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/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.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 "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
  82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  86                             struct nfs_fattr *fattr, struct iattr *sattr,
  87                             struct nfs4_state *state, struct nfs4_label *ilabel,
  88                             struct nfs4_label *olabel);
  89 #ifdef CONFIG_NFS_V4_1
  90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  91                 struct rpc_cred *);
  92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  93                 struct rpc_cred *);
  94 #endif
  95
  96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  97 static inline struct nfs4_label *
  98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  99         struct iattr *sattr, struct nfs4_label *label)
 100 {
 101         int err;
 102
 103         if (label == NULL)
 104                 return NULL;
 105
 106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 107                 return NULL;
 108
 109         err = security_dentry_init_security(dentry, sattr->ia_mode,
 110                                 &dentry->d_name, (void **)&label->label, &label->len);
 111         if (err == 0)
 112                 return label;
 113
 114         return NULL;
 115 }
 116 static inline void
 117 nfs4_label_release_security(struct nfs4_label *label)
 118 {
 119         if (label)
 120                 security_release_secctx(label->label, label->len);
 121 }
 122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 123 {
 124         if (label)
 125                 return server->attr_bitmask;
 126
 127         return server->attr_bitmask_nl;
 128 }
 129 #else
 130 static inline struct nfs4_label *
 131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 132         struct iattr *sattr, struct nfs4_label *l)
 133 { return NULL; }
 134 static inline void
 135 nfs4_label_release_security(struct nfs4_label *label)
 136 { return; }
 137 static inline u32 *
 138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 139 { return server->attr_bitmask; }
 140 #endif
 141
 142 /* Prevent leaks of NFSv4 errors into userland */
 143 static int nfs4_map_errors(int err)
 144 {
 145         if (err >= -1000)
 146                 return err;
 147         switch (err) {
 148         case -NFS4ERR_RESOURCE:
 149         case -NFS4ERR_LAYOUTTRYLATER:
 150         case -NFS4ERR_RECALLCONFLICT:
 151                 return -EREMOTEIO;
 152         case -NFS4ERR_WRONGSEC:
 153         case -NFS4ERR_WRONG_CRED:
 154                 return -EPERM;
 155         case -NFS4ERR_BADOWNER:
 156         case -NFS4ERR_BADNAME:
 157                 return -EINVAL;
 158         case -NFS4ERR_SHARE_DENIED:
 159                 return -EACCES;
 160         case -NFS4ERR_MINOR_VERS_MISMATCH:
 161                 return -EPROTONOSUPPORT;
 162         case -NFS4ERR_FILE_OPEN:
 163                 return -EBUSY;
 164         default:
 165                 dprintk("%s could not handle NFSv4 error %d\n",
 166                                 __func__, -err);
 167                 break;
 168         }
 169         return -EIO;
 170 }
 171
 172 /*
 173  * This is our standard bitmap for GETATTR requests.
