
3039                 RPC_AUTH_NULL,
3040         };
3041         int status = -EPERM;
3042         size_t i;
3043
3044         if (server->auth_info.flavor_len > 0) {
3045                 /* try each flavor specified by user */
3046                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3047                         status = nfs4_lookup_root_sec(server, fhandle, info,
3048                                                 server->auth_info.flavors[i]);
3049                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3050                                 continue;
3051                         break;
3052                 }
3053         } else {
3054                 /* no flavors specified by user, try default list */
3055                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3056                         status = nfs4_lookup_root_sec(server, fhandle, info,
3057                                                       flav_array[i]);
3058                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3059                                 continue;
3060                         break;
3061                 }
3062         }
3063
3064         /*
3065          * -EACCESS could mean that the user doesn't have correct permissions
3066          * to access the mount.  It could also mean that we tried to mount
3067          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3068          * existing mount programs don't handle -EACCES very well so it should
3069          * be mapped to -EPERM instead.
3070          */
3071         if (status == -EACCES)
3072                 status = -EPERM;
3073         return status;
3074 }
3075
3076 static int nfs4_do_find_root_sec(struct nfs_server *server,
3077                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3078 {
3079         int mv = server->nfs_client->cl_minorversion;
3080         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3081 }
3082
3083 /**
3084  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3085  * @server: initialized nfs_server handle
3086  * @fhandle: we fill in the pseudo-fs root file handle
3087  * @info: we fill in an FSINFO struct
3088  * @auth_probe: probe the auth flavours
3089  *
3090  * Returns zero on success, or a negative errno.
3091  */
3092 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3093                          struct nfs_fsinfo *info,
3094                          bool auth_probe)
3095 {
3096         int status;
3097
3098         switch (auth_probe) {
3099         case false:
3100                 status = nfs4_lookup_root(server, fhandle, info);
3101                 if (status != -NFS4ERR_WRONGSEC)
3102                         break;
3103         default:
3104                 status = nfs4_do_find_root_sec(server, fhandle, info);
3105         }
3106
3107         if (status == 0)
3108                 status = nfs4_server_capabilities(server, fhandle);
3109         if (status == 0)
3110                 status = nfs4_do_fsinfo(server, fhandle, info);
3111
3112         return nfs4_map_errors(status);
3113 }
3114
3115 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3116                               struct nfs_fsinfo *info)
3117 {
3118         int error;
3119         struct nfs_fattr *fattr = info->fattr;
3120         struct nfs4_label *label = NULL;
3121
3122         error = nfs4_server_capabilities(server, mntfh);
3123         if (error < 0) {
3124                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3125                 return error;
3126         }
3127
3128         label = nfs4_label_alloc(server, GFP_KERNEL);
3129         if (IS_ERR(label))
3130                 return PTR_ERR(label);
3131
3132         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3133         if (error < 0) {
3134                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3135                 goto err_free_label;
3136         }
3137
3138         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3139             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3140                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3141
3142 err_free_label:
3143         nfs4_label_free(label);
3144
3145         return error;
3146 }
3147
3148 /*
3149  * Get locations and (maybe) other attributes of a referral.
3150  * Note that we'll actually follow the referral later when
3151  * we detect fsid mismatch in inode revalidation
3152  */
3153 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3154                              const struct qstr *name, struct nfs_fattr *fattr,
3155                              struct nfs_fh *fhandle)
3156 {
3157         int status = -ENOMEM;
3158         struct page *page = NULL;
3159         struct nfs4_fs_locations *locations = NULL;
3160
3161         page = alloc_page(GFP_KERNEL);
3162         if (page == NULL)
3163                 goto out;
3164         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3165         if (locations == NULL)
3166                 goto out;
3167
3168         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3169         if (status != 0)
3170                 goto out;
3171
3172         /*
3173          * If the fsid didn't change, this is a migration event, not a
3174          * referral.  Cause us to drop into the exception handler, which
3175          * will kick off migration recovery.
3176          */
3177         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3178                 dprintk("%s: server did not return a different fsid for"
3179                         " a referral at %s\n", __func__, name->name);
3180                 status = -NFS4ERR_MOVED;
3181                 goto out;
3182         }
3183         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3184         nfs_fixup_referral_attributes(&locations->fattr);
3185
3186         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3187         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3188         memset(fhandle, 0, sizeof(struct nfs_fh));
3189 out:
3190         if (page)
3191                 __free_page(page);
3192         kfree(locations);
3193         return status;
3194 }
3195
3196 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3197                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3198 {
3199         struct nfs4_getattr_arg args = {
3200                 .fh = fhandle,
3201                 .bitmask = server->attr_bitmask,
3202         };
3203         struct nfs4_getattr_res res = {
3204                 .fattr = fattr,
3205                 .label = label,
3206                 .server = server,
3207         };
3208         struct rpc_message msg = {
3209                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3210                 .rpc_argp = &args,
3211                 .rpc_resp = &res,
3212         };
3213
3214         args.bitmask = nfs4_bitmask(server, label);
3215
3216         nfs_fattr_init(fattr);
3217         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3218 }
3219
3220 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3221                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3222 {
3223         struct nfs4_exception exception = { };
3224         int err;
3225         do {
3226                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3227                 trace_nfs4_getattr(server, fhandle, fattr, err);
3228                 err = nfs4_handle_exception(server, err,
3229                                 &exception);
3230         } while (exception.retry);
3231         return err;
3232 }
3233
3234 /*
3235  * The file is not closed if it is opened due to the a request to change
3236  * the size of the file. The open call will not be needed once the
3237  * VFS layer lookup-intents are implemented.
3238  *
3239  * Close is called when the inode is destroyed.
3240  * If we haven't opened the file for O_WRONLY, we
3241  * need to in the size_change case to obtain a stateid.
3242  *
3243  * Got race?
3244  * Because OPEN is always done by name in nfsv4, it is
3245  * possible that we opened a different file by the same
3246  * name.  We can recognize this race condition, but we
3247  * can't do anything about it besides returning an error.
3248  *
3249  * This will be fixed with VFS changes (lookup-intent).
3250  */
3251 static int
3252 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3253                   struct iattr *sattr)
3254 {
3255         struct inode *inode = dentry->d_inode;
3256         struct rpc_cred *cred = NULL;
3257         struct nfs4_state *state = NULL;
3258         struct nfs4_label *label = NULL;
3259         int status;
3260
3261         if (pnfs_ld_layoutret_on_setattr(inode) &&
3262             sattr->ia_valid & ATTR_SIZE &&
3263             sattr->ia_size < i_size_read(inode))
3264                 pnfs_commit_and_return_layout(inode);
3265
3266         nfs_fattr_init(fattr);
3267
3268         /* Deal with open(O_TRUNC) */
3269         if (sattr->ia_valid & ATTR_OPEN)
3270                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3271
3272         /* Optimization: if the end result is no change, don't RPC */
3273         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3274                 return 0;
3275
3276         /* Search for an existing open(O_WRITE) file */
3277         if (sattr->ia_valid & ATTR_FILE) {
3278                 struct nfs_open_context *ctx;
3279
3280                 ctx = nfs_file_open_context(sattr->ia_file);
3281                 if (ctx) {
3282                         cred = ctx->cred;
3283                         state = ctx->state;
3284                 }
3285         }
3286
3287         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3288         if (IS_ERR(label))
3289                 return PTR_ERR(label);
3290
3291         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3292         if (status == 0) {
3293                 nfs_setattr_update_inode(inode, sattr);
3294                 nfs_setsecurity(inode, fattr, label);
3295         }
3296         nfs4_label_free(label);
3297         return status;
3298 }
3299
3300 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3301                 const struct qstr *name, struct nfs_fh *fhandle,
3302                 struct nfs_fattr *fattr, struct nfs4_label *label)
3303 {
3304         struct nfs_server *server = NFS_SERVER(dir);
3305         int                    status;
3306         struct nfs4_lookup_arg args = {
3307                 .bitmask = server->attr_bitmask,
3308                 .dir_fh = NFS_FH(dir),
3309                 .name = name,
3310         };
3311         struct nfs4_lookup_res res = {
3312                 .server = server,
3313                 .fattr = fattr,
3314                 .label = label,
3315                 .fh = fhandle,
3316         };
3317         struct rpc_message msg = {
3318                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3319                 .rpc_argp = &args,
3320                 .rpc_resp = &res,
3321         };
3322
3323         args.bitmask = nfs4_bitmask(server, label);
3324
3325         nfs_fattr_init(fattr);
3326
3327         dprintk("NFS call  lookup %s\n", name->name);
3328         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3329         dprintk("NFS reply lookup: %d\n", status);
3330         return status;
3331 }
3332
3333 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3334 {
3335         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3336                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3337         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3338         fattr->nlink = 2;
3339 }
3340
3341 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3342                                    struct qstr *name, struct nfs_fh *fhandle,
3343                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3344 {
3345         struct nfs4_exception exception = { };
3346         struct rpc_clnt *client = *clnt;
3347         int err;
3348         do {
3349                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3350                 trace_nfs4_lookup(dir, name, err);
3351                 switch (err) {
3352                 case -NFS4ERR_BADNAME:
3353                         err = -ENOENT;
3354                         goto out;
3355                 case -NFS4ERR_MOVED:
3356                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3357                         goto out;
3358                 case -NFS4ERR_WRONGSEC:
3359                         err = -EPERM;
3360                         if (client != *clnt)
3361                                 goto out;
3362                         client = nfs4_negotiate_security(client, dir, name);
3363                         if (IS_ERR(client))
3364                                 return PTR_ERR(client);
3365
3366                         exception.retry = 1;
3367                         break;
3368                 default:
3369                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3370                 }
3371         } while (exception.retry);
3372
3373 out:
3374         if (err == 0)
3375                 *clnt = client;
3376         else if (client != *clnt)
3377                 rpc_shutdown_client(client);
3378
3379         return err;
3380 }
3381
3382 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3383                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3384                             struct nfs4_label *label)
3385 {
3386         int status;
3387         struct rpc_clnt *client = NFS_CLIENT(dir);
3388
3389         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3390         if (client != NFS_CLIENT(dir)) {
3391                 rpc_shutdown_client(client);
3392                 nfs_fixup_secinfo_attributes(fattr);
3393         }
3394         return status;
3395 }
3396
3397 struct rpc_clnt *
3398 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3399                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3400 {
3401         struct rpc_clnt *client = NFS_CLIENT(dir);
3402         int status;
3403
3404         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3405         if (status < 0)
3406                 return ERR_PTR(status);
3407         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3408 }
3409
3410 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3411 {
3412         struct nfs_server *server = NFS_SERVER(inode);
3413         struct nfs4_accessargs args = {
3414                 .fh = NFS_FH(inode),
3415                 .bitmask = server->cache_consistency_bitmask,
3416         };
3417         struct nfs4_accessres res = {
3418                 .server = server,
3419         };
3420         struct rpc_message msg = {
3421                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3422                 .rpc_argp = &args,
3423                 .rpc_resp = &res,
3424                 .rpc_cred = entry->cred,
3425         };
3426         int mode = entry->mask;
3427         int status = 0;
3428
3429         /*
3430          * Determine which access bits we want to ask for...
3431          */
3432         if (mode & MAY_READ)
3433                 args.access |= NFS4_ACCESS_READ;
3434         if (S_ISDIR(inode->i_mode)) {
3435                 if (mode & MAY_WRITE)
3436                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3437                 if (mode & MAY_EXEC)
3438                         args.access |= NFS4_ACCESS_LOOKUP;
3439         } else {
3440                 if (mode & MAY_WRITE)
3441                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3442                 if (mode & MAY_EXEC)
3443                         args.access |= NFS4_ACCESS_EXECUTE;
3444         }
3445
3446         res.fattr = nfs_alloc_fattr();
3447         if (res.fattr == NULL)
3448                 return -ENOMEM;
3449
3450         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3451         if (!status) {
3452                 nfs_access_set_mask(entry, res.access);
3453                 nfs_refresh_inode(inode, res.fattr);
3454         }
3455         nfs_free_fattr(res.fattr);
3456         return status;
3457 }
3458
3459 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3460 {
3461         struct nfs4_exception exception = { };
3462         int err;
3463         do {
3464                 err = _nfs4_proc_access(inode, entry);
3465                 trace_nfs4_access(inode, err);
3466                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3467                                 &exception);
3468         } while (exception.retry);
3469         return err;
3470 }
3471
3472 /*
3473  * TODO: For the time being, we don't try to get any attributes
3474  * along with any of the zero-copy operations READ, READDIR,
3475  * READLINK, WRITE.
3476  *
3477  * In the case of the first three, we want to put the GETATTR
3478  * after the read-type operation -- this is because it is hard
3479  * to predict the length of a GETATTR response in v4, and thus
3480  * align the READ data correctly.  This means that the GETATTR
3481  * may end up partially falling into the page cache, and we should
3482  * shift it into the 'tail' of the xdr_buf before processing.
3483  * To do this efficiently, we need to know the total length
3484  * of data received, which doesn't seem to be available outside
3485  * of the RPC layer.
3486  *
3487  * In the case of WRITE, we also want to put the GETATTR after
3488  * the operation -- in this case because we want to make sure
3489  * we get the post-operation mtime and size.
3490  *
3491  * Both of these changes to the XDR layer would in fact be quite
3492  * minor, but I decided to leave them for a subsequent patch.
3493  */
3494 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3495                 unsigned int pgbase, unsigned int pglen)
3496 {
3497         struct nfs4_readlink args = {
3498                 .fh       = NFS_FH(inode),
3499                 .pgbase   = pgbase,
3500                 .pglen    = pglen,
3501                 .pages    = &page,
3502         };
3503         struct nfs4_readlink_res res;
3504         struct rpc_message msg = {
3505                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3506                 .rpc_argp = &args,
3507                 .rpc_resp = &res,
3508         };
3509
3510         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3511 }
3512
3513 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3514                 unsigned int pgbase, unsigned int pglen)
3515 {
3516         struct nfs4_exception exception = { };
3517         int err;
3518         do {
3519                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3520                 trace_nfs4_readlink(inode, err);
3521                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3522                                 &exception);
3523         } while (exception.retry);
3524         return err;
3525 }
3526
3527 /*
3528  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3529  */
3530 static int
3531 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3532                  int flags)
3533 {
3534         struct nfs4_label l, *ilabel = NULL;
3535         struct nfs_open_context *ctx;
3536         struct nfs4_state *state;
3537         int opened = 0;
3538         int status = 0;
3539
3540         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3541         if (IS_ERR(ctx))
3542                 return PTR_ERR(ctx);
3543
3544         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3545
3546         sattr->ia_mode &= ~current_umask();
3547         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3548         if (IS_ERR(state)) {
3549                 status = PTR_ERR(state);
3550                 goto out;
3551         }
3552 out:
3553         nfs4_label_release_security(ilabel);
3554         put_nfs_open_context(ctx);
3555         return status;
3556 }
3557
3558 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3559 {
3560         struct nfs_server *server = NFS_SERVER(dir);
3561         struct nfs_removeargs args = {
3562                 .fh = NFS_FH(dir),
3563                 .name = *name,
3564         };
3565         struct nfs_removeres res = {
3566                 .server = server,
3567         };
3568         struct rpc_message msg = {
3569                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3570                 .rpc_argp = &args,
3571                 .rpc_resp = &res,
3572         };
3573         int status;
3574
3575         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3576         if (status == 0)
3577                 update_changeattr(dir, &res.cinfo);
3578         return status;
3579 }
3580
3581 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3582 {
3583         struct nfs4_exception exception = { };
3584         int err;
3585         do {
3586                 err = _nfs4_proc_remove(dir, name);
3587                 trace_nfs4_remove(dir, name, err);
3588                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3589                                 &exception);
3590         } while (exception.retry);
3591         return err;
3592 }
3593
3594 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3595 {
3596         struct nfs_server *server = NFS_SERVER(dir);
3597         struct nfs_removeargs *args = msg->rpc_argp;
3598         struct nfs_removeres *res = msg->rpc_resp;
3599
3600         res->server = server;
3601         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3602         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3603
3604         nfs_fattr_init(res->dir_attr);
3605 }
3606
3607 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3608 {
3609         nfs4_setup_sequence(NFS_SERVER(data->dir),
3610                         &data->args.seq_args,
3611                         &data->res.seq_res,
3612                         task);
3613 }
3614
3615 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3616 {
3617         struct nfs_unlinkdata *data = task->tk_calldata;
3618         struct nfs_removeres *res = &data->res;
3619
3620         if (!nfs4_sequence_done(task, &res->seq_res))
3621                 return 0;
3622         if (nfs4_async_handle_error(task, res->server, NULL,
3623                                     &data->timeout) == -EAGAIN)
3624                 return 0;
3625         update_changeattr(dir, &res->cinfo);
3626         return 1;
3627 }
3628
3629 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3630 {
3631         struct nfs_server *server = NFS_SERVER(dir);
3632         struct nfs_renameargs *arg = msg->rpc_argp;
3633         struct nfs_renameres *res = msg->rpc_resp;
3634
3635         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3636         res->server = server;
3637         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3638 }
3639
3640 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3641 {
3642         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3643                         &data->args.seq_args,
3644                         &data->res.seq_res,
3645                         task);
3646 }
3647
3648 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3649                                  struct inode *new_dir)
3650 {
3651         struct nfs_renamedata *data = task->tk_calldata;
3652         struct nfs_renameres *res = &data->res;
3653
3654         if (!nfs4_sequence_done(task, &res->seq_res))
3655                 return 0;
3656         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3657                 return 0;
3658
3659         update_changeattr(old_dir, &res->old_cinfo);
3660         update_changeattr(new_dir, &res->new_cinfo);
3661         return 1;
3662 }
3663
3664 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3665 {
3666         struct nfs_server *server = NFS_SERVER(inode);
3667         struct nfs4_link_arg arg = {
3668                 .fh     = NFS_FH(inode),
3669                 .dir_fh = NFS_FH(dir),
3670                 .name   = name,
3671                 .bitmask = server->attr_bitmask,
3672         };
3673         struct nfs4_link_res res = {
3674                 .server = server,
3675                 .label = NULL,
3676         };
3677         struct rpc_message msg = {
3678                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3679                 .rpc_argp = &arg,
3680                 .rpc_resp = &res,
3681         };
3682         int status = -ENOMEM;
3683
3684         res.fattr = nfs_alloc_fattr();
3685         if (res.fattr == NULL)
3686                 goto out;
3687
3688         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3689         if (IS_ERR(res.label)) {
3690                 status = PTR_ERR(res.label);
3691                 goto out;
3692         }
3693         arg.bitmask = nfs4_bitmask(server, res.label);
3694
3695         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3696         if (!status) {
3697                 update_changeattr(dir, &res.cinfo);
3698                 status = nfs_post_op_update_inode(inode, res.fattr);
3699                 if (!status)
3700                         nfs_setsecurity(inode, res.fattr, res.label);
3701         }
3702
3703
3704         nfs4_label_free(res.label);
3705
3706 out:
3707         nfs_free_fattr(res.fattr);
3708         return status;
3709 }
3710
3711 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3712 {
3713         struct nfs4_exception exception = { };
3714         int err;
3715         do {
3716                 err = nfs4_handle_exception(NFS_SERVER(inode),
3717                                 _nfs4_proc_link(inode, dir, name),
3718                                 &exception);
3719         } while (exception.retry);
3720         return err;
3721 }
3722
3723 struct nfs4_createdata {
3724         struct rpc_message msg;
3725         struct nfs4_create_arg arg;
3726         struct nfs4_create_res res;
3727         struct nfs_fh fh;
3728         struct nfs_fattr fattr;
3729         struct nfs4_label *label;
3730 };
3731
3732 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3733                 struct qstr *name, struct iattr *sattr, u32 ftype)
3734 {
3735         struct nfs4_createdata *data;
3736
3737         data = kzalloc(sizeof(*data), GFP_KERNEL);
3738         if (data != NULL) {
3739                 struct nfs_server *server = NFS_SERVER(dir);
3740
3741                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3742                 if (IS_ERR(data->label))
3743                         goto out_free;
3744
3745                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3746                 data->msg.rpc_argp = &data->arg;
3747                 data->msg.rpc_resp = &data->res;
3748                 data->arg.dir_fh = NFS_FH(dir);
3749                 data->arg.server = server;
3750                 data->arg.name = name;
3751                 data->arg.attrs = sattr;
3752                 data->arg.ftype = ftype;
3753                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3754                 data->res.server = server;
3755                 data->res.fh = &data->fh;
3756                 data->res.fattr = &data->fattr;
3757                 data->res.label = data->label;
3758                 nfs_fattr_init(data->res.fattr);
3759         }
3760         return data;
3761 out_free:
3762         kfree(data);
3763         return NULL;
3764 }
3765
3766 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3767 {
3768         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3769                                     &data->arg.seq_args, &data->res.seq_res, 1);
3770         if (status == 0) {
3771                 update_changeattr(dir, &data->res.dir_cinfo);