 174  */
 175 const u32 nfs4_fattr_bitmap[3] = {
 176         FATTR4_WORD0_TYPE
 177         | FATTR4_WORD0_CHANGE
 178         | FATTR4_WORD0_SIZE
 179         | FATTR4_WORD0_FSID
 180         | FATTR4_WORD0_FILEID,
 181         FATTR4_WORD1_MODE
 182         | FATTR4_WORD1_NUMLINKS
 183         | FATTR4_WORD1_OWNER
 184         | FATTR4_WORD1_OWNER_GROUP
 185         | FATTR4_WORD1_RAWDEV
 186         | FATTR4_WORD1_SPACE_USED
 187         | FATTR4_WORD1_TIME_ACCESS
 188         | FATTR4_WORD1_TIME_METADATA
 189         | FATTR4_WORD1_TIME_MODIFY
 190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 192         FATTR4_WORD2_SECURITY_LABEL
 193 #endif
 194 };
 195
 196 static const u32 nfs4_pnfs_open_bitmap[3] = {
 197         FATTR4_WORD0_TYPE
 198         | FATTR4_WORD0_CHANGE
 199         | FATTR4_WORD0_SIZE
 200         | FATTR4_WORD0_FSID
 201         | FATTR4_WORD0_FILEID,
 202         FATTR4_WORD1_MODE
 203         | FATTR4_WORD1_NUMLINKS
 204         | FATTR4_WORD1_OWNER
 205         | FATTR4_WORD1_OWNER_GROUP
 206         | FATTR4_WORD1_RAWDEV
 207         | FATTR4_WORD1_SPACE_USED
 208         | FATTR4_WORD1_TIME_ACCESS
 209         | FATTR4_WORD1_TIME_METADATA
 210         | FATTR4_WORD1_TIME_MODIFY,
 211         FATTR4_WORD2_MDSTHRESHOLD
 212 };
 213
 214 static const u32 nfs4_open_noattr_bitmap[3] = {
 215         FATTR4_WORD0_TYPE
 216         | FATTR4_WORD0_CHANGE
 217         | FATTR4_WORD0_FILEID,
 218 };
 219
 220 const u32 nfs4_statfs_bitmap[3] = {
 221         FATTR4_WORD0_FILES_AVAIL
 222         | FATTR4_WORD0_FILES_FREE
 223         | FATTR4_WORD0_FILES_TOTAL,
 224         FATTR4_WORD1_SPACE_AVAIL
 225         | FATTR4_WORD1_SPACE_FREE
 226         | FATTR4_WORD1_SPACE_TOTAL
 227 };
 228
 229 const u32 nfs4_pathconf_bitmap[3] = {
 230         FATTR4_WORD0_MAXLINK
 231         | FATTR4_WORD0_MAXNAME,
 232         0
 233 };
 234
 235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 236                         | FATTR4_WORD0_MAXREAD
 237                         | FATTR4_WORD0_MAXWRITE
 238                         | FATTR4_WORD0_LEASE_TIME,
 239                         FATTR4_WORD1_TIME_DELTA
 240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 241                         FATTR4_WORD2_LAYOUT_BLKSIZE
 242 };
 243
 244 const u32 nfs4_fs_locations_bitmap[3] = {
 245         FATTR4_WORD0_TYPE
 246         | FATTR4_WORD0_CHANGE
 247         | FATTR4_WORD0_SIZE
 248         | FATTR4_WORD0_FSID
 249         | FATTR4_WORD0_FILEID
 250         | FATTR4_WORD0_FS_LOCATIONS,
 251         FATTR4_WORD1_MODE
 252         | FATTR4_WORD1_NUMLINKS
 253         | FATTR4_WORD1_OWNER
 254         | FATTR4_WORD1_OWNER_GROUP
 255         | FATTR4_WORD1_RAWDEV
 256         | FATTR4_WORD1_SPACE_USED
 257         | FATTR4_WORD1_TIME_ACCESS
 258         | FATTR4_WORD1_TIME_METADATA
 259         | FATTR4_WORD1_TIME_MODIFY
 260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 261 };
 262
 263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 264                 struct nfs4_readdir_arg *readdir)
 265 {
 266         __be32 *start, *p;
 267
 268         if (cookie > 2) {
 269                 readdir->cookie = cookie;
 270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 271                 return;
 272         }
 273
 274         readdir->cookie = 0;
 275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 276         if (cookie == 2)
 277                 return;
 278
 279         /*
 280          * NFSv4 servers do not return entries for '.' and '..'
 281          * Therefore, we fake these entries here.  We let '.'
 282          * have cookie 0 and '..' have cookie 1.  Note that
 283          * when talking to the server, we always send cookie 0
 284          * instead of 1 or 2.