3772                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3773         }
3774         return status;
3775 }
3776
3777 static void nfs4_free_createdata(struct nfs4_createdata *data)
3778 {
3779         nfs4_label_free(data->label);
3780         kfree(data);
3781 }
3782
3783 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3784                 struct page *page, unsigned int len, struct iattr *sattr,
3785                 struct nfs4_label *label)
3786 {
3787         struct nfs4_createdata *data;
3788         int status = -ENAMETOOLONG;
3789
3790         if (len > NFS4_MAXPATHLEN)
3791                 goto out;
3792
3793         status = -ENOMEM;
3794         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3795         if (data == NULL)
3796                 goto out;
3797
3798         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3799         data->arg.u.symlink.pages = &page;
3800         data->arg.u.symlink.len = len;
3801         data->arg.label = label;
3802
3803         status = nfs4_do_create(dir, dentry, data);
3804
3805         nfs4_free_createdata(data);
3806 out:
3807         return status;
3808 }
3809
3810 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3811                 struct page *page, unsigned int len, struct iattr *sattr)
3812 {
3813         struct nfs4_exception exception = { };
3814         struct nfs4_label l, *label = NULL;
3815         int err;
3816
3817         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3818
3819         do {
3820                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3821                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3822                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3823                                 &exception);
3824         } while (exception.retry);
3825
3826         nfs4_label_release_security(label);
3827         return err;
3828 }
3829
3830 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3831                 struct iattr *sattr, struct nfs4_label *label)
3832 {
3833         struct nfs4_createdata *data;
3834         int status = -ENOMEM;
3835
3836         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3837         if (data == NULL)
3838                 goto out;
3839
3840         data->arg.label = label;
3841         status = nfs4_do_create(dir, dentry, data);
3842
3843         nfs4_free_createdata(data);
3844 out:
3845         return status;
3846 }
3847
3848 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3849                 struct iattr *sattr)
3850 {
3851         struct nfs4_exception exception = { };
3852         struct nfs4_label l, *label = NULL;
3853         int err;
3854
3855         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3856
3857         sattr->ia_mode &= ~current_umask();
3858         do {
3859                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3860                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3861                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3862                                 &exception);
3863         } while (exception.retry);
3864         nfs4_label_release_security(label);
3865
3866         return err;
3867 }
3868
3869 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3870                 u64 cookie, struct page **pages, unsigned int count, int plus)
3871 {
3872         struct inode            *dir = dentry->d_inode;
3873         struct nfs4_readdir_arg args = {
3874                 .fh = NFS_FH(dir),
3875                 .pages = pages,
3876                 .pgbase = 0,
3877                 .count = count,
3878                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3879                 .plus = plus,
3880         };
3881         struct nfs4_readdir_res res;
3882         struct rpc_message msg = {
3883                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3884                 .rpc_argp = &args,
3885                 .rpc_resp = &res,
3886                 .rpc_cred = cred,
3887         };
3888         int                     status;
3889
3890         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3891                         dentry,
3892                         (unsigned long long)cookie);
3893         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3894         res.pgbase = args.pgbase;
3895         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3896         if (status >= 0) {
3897                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3898                 status += args.pgbase;
3899         }
3900
3901         nfs_invalidate_atime(dir);
3902
3903         dprintk("%s: returns %d\n", __func__, status);
3904         return status;
3905 }
3906
3907 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3908                 u64 cookie, struct page **pages, unsigned int count, int plus)
3909 {
3910         struct nfs4_exception exception = { };
3911         int err;
3912         do {
3913                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3914                                 pages, count, plus);
3915                 trace_nfs4_readdir(dentry->d_inode, err);
3916                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3917                                 &exception);
3918         } while (exception.retry);
3919         return err;
3920 }
3921
3922 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3923                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3924 {
3925         struct nfs4_createdata *data;
3926         int mode = sattr->ia_mode;
3927         int status = -ENOMEM;
3928
3929         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3930         if (data == NULL)
3931                 goto out;
3932
3933         if (S_ISFIFO(mode))
3934                 data->arg.ftype = NF4FIFO;
3935         else if (S_ISBLK(mode)) {
3936                 data->arg.ftype = NF4BLK;
3937                 data->arg.u.device.specdata1 = MAJOR(rdev);
3938                 data->arg.u.device.specdata2 = MINOR(rdev);
3939         }
3940         else if (S_ISCHR(mode)) {
3941                 data->arg.ftype = NF4CHR;
3942                 data->arg.u.device.specdata1 = MAJOR(rdev);
3943                 data->arg.u.device.specdata2 = MINOR(rdev);
3944         } else if (!S_ISSOCK(mode)) {
3945                 status = -EINVAL;
3946                 goto out_free;
3947         }
3948
3949         data->arg.label = label;
3950         status = nfs4_do_create(dir, dentry, data);
3951 out_free:
3952         nfs4_free_createdata(data);
3953 out:
3954         return status;
3955 }
3956
3957 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3958                 struct iattr *sattr, dev_t rdev)
3959 {
3960         struct nfs4_exception exception = { };
3961         struct nfs4_label l, *label = NULL;
3962         int err;
3963
3964         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3965
3966         sattr->ia_mode &= ~current_umask();
3967         do {
3968                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3969                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3970                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3971                                 &exception);
3972         } while (exception.retry);
3973
3974         nfs4_label_release_security(label);
3975
3976         return err;
3977 }
3978
3979 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3980                  struct nfs_fsstat *fsstat)
3981 {
3982         struct nfs4_statfs_arg args = {
3983                 .fh = fhandle,
3984                 .bitmask = server->attr_bitmask,
3985         };
3986         struct nfs4_statfs_res res = {
3987                 .fsstat = fsstat,
3988         };
3989         struct rpc_message msg = {
3990                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3991                 .rpc_argp = &args,
3992                 .rpc_resp = &res,
3993         };
3994
3995         nfs_fattr_init(fsstat->fattr);
3996         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3997 }
3998
3999 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4000 {
4001         struct nfs4_exception exception = { };
4002         int err;
4003         do {
4004                 err = nfs4_handle_exception(server,
4005                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4006                                 &exception);
4007         } while (exception.retry);
4008         return err;
4009 }
4010
4011 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4012                 struct nfs_fsinfo *fsinfo)
4013 {
4014         struct nfs4_fsinfo_arg args = {
4015                 .fh = fhandle,
4016                 .bitmask = server->attr_bitmask,
4017         };
4018         struct nfs4_fsinfo_res res = {
4019                 .fsinfo = fsinfo,
4020         };
4021         struct rpc_message msg = {
4022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4023                 .rpc_argp = &args,
4024                 .rpc_resp = &res,
4025         };
4026
4027         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4028 }
4029
4030 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4031 {
4032         struct nfs4_exception exception = { };
4033         unsigned long now = jiffies;
4034         int err;
4035
4036         do {
4037                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4038                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4039                 if (err == 0) {
4040                         struct nfs_client *clp = server->nfs_client;
4041
4042                         spin_lock(&clp->cl_lock);
4043                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4044                         clp->cl_last_renewal = now;
4045                         spin_unlock(&clp->cl_lock);
4046                         break;
4047                 }
4048                 err = nfs4_handle_exception(server, err, &exception);
4049         } while (exception.retry);
4050         return err;
4051 }
4052
4053 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4054 {
4055         int error;
4056
4057         nfs_fattr_init(fsinfo->fattr);
4058         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4059         if (error == 0) {
4060                 /* block layout checks this! */
4061                 server->pnfs_blksize = fsinfo->blksize;
4062                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4063         }
4064
4065         return error;
4066 }
4067
4068 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4069                 struct nfs_pathconf *pathconf)
4070 {
4071         struct nfs4_pathconf_arg args = {
4072                 .fh = fhandle,
4073                 .bitmask = server->attr_bitmask,
4074         };
4075         struct nfs4_pathconf_res res = {
4076                 .pathconf = pathconf,
4077         };
4078         struct rpc_message msg = {
4079                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4080                 .rpc_argp = &args,
4081                 .rpc_resp = &res,
4082         };
4083
4084         /* None of the pathconf attributes are mandatory to implement */
4085         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4086                 memset(pathconf, 0, sizeof(*pathconf));
4087                 return 0;
4088         }
4089
4090         nfs_fattr_init(pathconf->fattr);
4091         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4092 }
4093
4094 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4095                 struct nfs_pathconf *pathconf)
4096 {
4097         struct nfs4_exception exception = { };
4098         int err;
4099
4100         do {
4101                 err = nfs4_handle_exception(server,
4102                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4103                                 &exception);
4104         } while (exception.retry);
4105         return err;
4106 }
4107
4108 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4109                 const struct nfs_open_context *ctx,
4110                 const struct nfs_lock_context *l_ctx,
4111                 fmode_t fmode)
4112 {
4113         const struct nfs_lockowner *lockowner = NULL;
4114
4115         if (l_ctx != NULL)
4116                 lockowner = &l_ctx->lockowner;
4117         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4118 }
4119 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4120
4121 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4122                 const struct nfs_open_context *ctx,
4123                 const struct nfs_lock_context *l_ctx,
4124                 fmode_t fmode)
4125 {
4126         nfs4_stateid current_stateid;
4127
4128         /* If the current stateid represents a lost lock, then exit */
4129         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4130                 return true;
4131         return nfs4_stateid_match(stateid, &current_stateid);
4132 }
4133
4134 static bool nfs4_error_stateid_expired(int err)
4135 {
4136         switch (err) {
4137         case -NFS4ERR_DELEG_REVOKED:
4138         case -NFS4ERR_ADMIN_REVOKED:
4139         case -NFS4ERR_BAD_STATEID:
4140         case -NFS4ERR_STALE_STATEID:
4141         case -NFS4ERR_OLD_STATEID:
4142         case -NFS4ERR_OPENMODE:
4143         case -NFS4ERR_EXPIRED:
4144                 return true;
4145         }
4146         return false;
4147 }
4148
4149 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4150 {
4151         nfs_invalidate_atime(hdr->inode);
4152 }
4153
4154 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4155 {
4156         struct nfs_server *server = NFS_SERVER(hdr->inode);
4157
4158         trace_nfs4_read(hdr, task->tk_status);
4159         if (nfs4_async_handle_error(task, server,
4160                                     hdr->args.context->state,
4161                                     NULL) == -EAGAIN) {
4162                 rpc_restart_call_prepare(task);
4163                 return -EAGAIN;
4164         }
4165
4166         __nfs4_read_done_cb(hdr);
4167         if (task->tk_status > 0)
4168                 renew_lease(server, hdr->timestamp);
4169         return 0;
4170 }
4171
4172 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4173                 struct nfs_pgio_args *args)
4174 {
4175
4176         if (!nfs4_error_stateid_expired(task->tk_status) ||
4177                 nfs4_stateid_is_current(&args->stateid,
4178                                 args->context,
4179                                 args->lock_context,
4180                                 FMODE_READ))
4181                 return false;
4182         rpc_restart_call_prepare(task);
4183         return true;
4184 }
4185
4186 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4187 {
4188
4189         dprintk("--> %s\n", __func__);
4190
4191         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4192                 return -EAGAIN;
4193         if (nfs4_read_stateid_changed(task, &hdr->args))
4194                 return -EAGAIN;
4195         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4196                                     nfs4_read_done_cb(task, hdr);
4197 }
4198
4199 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4200                                  struct rpc_message *msg)
4201 {
4202         hdr->timestamp   = jiffies;
4203         hdr->pgio_done_cb = nfs4_read_done_cb;
4204         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4205         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4206 }
4207
4208 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4209                                       struct nfs_pgio_header *hdr)
4210 {
4211         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4212                         &hdr->args.seq_args,
4213                         &hdr->res.seq_res,
4214                         task))
4215                 return 0;
4216         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4217                                 hdr->args.lock_context,
4218                                 hdr->rw_ops->rw_mode) == -EIO)
4219                 return -EIO;
4220         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4221                 return -EIO;
4222         return 0;
4223 }
4224
4225 static int nfs4_write_done_cb(struct rpc_task *task,
4226                               struct nfs_pgio_header *hdr)
4227 {
4228         struct inode *inode = hdr->inode;
4229
4230         trace_nfs4_write(hdr, task->tk_status);
4231         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4232                                     hdr->args.context->state,
4233                                     NULL) == -EAGAIN) {
4234                 rpc_restart_call_prepare(task);
4235                 return -EAGAIN;
4236         }
4237         if (task->tk_status >= 0) {
4238                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4239                 nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4240         }
4241         return 0;
4242 }
4243
4244 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4245                 struct nfs_pgio_args *args)
4246 {
4247
4248         if (!nfs4_error_stateid_expired(task->tk_status) ||
4249                 nfs4_stateid_is_current(&args->stateid,
4250                                 args->context,
4251                                 args->lock_context,
4252                                 FMODE_WRITE))
4253                 return false;
4254         rpc_restart_call_prepare(task);
4255         return true;
4256 }
4257
4258 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4259 {
4260         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4261                 return -EAGAIN;
4262         if (nfs4_write_stateid_changed(task, &hdr->args))
4263                 return -EAGAIN;
4264         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4265                 nfs4_write_done_cb(task, hdr);
4266 }
4267
4268 static
4269 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4270 {
4271         /* Don't request attributes for pNFS or O_DIRECT writes */
4272         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4273                 return false;
4274         /* Otherwise, request attributes if and only if we don't hold
4275          * a delegation
4276          */
4277         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4278 }
4279
4280 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4281                                   struct rpc_message *msg)
4282 {
4283         struct nfs_server *server = NFS_SERVER(hdr->inode);
4284
4285         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4286                 hdr->args.bitmask = NULL;
4287                 hdr->res.fattr = NULL;
4288         } else
4289                 hdr->args.bitmask = server->cache_consistency_bitmask;
4290
4291         if (!hdr->pgio_done_cb)
4292                 hdr->pgio_done_cb = nfs4_write_done_cb;
4293         hdr->res.server = server;
4294         hdr->timestamp   = jiffies;
4295
4296         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4297         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4298 }
4299
4300 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4301 {
4302         nfs4_setup_sequence(NFS_SERVER(data->inode),
4303                         &data->args.seq_args,
4304                         &data->res.seq_res,
4305                         task);
4306 }
4307
4308 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4309 {
4310         struct inode *inode = data->inode;
4311
4312         trace_nfs4_commit(data, task->tk_status);
4313         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4314                                     NULL, NULL) == -EAGAIN) {
4315                 rpc_restart_call_prepare(task);
4316                 return -EAGAIN;
4317         }
4318         return 0;
4319 }
4320
4321 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4322 {
4323         if (!nfs4_sequence_done(task, &data->res.seq_res))
4324                 return -EAGAIN;
4325         return data->commit_done_cb(task, data);
4326 }
4327
4328 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4329 {
4330         struct nfs_server *server = NFS_SERVER(data->inode);
4331
4332         if (data->commit_done_cb == NULL)
4333                 data->commit_done_cb = nfs4_commit_done_cb;
4334         data->res.server = server;
4335         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4336         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4337 }
4338
4339 struct nfs4_renewdata {
4340         struct nfs_client       *client;
4341         unsigned long           timestamp;
4342 };
4343
4344 /*
4345  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4346  * standalone procedure for queueing an asynchronous RENEW.
4347  */
4348 static void nfs4_renew_release(void *calldata)
4349 {
4350         struct nfs4_renewdata *data = calldata;
4351         struct nfs_client *clp = data->client;
4352
4353         if (atomic_read(&clp->cl_count) > 1)
4354                 nfs4_schedule_state_renewal(clp);
4355         nfs_put_client(clp);
4356         kfree(data);
4357 }
4358
4359 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4360 {
4361         struct nfs4_renewdata *data = calldata;
4362         struct nfs_client *clp = data->client;
4363         unsigned long timestamp = data->timestamp;
4364
4365         trace_nfs4_renew_async(clp, task->tk_status);
4366         switch (task->tk_status) {
4367         case 0:
4368                 break;
4369         case -NFS4ERR_LEASE_MOVED:
4370                 nfs4_schedule_lease_moved_recovery(clp);
4371                 break;
4372         default:
4373                 /* Unless we're shutting down, schedule state recovery! */
4374                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4375                         return;
4376                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4377                         nfs4_schedule_lease_recovery(clp);
4378                         return;
4379                 }
4380                 nfs4_schedule_path_down_recovery(clp);
4381         }
4382         do_renew_lease(clp, timestamp);
4383 }
4384
4385 static const struct rpc_call_ops nfs4_renew_ops = {
4386         .rpc_call_done = nfs4_renew_done,
4387         .rpc_release = nfs4_renew_release,
4388 };
4389
4390 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4391 {
4392         struct rpc_message msg = {
4393                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4394                 .rpc_argp       = clp,
4395                 .rpc_cred       = cred,
4396         };
4397         struct nfs4_renewdata *data;
4398
4399         if (renew_flags == 0)
4400                 return 0;
4401         if (!atomic_inc_not_zero(&clp->cl_count))
4402                 return -EIO;
4403         data = kmalloc(sizeof(*data), GFP_NOFS);
4404         if (data == NULL)
4405                 return -ENOMEM;
4406         data->client = clp;
4407         data->timestamp = jiffies;
4408         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4409                         &nfs4_renew_ops, data);
4410 }
4411
4412 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4413 {
4414         struct rpc_message msg = {
4415                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4416                 .rpc_argp       = clp,
4417                 .rpc_cred       = cred,
4418         };
4419         unsigned long now = jiffies;
4420         int status;
4421
4422         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4423         if (status < 0)
4424                 return status;
4425         do_renew_lease(clp, now);
4426         return 0;
4427 }
4428
4429 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4430 {
4431         return server->caps & NFS_CAP_ACLS;
4432 }
4433
4434 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4435  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4436  * the stack.
4437  */
4438 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4439
4440 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4441                 struct page **pages, unsigned int *pgbase)
4442 {
4443         struct page *newpage, **spages;
4444         int rc = 0;
4445         size_t len;
4446         spages = pages;
4447
4448         do {
4449                 len = min_t(size_t, PAGE_SIZE, buflen);
4450                 newpage = alloc_page(GFP_KERNEL);
4451
4452                 if (newpage == NULL)
4453                         goto unwind;
4454                 memcpy(page_address(newpage), buf, len);
4455                 buf += len;
4456                 buflen -= len;
4457                 *pages++ = newpage;
4458                 rc++;
4459         } while (buflen != 0);
4460
4461         return rc;
4462
4463 unwind:
4464         for(; rc > 0; rc--)
4465                 __free_page(spages[rc-1]);
4466         return -ENOMEM;
4467 }
4468
4469 struct nfs4_cached_acl {
4470         int cached;
4471         size_t len;
4472         char data[0];
4473 };
4474
4475 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4476 {
4477         struct nfs_inode *nfsi = NFS_I(inode);
4478
4479         spin_lock(&inode->i_lock);
4480         kfree(nfsi->nfs4_acl);
4481         nfsi->nfs4_acl = acl;
4482         spin_unlock(&inode->i_lock);
4483 }
4484
4485 static void nfs4_zap_acl_attr(struct inode *inode)
4486 {
4487         nfs4_set_cached_acl(inode, NULL);
4488 }
4489
4490 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4491 {
4492         struct nfs_inode *nfsi = NFS_I(inode);
4493         struct nfs4_cached_acl *acl;
4494         int ret = -ENOENT;
4495
4496         spin_lock(&inode->i_lock);
4497         acl = nfsi->nfs4_acl;
4498         if (acl == NULL)
4499                 goto out;
4500         if (buf == NULL) /* user is just asking for length */
4501                 goto out_len;
4502         if (acl->cached == 0)
4503                 goto out;
4504         ret = -ERANGE; /* see getxattr(2) man page */
4505         if (acl->len > buflen)
4506                 goto out;
4507         memcpy(buf, acl->data, acl->len);
4508 out_len:
4509         ret = acl->len;
4510 out:
4511         spin_unlock(&inode->i_lock);
4512         return ret;
4513 }
4514
4515 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4516 {
4517         struct nfs4_cached_acl *acl;
4518         size_t buflen = sizeof(*acl) + acl_len;
4519
4520         if (buflen <= PAGE_SIZE) {
4521                 acl = kmalloc(buflen, GFP_KERNEL);
4522                 if (acl == NULL)
4523                         goto out;
4524                 acl->cached = 1;
4525                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4526         } else {
4527                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4528                 if (acl == NULL)
4529                         goto out;
4530                 acl->cached = 0;
4531         }
4532         acl->len = acl_len;
4533 out:
4534         nfs4_set_cached_acl(inode, acl);
4535 }
4536
4537 /*
4538  * The getxattr API returns the required buffer length when called with a
4539  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4540  * the required buf.  On a NULL buf, we send a page of data to the server
4541  * guessing that the ACL request can be serviced by a page. If so, we cache
4542  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4543  * the cache. If not so, we throw away the page, and cache the required
4544  * length. The next getxattr call will then produce another round trip to
4545  * the server, this time with the input buf of the required size.