 285          */
 286         start = p = kmap_atomic(*readdir->pages);
 287
 288         if (cookie == 0) {
 289                 *p++ = xdr_one;                                  /* next */
 290                 *p++ = xdr_zero;                   /* cookie, first word */
 291                 *p++ = xdr_one;                   /* cookie, second word */
 292                 *p++ = xdr_one;                             /* entry len */
 293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 294                 p++;
 295                 *p++ = xdr_one;                         /* bitmap length */
 296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 297                 *p++ = htonl(8);              /* attribute buffer length */
 298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 299         }
 300
 301         *p++ = xdr_one;                                  /* next */
 302         *p++ = xdr_zero;                   /* cookie, first word */
 303         *p++ = xdr_two;                   /* cookie, second word */
 304         *p++ = xdr_two;                             /* entry len */
 305         memcpy(p, "..\0\0", 4);                         /* entry */
 306         p++;
 307         *p++ = xdr_one;                         /* bitmap length */
 308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309         *p++ = htonl(8);              /* attribute buffer length */
 310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 311
 312         readdir->pgbase = (char *)p - (char *)start;
 313         readdir->count -= readdir->pgbase;
 314         kunmap_atomic(start);
 315 }
 316
 317 static long nfs4_update_delay(long *timeout)
 318 {
 319         long ret;
 320         if (!timeout)
 321                 return NFS4_POLL_RETRY_MAX;
 322         if (*timeout <= 0)
 323                 *timeout = NFS4_POLL_RETRY_MIN;
 324         if (*timeout > NFS4_POLL_RETRY_MAX)
 325                 *timeout = NFS4_POLL_RETRY_MAX;
 326         ret = *timeout;
 327         *timeout <<= 1;
 328         return ret;
 329 }
 330
 331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 332 {
 333         int res = 0;
 334
 335         might_sleep();
 336
 337         freezable_schedule_timeout_killable_unsafe(
 338                 nfs4_update_delay(timeout));
 339         if (fatal_signal_pending(current))
 340                 res = -ERESTARTSYS;
 341         return res;
 342 }
 343
 344 /* This is the error handling routine for processes that are allowed
 345  * to sleep.
 346  */
 347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 348 {
 349         struct nfs_client *clp = server->nfs_client;
 350         struct nfs4_state *state = exception->state;
 351         struct inode *inode = exception->inode;
 352         int ret = errorcode;
 353
 354         exception->retry = 0;
 355         switch(errorcode) {
 356                 case 0:
 357                         return 0;
 358                 case -NFS4ERR_OPENMODE:
 359                 case -NFS4ERR_DELEG_REVOKED:
 360                 case -NFS4ERR_ADMIN_REVOKED:
 361                 case -NFS4ERR_BAD_STATEID:
 362                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 363                                 nfs4_inode_return_delegation(inode);
 364                                 exception->retry = 1;
 365                                 return 0;
 366                         }
 367                         if (state == NULL)
 368                                 break;
 369                         ret = nfs4_schedule_stateid_recovery(server, state);
 370                         if (ret < 0)
 371                                 break;
 372                         goto wait_on_recovery;
 373                 case -NFS4ERR_EXPIRED:
 374                         if (state != NULL) {
 375                                 ret = nfs4_schedule_stateid_recovery(server, state);
 376                                 if (ret < 0)
 377                                         break;
 378                         }
 379                 case -NFS4ERR_STALE_STATEID:
 380                 case -NFS4ERR_STALE_CLIENTID:
 381                         nfs4_schedule_lease_recovery(clp);
 382                         goto wait_on_recovery;
 383                 case -NFS4ERR_MOVED:
 384                         ret = nfs4_schedule_migration_recovery(server);
 385                         if (ret < 0)
 386                                 break;
 387                         goto wait_on_recovery;
 388                 case -NFS4ERR_LEASE_MOVED:
 389                         nfs4_schedule_lease_moved_recovery(clp);
 390                         goto wait_on_recovery;
 391 #if defined(CONFIG_NFS_V4_1)
 392                 case -NFS4ERR_BADSESSION:
 393                 case -NFS4ERR_BADSLOT:
 394                 case -NFS4ERR_BAD_HIGH_SLOT:
 395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 396                 case -NFS4ERR_DEADSESSION:
 397                 case -NFS4ERR_SEQ_FALSE_RETRY:
 398                 case -NFS4ERR_SEQ_MISORDERED:
 399                         dprintk("%s ERROR: %d Reset session\n", __func__,
 400                                 errorcode);
 401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 402                         goto wait_on_recovery;
 403 #endif /* defined(CONFIG_NFS_V4_1) */
 404                 case -NFS4ERR_FILE_OPEN:
 405                         if (exception->timeout > HZ) {
 406                                 /* We have retried a decent amount, time to
 407                                  * fail
 408                                  */
 409                                 ret = -EBUSY;
 410                                 break;
 411                         }
 412                 case -NFS4ERR_GRACE:
 413                 case -NFS4ERR_DELAY:
 414                         ret = nfs4_delay(server->client, &exception->timeout);
 415                         if (ret != 0)
 416                                 break;
 417                 case -NFS4ERR_RETRY_UNCACHED_REP:
 418                 case -NFS4ERR_OLD_STATEID:
 419                         exception->retry = 1;
 420                         break;
 421                 case -NFS4ERR_BADOWNER:
 422                         /* The following works around a Linux server bug! */
 423                 case -NFS4ERR_BADNAME:
 424                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 425                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 426                                 exception->retry = 1;
 427                                 printk(KERN_WARNING "NFS: v4 server %s "
 428                                                 "does not accept raw "
 429                                                 "uid/gids. "
 430                                                 "Reenabling the idmapper.\n",
 431                                                 server->nfs_client->cl_hostname);
 432                         }
 433         }
 434         /* We failed to handle the error */
 435         return nfs4_map_errors(ret);
 436 wait_on_recovery:
 437         ret = nfs4_wait_clnt_recover(clp);
 438         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 439                 return -EIO;
 440         if (ret == 0)
 441                 exception->retry = 1;
 442         return ret;
 443 }
 444
 445 /*
 446  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 447  * or 'false' otherwise.
 448  */
 449 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 450 {
 451         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 452
 453         if (flavor == RPC_AUTH_GSS_KRB5I ||
 454             flavor == RPC_AUTH_GSS_KRB5P)
 455                 return true;
 456
 457         return false;
 458 }
 459
 460 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 461 {
 462         spin_lock(&clp->cl_lock);
 463         if (time_before(clp->cl_last_renewal,timestamp))
 464                 clp->cl_last_renewal = timestamp;
 465         spin_unlock(&clp->cl_lock);
 466 }
 467
 468 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 469 {
 470         do_renew_lease(server->nfs_client, timestamp);
 471 }
 472
 473 struct nfs4_call_sync_data {
 474         const struct nfs_server *seq_server;
 475         struct nfs4_sequence_args *seq_args;
 476         struct nfs4_sequence_res *seq_res;
 477 };
 478
 479 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 480                         struct nfs4_sequence_res *res, int cache_reply)
 481 {
 482         args->sa_slot = NULL;
 483         args->sa_cache_this = cache_reply;
 484         args->sa_privileged = 0;
 485
 486         res->sr_slot = NULL;
 487 }
 488
 489 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 490 {
 491         args->sa_privileged = 1;
 492 }
 493
 494 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 495                          struct nfs4_sequence_args *args,
 496                          struct nfs4_sequence_res *res,
 497                          struct rpc_task *task)
 498 {
 499         struct nfs4_slot *slot;
 500
 501         /* slot already allocated? */
 502         if (res->sr_slot != NULL)
 503                 goto out_start;
 504
 505         spin_lock(&tbl->slot_tbl_lock);
 506         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 507                 goto out_sleep;
 508
 509         slot = nfs4_alloc_slot(tbl);
 510         if (IS_ERR(slot)) {
 511                 if (slot == ERR_PTR(-ENOMEM))
 512                         task->tk_timeout = HZ >> 2;
 513                 goto out_sleep;
 514         }
 515         spin_unlock(&tbl->slot_tbl_lock);
 516
 517         args->sa_slot = slot;
 518         res->sr_slot = slot;