4546  */
4547 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4548 {
4549         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4550         struct nfs_getaclargs args = {
4551                 .fh = NFS_FH(inode),
4552                 .acl_pages = pages,
4553                 .acl_len = buflen,
4554         };
4555         struct nfs_getaclres res = {
4556                 .acl_len = buflen,
4557         };
4558         struct rpc_message msg = {
4559                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4560                 .rpc_argp = &args,
4561                 .rpc_resp = &res,
4562         };
4563         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4564         int ret = -ENOMEM, i;
4565
4566         /* As long as we're doing a round trip to the server anyway,
4567          * let's be prepared for a page of acl data. */
4568         if (npages == 0)
4569                 npages = 1;
4570         if (npages > ARRAY_SIZE(pages))
4571                 return -ERANGE;
4572
4573         for (i = 0; i < npages; i++) {
4574                 pages[i] = alloc_page(GFP_KERNEL);
4575                 if (!pages[i])
4576                         goto out_free;
4577         }
4578
4579         /* for decoding across pages */
4580         res.acl_scratch = alloc_page(GFP_KERNEL);
4581         if (!res.acl_scratch)
4582                 goto out_free;
4583
4584         args.acl_len = npages * PAGE_SIZE;
4585         args.acl_pgbase = 0;
4586
4587         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4588                 __func__, buf, buflen, npages, args.acl_len);
4589         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4590                              &msg, &args.seq_args, &res.seq_res, 0);
4591         if (ret)
4592                 goto out_free;
4593
4594         /* Handle the case where the passed-in buffer is too short */
4595         if (res.acl_flags & NFS4_ACL_TRUNC) {
4596                 /* Did the user only issue a request for the acl length? */
4597                 if (buf == NULL)
4598                         goto out_ok;
4599                 ret = -ERANGE;
4600                 goto out_free;
4601         }
4602         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4603         if (buf) {
4604                 if (res.acl_len > buflen) {
4605                         ret = -ERANGE;
4606                         goto out_free;
4607                 }
4608                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4609         }
4610 out_ok:
4611         ret = res.acl_len;
4612 out_free:
4613         for (i = 0; i < npages; i++)
4614                 if (pages[i])
4615                         __free_page(pages[i]);
4616         if (res.acl_scratch)
4617                 __free_page(res.acl_scratch);
4618         return ret;
4619 }
4620
4621 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4622 {
4623         struct nfs4_exception exception = { };
4624         ssize_t ret;
4625         do {
4626                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4627                 trace_nfs4_get_acl(inode, ret);
4628                 if (ret >= 0)
4629                         break;
4630                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4631         } while (exception.retry);
4632         return ret;
4633 }
4634
4635 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4636 {
4637         struct nfs_server *server = NFS_SERVER(inode);
4638         int ret;
4639
4640         if (!nfs4_server_supports_acls(server))
4641                 return -EOPNOTSUPP;
4642         ret = nfs_revalidate_inode(server, inode);
4643         if (ret < 0)
4644                 return ret;
4645         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4646                 nfs_zap_acl_cache(inode);
4647         ret = nfs4_read_cached_acl(inode, buf, buflen);
4648         if (ret != -ENOENT)
4649                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4650                  * but no cached acl data, just the acl length */
4651                 return ret;
4652         return nfs4_get_acl_uncached(inode, buf, buflen);
4653 }
4654
4655 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4656 {
4657         struct nfs_server *server = NFS_SERVER(inode);
4658         struct page *pages[NFS4ACL_MAXPAGES];
4659         struct nfs_setaclargs arg = {
4660                 .fh             = NFS_FH(inode),
4661                 .acl_pages      = pages,
4662                 .acl_len        = buflen,
4663         };
4664         struct nfs_setaclres res;
4665         struct rpc_message msg = {
4666                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4667                 .rpc_argp       = &arg,
4668                 .rpc_resp       = &res,
4669         };
4670         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4671         int ret, i;
4672
4673         if (!nfs4_server_supports_acls(server))
4674                 return -EOPNOTSUPP;
4675         if (npages > ARRAY_SIZE(pages))
4676                 return -ERANGE;
4677         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4678         if (i < 0)
4679                 return i;
4680         nfs4_inode_return_delegation(inode);
4681         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4682
4683         /*
4684          * Free each page after tx, so the only ref left is
4685          * held by the network stack
4686          */
4687         for (; i > 0; i--)
4688                 put_page(pages[i-1]);
4689
4690         /*
4691          * Acl update can result in inode attribute update.
4692          * so mark the attribute cache invalid.
4693          */
4694         spin_lock(&inode->i_lock);
4695         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4696         spin_unlock(&inode->i_lock);
4697         nfs_access_zap_cache(inode);
4698         nfs_zap_acl_cache(inode);
4699         return ret;
4700 }
4701
4702 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4703 {
4704         struct nfs4_exception exception = { };
4705         int err;
4706         do {
4707                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4708                 trace_nfs4_set_acl(inode, err);
4709                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4710                                 &exception);
4711         } while (exception.retry);
4712         return err;
4713 }
4714
4715 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4716 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4717                                         size_t buflen)
4718 {
4719         struct nfs_server *server = NFS_SERVER(inode);
4720         struct nfs_fattr fattr;
4721         struct nfs4_label label = {0, 0, buflen, buf};
4722
4723         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4724         struct nfs4_getattr_arg arg = {
4725                 .fh             = NFS_FH(inode),
4726                 .bitmask        = bitmask,
4727         };
4728         struct nfs4_getattr_res res = {
4729                 .fattr          = &fattr,
4730                 .label          = &label,
4731                 .server         = server,
4732         };
4733         struct rpc_message msg = {
4734                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4735                 .rpc_argp       = &arg,
4736                 .rpc_resp       = &res,
4737         };
4738         int ret;
4739
4740         nfs_fattr_init(&fattr);
4741
4742         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4743         if (ret)
4744                 return ret;
4745         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4746                 return -ENOENT;
4747         if (buflen < label.len)
4748                 return -ERANGE;
4749         return 0;
4750 }
4751
4752 static int nfs4_get_security_label(struct inode *inode, void *buf,
4753                                         size_t buflen)
4754 {
4755         struct nfs4_exception exception = { };
4756         int err;
4757
4758         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4759                 return -EOPNOTSUPP;
4760
4761         do {
4762                 err = _nfs4_get_security_label(inode, buf, buflen);
4763                 trace_nfs4_get_security_label(inode, err);
4764                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4765                                 &exception);
4766         } while (exception.retry);
4767         return err;
4768 }
4769
4770 static int _nfs4_do_set_security_label(struct inode *inode,
4771                 struct nfs4_label *ilabel,
4772                 struct nfs_fattr *fattr,
4773                 struct nfs4_label *olabel)
4774 {
4775
4776         struct iattr sattr = {0};
4777         struct nfs_server *server = NFS_SERVER(inode);
4778         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4779         struct nfs_setattrargs arg = {
4780                 .fh             = NFS_FH(inode),
4781                 .iap            = &sattr,
4782                 .server         = server,
4783                 .bitmask        = bitmask,
4784                 .label          = ilabel,
4785         };
4786         struct nfs_setattrres res = {
4787                 .fattr          = fattr,
4788                 .label          = olabel,
4789                 .server         = server,
4790         };
4791         struct rpc_message msg = {
4792                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4793                 .rpc_argp       = &arg,
4794                 .rpc_resp       = &res,
4795         };
4796         int status;
4797
4798         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4799
4800         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4801         if (status)
4802                 dprintk("%s failed: %d\n", __func__, status);
4803
4804         return status;
4805 }
4806
4807 static int nfs4_do_set_security_label(struct inode *inode,
4808                 struct nfs4_label *ilabel,
4809                 struct nfs_fattr *fattr,
4810                 struct nfs4_label *olabel)
4811 {
4812         struct nfs4_exception exception = { };
4813         int err;
4814
4815         do {
4816                 err = _nfs4_do_set_security_label(inode, ilabel,
4817                                 fattr, olabel);
4818                 trace_nfs4_set_security_label(inode, err);
4819                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4820                                 &exception);
4821         } while (exception.retry);
4822         return err;
4823 }
4824
4825 static int
4826 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4827 {
4828         struct nfs4_label ilabel, *olabel = NULL;
4829         struct nfs_fattr fattr;
4830         struct rpc_cred *cred;
4831         struct inode *inode = dentry->d_inode;
4832         int status;
4833
4834         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4835                 return -EOPNOTSUPP;
4836
4837         nfs_fattr_init(&fattr);
4838
4839         ilabel.pi = 0;
4840         ilabel.lfs = 0;
4841         ilabel.label = (char *)buf;
4842         ilabel.len = buflen;
4843
4844         cred = rpc_lookup_cred();
4845         if (IS_ERR(cred))
4846                 return PTR_ERR(cred);
4847
4848         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4849         if (IS_ERR(olabel)) {
4850                 status = -PTR_ERR(olabel);
4851                 goto out;
4852         }
4853
4854         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4855         if (status == 0)
4856                 nfs_setsecurity(inode, &fattr, olabel);
4857
4858         nfs4_label_free(olabel);
4859 out:
4860         put_rpccred(cred);
4861         return status;
4862 }
4863 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4864
4865
4866 static int
4867 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4868                         struct nfs4_state *state, long *timeout)
4869 {
4870         struct nfs_client *clp = server->nfs_client;
4871
4872         if (task->tk_status >= 0)
4873                 return 0;
4874         switch(task->tk_status) {
4875                 case -NFS4ERR_DELEG_REVOKED:
4876                 case -NFS4ERR_ADMIN_REVOKED:
4877                 case -NFS4ERR_BAD_STATEID:
4878                 case -NFS4ERR_OPENMODE:
4879                         if (state == NULL)
4880                                 break;
4881                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4882                                 goto recovery_failed;
4883                         goto wait_on_recovery;
4884                 case -NFS4ERR_EXPIRED:
4885                         if (state != NULL) {
4886                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4887                                         goto recovery_failed;
4888                         }
4889                 case -NFS4ERR_STALE_STATEID:
4890                 case -NFS4ERR_STALE_CLIENTID:
4891                         nfs4_schedule_lease_recovery(clp);
4892                         goto wait_on_recovery;
4893                 case -NFS4ERR_MOVED:
4894                         if (nfs4_schedule_migration_recovery(server) < 0)
4895                                 goto recovery_failed;
4896                         goto wait_on_recovery;
4897                 case -NFS4ERR_LEASE_MOVED:
4898                         nfs4_schedule_lease_moved_recovery(clp);
4899                         goto wait_on_recovery;
4900 #if defined(CONFIG_NFS_V4_1)
4901                 case -NFS4ERR_BADSESSION:
4902                 case -NFS4ERR_BADSLOT:
4903                 case -NFS4ERR_BAD_HIGH_SLOT:
4904                 case -NFS4ERR_DEADSESSION:
4905                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4906                 case -NFS4ERR_SEQ_FALSE_RETRY:
4907                 case -NFS4ERR_SEQ_MISORDERED:
4908                         dprintk("%s ERROR %d, Reset session\n", __func__,
4909                                 task->tk_status);
4910                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4911                         goto wait_on_recovery;
4912 #endif /* CONFIG_NFS_V4_1 */
4913                 case -NFS4ERR_DELAY:
4914                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4915                         rpc_delay(task, nfs4_update_delay(timeout));
4916                         goto restart_call;
4917                 case -NFS4ERR_GRACE:
4918                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4919                 case -NFS4ERR_RETRY_UNCACHED_REP:
4920                 case -NFS4ERR_OLD_STATEID:
4921                         goto restart_call;
4922         }
4923         task->tk_status = nfs4_map_errors(task->tk_status);
4924         return 0;
4925 recovery_failed:
4926         task->tk_status = -EIO;
4927         return 0;
4928 wait_on_recovery:
4929         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4930         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4931                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4932         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4933                 goto recovery_failed;
4934 restart_call:
4935         task->tk_status = 0;
4936         return -EAGAIN;
4937 }
4938
4939 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4940                                     nfs4_verifier *bootverf)
4941 {
4942         __be32 verf[2];
4943
4944         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4945                 /* An impossible timestamp guarantees this value
4946                  * will never match a generated boot time. */
4947                 verf[0] = 0;
4948                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4949         } else {
4950                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4951                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4952                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4953         }
4954         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4955 }
4956
4957 static unsigned int
4958 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4959                                    char *buf, size_t len)
4960 {
4961         unsigned int result;
4962
4963         if (clp->cl_owner_id != NULL)
4964                 return strlcpy(buf, clp->cl_owner_id, len);
4965
4966         rcu_read_lock();
4967         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4968                                 clp->cl_ipaddr,
4969                                 rpc_peeraddr2str(clp->cl_rpcclient,
4970                                                         RPC_DISPLAY_ADDR),
4971                                 rpc_peeraddr2str(clp->cl_rpcclient,
4972                                                         RPC_DISPLAY_PROTO));
4973         rcu_read_unlock();
4974         clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4975         return result;
4976 }
4977
4978 static unsigned int
4979 nfs4_init_uniform_client_string(struct nfs_client *clp,
4980                                 char *buf, size_t len)
4981 {
4982         const char *nodename = clp->cl_rpcclient->cl_nodename;
4983         unsigned int result;
4984
4985         if (clp->cl_owner_id != NULL)
4986                 return strlcpy(buf, clp->cl_owner_id, len);
4987
4988         if (nfs4_client_id_uniquifier[0] != '\0')
4989                 result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4990                                 clp->rpc_ops->version,
4991                                 clp->cl_minorversion,
4992                                 nfs4_client_id_uniquifier,
4993                                 nodename);
4994         else
4995                 result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4996                                 clp->rpc_ops->version, clp->cl_minorversion,
4997                                 nodename);
4998         clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4999         return result;
5000 }
5001
5002 /*
5003  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5004  * services.  Advertise one based on the address family of the
5005  * clientaddr.
5006  */
5007 static unsigned int
5008 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5009 {
5010         if (strchr(clp->cl_ipaddr, ':') != NULL)
5011                 return scnprintf(buf, len, "tcp6");
5012         else
5013                 return scnprintf(buf, len, "tcp");
5014 }
5015
5016 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5017 {
5018         struct nfs4_setclientid *sc = calldata;
5019
5020         if (task->tk_status == 0)
5021                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5022 }
5023
5024 static const struct rpc_call_ops nfs4_setclientid_ops = {
5025         .rpc_call_done = nfs4_setclientid_done,
5026 };
5027
5028 /**
5029  * nfs4_proc_setclientid - Negotiate client ID
5030  * @clp: state data structure
5031  * @program: RPC program for NFSv4 callback service
5032  * @port: IP port number for NFS4 callback service
5033  * @cred: RPC credential to use for this call
5034  * @res: where to place the result
5035  *
5036  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5037  */
5038 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5039                 unsigned short port, struct rpc_cred *cred,
5040                 struct nfs4_setclientid_res *res)
5041 {
5042         nfs4_verifier sc_verifier;
5043         struct nfs4_setclientid setclientid = {
5044                 .sc_verifier = &sc_verifier,
5045                 .sc_prog = program,
5046                 .sc_cb_ident = clp->cl_cb_ident,
5047         };
5048         struct rpc_message msg = {
5049                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5050                 .rpc_argp = &setclientid,
5051                 .rpc_resp = res,
5052                 .rpc_cred = cred,
5053         };
5054         struct rpc_task *task;
5055         struct rpc_task_setup task_setup_data = {
5056                 .rpc_client = clp->cl_rpcclient,
5057                 .rpc_message = &msg,
5058                 .callback_ops = &nfs4_setclientid_ops,
5059                 .callback_data = &setclientid,
5060                 .flags = RPC_TASK_TIMEOUT,
5061         };
5062         int status;
5063
5064         /* nfs_client_id4 */
5065         nfs4_init_boot_verifier(clp, &sc_verifier);
5066         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5067                 setclientid.sc_name_len =
5068                                 nfs4_init_uniform_client_string(clp,
5069                                                 setclientid.sc_name,
5070                                                 sizeof(setclientid.sc_name));
5071         else
5072                 setclientid.sc_name_len =
5073                                 nfs4_init_nonuniform_client_string(clp,
5074                                                 setclientid.sc_name,
5075                                                 sizeof(setclientid.sc_name));
5076         /* cb_client4 */
5077         setclientid.sc_netid_len =
5078                                 nfs4_init_callback_netid(clp,
5079                                                 setclientid.sc_netid,
5080                                                 sizeof(setclientid.sc_netid));
5081         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5082                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5083                                 clp->cl_ipaddr, port >> 8, port & 255);
5084
5085         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5086                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5087                 setclientid.sc_name_len, setclientid.sc_name);
5088         task = rpc_run_task(&task_setup_data);
5089         if (IS_ERR(task)) {
5090                 status = PTR_ERR(task);
5091                 goto out;
5092         }
5093         status = task->tk_status;
5094         if (setclientid.sc_cred) {
5095                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5096                 put_rpccred(setclientid.sc_cred);
5097         }
5098         rpc_put_task(task);
5099 out:
5100         trace_nfs4_setclientid(clp, status);
5101         dprintk("NFS reply setclientid: %d\n", status);
5102         return status;
5103 }
5104
5105 /**
5106  * nfs4_proc_setclientid_confirm - Confirm client ID
5107  * @clp: state data structure
5108  * @res: result of a previous SETCLIENTID
5109  * @cred: RPC credential to use for this call
5110  *
5111  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5112  */
5113 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5114                 struct nfs4_setclientid_res *arg,
5115                 struct rpc_cred *cred)
5116 {
5117         struct rpc_message msg = {
5118                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5119                 .rpc_argp = arg,
5120                 .rpc_cred = cred,
5121         };
5122         int status;
5123
5124         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5125                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5126                 clp->cl_clientid);
5127         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5128         trace_nfs4_setclientid_confirm(clp, status);
5129         dprintk("NFS reply setclientid_confirm: %d\n", status);
5130         return status;
5131 }
5132
5133 struct nfs4_delegreturndata {
5134         struct nfs4_delegreturnargs args;
5135         struct nfs4_delegreturnres res;
5136         struct nfs_fh fh;
5137         nfs4_stateid stateid;
5138         unsigned long timestamp;
5139         struct nfs_fattr fattr;
5140         int rpc_status;
5141         struct inode *inode;
5142         bool roc;
5143         u32 roc_barrier;
5144 };
5145
5146 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5147 {
5148         struct nfs4_delegreturndata *data = calldata;
5149
5150         if (!nfs4_sequence_done(task, &data->res.seq_res))
5151                 return;
5152
5153         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5154         switch (task->tk_status) {
5155         case 0:
5156                 renew_lease(data->res.server, data->timestamp);
5157         case -NFS4ERR_ADMIN_REVOKED:
5158         case -NFS4ERR_DELEG_REVOKED:
5159         case -NFS4ERR_BAD_STATEID:
5160         case -NFS4ERR_OLD_STATEID:
5161         case -NFS4ERR_STALE_STATEID:
5162         case -NFS4ERR_EXPIRED:
5163                 task->tk_status = 0;
5164                 if (data->roc)
5165                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5166                 break;
5167         default:
5168                 if (nfs4_async_handle_error(task, data->res.server,
5169                                             NULL, NULL) == -EAGAIN) {
5170                         rpc_restart_call_prepare(task);
5171                         return;
5172                 }
5173         }
5174         data->rpc_status = task->tk_status;
5175 }
5176
5177 static void nfs4_delegreturn_release(void *calldata)
5178 {
5179         struct nfs4_delegreturndata *data = calldata;
5180
5181         if (data->roc)
5182                 pnfs_roc_release(data->inode);
5183         kfree(calldata);
5184 }
5185
5186 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5187 {
5188         struct nfs4_delegreturndata *d_data;
5189
5190         d_data = (struct nfs4_delegreturndata *)data;
5191
5192         if (d_data->roc &&
5193             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5194                 return;
5195
5196         nfs4_setup_sequence(d_data->res.server,
5197                         &d_data->args.seq_args,
5198                         &d_data->res.seq_res,
5199                         task);
5200 }
5201
5202 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5203         .rpc_call_prepare = nfs4_delegreturn_prepare,
5204         .rpc_call_done = nfs4_delegreturn_done,
5205         .rpc_release = nfs4_delegreturn_release,
5206 };
5207
5208 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5209 {
5210         struct nfs4_delegreturndata *data;
5211         struct nfs_server *server = NFS_SERVER(inode);
5212         struct rpc_task *task;
5213         struct rpc_message msg = {
5214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5215                 .rpc_cred = cred,
5216         };
5217         struct rpc_task_setup task_setup_data = {
5218                 .rpc_client = server->client,
5219                 .rpc_message = &msg,
5220                 .callback_ops = &nfs4_delegreturn_ops,
5221                 .flags = RPC_TASK_ASYNC,
5222         };
5223         int status = 0;
5224
5225         data = kzalloc(sizeof(*data), GFP_NOFS);
5226         if (data == NULL)
5227                 return -ENOMEM;
5228         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5229         data->args.fhandle = &data->fh;
5230         data->args.stateid = &data->stateid;
5231         data->args.bitmask = server->cache_consistency_bitmask;
5232         nfs_copy_fh(&data->fh, NFS_FH(inode));
5233         nfs4_stateid_copy(&data->stateid, stateid);
5234         data->res.fattr = &data->fattr;
5235         data->res.server = server;
5236         nfs_fattr_init(data->res.fattr);
5237         data->timestamp = jiffies;
5238         data->rpc_status = 0;
5239         data->inode = inode;
5240         data->roc = list_empty(&NFS_I(inode)->open_files) ?