 519
 520 out_start:
 521         rpc_call_start(task);
 522         return 0;
 523
 524 out_sleep:
 525         if (args->sa_privileged)
 526                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 527                                 NULL, RPC_PRIORITY_PRIVILEGED);
 528         else
 529                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 530         spin_unlock(&tbl->slot_tbl_lock);
 531         return -EAGAIN;
 532 }
 533 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 534
 535 static int nfs40_sequence_done(struct rpc_task *task,
 536                                struct nfs4_sequence_res *res)
 537 {
 538         struct nfs4_slot *slot = res->sr_slot;
 539         struct nfs4_slot_table *tbl;
 540
 541         if (slot == NULL)
 542                 goto out;
 543
 544         tbl = slot->table;
 545         spin_lock(&tbl->slot_tbl_lock);
 546         if (!nfs41_wake_and_assign_slot(tbl, slot))
 547                 nfs4_free_slot(tbl, slot);
 548         spin_unlock(&tbl->slot_tbl_lock);
 549
 550         res->sr_slot = NULL;
 551 out:
 552         return 1;
 553 }
 554
 555 #if defined(CONFIG_NFS_V4_1)
 556
 557 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 558 {
 559         struct nfs4_session *session;
 560         struct nfs4_slot_table *tbl;
 561         struct nfs4_slot *slot = res->sr_slot;
 562         bool send_new_highest_used_slotid = false;
 563
 564         tbl = slot->table;
 565         session = tbl->session;
 566
 567         spin_lock(&tbl->slot_tbl_lock);
 568         /* Be nice to the server: try to ensure that the last transmitted
 569          * value for highest_user_slotid <= target_highest_slotid
 570          */
 571         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 572                 send_new_highest_used_slotid = true;
 573
 574         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 575                 send_new_highest_used_slotid = false;
 576                 goto out_unlock;
 577         }
 578         nfs4_free_slot(tbl, slot);
 579
 580         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 581                 send_new_highest_used_slotid = false;
 582 out_unlock:
 583         spin_unlock(&tbl->slot_tbl_lock);
 584         res->sr_slot = NULL;
 585         if (send_new_highest_used_slotid)
 586                 nfs41_server_notify_highest_slotid_update(session->clp);
 587 }
 588
 589 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 590 {
 591         struct nfs4_session *session;
 592         struct nfs4_slot *slot = res->sr_slot;
 593         struct nfs_client *clp;
 594         bool interrupted = false;
 595         int ret = 1;
 596
 597         if (slot == NULL)
 598                 goto out_noaction;
 599         /* don't increment the sequence number if the task wasn't sent */
 600         if (!RPC_WAS_SENT(task))
 601                 goto out;
 602
 603         session = slot->table->session;
 604
 605         if (slot->interrupted) {
 606                 slot->interrupted = 0;
 607                 interrupted = true;
 608         }
 609
 610         trace_nfs4_sequence_done(session, res);
 611         /* Check the SEQUENCE operation status */
 612         switch (res->sr_status) {
 613         case 0:
 614                 /* Update the slot's sequence and clientid lease timer */
 615                 ++slot->seq_nr;
 616                 clp = session->clp;
 617                 do_renew_lease(clp, res->sr_timestamp);
 618                 /* Check sequence flags */
 619                 if (res->sr_status_flags != 0)
 620                         nfs4_schedule_lease_recovery(clp);
 621                 nfs41_update_target_slotid(slot->table, slot, res);
 622                 break;
 623         case 1:
 624                 /*
 625                  * sr_status remains 1 if an RPC level error occurred.
 626                  * The server may or may not have processed the sequence
 627                  * operation..
 628                  * Mark the slot as having hosted an interrupted RPC call.
 629                  */
 630                 slot->interrupted = 1;
 631                 goto out;
 632         case -NFS4ERR_DELAY:
 633                 /* The server detected a resend of the RPC call and
 634                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 635                  * of RFC5661.
 636                  */
 637                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 638                         __func__,
 639                         slot->slot_nr,
 640                         slot->seq_nr);
 641                 goto out_retry;
 642         case -NFS4ERR_BADSLOT:
 643                 /*
 644                  * The slot id we used was probably retired. Try again
 645                  * using a different slot id.