5241                     pnfs_roc(inode) : false;
5242
5243         task_setup_data.callback_data = data;
5244         msg.rpc_argp = &data->args;
5245         msg.rpc_resp = &data->res;
5246         task = rpc_run_task(&task_setup_data);
5247         if (IS_ERR(task))
5248                 return PTR_ERR(task);
5249         if (!issync)
5250                 goto out;
5251         status = nfs4_wait_for_completion_rpc_task(task);
5252         if (status != 0)
5253                 goto out;
5254         status = data->rpc_status;
5255         if (status == 0)
5256                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5257         else
5258                 nfs_refresh_inode(inode, &data->fattr);
5259 out:
5260         rpc_put_task(task);
5261         return status;
5262 }
5263
5264 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5265 {
5266         struct nfs_server *server = NFS_SERVER(inode);
5267         struct nfs4_exception exception = { };
5268         int err;
5269         do {
5270                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5271                 trace_nfs4_delegreturn(inode, err);
5272                 switch (err) {
5273                         case -NFS4ERR_STALE_STATEID:
5274                         case -NFS4ERR_EXPIRED:
5275                         case 0:
5276                                 return 0;
5277                 }
5278                 err = nfs4_handle_exception(server, err, &exception);
5279         } while (exception.retry);
5280         return err;
5281 }
5282
5283 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5284 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5285
5286 /*
5287  * sleep, with exponential backoff, and retry the LOCK operation.
5288  */
5289 static unsigned long
5290 nfs4_set_lock_task_retry(unsigned long timeout)
5291 {
5292         freezable_schedule_timeout_killable_unsafe(timeout);
5293         timeout <<= 1;
5294         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5295                 return NFS4_LOCK_MAXTIMEOUT;
5296         return timeout;
5297 }
5298
5299 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5300 {
5301         struct inode *inode = state->inode;
5302         struct nfs_server *server = NFS_SERVER(inode);
5303         struct nfs_client *clp = server->nfs_client;
5304         struct nfs_lockt_args arg = {
5305                 .fh = NFS_FH(inode),
5306                 .fl = request,
5307         };
5308         struct nfs_lockt_res res = {
5309                 .denied = request,
5310         };
5311         struct rpc_message msg = {
5312                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5313                 .rpc_argp       = &arg,
5314                 .rpc_resp       = &res,
5315                 .rpc_cred       = state->owner->so_cred,
5316         };
5317         struct nfs4_lock_state *lsp;
5318         int status;
5319
5320         arg.lock_owner.clientid = clp->cl_clientid;
5321         status = nfs4_set_lock_state(state, request);
5322         if (status != 0)
5323                 goto out;
5324         lsp = request->fl_u.nfs4_fl.owner;
5325         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5326         arg.lock_owner.s_dev = server->s_dev;
5327         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5328         switch (status) {
5329                 case 0:
5330                         request->fl_type = F_UNLCK;
5331                         break;
5332                 case -NFS4ERR_DENIED:
5333                         status = 0;
5334         }
5335         request->fl_ops->fl_release_private(request);
5336         request->fl_ops = NULL;
5337 out:
5338         return status;
5339 }
5340
5341 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5342 {
5343         struct nfs4_exception exception = { };
5344         int err;
5345
5346         do {
5347                 err = _nfs4_proc_getlk(state, cmd, request);
5348                 trace_nfs4_get_lock(request, state, cmd, err);
5349                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5350                                 &exception);
5351         } while (exception.retry);
5352         return err;
5353 }
5354
5355 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5356 {
5357         int res = 0;
5358         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5359                 case FL_POSIX:
5360                         res = posix_lock_file_wait(file, fl);
5361                         break;
5362                 case FL_FLOCK:
5363                         res = flock_lock_file_wait(file, fl);
5364                         break;
5365                 default:
5366                         BUG();
5367         }
5368         return res;
5369 }
5370
5371 struct nfs4_unlockdata {
5372         struct nfs_locku_args arg;
5373         struct nfs_locku_res res;
5374         struct nfs4_lock_state *lsp;
5375         struct nfs_open_context *ctx;
5376         struct file_lock fl;
5377         const struct nfs_server *server;
5378         unsigned long timestamp;
5379 };
5380
5381 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5382                 struct nfs_open_context *ctx,
5383                 struct nfs4_lock_state *lsp,
5384                 struct nfs_seqid *seqid)
5385 {
5386         struct nfs4_unlockdata *p;
5387         struct inode *inode = lsp->ls_state->inode;
5388
5389         p = kzalloc(sizeof(*p), GFP_NOFS);
5390         if (p == NULL)
5391                 return NULL;
5392         p->arg.fh = NFS_FH(inode);
5393         p->arg.fl = &p->fl;
5394         p->arg.seqid = seqid;
5395         p->res.seqid = seqid;
5396         p->arg.stateid = &lsp->ls_stateid;
5397         p->lsp = lsp;
5398         atomic_inc(&lsp->ls_count);
5399         /* Ensure we don't close file until we're done freeing locks! */
5400         p->ctx = get_nfs_open_context(ctx);
5401         memcpy(&p->fl, fl, sizeof(p->fl));
5402         p->server = NFS_SERVER(inode);
5403         return p;
5404 }
5405
5406 static void nfs4_locku_release_calldata(void *data)
5407 {
5408         struct nfs4_unlockdata *calldata = data;
5409         nfs_free_seqid(calldata->arg.seqid);
5410         nfs4_put_lock_state(calldata->lsp);
5411         put_nfs_open_context(calldata->ctx);
5412         kfree(calldata);
5413 }
5414
5415 static void nfs4_locku_done(struct rpc_task *task, void *data)
5416 {
5417         struct nfs4_unlockdata *calldata = data;
5418
5419         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5420                 return;
5421         switch (task->tk_status) {
5422                 case 0:
5423                         renew_lease(calldata->server, calldata->timestamp);
5424                         nfs4_update_lock_stateid(calldata->lsp,
5425                                         &calldata->res.stateid);
5426                         break;
5427                 case -NFS4ERR_BAD_STATEID:
5428                 case -NFS4ERR_OLD_STATEID:
5429                 case -NFS4ERR_STALE_STATEID:
5430                 case -NFS4ERR_EXPIRED:
5431                         break;
5432                 default:
5433                         if (nfs4_async_handle_error(task, calldata->server,
5434                                                     NULL, NULL) == -EAGAIN)
5435                                 rpc_restart_call_prepare(task);
5436         }
5437         nfs_release_seqid(calldata->arg.seqid);
5438 }
5439
5440 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5441 {
5442         struct nfs4_unlockdata *calldata = data;
5443
5444         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5445                 goto out_wait;
5446         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5447                 /* Note: exit _without_ running nfs4_locku_done */
5448                 goto out_no_action;
5449         }
5450         calldata->timestamp = jiffies;
5451         if (nfs4_setup_sequence(calldata->server,
5452                                 &calldata->arg.seq_args,
5453                                 &calldata->res.seq_res,
5454                                 task) != 0)
5455                 nfs_release_seqid(calldata->arg.seqid);
5456         return;
5457 out_no_action:
5458         task->tk_action = NULL;
5459 out_wait:
5460         nfs4_sequence_done(task, &calldata->res.seq_res);
5461 }
5462
5463 static const struct rpc_call_ops nfs4_locku_ops = {
5464         .rpc_call_prepare = nfs4_locku_prepare,
5465         .rpc_call_done = nfs4_locku_done,
5466         .rpc_release = nfs4_locku_release_calldata,
5467 };
5468
5469 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5470                 struct nfs_open_context *ctx,
5471                 struct nfs4_lock_state *lsp,
5472                 struct nfs_seqid *seqid)
5473 {
5474         struct nfs4_unlockdata *data;
5475         struct rpc_message msg = {
5476                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5477                 .rpc_cred = ctx->cred,
5478         };
5479         struct rpc_task_setup task_setup_data = {
5480                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5481                 .rpc_message = &msg,
5482                 .callback_ops = &nfs4_locku_ops,
5483                 .workqueue = nfsiod_workqueue,
5484                 .flags = RPC_TASK_ASYNC,
5485         };
5486
5487         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5488                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5489
5490         /* Ensure this is an unlock - when canceling a lock, the
5491          * canceled lock is passed in, and it won't be an unlock.
5492          */
5493         fl->fl_type = F_UNLCK;
5494
5495         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5496         if (data == NULL) {
5497                 nfs_free_seqid(seqid);
5498                 return ERR_PTR(-ENOMEM);
5499         }
5500
5501         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5502         msg.rpc_argp = &data->arg;
5503         msg.rpc_resp = &data->res;
5504         task_setup_data.callback_data = data;
5505         return rpc_run_task(&task_setup_data);
5506 }
5507
5508 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5509 {
5510         struct inode *inode = state->inode;
5511         struct nfs4_state_owner *sp = state->owner;
5512         struct nfs_inode *nfsi = NFS_I(inode);
5513         struct nfs_seqid *seqid;
5514         struct nfs4_lock_state *lsp;
5515         struct rpc_task *task;
5516         int status = 0;
5517         unsigned char fl_flags = request->fl_flags;
5518
5519         status = nfs4_set_lock_state(state, request);
5520         /* Unlock _before_ we do the RPC call */
5521         request->fl_flags |= FL_EXISTS;
5522         /* Exclude nfs_delegation_claim_locks() */
5523         mutex_lock(&sp->so_delegreturn_mutex);
5524         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5525         down_read(&nfsi->rwsem);
5526         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5527                 up_read(&nfsi->rwsem);
5528                 mutex_unlock(&sp->so_delegreturn_mutex);
5529                 goto out;
5530         }
5531         up_read(&nfsi->rwsem);
5532         mutex_unlock(&sp->so_delegreturn_mutex);
5533         if (status != 0)
5534                 goto out;
5535         /* Is this a delegated lock? */
5536         lsp = request->fl_u.nfs4_fl.owner;
5537         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5538                 goto out;
5539         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5540         status = -ENOMEM;
5541         if (IS_ERR(seqid))
5542                 goto out;
5543         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5544         status = PTR_ERR(task);
5545         if (IS_ERR(task))
5546                 goto out;
5547         status = nfs4_wait_for_completion_rpc_task(task);
5548         rpc_put_task(task);
5549 out:
5550         request->fl_flags = fl_flags;
5551         trace_nfs4_unlock(request, state, F_SETLK, status);
5552         return status;
5553 }
5554
5555 struct nfs4_lockdata {
5556         struct nfs_lock_args arg;
5557         struct nfs_lock_res res;
5558         struct nfs4_lock_state *lsp;
5559         struct nfs_open_context *ctx;
5560         struct file_lock fl;
5561         unsigned long timestamp;
5562         int rpc_status;
5563         int cancelled;
5564         struct nfs_server *server;
5565 };
5566
5567 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5568                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5569                 gfp_t gfp_mask)
5570 {
5571         struct nfs4_lockdata *p;
5572         struct inode *inode = lsp->ls_state->inode;
5573         struct nfs_server *server = NFS_SERVER(inode);
5574
5575         p = kzalloc(sizeof(*p), gfp_mask);
5576         if (p == NULL)
5577                 return NULL;
5578
5579         p->arg.fh = NFS_FH(inode);
5580         p->arg.fl = &p->fl;
5581         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5582         if (IS_ERR(p->arg.open_seqid))
5583                 goto out_free;
5584         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5585         if (IS_ERR(p->arg.lock_seqid))
5586                 goto out_free_seqid;
5587         p->arg.lock_stateid = &lsp->ls_stateid;
5588         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5589         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5590         p->arg.lock_owner.s_dev = server->s_dev;
5591         p->res.lock_seqid = p->arg.lock_seqid;
5592         p->lsp = lsp;
5593         p->server = server;
5594         atomic_inc(&lsp->ls_count);
5595         p->ctx = get_nfs_open_context(ctx);
5596         memcpy(&p->fl, fl, sizeof(p->fl));
5597         return p;
5598 out_free_seqid:
5599         nfs_free_seqid(p->arg.open_seqid);
5600 out_free:
5601         kfree(p);
5602         return NULL;
5603 }
5604
5605 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5606 {
5607         struct nfs4_lockdata *data = calldata;
5608         struct nfs4_state *state = data->lsp->ls_state;
5609
5610         dprintk("%s: begin!\n", __func__);
5611         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5612                 goto out_wait;
5613         /* Do we need to do an open_to_lock_owner? */
5614         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5615                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5616                         goto out_release_lock_seqid;
5617                 }
5618                 data->arg.open_stateid = &state->open_stateid;
5619                 data->arg.new_lock_owner = 1;
5620                 data->res.open_seqid = data->arg.open_seqid;
5621         } else
5622                 data->arg.new_lock_owner = 0;
5623         if (!nfs4_valid_open_stateid(state)) {
5624                 data->rpc_status = -EBADF;
5625                 task->tk_action = NULL;
5626                 goto out_release_open_seqid;
5627         }
5628         data->timestamp = jiffies;
5629         if (nfs4_setup_sequence(data->server,
5630                                 &data->arg.seq_args,
5631                                 &data->res.seq_res,
5632                                 task) == 0)
5633                 return;
5634 out_release_open_seqid:
5635         nfs_release_seqid(data->arg.open_seqid);
5636 out_release_lock_seqid:
5637         nfs_release_seqid(data->arg.lock_seqid);
5638 out_wait:
5639         nfs4_sequence_done(task, &data->res.seq_res);
5640         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5641 }
5642
5643 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5644 {
5645         struct nfs4_lockdata *data = calldata;
5646         struct nfs4_lock_state *lsp = data->lsp;
5647
5648         dprintk("%s: begin!\n", __func__);
5649
5650         if (!nfs4_sequence_done(task, &data->res.seq_res))
5651                 return;
5652
5653         data->rpc_status = task->tk_status;
5654         if (task->tk_status == 0) {
5655                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode),
5656                                 data->timestamp);
5657                 if (data->arg.new_lock_owner != 0) {
5658                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5659                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5660                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5661                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5662                         rpc_restart_call_prepare(task);
5663         }
5664         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5665 }
5666
5667 static void nfs4_lock_release(void *calldata)
5668 {
5669         struct nfs4_lockdata *data = calldata;
5670
5671         dprintk("%s: begin!\n", __func__);
5672         nfs_free_seqid(data->arg.open_seqid);
5673         if (data->cancelled != 0) {
5674                 struct rpc_task *task;
5675                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5676                                 data->arg.lock_seqid);
5677                 if (!IS_ERR(task))
5678                         rpc_put_task_async(task);
5679                 dprintk("%s: cancelling lock!\n", __func__);
5680         } else
5681                 nfs_free_seqid(data->arg.lock_seqid);
5682         nfs4_put_lock_state(data->lsp);
5683         put_nfs_open_context(data->ctx);
5684         kfree(data);
5685         dprintk("%s: done!\n", __func__);
5686 }
5687
5688 static const struct rpc_call_ops nfs4_lock_ops = {
5689         .rpc_call_prepare = nfs4_lock_prepare,
5690         .rpc_call_done = nfs4_lock_done,
5691         .rpc_release = nfs4_lock_release,
5692 };
5693
5694 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5695 {
5696         switch (error) {
5697         case -NFS4ERR_ADMIN_REVOKED:
5698         case -NFS4ERR_BAD_STATEID:
5699                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5700                 if (new_lock_owner != 0 ||
5701                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5702                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5703                 break;
5704         case -NFS4ERR_STALE_STATEID:
5705                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5706         case -NFS4ERR_EXPIRED:
5707                 nfs4_schedule_lease_recovery(server->nfs_client);
5708         };
5709 }
5710
5711 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5712 {
5713         struct nfs4_lockdata *data;
5714         struct rpc_task *task;
5715         struct rpc_message msg = {
5716                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5717                 .rpc_cred = state->owner->so_cred,
5718         };
5719         struct rpc_task_setup task_setup_data = {
5720                 .rpc_client = NFS_CLIENT(state->inode),
5721                 .rpc_message = &msg,
5722                 .callback_ops = &nfs4_lock_ops,
5723                 .workqueue = nfsiod_workqueue,
5724                 .flags = RPC_TASK_ASYNC,
5725         };
5726         int ret;
5727
5728         dprintk("%s: begin!\n", __func__);
5729         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5730                         fl->fl_u.nfs4_fl.owner,
5731                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5732         if (data == NULL)
5733                 return -ENOMEM;
5734         if (IS_SETLKW(cmd))
5735                 data->arg.block = 1;
5736         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5737         msg.rpc_argp = &data->arg;
5738         msg.rpc_resp = &data->res;
5739         task_setup_data.callback_data = data;
5740         if (recovery_type > NFS_LOCK_NEW) {
5741                 if (recovery_type == NFS_LOCK_RECLAIM)
5742                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5743                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5744         }
5745         task = rpc_run_task(&task_setup_data);
5746         if (IS_ERR(task))
5747                 return PTR_ERR(task);
5748         ret = nfs4_wait_for_completion_rpc_task(task);
5749         if (ret == 0) {
5750                 ret = data->rpc_status;
5751                 if (ret)
5752                         nfs4_handle_setlk_error(data->server, data->lsp,
5753                                         data->arg.new_lock_owner, ret);
5754         } else
5755                 data->cancelled = 1;
5756         rpc_put_task(task);
5757         dprintk("%s: done, ret = %d!\n", __func__, ret);
5758         return ret;
5759 }
5760
5761 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5762 {
5763         struct nfs_server *server = NFS_SERVER(state->inode);
5764         struct nfs4_exception exception = {
5765                 .inode = state->inode,
5766         };
5767         int err;
5768
5769         do {
5770                 /* Cache the lock if possible... */
5771                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5772                         return 0;
5773                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5774                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5775                 if (err != -NFS4ERR_DELAY)
5776                         break;
5777                 nfs4_handle_exception(server, err, &exception);
5778         } while (exception.retry);
5779         return err;
5780 }
5781
5782 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5783 {
5784         struct nfs_server *server = NFS_SERVER(state->inode);
5785         struct nfs4_exception exception = {
5786                 .inode = state->inode,
5787         };
5788         int err;
5789
5790         err = nfs4_set_lock_state(state, request);
5791         if (err != 0)
5792                 return err;
5793         if (!recover_lost_locks) {
5794                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5795                 return 0;
5796         }
5797         do {
5798                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5799                         return 0;
5800                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5801                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5802                 switch (err) {
5803                 default:
5804                         goto out;
5805                 case -NFS4ERR_GRACE:
5806                 case -NFS4ERR_DELAY:
5807                         nfs4_handle_exception(server, err, &exception);
5808                         err = 0;
5809                 }
5810         } while (exception.retry);
5811 out:
5812         return err;
5813 }
5814
5815 #if defined(CONFIG_NFS_V4_1)
5816 /**
5817  * nfs41_check_expired_locks - possibly free a lock stateid
5818  *
5819  * @state: NFSv4 state for an inode
5820  *
5821  * Returns NFS_OK if recovery for this stateid is now finished.
5822  * Otherwise a negative NFS4ERR value is returned.
5823  */
5824 static int nfs41_check_expired_locks(struct nfs4_state *state)
5825 {
5826         int status, ret = -NFS4ERR_BAD_STATEID;
5827         struct nfs4_lock_state *lsp;
5828         struct nfs_server *server = NFS_SERVER(state->inode);
5829
5830         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5831                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5832                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5833
5834                         status = nfs41_test_stateid(server,
5835                                         &lsp->ls_stateid,
5836                                         cred);
5837                         trace_nfs4_test_lock_stateid(state, lsp, status);
5838                         if (status != NFS_OK) {
5839                                 /* Free the stateid unless the server
5840                                  * informs us the stateid is unrecognized. */
5841                                 if (status != -NFS4ERR_BAD_STATEID)
5842                                         nfs41_free_stateid(server,
5843                                                         &lsp->ls_stateid,
5844                                                         cred);
5845                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5846                                 ret = status;
5847                         }
5848                 }
5849         };
5850
5851         return ret;
5852 }
5853
5854 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5855 {
5856         int status = NFS_OK;
5857
5858         if (test_bit(LK_STATE_IN_USE, &state->flags))
5859                 status = nfs41_check_expired_locks(state);
5860         if (status != NFS_OK)
5861                 status = nfs4_lock_expired(state, request);
5862         return status;
5863 }
5864 #endif
5865
5866 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5867 {
5868         struct nfs4_state_owner *sp = state->owner;
5869         struct nfs_inode *nfsi = NFS_I(state->inode);
5870         unsigned char fl_flags = request->fl_flags;
5871         unsigned int seq;
5872         int status = -ENOLCK;
5873
5874         if ((fl_flags & FL_POSIX) &&
5875                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5876                 goto out;
5877         /* Is this a delegated open? */
5878         status = nfs4_set_lock_state(state, request);
5879         if (status != 0)
5880                 goto out;
5881         request->fl_flags |= FL_ACCESS;
5882         status = do_vfs_lock(request->fl_file, request);
5883         if (status < 0)
5884                 goto out;
5885         down_read(&nfsi->rwsem);
5886         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5887                 /* Yes: cache locks! */
5888                 /* ...but avoid races with delegation recall... */
5889                 request->fl_flags = fl_flags & ~FL_SLEEP;
5890                 status = do_vfs_lock(request->fl_file, request);
5891                 goto out_unlock;
5892         }
5893         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5894         up_read(&nfsi->rwsem);
5895         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5896         if (status != 0)
5897                 goto out;
5898         down_read(&nfsi->rwsem);
5899         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5900                 status = -NFS4ERR_DELAY;
5901                 goto out_unlock;
5902         }
5903         /* Note: we always want to sleep here! */
5904         request->fl_flags = fl_flags | FL_SLEEP;
5905         if (do_vfs_lock(request->fl_file, request) < 0)
5906                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5907                         "manager!\n", __func__);
5908 out_unlock:
5909         up_read(&nfsi->rwsem);
5910 out:
5911         request->fl_flags = fl_flags;
5912         return status;
5913 }
5914
5915 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5916 {
5917         struct nfs4_exception exception = {
5918                 .state = state,
5919                 .inode = state->inode,
5920         };
5921         int err;
5922
5923         do {
5924                 err = _nfs4_proc_setlk(state, cmd, request);
5925                 trace_nfs4_set_lock(request, state, cmd, err);
5926                 if (err == -NFS4ERR_DENIED)
5927                         err = -EAGAIN;
5928                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5929                                 err, &exception);
5930         } while (exception.retry);
5931         return err;
5932 }
5933
5934 static int
5935 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5936 {
5937         struct nfs_open_context *ctx;
5938         struct nfs4_state *state;
5939         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5940         int status;
5941
5942         /* verify open state */
5943         ctx = nfs_file_open_context(filp);
5944         state = ctx->state;
5945
5946         if (request->fl_start < 0 || request->fl_end < 0)
5947                 return -EINVAL;
5948
5949         if (IS_GETLK(cmd)) {
5950                 if (state != NULL)
5951                         return nfs4_proc_getlk(state, F_GETLK, request);
5952                 return 0;
5953         }
5954
5955         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5956                 return -EINVAL;
5957
5958         if (request->fl_type == F_UNLCK) {
5959                 if (state != NULL)
5960                         return nfs4_proc_unlck(state, cmd, request);
5961                 return 0;
5962         }
5963
5964         if (state == NULL)
5965                 return -ENOLCK;
5966         /*
5967          * Don't rely on the VFS having checked the file open mode,
5968          * since it won't do this for flock() locks.
5969          */
5970         switch (request->fl_type) {
5971         case F_RDLCK:
5972                 if (!(filp->f_mode & FMODE_READ))
5973                         return -EBADF;
5974                 break;
5975         case F_WRLCK:
5976                 if (!(filp->f_mode & FMODE_WRITE))
5977                         return -EBADF;
5978         }
5979
5980         do {
5981                 status = nfs4_proc_setlk(state, cmd, request);
5982                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5983                         break;
5984                 timeout = nfs4_set_lock_task_retry(timeout);
5985                 status = -ERESTARTSYS;
5986                 if (signalled())
5987                         break;
5988         } while(status < 0);
5989         return status;
5990 }
5991
5992 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5993 {
5994         struct nfs_server *server = NFS_SERVER(state->inode);
5995         int err;
5996
5997         err = nfs4_set_lock_state(state, fl);
5998         if (err != 0)
5999                 return err;
6000         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6001         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6002 }
6003
6004 struct nfs_release_lockowner_data {
6005         struct nfs4_lock_state *lsp;
6006         struct nfs_server *server;
6007         struct nfs_release_lockowner_args args;
6008         struct nfs_release_lockowner_res res;
6009         unsigned long timestamp;
6010 };
6011
6012 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6013 {
6014         struct nfs_release_lockowner_data *data = calldata;
6015         struct nfs_server *server = data->server;
6016         nfs40_setup_sequence(server, &data->args.seq_args,
6017                                 &data->res.seq_res, task);
6018         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6019         data->timestamp = jiffies;
6020 }
6021
6022 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6023 {
6024         struct nfs_release_lockowner_data *data = calldata;
6025         struct nfs_server *server = data->server;
6026
6027         nfs40_sequence_done(task, &data->res.seq_res);
6028
6029         switch (task->tk_status) {
6030         case 0:
6031                 renew_lease(server, data->timestamp);
6032                 break;
6033         case -NFS4ERR_STALE_CLIENTID:
6034         case -NFS4ERR_EXPIRED:
6035                 nfs4_schedule_lease_recovery(server->nfs_client);
6036                 break;
6037         case -NFS4ERR_LEASE_MOVED:
6038         case -NFS4ERR_DELAY:
6039                 if (nfs4_async_handle_error(task, server,
6040                                             NULL, NULL) == -EAGAIN)
6041                         rpc_restart_call_prepare(task);
6042         }
6043 }
6044
6045 static void nfs4_release_lockowner_release(void *calldata)
6046 {
6047         struct nfs_release_lockowner_data *data = calldata;
6048         nfs4_free_lock_state(data->server, data->lsp);
6049         kfree(calldata);
6050 }
6051
6052 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6053         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6054         .rpc_call_done = nfs4_release_lockowner_done,
6055         .rpc_release = nfs4_release_lockowner_release,
6056 };
6057
6058 static void
6059 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6060 {
6061         struct nfs_release_lockowner_data *data;
6062         struct rpc_message msg = {
6063                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6064         };
6065
6066         if (server->nfs_client->cl_mvops->minor_version != 0)
6067                 return;
6068
6069         data = kmalloc(sizeof(*data), GFP_NOFS);
6070         if (!data)
6071                 return;
6072         data->lsp = lsp;
6073         data->server = server;
6074         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6075         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6076         data->args.lock_owner.s_dev = server->s_dev;
6077
6078         msg.rpc_argp = &data->args;
6079         msg.rpc_resp = &data->res;
6080         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6081         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6082 }
6083
6084 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6085
6086 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6087                                    const void *buf, size_t buflen,
6088                                    int flags, int type)
6089 {
6090         if (strcmp(key, "") != 0)
6091                 return -EINVAL;
6092
6093         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6094 }
6095
6096 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6097                                    void *buf, size_t buflen, int type)
6098 {
6099         if (strcmp(key, "") != 0)
6100                 return -EINVAL;
6101
6102         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6103 }
6104
6105 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6106                                        size_t list_len, const char *name,
6107                                        size_t name_len, int type)
6108 {
6109         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6110
6111         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6112                 return 0;
6113
6114         if (list && len <= list_len)
6115                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6116         return len;
6117 }
6118
6119 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6120 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6121 {
6122         return server->caps & NFS_CAP_SECURITY_LABEL;
6123 }
6124
6125 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6126                                    const void *buf, size_t buflen,
6127                                    int flags, int type)
6128 {
6129         if (security_ismaclabel(key))
6130                 return nfs4_set_security_label(dentry, buf, buflen);
6131
6132         return -EOPNOTSUPP;
6133 }
6134
6135 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6136                                    void *buf, size_t buflen, int type)
6137 {
6138         if (security_ismaclabel(key))
6139                 return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6140         return -EOPNOTSUPP;
6141 }
6142
6143 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6144                                        size_t list_len, const char *name,
6145                                        size_t name_len, int type)
6146 {
6147         size_t len = 0;
6148
6149         if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6150                 len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6151                 if (list && len <= list_len)
6152                         security_inode_listsecurity(dentry->d_inode, list, len);
6153         }
6154         return len;
6155 }
6156
6157 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6158         .prefix = XATTR_SECURITY_PREFIX,
6159         .list   = nfs4_xattr_list_nfs4_label,
6160         .get    = nfs4_xattr_get_nfs4_label,
6161         .set    = nfs4_xattr_set_nfs4_label,
6162 };
6163 #endif
6164
6165
6166 /*
6167  * nfs_fhget will use either the mounted_on_fileid or the fileid
6168  */
6169 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6170 {
6171         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6172                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6173               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6174               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6175                 return;
6176
6177         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6178                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6179         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6180         fattr->nlink = 2;
6181 }
6182
6183 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6184                                    const struct qstr *name,
6185                                    struct nfs4_fs_locations *fs_locations,
6186                                    struct page *page)
6187 {
6188         struct nfs_server *server = NFS_SERVER(dir);
6189         u32 bitmask[3] = {
6190                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6191         };
6192         struct nfs4_fs_locations_arg args = {
6193                 .dir_fh = NFS_FH(dir),
6194                 .name = name,
6195                 .page = page,
6196                 .bitmask = bitmask,
6197         };
6198         struct nfs4_fs_locations_res res = {
6199                 .fs_locations = fs_locations,
6200         };
6201         struct rpc_message msg = {
6202                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6203                 .rpc_argp = &args,
6204                 .rpc_resp = &res,
6205         };
6206         int status;
6207
6208         dprintk("%s: start\n", __func__);
6209
6210         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6211          * is not supported */
6212         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6213                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6214         else
6215                 bitmask[0] |= FATTR4_WORD0_FILEID;
6216
6217         nfs_fattr_init(&fs_locations->fattr);
6218         fs_locations->server = server;
6219         fs_locations->nlocations = 0;
6220         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6221         dprintk("%s: returned status = %d\n", __func__, status);
6222         return status;
6223 }
6224
6225 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6226                            const struct qstr *name,
6227                            struct nfs4_fs_locations *fs_locations,
6228                            struct page *page)
6229 {
6230         struct nfs4_exception exception = { };
6231         int err;
6232         do {
6233                 err = _nfs4_proc_fs_locations(client, dir, name,
6234                                 fs_locations, page);
6235                 trace_nfs4_get_fs_locations(dir, name, err);
6236                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6237                                 &exception);
6238         } while (exception.retry);
6239         return err;
6240 }
6241
6242 /*
6243  * This operation also signals the server that this client is
6244  * performing migration recovery.  The server can stop returning
6245  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6246  * appended to this compound to identify the client ID which is
6247  * performing recovery.
6248  */
6249 static int _nfs40_proc_get_locations(struct inode *inode,
6250                                      struct nfs4_fs_locations *locations,
6251                                      struct page *page, struct rpc_cred *cred)
6252 {
6253         struct nfs_server *server = NFS_SERVER(inode);
6254         struct rpc_clnt *clnt = server->client;
6255         u32 bitmask[2] = {
6256                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6257         };
6258         struct nfs4_fs_locations_arg args = {
6259                 .clientid       = server->nfs_client->cl_clientid,
6260                 .fh             = NFS_FH(inode),
6261                 .page           = page,
6262                 .bitmask        = bitmask,
6263                 .migration      = 1,            /* skip LOOKUP */
6264                 .renew          = 1,            /* append RENEW */
6265         };
6266         struct nfs4_fs_locations_res res = {
6267                 .fs_locations   = locations,
6268                 .migration      = 1,
6269                 .renew          = 1,
6270         };
6271         struct rpc_message msg = {
6272                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6273                 .rpc_argp       = &args,
6274                 .rpc_resp       = &res,
6275                 .rpc_cred       = cred,
6276         };
6277         unsigned long now = jiffies;
6278         int status;
6279
6280         nfs_fattr_init(&locations->fattr);
6281         locations->server = server;
6282         locations->nlocations = 0;
6283
6284         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6285         nfs4_set_sequence_privileged(&args.seq_args);
6286         status = nfs4_call_sync_sequence(clnt, server, &msg,
6287                                         &args.seq_args, &res.seq_res);
6288         if (status)
6289                 return status;
6290
6291         renew_lease(server, now);
6292         return 0;
6293 }
6294
6295 #ifdef CONFIG_NFS_V4_1
6296
6297 /*
6298  * This operation also signals the server that this client is
6299  * performing migration recovery.  The server can stop asserting
6300  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6301  * performing this operation is identified in the SEQUENCE
6302  * operation in this compound.
6303  *
6304  * When the client supports GETATTR(fs_locations_info), it can
6305  * be plumbed in here.
6306  */
6307 static int _nfs41_proc_get_locations(struct inode *inode,
6308                                      struct nfs4_fs_locations *locations,
6309                                      struct page *page, struct rpc_cred *cred)
6310 {
6311         struct nfs_server *server = NFS_SERVER(inode);
6312         struct rpc_clnt *clnt = server->client;
6313         u32 bitmask[2] = {
6314                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6315         };
6316         struct nfs4_fs_locations_arg args = {
6317                 .fh             = NFS_FH(inode),
6318                 .page           = page,
6319                 .bitmask        = bitmask,
6320                 .migration      = 1,            /* skip LOOKUP */
6321         };
6322         struct nfs4_fs_locations_res res = {
6323                 .fs_locations   = locations,
6324                 .migration      = 1,
6325         };
6326         struct rpc_message msg = {
6327                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6328                 .rpc_argp       = &args,
6329                 .rpc_resp       = &res,
6330                 .rpc_cred       = cred,
6331         };
6332         int status;
6333
6334         nfs_fattr_init(&locations->fattr);
6335         locations->server = server;
6336         locations->nlocations = 0;
6337
6338         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6339         nfs4_set_sequence_privileged(&args.seq_args);
6340         status = nfs4_call_sync_sequence(clnt, server, &msg,
6341                                         &args.seq_args, &res.seq_res);
6342         if (status == NFS4_OK &&
6343             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6344                 status = -NFS4ERR_LEASE_MOVED;
6345         return status;
6346 }
6347
6348 #endif  /* CONFIG_NFS_V4_1 */
6349
6350 /**
6351  * nfs4_proc_get_locations - discover locations for a migrated FSID
6352  * @inode: inode on FSID that is migrating
6353  * @locations: result of query
6354  * @page: buffer
6355  * @cred: credential to use for this operation
6356  *
6357  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6358  * operation failed, or a negative errno if a local error occurred.
6359  *
6360  * On success, "locations" is filled in, but if the server has
6361  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6362  * asserted.
6363  *
6364  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6365  * from this client that require migration recovery.
6366  */
6367 int nfs4_proc_get_locations(struct inode *inode,
6368                             struct nfs4_fs_locations *locations,
6369                             struct page *page, struct rpc_cred *cred)
6370 {
6371         struct nfs_server *server = NFS_SERVER(inode);
6372         struct nfs_client *clp = server->nfs_client;
6373         const struct nfs4_mig_recovery_ops *ops =
6374                                         clp->cl_mvops->mig_recovery_ops;
6375         struct nfs4_exception exception = { };
6376         int status;
6377
6378         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6379                 (unsigned long long)server->fsid.major,
6380                 (unsigned long long)server->fsid.minor,
6381                 clp->cl_hostname);
6382         nfs_display_fhandle(NFS_FH(inode), __func__);
6383
6384         do {
6385                 status = ops->get_locations(inode, locations, page, cred);
6386                 if (status != -NFS4ERR_DELAY)
6387                         break;
6388                 nfs4_handle_exception(server, status, &exception);
6389         } while (exception.retry);
6390         return status;
6391 }
6392
6393 /*
6394  * This operation also signals the server that this client is
6395  * performing "lease moved" recovery.  The server can stop
6396  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6397  * is appended to this compound to identify the client ID which is
6398  * performing recovery.
6399  */
6400 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6401 {
6402         struct nfs_server *server = NFS_SERVER(inode);
6403         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6404         struct rpc_clnt *clnt = server->client;
6405         struct nfs4_fsid_present_arg args = {
6406                 .fh             = NFS_FH(inode),
6407                 .clientid       = clp->cl_clientid,
6408                 .renew          = 1,            /* append RENEW */
6409         };
6410         struct nfs4_fsid_present_res res = {
6411                 .renew          = 1,
6412         };
6413         struct rpc_message msg = {
6414                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6415                 .rpc_argp       = &args,
6416                 .rpc_resp       = &res,
6417                 .rpc_cred       = cred,
6418         };
6419         unsigned long now = jiffies;
6420         int status;
6421
6422         res.fh = nfs_alloc_fhandle();
6423         if (res.fh == NULL)
6424                 return -ENOMEM;
6425
6426         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6427         nfs4_set_sequence_privileged(&args.seq_args);
6428         status = nfs4_call_sync_sequence(clnt, server, &msg,
6429                                                 &args.seq_args, &res.seq_res);
6430         nfs_free_fhandle(res.fh);
6431         if (status)
6432                 return status;
6433
6434         do_renew_lease(clp, now);
6435         return 0;
6436 }
6437
6438 #ifdef CONFIG_NFS_V4_1
6439
6440 /*
6441  * This operation also signals the server that this client is
6442  * performing "lease moved" recovery.  The server can stop asserting
6443  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6444  * this operation is identified in the SEQUENCE operation in this
6445  * compound.
6446  */
6447 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6448 {
6449         struct nfs_server *server = NFS_SERVER(inode);
6450         struct rpc_clnt *clnt = server->client;
6451         struct nfs4_fsid_present_arg args = {
6452                 .fh             = NFS_FH(inode),
6453         };
6454         struct nfs4_fsid_present_res res = {
6455         };
6456         struct rpc_message msg = {
6457                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6458                 .rpc_argp       = &args,
6459                 .rpc_resp       = &res,
6460                 .rpc_cred       = cred,
6461         };
6462         int status;
6463
6464         res.fh = nfs_alloc_fhandle();
6465         if (res.fh == NULL)
6466                 return -ENOMEM;
6467
6468         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6469         nfs4_set_sequence_privileged(&args.seq_args);
6470         status = nfs4_call_sync_sequence(clnt, server, &msg,
6471                                                 &args.seq_args, &res.seq_res);
6472         nfs_free_fhandle(res.fh);
6473         if (status == NFS4_OK &&
6474             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6475                 status = -NFS4ERR_LEASE_MOVED;
6476         return status;
6477 }
6478
6479 #endif  /* CONFIG_NFS_V4_1 */
6480
6481 /**
6482  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6483  * @inode: inode on FSID to check
6484  * @cred: credential to use for this operation
6485  *
6486  * Server indicates whether the FSID is present, moved, or not
6487  * recognized.  This operation is necessary to clear a LEASE_MOVED
6488  * condition for this client ID.
6489  *
6490  * Returns NFS4_OK if the FSID is present on this server,
6491  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6492  *  NFS4ERR code if some error occurred on the server, or a
6493  *  negative errno if a local failure occurred.
6494  */
6495 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6496 {
6497         struct nfs_server *server = NFS_SERVER(inode);
6498         struct nfs_client *clp = server->nfs_client;
6499         const struct nfs4_mig_recovery_ops *ops =
6500                                         clp->cl_mvops->mig_recovery_ops;
6501         struct nfs4_exception exception = { };
6502         int status;
6503
6504         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6505                 (unsigned long long)server->fsid.major,
6506                 (unsigned long long)server->fsid.minor,
6507                 clp->cl_hostname);
6508         nfs_display_fhandle(NFS_FH(inode), __func__);
6509
6510         do {
6511                 status = ops->fsid_present(inode, cred);
6512                 if (status != -NFS4ERR_DELAY)
6513                         break;
6514                 nfs4_handle_exception(server, status, &exception);
6515         } while (exception.retry);
6516         return status;
6517 }
6518
6519 /**
6520  * If 'use_integrity' is true and the state managment nfs_client
6521  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6522  * and the machine credential as per RFC3530bis and RFC5661 Security
6523  * Considerations sections. Otherwise, just use the user cred with the
6524  * filesystem's rpc_client.
6525  */
6526 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6527 {
6528         int status;
6529         struct nfs4_secinfo_arg args = {
6530                 .dir_fh = NFS_FH(dir),
6531                 .name   = name,
6532         };
6533         struct nfs4_secinfo_res res = {
6534                 .flavors     = flavors,
6535         };
6536         struct rpc_message msg = {
6537                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6538                 .rpc_argp = &args,
6539                 .rpc_resp = &res,
6540         };
6541         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6542         struct rpc_cred *cred = NULL;
6543
6544         if (use_integrity) {
6545                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6546                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6547                 msg.rpc_cred = cred;
6548         }
6549
6550         dprintk("NFS call  secinfo %s\n", name->name);
6551
6552         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6553                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6554
6555         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6556                                 &res.seq_res, 0);
6557         dprintk("NFS reply  secinfo: %d\n", status);
6558
6559         if (cred)
6560                 put_rpccred(cred);
6561
6562         return status;
6563 }
6564
6565 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6566                       struct nfs4_secinfo_flavors *flavors)
6567 {
6568         struct nfs4_exception exception = { };
6569         int err;
6570         do {
6571                 err = -NFS4ERR_WRONGSEC;
6572
6573                 /* try to use integrity protection with machine cred */
6574                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6575                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6576
6577                 /*
6578                  * if unable to use integrity protection, or SECINFO with
6579                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6580                  * disallowed by spec, but exists in deployed servers) use
6581                  * the current filesystem's rpc_client and the user cred.