 646                  */
 647                 goto retry_nowait;
 648         case -NFS4ERR_SEQ_MISORDERED:
 649                 /*
 650                  * Was the last operation on this sequence interrupted?
 651                  * If so, retry after bumping the sequence number.
 652                  */
 653                 if (interrupted) {
 654                         ++slot->seq_nr;
 655                         goto retry_nowait;
 656                 }
 657                 /*
 658                  * Could this slot have been previously retired?
 659                  * If so, then the server may be expecting seq_nr = 1!
 660                  */
 661                 if (slot->seq_nr != 1) {
 662                         slot->seq_nr = 1;
 663                         goto retry_nowait;
 664                 }
 665                 break;
 666         case -NFS4ERR_SEQ_FALSE_RETRY:
 667                 ++slot->seq_nr;
 668                 goto retry_nowait;
 669         default:
 670                 /* Just update the slot sequence no. */
 671                 ++slot->seq_nr;
 672         }
 673 out:
 674         /* The session may be reset by one of the error handlers. */
 675         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 676         nfs41_sequence_free_slot(res);
 677 out_noaction:
 678         return ret;
 679 retry_nowait:
 680         if (rpc_restart_call_prepare(task)) {
 681                 task->tk_status = 0;
 682                 ret = 0;
 683         }
 684         goto out;
 685 out_retry:
 686         if (!rpc_restart_call(task))
 687                 goto out;
 688         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 689         return 0;
 690 }
 691 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 692
 693 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 694 {
 695         if (res->sr_slot == NULL)
 696                 return 1;
 697         if (!res->sr_slot->table->session)
 698                 return nfs40_sequence_done(task, res);
 699         return nfs41_sequence_done(task, res);
 700 }
 701 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 702
 703 int nfs41_setup_sequence(struct nfs4_session *session,
 704                                 struct nfs4_sequence_args *args,
 705                                 struct nfs4_sequence_res *res,
 706                                 struct rpc_task *task)
 707 {
 708         struct nfs4_slot *slot;
 709         struct nfs4_slot_table *tbl;
 710
 711         dprintk("--> %s\n", __func__);
 712         /* slot already allocated? */
 713         if (res->sr_slot != NULL)
 714                 goto out_success;
 715
 716         tbl = &session->fc_slot_table;
 717
 718         task->tk_timeout = 0;
 719
 720         spin_lock(&tbl->slot_tbl_lock);
 721         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 722             !args->sa_privileged) {
 723                 /* The state manager will wait until the slot table is empty */
 724                 dprintk("%s session is draining\n", __func__);
 725                 goto out_sleep;
 726         }
 727
 728         slot = nfs4_alloc_slot(tbl);
 729         if (IS_ERR(slot)) {
 730                 /* If out of memory, try again in 1/4 second */
 731                 if (slot == ERR_PTR(-ENOMEM))
 732                         task->tk_timeout = HZ >> 2;
 733                 dprintk("<-- %s: no free slots\n", __func__);
 734                 goto out_sleep;
 735         }
 736         spin_unlock(&tbl->slot_tbl_lock);
 737
 738         args->sa_slot = slot;
 739
 740         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 741                         slot->slot_nr, slot->seq_nr);
 742
 743         res->sr_slot = slot;
 744         res->sr_timestamp = jiffies;
 745         res->sr_status_flags = 0;
 746         /*
 747          * sr_status is only set in decode_sequence, and so will remain
 748          * set to 1 if an rpc level failure occurs.