6582                  */
6583                 if (err == -NFS4ERR_WRONGSEC)
6584                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6585
6586                 trace_nfs4_secinfo(dir, name, err);
6587                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6588                                 &exception);
6589         } while (exception.retry);
6590         return err;
6591 }
6592
6593 #ifdef CONFIG_NFS_V4_1
6594 /*
6595  * Check the exchange flags returned by the server for invalid flags, having
6596  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6597  * DS flags set.
6598  */
6599 static int nfs4_check_cl_exchange_flags(u32 flags)
6600 {
6601         if (flags & ~EXCHGID4_FLAG_MASK_R)
6602                 goto out_inval;
6603         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6604             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6605                 goto out_inval;
6606         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6607                 goto out_inval;
6608         return NFS_OK;
6609 out_inval:
6610         return -NFS4ERR_INVAL;
6611 }
6612
6613 static bool
6614 nfs41_same_server_scope(struct nfs41_server_scope *a,
6615                         struct nfs41_server_scope *b)
6616 {
6617         if (a->server_scope_sz == b->server_scope_sz &&
6618             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6619                 return true;
6620
6621         return false;
6622 }
6623
6624 /*
6625  * nfs4_proc_bind_conn_to_session()
6626  *
6627  * The 4.1 client currently uses the same TCP connection for the
6628  * fore and backchannel.
6629  */
6630 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6631 {
6632         int status;
6633         struct nfs41_bind_conn_to_session_res res;
6634         struct rpc_message msg = {
6635                 .rpc_proc =
6636                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6637                 .rpc_argp = clp,
6638                 .rpc_resp = &res,
6639                 .rpc_cred = cred,
6640         };
6641
6642         dprintk("--> %s\n", __func__);
6643
6644         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6645         if (unlikely(res.session == NULL)) {
6646                 status = -ENOMEM;
6647                 goto out;
6648         }
6649
6650         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6651         trace_nfs4_bind_conn_to_session(clp, status);
6652         if (status == 0) {
6653                 if (memcmp(res.session->sess_id.data,
6654                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6655                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6656                         status = -EIO;
6657                         goto out_session;
6658                 }
6659                 if (res.dir != NFS4_CDFS4_BOTH) {
6660                         dprintk("NFS: %s: Unexpected direction from server\n",
6661                                 __func__);
6662                         status = -EIO;
6663                         goto out_session;
6664                 }
6665                 if (res.use_conn_in_rdma_mode) {
6666                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6667                                 __func__);
6668                         status = -EIO;
6669                         goto out_session;
6670                 }
6671         }
6672 out_session:
6673         kfree(res.session);
6674 out:
6675         dprintk("<-- %s status= %d\n", __func__, status);
6676         return status;
6677 }
6678
6679 /*
6680  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6681  * and operations we'd like to see to enable certain features in the allow map
6682  */
6683 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6684         .how = SP4_MACH_CRED,
6685         .enforce.u.words = {
6686                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6687                       1 << (OP_EXCHANGE_ID - 32) |
6688                       1 << (OP_CREATE_SESSION - 32) |
6689                       1 << (OP_DESTROY_SESSION - 32) |
6690                       1 << (OP_DESTROY_CLIENTID - 32)
6691         },
6692         .allow.u.words = {
6693                 [0] = 1 << (OP_CLOSE) |
6694                       1 << (OP_LOCKU) |
6695                       1 << (OP_COMMIT),
6696                 [1] = 1 << (OP_SECINFO - 32) |
6697                       1 << (OP_SECINFO_NO_NAME - 32) |
6698                       1 << (OP_TEST_STATEID - 32) |
6699                       1 << (OP_FREE_STATEID - 32) |
6700                       1 << (OP_WRITE - 32)
6701         }
6702 };
6703
6704 /*
6705  * Select the state protection mode for client `clp' given the server results
6706  * from exchange_id in `sp'.
6707  *
6708  * Returns 0 on success, negative errno otherwise.
6709  */
6710 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6711                                  struct nfs41_state_protection *sp)
6712 {
6713         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6714                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6715                       1 << (OP_EXCHANGE_ID - 32) |
6716                       1 << (OP_CREATE_SESSION - 32) |
6717                       1 << (OP_DESTROY_SESSION - 32) |
6718                       1 << (OP_DESTROY_CLIENTID - 32)
6719         };
6720         unsigned int i;
6721
6722         if (sp->how == SP4_MACH_CRED) {
6723                 /* Print state protect result */
6724                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6725                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6726                         if (test_bit(i, sp->enforce.u.longs))
6727                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6728                         if (test_bit(i, sp->allow.u.longs))
6729                                 dfprintk(MOUNT, "  allow op %d\n", i);
6730                 }
6731
6732                 /* make sure nothing is on enforce list that isn't supported */
6733                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6734                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6735                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6736                                 return -EINVAL;
6737                         }
6738                 }
6739
6740                 /*
6741                  * Minimal mode - state operations are allowed to use machine
6742                  * credential.  Note this already happens by default, so the
6743                  * client doesn't have to do anything more than the negotiation.
6744                  *
6745                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6746                  *       we're already using the machine cred for exchange_id
6747                  *       and will never use a different cred.
6748                  */
6749                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6750                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6751                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6752                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6753                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6754                         dfprintk(MOUNT, "  minimal mode enabled\n");
6755                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6756                 } else {
6757                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6758                         return -EINVAL;
6759                 }
6760
6761                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6762                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6763                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6764                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6765                 }
6766
6767                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6768                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6769                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6770                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6771                 }
6772
6773                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6774                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6775                         dfprintk(MOUNT, "  stateid mode enabled\n");
6776                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6777                 }
6778
6779                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6780                         dfprintk(MOUNT, "  write mode enabled\n");
6781                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6782                 }
6783
6784                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6785                         dfprintk(MOUNT, "  commit mode enabled\n");
6786                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6787                 }
6788         }
6789
6790         return 0;
6791 }
6792
6793 /*
6794  * _nfs4_proc_exchange_id()
6795  *
6796  * Wrapper for EXCHANGE_ID operation.
6797  */
6798 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6799         u32 sp4_how)
6800 {
6801         nfs4_verifier verifier;
6802         struct nfs41_exchange_id_args args = {
6803                 .verifier = &verifier,
6804                 .client = clp,
6805 #ifdef CONFIG_NFS_V4_1_MIGRATION
6806                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6807                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6808                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6809 #else
6810                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6811                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6812 #endif
6813         };
6814         struct nfs41_exchange_id_res res = {
6815                 0
6816         };
6817         int status;
6818         struct rpc_message msg = {
6819                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6820                 .rpc_argp = &args,
6821                 .rpc_resp = &res,
6822                 .rpc_cred = cred,
6823         };
6824
6825         nfs4_init_boot_verifier(clp, &verifier);
6826         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6827                                                         sizeof(args.id));
6828         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6829                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6830                 args.id_len, args.id);
6831
6832         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6833                                         GFP_NOFS);
6834         if (unlikely(res.server_owner == NULL)) {
6835                 status = -ENOMEM;
6836                 goto out;
6837         }
6838
6839         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6840                                         GFP_NOFS);
6841         if (unlikely(res.server_scope == NULL)) {
6842                 status = -ENOMEM;
6843                 goto out_server_owner;
6844         }
6845
6846         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6847         if (unlikely(res.impl_id == NULL)) {
6848                 status = -ENOMEM;
6849                 goto out_server_scope;
6850         }
6851
6852         switch (sp4_how) {
6853         case SP4_NONE:
6854                 args.state_protect.how = SP4_NONE;
6855                 break;
6856
6857         case SP4_MACH_CRED:
6858                 args.state_protect = nfs4_sp4_mach_cred_request;
6859                 break;
6860
6861         default:
6862                 /* unsupported! */
6863                 WARN_ON_ONCE(1);
6864                 status = -EINVAL;
6865                 goto out_server_scope;
6866         }
6867
6868         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6869         trace_nfs4_exchange_id(clp, status);
6870         if (status == 0)
6871                 status = nfs4_check_cl_exchange_flags(res.flags);
6872
6873         if (status == 0)
6874                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6875
6876         if (status == 0) {
6877                 clp->cl_clientid = res.clientid;
6878                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6879                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6880                         clp->cl_seqid = res.seqid;
6881
6882                 kfree(clp->cl_serverowner);
6883                 clp->cl_serverowner = res.server_owner;
6884                 res.server_owner = NULL;
6885
6886                 /* use the most recent implementation id */
6887                 kfree(clp->cl_implid);
6888                 clp->cl_implid = res.impl_id;
6889
6890                 if (clp->cl_serverscope != NULL &&
6891                     !nfs41_same_server_scope(clp->cl_serverscope,
6892                                              res.server_scope)) {
6893                         dprintk("%s: server_scope mismatch detected\n",
6894                                 __func__);
6895                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6896                         kfree(clp->cl_serverscope);
6897                         clp->cl_serverscope = NULL;
6898                 }
6899
6900                 if (clp->cl_serverscope == NULL) {
6901                         clp->cl_serverscope = res.server_scope;
6902                         goto out;
6903                 }
6904         } else
6905                 kfree(res.impl_id);
6906
6907 out_server_owner:
6908         kfree(res.server_owner);
6909 out_server_scope:
6910         kfree(res.server_scope);
6911 out:
6912         if (clp->cl_implid != NULL)
6913                 dprintk("NFS reply exchange_id: Server Implementation ID: "
6914                         "domain: %s, name: %s, date: %llu,%u\n",
6915                         clp->cl_implid->domain, clp->cl_implid->name,
6916                         clp->cl_implid->date.seconds,
6917                         clp->cl_implid->date.nseconds);
6918         dprintk("NFS reply exchange_id: %d\n", status);
6919         return status;
6920 }
6921
6922 /*
6923  * nfs4_proc_exchange_id()
6924  *
6925  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6926  *
6927  * Since the clientid has expired, all compounds using sessions
6928  * associated with the stale clientid will be returning
6929  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6930  * be in some phase of session reset.
6931  *
6932  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6933  */
6934 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6935 {
6936         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6937         int status;
6938
6939         /* try SP4_MACH_CRED if krb5i/p */
6940         if (authflavor == RPC_AUTH_GSS_KRB5I ||
6941             authflavor == RPC_AUTH_GSS_KRB5P) {
6942                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6943                 if (!status)
6944                         return 0;
6945         }
6946
6947         /* try SP4_NONE */
6948         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6949 }
6950
6951 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6952                 struct rpc_cred *cred)
6953 {
6954         struct rpc_message msg = {
6955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6956                 .rpc_argp = clp,
6957                 .rpc_cred = cred,
6958         };
6959         int status;
6960
6961         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6962         trace_nfs4_destroy_clientid(clp, status);
6963         if (status)
6964                 dprintk("NFS: Got error %d from the server %s on "
6965                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
6966         return status;
6967 }
6968
6969 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6970                 struct rpc_cred *cred)
6971 {
6972         unsigned int loop;
6973         int ret;
6974
6975         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6976                 ret = _nfs4_proc_destroy_clientid(clp, cred);
6977                 switch (ret) {
6978                 case -NFS4ERR_DELAY:
6979                 case -NFS4ERR_CLIENTID_BUSY:
6980                         ssleep(1);
6981                         break;
6982                 default:
6983                         return ret;
6984                 }
6985         }
6986         return 0;
6987 }
6988
6989 int nfs4_destroy_clientid(struct nfs_client *clp)
6990 {
6991         struct rpc_cred *cred;
6992         int ret = 0;
6993
6994         if (clp->cl_mvops->minor_version < 1)
6995                 goto out;
6996         if (clp->cl_exchange_flags == 0)
6997                 goto out;
6998         if (clp->cl_preserve_clid)
6999                 goto out;
7000         cred = nfs4_get_clid_cred(clp);
7001         ret = nfs4_proc_destroy_clientid(clp, cred);
7002         if (cred)
7003                 put_rpccred(cred);
7004         switch (ret) {
7005         case 0:
7006         case -NFS4ERR_STALE_CLIENTID:
7007                 clp->cl_exchange_flags = 0;
7008         }
7009 out:
7010         return ret;
7011 }
7012
7013 struct nfs4_get_lease_time_data {
7014         struct nfs4_get_lease_time_args *args;
7015         struct nfs4_get_lease_time_res *res;
7016         struct nfs_client *clp;
7017 };
7018
7019 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7020                                         void *calldata)
7021 {
7022         struct nfs4_get_lease_time_data *data =
7023                         (struct nfs4_get_lease_time_data *)calldata;
7024
7025         dprintk("--> %s\n", __func__);
7026         /* just setup sequence, do not trigger session recovery
7027            since we're invoked within one */
7028         nfs41_setup_sequence(data->clp->cl_session,
7029                         &data->args->la_seq_args,
7030                         &data->res->lr_seq_res,
7031                         task);
7032         dprintk("<-- %s\n", __func__);
7033 }
7034
7035 /*
7036  * Called from nfs4_state_manager thread for session setup, so don't recover
7037  * from sequence operation or clientid errors.
7038  */
7039 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7040 {
7041         struct nfs4_get_lease_time_data *data =
7042                         (struct nfs4_get_lease_time_data *)calldata;
7043
7044         dprintk("--> %s\n", __func__);
7045         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7046                 return;
7047         switch (task->tk_status) {
7048         case -NFS4ERR_DELAY:
7049         case -NFS4ERR_GRACE:
7050                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7051                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7052                 task->tk_status = 0;
7053                 /* fall through */
7054         case -NFS4ERR_RETRY_UNCACHED_REP:
7055                 rpc_restart_call_prepare(task);
7056                 return;
7057         }
7058         dprintk("<-- %s\n", __func__);
7059 }
7060
7061 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7062         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7063         .rpc_call_done = nfs4_get_lease_time_done,
7064 };
7065
7066 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7067 {
7068         struct rpc_task *task;
7069         struct nfs4_get_lease_time_args args;
7070         struct nfs4_get_lease_time_res res = {
7071                 .lr_fsinfo = fsinfo,
7072         };
7073         struct nfs4_get_lease_time_data data = {
7074                 .args = &args,
7075                 .res = &res,
7076                 .clp = clp,
7077         };
7078         struct rpc_message msg = {
7079                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7080                 .rpc_argp = &args,
7081                 .rpc_resp = &res,
7082         };
7083         struct rpc_task_setup task_setup = {
7084                 .rpc_client = clp->cl_rpcclient,
7085                 .rpc_message = &msg,
7086                 .callback_ops = &nfs4_get_lease_time_ops,
7087                 .callback_data = &data,
7088                 .flags = RPC_TASK_TIMEOUT,
7089         };
7090         int status;
7091
7092         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7093         nfs4_set_sequence_privileged(&args.la_seq_args);
7094         dprintk("--> %s\n", __func__);
7095         task = rpc_run_task(&task_setup);
7096
7097         if (IS_ERR(task))
7098                 status = PTR_ERR(task);
7099         else {
7100                 status = task->tk_status;
7101                 rpc_put_task(task);
7102         }
7103         dprintk("<-- %s return %d\n", __func__, status);
7104
7105         return status;
7106 }
7107
7108 /*
7109  * Initialize the values to be used by the client in CREATE_SESSION
7110  * If nfs4_init_session set the fore channel request and response sizes,
7111  * use them.
7112  *
7113  * Set the back channel max_resp_sz_cached to zero to force the client to
7114  * always set csa_cachethis to FALSE because the current implementation
7115  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7116  */
7117 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7118 {
7119         unsigned int max_rqst_sz, max_resp_sz;
7120
7121         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7122         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7123
7124         /* Fore channel attributes */
7125         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7126         args->fc_attrs.max_resp_sz = max_resp_sz;
7127         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7128         args->fc_attrs.max_reqs = max_session_slots;
7129
7130         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7131                 "max_ops=%u max_reqs=%u\n",
7132                 __func__,
7133                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7134                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7135
7136         /* Back channel attributes */
7137         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7138         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7139         args->bc_attrs.max_resp_sz_cached = 0;
7140         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7141         args->bc_attrs.max_reqs = 1;
7142
7143         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7144                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7145                 __func__,
7146                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7147                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7148                 args->bc_attrs.max_reqs);
7149 }
7150
7151 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7152 {
7153         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7154         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7155
7156         if (rcvd->max_resp_sz > sent->max_resp_sz)
7157                 return -EINVAL;
7158         /*
7159          * Our requested max_ops is the minimum we need; we're not
7160          * prepared to break up compounds into smaller pieces than that.
7161          * So, no point even trying to continue if the server won't
7162          * cooperate:
7163          */
7164         if (rcvd->max_ops < sent->max_ops)
7165                 return -EINVAL;
7166         if (rcvd->max_reqs == 0)
7167                 return -EINVAL;
7168         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7169                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7170         return 0;
7171 }
7172
7173 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7174 {
7175         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7176         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7177
7178         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7179                 return -EINVAL;
7180         if (rcvd->max_resp_sz < sent->max_resp_sz)
7181                 return -EINVAL;
7182         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7183                 return -EINVAL;
7184         /* These would render the backchannel useless: */
7185         if (rcvd->max_ops != sent->max_ops)
7186                 return -EINVAL;
7187         if (rcvd->max_reqs != sent->max_reqs)
7188                 return -EINVAL;
7189         return 0;
7190 }
7191
7192 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7193                                      struct nfs4_session *session)
7194 {
7195         int ret;
7196
7197         ret = nfs4_verify_fore_channel_attrs(args, session);
7198         if (ret)
7199                 return ret;
7200         return nfs4_verify_back_channel_attrs(args, session);
7201 }
7202
7203 static int _nfs4_proc_create_session(struct nfs_client *clp,
7204                 struct rpc_cred *cred)
7205 {
7206         struct nfs4_session *session = clp->cl_session;
7207         struct nfs41_create_session_args args = {
7208                 .client = clp,
7209                 .cb_program = NFS4_CALLBACK,
7210         };
7211         struct nfs41_create_session_res res = {
7212                 .client = clp,
7213         };
7214         struct rpc_message msg = {
7215                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7216                 .rpc_argp = &args,
7217                 .rpc_resp = &res,
7218                 .rpc_cred = cred,
7219         };
7220         int status;
7221
7222         nfs4_init_channel_attrs(&args);
7223         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7224
7225         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7226         trace_nfs4_create_session(clp, status);
7227
7228         if (!status) {
7229                 /* Verify the session's negotiated channel_attrs values */
7230                 status = nfs4_verify_channel_attrs(&args, session);
7231                 /* Increment the clientid slot sequence id */
7232                 clp->cl_seqid++;
7233         }
7234
7235         return status;
7236 }
7237
7238 /*
7239  * Issues a CREATE_SESSION operation to the server.
7240  * It is the responsibility of the caller to verify the session is
7241  * expired before calling this routine.
7242  */
7243 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7244 {
7245         int status;
7246         unsigned *ptr;
7247         struct nfs4_session *session = clp->cl_session;
7248
7249         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7250
7251         status = _nfs4_proc_create_session(clp, cred);
7252         if (status)
7253                 goto out;
7254
7255         /* Init or reset the session slot tables */
7256         status = nfs4_setup_session_slot_tables(session);
7257         dprintk("slot table setup returned %d\n", status);
7258         if (status)
7259                 goto out;
7260
7261         ptr = (unsigned *)&session->sess_id.data[0];
7262         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7263                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7264 out:
7265         dprintk("<-- %s\n", __func__);
7266         return status;
7267 }
7268
7269 /*
7270  * Issue the over-the-wire RPC DESTROY_SESSION.
7271  * The caller must serialize access to this routine.
7272  */
7273 int nfs4_proc_destroy_session(struct nfs4_session *session,
7274                 struct rpc_cred *cred)
7275 {
7276         struct rpc_message msg = {
7277                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7278                 .rpc_argp = session,
7279                 .rpc_cred = cred,
7280         };
7281         int status = 0;
7282
7283         dprintk("--> nfs4_proc_destroy_session\n");
7284
7285         /* session is still being setup */
7286         if (session->clp->cl_cons_state != NFS_CS_READY)
7287                 return status;
7288
7289         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7290         trace_nfs4_destroy_session(session->clp, status);
7291
7292         if (status)
7293                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7294                         "Session has been destroyed regardless...\n", status);
7295
7296         dprintk("<-- nfs4_proc_destroy_session\n");
7297         return status;
7298 }
7299
7300 /*
7301  * Renew the cl_session lease.
7302  */
7303 struct nfs4_sequence_data {
7304         struct nfs_client *clp;
7305         struct nfs4_sequence_args args;
7306         struct nfs4_sequence_res res;
7307 };
7308
7309 static void nfs41_sequence_release(void *data)
7310 {
7311         struct nfs4_sequence_data *calldata = data;
7312         struct nfs_client *clp = calldata->clp;
7313
7314         if (atomic_read(&clp->cl_count) > 1)
7315                 nfs4_schedule_state_renewal(clp);
7316         nfs_put_client(clp);
7317         kfree(calldata);
7318 }
7319
7320 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7321 {
7322         switch(task->tk_status) {
7323         case -NFS4ERR_DELAY:
7324                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7325                 return -EAGAIN;
7326         default:
7327                 nfs4_schedule_lease_recovery(clp);
7328         }
7329         return 0;
7330 }
7331
7332 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7333 {
7334         struct nfs4_sequence_data *calldata = data;
7335         struct nfs_client *clp = calldata->clp;
7336
7337         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7338                 return;
7339
7340         trace_nfs4_sequence(clp, task->tk_status);
7341         if (task->tk_status < 0) {
7342                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7343                 if (atomic_read(&clp->cl_count) == 1)
7344                         goto out;
7345
7346                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7347                         rpc_restart_call_prepare(task);
7348                         return;
7349                 }
7350         }
7351         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7352 out:
7353         dprintk("<-- %s\n", __func__);
7354 }
7355
7356 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7357 {
7358         struct nfs4_sequence_data *calldata = data;
7359         struct nfs_client *clp = calldata->clp;
7360         struct nfs4_sequence_args *args;
7361         struct nfs4_sequence_res *res;
7362
7363         args = task->tk_msg.rpc_argp;
7364         res = task->tk_msg.rpc_resp;
7365
7366         nfs41_setup_sequence(clp->cl_session, args, res, task);
7367 }
7368
7369 static const struct rpc_call_ops nfs41_sequence_ops = {
7370         .rpc_call_done = nfs41_sequence_call_done,
7371         .rpc_call_prepare = nfs41_sequence_prepare,
7372         .rpc_release = nfs41_sequence_release,
7373 };
7374
7375 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7376                 struct rpc_cred *cred,
7377                 bool is_privileged)
7378 {
7379         struct nfs4_sequence_data *calldata;
7380         struct rpc_message msg = {
7381                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7382                 .rpc_cred = cred,
7383         };
7384         struct rpc_task_setup task_setup_data = {
7385                 .rpc_client = clp->cl_rpcclient,
7386                 .rpc_message = &msg,
7387                 .callback_ops = &nfs41_sequence_ops,
7388                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7389         };
7390
7391         if (!atomic_inc_not_zero(&clp->cl_count))
7392                 return ERR_PTR(-EIO);
7393         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7394         if (calldata == NULL) {
7395                 nfs_put_client(clp);
7396                 return ERR_PTR(-ENOMEM);
7397         }
7398         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7399         if (is_privileged)
7400                 nfs4_set_sequence_privileged(&calldata->args);
7401         msg.rpc_argp = &calldata->args;
7402         msg.rpc_resp = &calldata->res;
7403         calldata->clp = clp;
7404         task_setup_data.callback_data = calldata;
7405
7406         return rpc_run_task(&task_setup_data);
7407 }
7408
7409 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7410 {
7411         struct rpc_task *task;
7412         int ret = 0;
7413
7414         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7415                 return -EAGAIN;
7416         task = _nfs41_proc_sequence(clp, cred, false);
7417         if (IS_ERR(task))
7418                 ret = PTR_ERR(task);
7419         else
7420                 rpc_put_task_async(task);
7421         dprintk("<-- %s status=%d\n", __func__, ret);
7422         return ret;
7423 }
7424
7425 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7426 {
7427         struct rpc_task *task;
7428         int ret;
7429
7430         task = _nfs41_proc_sequence(clp, cred, true);
7431         if (IS_ERR(task)) {
7432                 ret = PTR_ERR(task);
7433                 goto out;
7434         }
7435         ret = rpc_wait_for_completion_task(task);
7436         if (!ret) {
7437                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7438
7439                 if (task->tk_status == 0)
7440                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7441                 ret = task->tk_status;
7442         }
7443         rpc_put_task(task);
7444 out:
7445         dprintk("<-- %s status=%d\n", __func__, ret);
7446         return ret;
7447 }
7448
7449 struct nfs4_reclaim_complete_data {
7450         struct nfs_client *clp;
7451         struct nfs41_reclaim_complete_args arg;
7452         struct nfs41_reclaim_complete_res res;
7453 };
7454
7455 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7456 {
7457         struct nfs4_reclaim_complete_data *calldata = data;
7458
7459         nfs41_setup_sequence(calldata->clp->cl_session,
7460                         &calldata->arg.seq_args,
7461                         &calldata->res.seq_res,
7462                         task);
7463 }
7464
7465 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7466 {
7467         switch(task->tk_status) {
7468         case 0:
7469         case -NFS4ERR_COMPLETE_ALREADY:
7470         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7471                 break;
7472         case -NFS4ERR_DELAY:
7473                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7474                 /* fall through */
7475         case -NFS4ERR_RETRY_UNCACHED_REP:
7476                 return -EAGAIN;
7477         default:
7478                 nfs4_schedule_lease_recovery(clp);
7479         }
7480         return 0;
7481 }
7482
7483 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7484 {
7485         struct nfs4_reclaim_complete_data *calldata = data;
7486         struct nfs_client *clp = calldata->clp;
7487         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7488
7489         dprintk("--> %s\n", __func__);
7490         if (!nfs41_sequence_done(task, res))
7491                 return;
7492
7493         trace_nfs4_reclaim_complete(clp, task->tk_status);
7494         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7495                 rpc_restart_call_prepare(task);
7496                 return;
7497         }
7498         dprintk("<-- %s\n", __func__);
7499 }
7500
7501 static void nfs4_free_reclaim_complete_data(void *data)
7502 {
7503         struct nfs4_reclaim_complete_data *calldata = data;
7504
7505         kfree(calldata);
7506 }
7507
7508 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7509         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7510         .rpc_call_done = nfs4_reclaim_complete_done,
7511         .rpc_release = nfs4_free_reclaim_complete_data,
7512 };
7513
7514 /*
7515  * Issue a global reclaim complete.
7516  */
7517 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7518                 struct rpc_cred *cred)
7519 {
7520         struct nfs4_reclaim_complete_data *calldata;
7521         struct rpc_task *task;
7522         struct rpc_message msg = {
7523                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7524                 .rpc_cred = cred,
7525         };
7526         struct rpc_task_setup task_setup_data = {
7527                 .rpc_client = clp->cl_rpcclient,
7528                 .rpc_message = &msg,
7529                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7530                 .flags = RPC_TASK_ASYNC,
7531         };
7532         int status = -ENOMEM;
7533
7534         dprintk("--> %s\n", __func__);
7535         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7536         if (calldata == NULL)
7537                 goto out;
7538         calldata->clp = clp;
7539         calldata->arg.one_fs = 0;
7540
7541         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7542         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7543         msg.rpc_argp = &calldata->arg;
7544         msg.rpc_resp = &calldata->res;
7545         task_setup_data.callback_data = calldata;
7546         task = rpc_run_task(&task_setup_data);
7547         if (IS_ERR(task)) {
7548                 status = PTR_ERR(task);
7549                 goto out;
7550         }
7551         status = nfs4_wait_for_completion_rpc_task(task);
7552         if (status == 0)
7553                 status = task->tk_status;
7554         rpc_put_task(task);
7555         return 0;
7556 out:
7557         dprintk("<-- %s status=%d\n", __func__, status);
7558         return status;
7559 }
7560
7561 static void
7562 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7563 {
7564         struct nfs4_layoutget *lgp = calldata;
7565         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7566         struct nfs4_session *session = nfs4_get_session(server);
7567
7568         dprintk("--> %s\n", __func__);
7569         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7570          * right now covering the LAYOUTGET we are about to send.
7571          * However, that is not so catastrophic, and there seems
7572          * to be no way to prevent it completely.
7573          */
7574         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7575                                 &lgp->res.seq_res, task))
7576                 return;
7577         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7578                                           NFS_I(lgp->args.inode)->layout,
7579                                           lgp->args.ctx->state)) {
7580                 rpc_exit(task, NFS4_OK);
7581         }
7582 }
7583
7584 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7585 {
7586         struct nfs4_layoutget *lgp = calldata;
7587         struct inode *inode = lgp->args.inode;
7588         struct nfs_server *server = NFS_SERVER(inode);
7589         struct pnfs_layout_hdr *lo;
7590         struct nfs4_state *state = NULL;
7591         unsigned long timeo, now, giveup;
7592
7593         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7594
7595         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7596                 goto out;
7597
7598         switch (task->tk_status) {
7599         case 0:
7600                 goto out;
7601         /*
7602          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7603          * (or clients) writing to the same RAID stripe
7604          */
7605         case -NFS4ERR_LAYOUTTRYLATER:
7606         /*
7607          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7608          * existing layout before getting a new one).
7609          */
7610         case -NFS4ERR_RECALLCONFLICT:
7611                 timeo = rpc_get_timeout(task->tk_client);
7612                 giveup = lgp->args.timestamp + timeo;
7613                 now = jiffies;
7614                 if (time_after(giveup, now)) {
7615                         unsigned long delay;
7616
7617                         /* Delay for:
7618                          * - Not less then NFS4_POLL_RETRY_MIN.
7619                          * - One last time a jiffie before we give up
7620                          * - exponential backoff (time_now minus start_attempt)
7621                          */
7622                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7623                                     min((giveup - now - 1),
7624                                         now - lgp->args.timestamp));
7625
7626                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7627                                 __func__, delay);
7628                         rpc_delay(task, delay);
7629                         task->tk_status = 0;
7630                         rpc_restart_call_prepare(task);
7631                         goto out; /* Do not call nfs4_async_handle_error() */
7632                 }
7633                 break;
7634         case -NFS4ERR_EXPIRED:
7635         case -NFS4ERR_BAD_STATEID:
7636                 spin_lock(&inode->i_lock);
7637                 lo = NFS_I(inode)->layout;
7638                 if (!lo || list_empty(&lo->plh_segs)) {
7639                         spin_unlock(&inode->i_lock);
7640                         /* If the open stateid was bad, then recover it. */
7641                         state = lgp->args.ctx->state;
7642                 } else {
7643                         LIST_HEAD(head);
7644
7645                         /*
7646                          * Mark the bad layout state as invalid, then retry
7647                          * with the current stateid.
7648                          */
7649                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7650                         spin_unlock(&inode->i_lock);
7651                         pnfs_free_lseg_list(&head);
7652
7653                         task->tk_status = 0;
7654                         rpc_restart_call_prepare(task);
7655                 }
7656         }
7657         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7658                 rpc_restart_call_prepare(task);
7659 out:
7660         dprintk("<-- %s\n", __func__);
7661 }
7662
7663 static size_t max_response_pages(struct nfs_server *server)
7664 {
7665         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7666         return nfs_page_array_len(0, max_resp_sz);
7667 }
7668
7669 static void nfs4_free_pages(struct page **pages, size_t size)
7670 {
7671         int i;
7672
7673         if (!pages)
7674                 return;
7675
7676         for (i = 0; i < size; i++) {
7677                 if (!pages[i])
7678                         break;
7679                 __free_page(pages[i]);
7680         }
7681         kfree(pages);
7682 }
7683
7684 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7685 {
7686         struct page **pages;
7687         int i;
7688
7689         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7690         if (!pages) {
7691                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7692                 return NULL;
7693         }
7694
7695         for (i = 0; i < size; i++) {
7696                 pages[i] = alloc_page(gfp_flags);
7697                 if (!pages[i]) {
7698                         dprintk("%s: failed to allocate page\n", __func__);
7699                         nfs4_free_pages(pages, size);
7700                         return NULL;
7701                 }
7702         }
7703
7704         return pages;
7705 }
7706
7707 static void nfs4_layoutget_release(void *calldata)
7708 {
7709         struct nfs4_layoutget *lgp = calldata;
7710         struct inode *inode = lgp->args.inode;
7711         struct nfs_server *server = NFS_SERVER(inode);
7712         size_t max_pages = max_response_pages(server);
7713
7714         dprintk("--> %s\n", __func__);
7715         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7716         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7717         put_nfs_open_context(lgp->args.ctx);
7718         kfree(calldata);
7719         dprintk("<-- %s\n", __func__);
7720 }
7721
7722 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7723         .rpc_call_prepare = nfs4_layoutget_prepare,
7724         .rpc_call_done = nfs4_layoutget_done,
7725         .rpc_release = nfs4_layoutget_release,
7726 };
7727
7728 struct pnfs_layout_segment *
7729 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7730 {
7731         struct inode *inode = lgp->args.inode;
7732         struct nfs_server *server = NFS_SERVER(inode);
7733         size_t max_pages = max_response_pages(server);
7734         struct rpc_task *task;
7735         struct rpc_message msg = {
7736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7737                 .rpc_argp = &lgp->args,
7738                 .rpc_resp = &lgp->res,
7739                 .rpc_cred = lgp->cred,
7740         };
7741         struct rpc_task_setup task_setup_data = {
7742                 .rpc_client = server->client,
7743                 .rpc_message = &msg,
7744                 .callback_ops = &nfs4_layoutget_call_ops,
7745                 .callback_data = lgp,
7746                 .flags = RPC_TASK_ASYNC,
7747         };
7748         struct pnfs_layout_segment *lseg = NULL;
7749         int status = 0;
7750
7751         dprintk("--> %s\n", __func__);
7752
7753         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7754         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7755
7756         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7757         if (!lgp->args.layout.pages) {
7758                 nfs4_layoutget_release(lgp);
7759                 return ERR_PTR(-ENOMEM);
7760         }
7761         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7762         lgp->args.timestamp = jiffies;
7763
7764         lgp->res.layoutp = &lgp->args.layout;
7765         lgp->res.seq_res.sr_slot = NULL;
7766         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7767
7768         task = rpc_run_task(&task_setup_data);
7769         if (IS_ERR(task))
7770                 return ERR_CAST(task);
7771         status = nfs4_wait_for_completion_rpc_task(task);
7772         if (status == 0)
7773                 status = task->tk_status;
7774         trace_nfs4_layoutget(lgp->args.ctx,
7775                         &lgp->args.range,
7776                         &lgp->res.range,
7777                         status);
7778         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7779         if (status == 0 && lgp->res.layoutp->len)
7780                 lseg = pnfs_layout_process(lgp);
7781         rpc_put_task(task);
7782         dprintk("<-- %s status=%d\n", __func__, status);
7783         if (status)
7784                 return ERR_PTR(status);
7785         return lseg;
7786 }
7787
7788 static void
7789 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7790 {
7791         struct nfs4_layoutreturn *lrp = calldata;
7792
7793         dprintk("--> %s\n", __func__);
7794         nfs41_setup_sequence(lrp->clp->cl_session,
7795                         &lrp->args.seq_args,
7796                         &lrp->res.seq_res,
7797                         task);
7798 }
7799
7800 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7801 {
7802         struct nfs4_layoutreturn *lrp = calldata;
7803         struct nfs_server *server;
7804
7805         dprintk("--> %s\n", __func__);
7806
7807         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7808                 return;
7809
7810         server = NFS_SERVER(lrp->args.inode);
7811         switch (task->tk_status) {
7812         default:
7813                 task->tk_status = 0;
7814         case 0:
7815                 break;
7816         case -NFS4ERR_DELAY:
7817                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7818                         break;
7819                 rpc_restart_call_prepare(task);
7820                 return;
7821         }
7822         dprintk("<-- %s\n", __func__);
7823 }
7824
7825 static void nfs4_layoutreturn_release(void *calldata)
7826 {
7827         struct nfs4_layoutreturn *lrp = calldata;
7828         struct pnfs_layout_hdr *lo = lrp->args.layout;
7829
7830         dprintk("--> %s\n", __func__);
7831         spin_lock(&lo->plh_inode->i_lock);
7832         if (lrp->res.lrs_present)
7833                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7834         lo->plh_block_lgets--;
7835         spin_unlock(&lo->plh_inode->i_lock);
7836         pnfs_put_layout_hdr(lrp->args.layout);
7837         kfree(calldata);
7838         dprintk("<-- %s\n", __func__);
7839 }
7840
7841 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7842         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7843         .rpc_call_done = nfs4_layoutreturn_done,
7844         .rpc_release = nfs4_layoutreturn_release,
7845 };
7846
7847 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7848 {
7849         struct rpc_task *task;
7850         struct rpc_message msg = {
7851                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7852                 .rpc_argp = &lrp->args,
7853                 .rpc_resp = &lrp->res,
7854                 .rpc_cred = lrp->cred,
7855         };
7856         struct rpc_task_setup task_setup_data = {
7857                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
7858                 .rpc_message = &msg,
7859                 .callback_ops = &nfs4_layoutreturn_call_ops,
7860                 .callback_data = lrp,
7861         };
7862         int status;
7863
7864         dprintk("--> %s\n", __func__);
7865         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7866         task = rpc_run_task(&task_setup_data);
7867         if (IS_ERR(task))
7868                 return PTR_ERR(task);
7869         status = task->tk_status;
7870         trace_nfs4_layoutreturn(lrp->args.inode, status);
7871         dprintk("<-- %s status=%d\n", __func__, status);
7872         rpc_put_task(task);
7873         return status;
7874 }
7875
7876 static int
7877 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7878                 struct pnfs_device *pdev,
7879                 struct rpc_cred *cred)
7880 {
7881         struct nfs4_getdeviceinfo_args args = {
7882                 .pdev = pdev,
7883         };
7884         struct nfs4_getdeviceinfo_res res = {
7885                 .pdev = pdev,
7886         };
7887         struct rpc_message msg = {
7888                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7889                 .rpc_argp = &args,
7890                 .rpc_resp = &res,
7891                 .rpc_cred = cred,
7892         };
7893         int status;
7894
7895         dprintk("--> %s\n", __func__);
7896         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7897         dprintk("<-- %s status=%d\n", __func__, status);
7898
7899         return status;
7900 }
7901
7902 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7903                 struct pnfs_device *pdev,
7904                 struct rpc_cred *cred)
7905 {
7906         struct nfs4_exception exception = { };
7907         int err;
7908
7909         do {
7910                 err = nfs4_handle_exception(server,
7911                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
7912                                         &exception);
7913         } while (exception.retry);
7914         return err;
7915 }
7916 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7917
7918 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7919 {
7920         struct nfs4_layoutcommit_data *data = calldata;
7921         struct nfs_server *server = NFS_SERVER(data->args.inode);
7922         struct nfs4_session *session = nfs4_get_session(server);
7923
7924         nfs41_setup_sequence(session,
7925                         &data->args.seq_args,
7926                         &data->res.seq_res,
7927                         task);
7928 }
7929
7930 static void
7931 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7932 {
7933         struct nfs4_layoutcommit_data *data = calldata;
7934         struct nfs_server *server = NFS_SERVER(data->args.inode);
7935
7936         if (!nfs41_sequence_done(task, &data->res.seq_res))
7937                 return;
7938
7939         switch (task->tk_status) { /* Just ignore these failures */
7940         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7941         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7942         case -NFS4ERR_BADLAYOUT:     /* no layout */
7943         case -NFS4ERR_GRACE:        /* loca_recalim always false */
7944                 task->tk_status = 0;
7945         case 0:
7946                 break;
7947         default:
7948                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
7949                         rpc_restart_call_prepare(task);
7950                         return;
7951                 }
7952         }
7953 }
7954
7955 static void nfs4_layoutcommit_release(void *calldata)
7956 {
7957         struct nfs4_layoutcommit_data *data = calldata;
7958
7959         pnfs_cleanup_layoutcommit(data);
7960         nfs_post_op_update_inode_force_wcc(data->args.inode,
7961                                            data->res.fattr);
7962         put_rpccred(data->cred);
7963         kfree(data);
7964 }
7965
7966 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7967         .rpc_call_prepare = nfs4_layoutcommit_prepare,
7968         .rpc_call_done = nfs4_layoutcommit_done,
7969         .rpc_release = nfs4_layoutcommit_release,
7970 };
7971
7972 int
7973 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7974 {
7975         struct rpc_message msg = {
7976                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7977                 .rpc_argp = &data->args,
7978                 .rpc_resp = &data->res,
7979                 .rpc_cred = data->cred,
7980         };
7981         struct rpc_task_setup task_setup_data = {
7982                 .task = &data->task,
7983                 .rpc_client = NFS_CLIENT(data->args.inode),
7984                 .rpc_message = &msg,
7985                 .callback_ops = &nfs4_layoutcommit_ops,
7986                 .callback_data = data,
7987                 .flags = RPC_TASK_ASYNC,
7988         };
7989         struct rpc_task *task;
7990         int status = 0;
7991
7992         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7993                 "lbw: %llu inode %lu\n",
7994                 data->task.tk_pid, sync,
7995                 data->args.lastbytewritten,
7996                 data->args.inode->i_ino);
7997
7998         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7999         task = rpc_run_task(&task_setup_data);
8000         if (IS_ERR(task))
8001                 return PTR_ERR(task);
8002         if (sync == false)
8003                 goto out;
8004         status = nfs4_wait_for_completion_rpc_task(task);
8005         if (status != 0)
8006                 goto out;
8007         status = task->tk_status;
8008         trace_nfs4_layoutcommit(data->args.inode, status);
8009 out:
8010         dprintk("%s: status %d\n", __func__, status);
8011         rpc_put_task(task);
8012         return status;
8013 }
8014
8015 /**
8016  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8017  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8018  */
8019 static int
8020 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8021                     struct nfs_fsinfo *info,
8022                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8023 {
8024         struct nfs41_secinfo_no_name_args args = {
8025                 .style = SECINFO_STYLE_CURRENT_FH,
8026         };
8027         struct nfs4_secinfo_res res = {
8028                 .flavors = flavors,
8029         };
8030         struct rpc_message msg = {
8031                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8032                 .rpc_argp = &args,
8033                 .rpc_resp = &res,
8034         };
8035         struct rpc_clnt *clnt = server->client;
8036         struct rpc_cred *cred = NULL;
8037         int status;
8038
8039         if (use_integrity) {
8040                 clnt = server->nfs_client->cl_rpcclient;
8041                 cred = nfs4_get_clid_cred(server->nfs_client);
8042                 msg.rpc_cred = cred;
8043         }
8044
8045         dprintk("--> %s\n", __func__);
8046         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8047                                 &res.seq_res, 0);
8048         dprintk("<-- %s status=%d\n", __func__, status);
8049
8050         if (cred)
8051                 put_rpccred(cred);
8052
8053         return status;
8054 }
8055
8056 static int
8057 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8058                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8059 {
8060         struct nfs4_exception exception = { };
8061         int err;
8062         do {
8063                 /* first try using integrity protection */
8064                 err = -NFS4ERR_WRONGSEC;
8065
8066                 /* try to use integrity protection with machine cred */
8067                 if (_nfs4_is_integrity_protected(server->nfs_client))
8068                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8069                                                           flavors, true);
8070
8071                 /*
8072                  * if unable to use integrity protection, or SECINFO with
8073                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8074                  * disallowed by spec, but exists in deployed servers) use
8075                  * the current filesystem's rpc_client and the user cred.
8076                  */
8077                 if (err == -NFS4ERR_WRONGSEC)
8078                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8079                                                           flavors, false);
8080
8081                 switch (err) {
8082                 case 0:
8083                 case -NFS4ERR_WRONGSEC:
8084                 case -ENOTSUPP:
8085                         goto out;
8086                 default:
8087                         err = nfs4_handle_exception(server, err, &exception);
8088                 }
8089         } while (exception.retry);
8090 out:
8091         return err;
8092 }
8093
8094 static int
8095 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8096                     struct nfs_fsinfo *info)
8097 {
8098         int err;
8099         struct page *page;
8100         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8101         struct nfs4_secinfo_flavors *flavors;
8102         struct nfs4_secinfo4 *secinfo;
8103         int i;
8104
8105         page = alloc_page(GFP_KERNEL);
8106         if (!page) {
8107                 err = -ENOMEM;
8108                 goto out;
8109         }
8110
8111         flavors = page_address(page);
8112         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8113
8114         /*
8115          * Fall back on "guess and check" method if
8116          * the server doesn't support SECINFO_NO_NAME
8117          */
8118         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8119                 err = nfs4_find_root_sec(server, fhandle, info);
8120                 goto out_freepage;
8121         }
8122         if (err)
8123                 goto out_freepage;
8124
8125         for (i = 0; i < flavors->num_flavors; i++) {
8126                 secinfo = &flavors->flavors[i];
8127
8128                 switch (secinfo->flavor) {
8129                 case RPC_AUTH_NULL:
8130                 case RPC_AUTH_UNIX:
8131                 case RPC_AUTH_GSS:
8132                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8133                                         &secinfo->flavor_info);
8134                         break;
8135                 default:
8136                         flavor = RPC_AUTH_MAXFLAVOR;
8137                         break;
8138                 }
8139
8140                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8141                         flavor = RPC_AUTH_MAXFLAVOR;
8142
8143                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8144                         err = nfs4_lookup_root_sec(server, fhandle,
8145                                                    info, flavor);
8146                         if (!err)
8147                                 break;
8148                 }
8149         }
8150
8151         if (flavor == RPC_AUTH_MAXFLAVOR)
8152                 err = -EPERM;
8153
8154 out_freepage:
8155         put_page(page);
8156         if (err == -EACCES)
8157                 return -EPERM;
8158 out:
8159         return err;
8160 }
8161
8162 static int _nfs41_test_stateid(struct nfs_server *server,
8163                 nfs4_stateid *stateid,
8164                 struct rpc_cred *cred)
8165 {
8166         int status;
8167         struct nfs41_test_stateid_args args = {
8168                 .stateid = stateid,
8169         };
8170         struct nfs41_test_stateid_res res;
8171         struct rpc_message msg = {
8172                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8173                 .rpc_argp = &args,
8174                 .rpc_resp = &res,
8175                 .rpc_cred = cred,
8176         };
8177         struct rpc_clnt *rpc_client = server->client;
8178
8179         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8180                 &rpc_client, &msg);
8181
8182         dprintk("NFS call  test_stateid %p\n", stateid);
8183         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8184         nfs4_set_sequence_privileged(&args.seq_args);
8185         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8186                         &args.seq_args, &res.seq_res);
8187         if (status != NFS_OK) {
8188                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8189                 return status;
8190         }
8191         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8192         return -res.status;
8193 }
8194
8195 /**
8196  * nfs41_test_stateid - perform a TEST_STATEID operation
8197  *
8198  * @server: server / transport on which to perform the operation
8199  * @stateid: state ID to test
8200  * @cred: credential
8201  *
8202  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8203  * Otherwise a negative NFS4ERR value is returned if the operation
8204  * failed or the state ID is not currently valid.
8205  */
8206 static int nfs41_test_stateid(struct nfs_server *server,
8207                 nfs4_stateid *stateid,
8208                 struct rpc_cred *cred)
8209 {
8210         struct nfs4_exception exception = { };
8211         int err;
8212         do {
8213                 err = _nfs41_test_stateid(server, stateid, cred);
8214                 if (err != -NFS4ERR_DELAY)
8215                         break;
8216                 nfs4_handle_exception(server, err, &exception);
8217         } while (exception.retry);
8218         return err;
8219 }
8220
8221 struct nfs_free_stateid_data {
8222         struct nfs_server *server;
8223         struct nfs41_free_stateid_args args;
8224         struct nfs41_free_stateid_res res;
8225 };
8226
8227 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8228 {
8229         struct nfs_free_stateid_data *data = calldata;
8230         nfs41_setup_sequence(nfs4_get_session(data->server),
8231                         &data->args.seq_args,
8232                         &data->res.seq_res,
8233                         task);
8234 }
8235
8236 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8237 {
8238         struct nfs_free_stateid_data *data = calldata;
8239
8240         nfs41_sequence_done(task, &data->res.seq_res);
8241
8242         switch (task->tk_status) {
8243         case -NFS4ERR_DELAY:
8244                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8245                         rpc_restart_call_prepare(task);
8246         }
8247 }
8248
8249 static void nfs41_free_stateid_release(void *calldata)
8250 {
8251         kfree(calldata);
8252 }
8253
8254 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8255         .rpc_call_prepare = nfs41_free_stateid_prepare,
8256         .rpc_call_done = nfs41_free_stateid_done,
8257         .rpc_release = nfs41_free_stateid_release,
8258 };
8259
8260 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8261                 nfs4_stateid *stateid,
8262                 struct rpc_cred *cred,
8263                 bool privileged)
8264 {
8265         struct rpc_message msg = {
8266                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8267                 .rpc_cred = cred,
8268         };
8269         struct rpc_task_setup task_setup = {
8270                 .rpc_client = server->client,
8271                 .rpc_message = &msg,
8272                 .callback_ops = &nfs41_free_stateid_ops,
8273                 .flags = RPC_TASK_ASYNC,
8274         };
8275         struct nfs_free_stateid_data *data;
8276
8277         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8278                 &task_setup.rpc_client, &msg);
8279
8280         dprintk("NFS call  free_stateid %p\n", stateid);
8281         data = kmalloc(sizeof(*data), GFP_NOFS);
8282         if (!data)
8283                 return ERR_PTR(-ENOMEM);
8284         data->server = server;
8285         nfs4_stateid_copy(&data->args.stateid, stateid);
8286
8287         task_setup.callback_data = data;
8288
8289         msg.rpc_argp = &data->args;
8290         msg.rpc_resp = &data->res;
8291         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8292         if (privileged)
8293                 nfs4_set_sequence_privileged(&data->args.seq_args);
8294
8295         return rpc_run_task(&task_setup);
8296 }
8297
8298 /**
8299  * nfs41_free_stateid - perform a FREE_STATEID operation
8300  *
8301  * @server: server / transport on which to perform the operation
8302  * @stateid: state ID to release
8303  * @cred: credential
8304  *
8305  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8306  * negative NFS4ERR value is returned.
8307  */
8308 static int nfs41_free_stateid(struct nfs_server *server,
8309                 nfs4_stateid *stateid,
8310                 struct rpc_cred *cred)
8311 {
8312         struct rpc_task *task;
8313         int ret;
8314
8315         task = _nfs41_free_stateid(server, stateid, cred, true);
8316         if (IS_ERR(task))
8317                 return PTR_ERR(task);
8318         ret = rpc_wait_for_completion_task(task);
8319         if (!ret)
8320                 ret = task->tk_status;
8321         rpc_put_task(task);
8322         return ret;
8323 }
8324
8325 static void
8326 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8327 {
8328         struct rpc_task *task;
8329         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8330
8331         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8332         nfs4_free_lock_state(server, lsp);
8333         if (IS_ERR(task))
8334                 return;
8335         rpc_put_task(task);
8336 }
8337
8338 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8339                 const nfs4_stateid *s2)
8340 {
8341         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8342                 return false;
8343
8344         if (s1->seqid == s2->seqid)
8345                 return true;
8346         if (s1->seqid == 0 || s2->seqid == 0)
8347                 return true;
8348
8349         return false;
8350 }
8351
8352 #endif /* CONFIG_NFS_V4_1 */
8353
8354 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8355                 const nfs4_stateid *s2)
8356 {
8357         return nfs4_stateid_match(s1, s2);
8358 }
8359
8360
8361 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8362         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8363         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8364         .recover_open   = nfs4_open_reclaim,
8365         .recover_lock   = nfs4_lock_reclaim,
8366         .establish_clid = nfs4_init_clientid,
8367         .detect_trunking = nfs40_discover_server_trunking,
8368 };
8369
8370 #if defined(CONFIG_NFS_V4_1)
8371 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8372         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8373         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8374         .recover_open   = nfs4_open_reclaim,
8375         .recover_lock   = nfs4_lock_reclaim,
8376         .establish_clid = nfs41_init_clientid,
8377         .reclaim_complete = nfs41_proc_reclaim_complete,
8378         .detect_trunking = nfs41_discover_server_trunking,
8379 };
8380 #endif /* CONFIG_NFS_V4_1 */
8381
8382 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8383         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8384         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8385         .recover_open   = nfs40_open_expired,
8386         .recover_lock   = nfs4_lock_expired,
8387         .establish_clid = nfs4_init_clientid,
8388 };
8389
8390 #if defined(CONFIG_NFS_V4_1)
8391 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8392         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8393         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8394         .recover_open   = nfs41_open_expired,
8395         .recover_lock   = nfs41_lock_expired,
8396         .establish_clid = nfs41_init_clientid,
8397 };
8398 #endif /* CONFIG_NFS_V4_1 */
8399
8400 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8401         .sched_state_renewal = nfs4_proc_async_renew,
8402         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8403         .renew_lease = nfs4_proc_renew,
8404 };
8405
8406 #if defined(CONFIG_NFS_V4_1)
8407 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8408         .sched_state_renewal = nfs41_proc_async_sequence,
8409         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8410         .renew_lease = nfs4_proc_sequence,
8411 };
8412 #endif
8413
8414 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8415         .get_locations = _nfs40_proc_get_locations,
8416         .fsid_present = _nfs40_proc_fsid_present,
8417 };
8418
8419 #if defined(CONFIG_NFS_V4_1)
8420 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8421         .get_locations = _nfs41_proc_get_locations,
8422         .fsid_present = _nfs41_proc_fsid_present,
8423 };
8424 #endif  /* CONFIG_NFS_V4_1 */
8425
8426 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8427         .minor_version = 0,
8428         .init_caps = NFS_CAP_READDIRPLUS
8429                 | NFS_CAP_ATOMIC_OPEN
8430                 | NFS_CAP_CHANGE_ATTR
8431                 | NFS_CAP_POSIX_LOCK,
8432         .init_client = nfs40_init_client,
8433         .shutdown_client = nfs40_shutdown_client,
8434         .match_stateid = nfs4_match_stateid,
8435         .find_root_sec = nfs4_find_root_sec,
8436         .free_lock_state = nfs4_release_lockowner,
8437         .alloc_seqid = nfs_alloc_seqid,
8438         .call_sync_ops = &nfs40_call_sync_ops,
8439         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8440         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8441         .state_renewal_ops = &nfs40_state_renewal_ops,
8442         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8443 };
8444
8445 #if defined(CONFIG_NFS_V4_1)
8446 static struct nfs_seqid *
8447 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8448 {
8449         return NULL;
8450 }
8451
8452 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8453         .minor_version = 1,
8454         .init_caps = NFS_CAP_READDIRPLUS
8455                 | NFS_CAP_ATOMIC_OPEN
8456                 | NFS_CAP_CHANGE_ATTR
8457                 | NFS_CAP_POSIX_LOCK
8458                 | NFS_CAP_STATEID_NFSV41
8459                 | NFS_CAP_ATOMIC_OPEN_V1,
8460         .init_client = nfs41_init_client,
8461         .shutdown_client = nfs41_shutdown_client,
8462         .match_stateid = nfs41_match_stateid,
8463         .find_root_sec = nfs41_find_root_sec,
8464         .free_lock_state = nfs41_free_lock_state,
8465         .alloc_seqid = nfs_alloc_no_seqid,
8466         .call_sync_ops = &nfs41_call_sync_ops,
8467         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8468         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8469         .state_renewal_ops = &nfs41_state_renewal_ops,
8470         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8471 };
8472 #endif
8473
8474 #if defined(CONFIG_NFS_V4_2)
8475 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8476         .minor_version = 2,
8477         .init_caps = NFS_CAP_READDIRPLUS
8478                 | NFS_CAP_ATOMIC_OPEN
8479                 | NFS_CAP_CHANGE_ATTR
8480                 | NFS_CAP_POSIX_LOCK
8481                 | NFS_CAP_STATEID_NFSV41
8482                 | NFS_CAP_ATOMIC_OPEN_V1
8483                 | NFS_CAP_ALLOCATE
8484                 | NFS_CAP_DEALLOCATE
8485                 | NFS_CAP_SEEK,
8486         .init_client = nfs41_init_client,
8487         .shutdown_client = nfs41_shutdown_client,
8488         .match_stateid = nfs41_match_stateid,
8489         .find_root_sec = nfs41_find_root_sec,
8490         .free_lock_state = nfs41_free_lock_state,
8491         .call_sync_ops = &nfs41_call_sync_ops,
8492         .alloc_seqid = nfs_alloc_no_seqid,
8493         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8494         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8495         .state_renewal_ops = &nfs41_state_renewal_ops,
8496 };
8497 #endif
8498
8499 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8500         [0] = &nfs_v4_0_minor_ops,
8501 #if defined(CONFIG_NFS_V4_1)
8502         [1] = &nfs_v4_1_minor_ops,
8503 #endif
8504 #if defined(CONFIG_NFS_V4_2)
8505         [2] = &nfs_v4_2_minor_ops,
8506 #endif
8507 };
8508
8509 static const struct inode_operations nfs4_dir_inode_operations = {
8510         .create         = nfs_create,
8511         .lookup         = nfs_lookup,
8512         .atomic_open    = nfs_atomic_open,
8513         .link           = nfs_link,
8514         .unlink         = nfs_unlink,
8515         .symlink        = nfs_symlink,
8516         .mkdir          = nfs_mkdir,
8517         .rmdir          = nfs_rmdir,
8518         .mknod          = nfs_mknod,
8519         .rename         = nfs_rename,
8520         .permission     = nfs_permission,
8521         .getattr        = nfs_getattr,
8522         .setattr        = nfs_setattr,
8523         .getxattr       = generic_getxattr,
8524         .setxattr       = generic_setxattr,
8525         .listxattr      = generic_listxattr,
8526         .removexattr    = generic_removexattr,
8527 };
8528
8529 static const struct inode_operations nfs4_file_inode_operations = {
8530         .permission     = nfs_permission,
8531         .getattr        = nfs_getattr,
8532         .setattr        = nfs_setattr,
8533         .getxattr       = generic_getxattr,
8534         .setxattr       = generic_setxattr,
8535         .listxattr      = generic_listxattr,
8536         .removexattr    = generic_removexattr,
8537 };
8538
8539 const struct nfs_rpc_ops nfs_v4_clientops = {
8540         .version        = 4,                    /* protocol version */
8541         .dentry_ops     = &nfs4_dentry_operations,
8542         .dir_inode_ops  = &nfs4_dir_inode_operations,
8543         .file_inode_ops = &nfs4_file_inode_operations,
8544         .file_ops       = &nfs4_file_operations,
8545         .getroot        = nfs4_proc_get_root,
8546         .submount       = nfs4_submount,
8547         .try_mount      = nfs4_try_mount,
8548         .getattr        = nfs4_proc_getattr,
8549         .setattr        = nfs4_proc_setattr,
8550         .lookup         = nfs4_proc_lookup,
8551         .access         = nfs4_proc_access,
8552         .readlink       = nfs4_proc_readlink,
8553         .create         = nfs4_proc_create,
8554         .remove         = nfs4_proc_remove,
8555         .unlink_setup   = nfs4_proc_unlink_setup,
8556         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8557         .unlink_done    = nfs4_proc_unlink_done,
8558         .rename_setup   = nfs4_proc_rename_setup,
8559         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8560         .rename_done    = nfs4_proc_rename_done,
8561         .link           = nfs4_proc_link,
8562         .symlink        = nfs4_proc_symlink,
8563         .mkdir          = nfs4_proc_mkdir,
8564         .rmdir          = nfs4_proc_remove,
8565         .readdir        = nfs4_proc_readdir,
8566         .mknod          = nfs4_proc_mknod,
8567         .statfs         = nfs4_proc_statfs,
8568         .fsinfo         = nfs4_proc_fsinfo,
8569         .pathconf       = nfs4_proc_pathconf,
8570         .set_capabilities = nfs4_server_capabilities,
8571         .decode_dirent  = nfs4_decode_dirent,
8572         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8573         .read_setup     = nfs4_proc_read_setup,
8574         .read_done      = nfs4_read_done,
8575         .write_setup    = nfs4_proc_write_setup,
8576         .write_done     = nfs4_write_done,
8577         .commit_setup   = nfs4_proc_commit_setup,
8578         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8579         .commit_done    = nfs4_commit_done,
8580         .lock           = nfs4_proc_lock,
8581         .clear_acl_cache = nfs4_zap_acl_attr,
8582         .close_context  = nfs4_close_context,
8583         .open_context   = nfs4_atomic_open,
8584         .have_delegation = nfs4_have_delegation,
8585         .return_delegation = nfs4_inode_return_delegation,
8586         .alloc_client   = nfs4_alloc_client,
8587         .init_client    = nfs4_init_client,
8588         .free_client    = nfs4_free_client,
8589         .create_server  = nfs4_create_server,
8590         .clone_server   = nfs_clone_server,
8591 };
8592
8593 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8594         .prefix = XATTR_NAME_NFSV4_ACL,
8595         .list   = nfs4_xattr_list_nfs4_acl,
8596         .get    = nfs4_xattr_get_nfs4_acl,
8597         .set    = nfs4_xattr_set_nfs4_acl,
8598 };
8599
8600 const struct xattr_handler *nfs4_xattr_handlers[] = {
8601         &nfs4_xattr_nfs4_acl_handler,
8602 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8603         &nfs4_xattr_nfs4_label_handler,
8604 #endif
8605         NULL
8606 };
8607
8608 /*
8609  * Local variables:
8610  *  c-basic-offset: 8
8611  * End:
8612  */