 749          */
 750         res->sr_status = 1;
 751         trace_nfs4_setup_sequence(session, args);
 752 out_success:
 753         rpc_call_start(task);
 754         return 0;
 755 out_sleep:
 756         /* Privileged tasks are queued with top priority */
 757         if (args->sa_privileged)
 758                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 759                                 NULL, RPC_PRIORITY_PRIVILEGED);
 760         else
 761                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 762         spin_unlock(&tbl->slot_tbl_lock);
 763         return -EAGAIN;
 764 }
 765 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 766
 767 static int nfs4_setup_sequence(const struct nfs_server *server,
 768                                struct nfs4_sequence_args *args,
 769                                struct nfs4_sequence_res *res,
 770                                struct rpc_task *task)
 771 {
 772         struct nfs4_session *session = nfs4_get_session(server);
 773         int ret = 0;
 774
 775         if (!session)
 776                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 777                                             args, res, task);
 778
 779         dprintk("--> %s clp %p session %p sr_slot %u\n",
 780                 __func__, session->clp, session, res->sr_slot ?
 781                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 782
 783         ret = nfs41_setup_sequence(session, args, res, task);
 784
 785         dprintk("<-- %s status=%d\n", __func__, ret);
 786         return ret;
 787 }
 788
 789 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 790 {
 791         struct nfs4_call_sync_data *data = calldata;
 792         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 793
 794         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 795
 796         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 797 }
 798
 799 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 800 {
 801         struct nfs4_call_sync_data *data = calldata;
 802
 803         nfs41_sequence_done(task, data->seq_res);
 804 }
 805
 806 static const struct rpc_call_ops nfs41_call_sync_ops = {
 807         .rpc_call_prepare = nfs41_call_sync_prepare,
 808         .rpc_call_done = nfs41_call_sync_done,
 809 };
 810
 811 #else   /* !CONFIG_NFS_V4_1 */
 812
 813 static int nfs4_setup_sequence(const struct nfs_server *server,
 814                                struct nfs4_sequence_args *args,
 815                                struct nfs4_sequence_res *res,
 816                                struct rpc_task *task)
 817 {
 818         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 819                                     args, res, task);
 820 }
 821
 822 int nfs4_sequence_done(struct rpc_task *task,
 823                        struct nfs4_sequence_res *res)
 824 {
 825         return nfs40_sequence_done(task, res);
 826 }
 827 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 828
 829 #endif  /* !CONFIG_NFS_V4_1 */
 830
 831 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 832 {
 833         struct nfs4_call_sync_data *data = calldata;
 834         nfs4_setup_sequence(data->seq_server,
 835                                 data->seq_args, data->seq_res, task);
 836 }
 837
 838 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 839 {
 840         struct nfs4_call_sync_data *data = calldata;
 841         nfs4_sequence_done(task, data->seq_res);
 842 }
 843
 844 static const struct rpc_call_ops nfs40_call_sync_ops = {
 845         .rpc_call_prepare = nfs40_call_sync_prepare,
 846         .rpc_call_done = nfs40_call_sync_done,
 847 };
 848
 849 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 850                                    struct nfs_server *server,
 851                                    struct rpc_message *msg,
 852                                    struct nfs4_sequence_args *args,
 853                                    struct nfs4_sequence_res *res)
 854 {
 855         int ret;
 856         struct rpc_task *task;
 857         struct nfs_client *clp = server->nfs_client;
 858         struct nfs4_call_sync_data data = {
 859                 .seq_server = server,
 860                 .seq_args = args,
 861                 .seq_res = res,
 862         };
 863         struct rpc_task_setup task_setup = {
 864                 .rpc_client = clnt,
 865                 .rpc_message = msg,
 866                 .callback_ops = clp->cl_mvops->call_sync_ops,
 867                 .callback_data = &data
 868         };
 869
 870         task = rpc_run_task(&task_setup);
 871         if (IS_ERR(task))
 872                 ret = PTR_ERR(task);
 873         else {
 874                 ret = task->tk_status;
 875                 rpc_put_task(task);
 876         }
 877         return ret;
 878 }
 879
 880 int nfs4_call_sync(struct rpc_clnt *clnt,
 881                    struct nfs_server *server,
 882                    struct rpc_message *msg,
 883                    struct nfs4_sequence_args *args,
 884                    struct nfs4_sequence_res *res,
 885                    int cache_reply)
 886 {
 887         nfs4_init_sequence(args, res, cache_reply);
 888         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 889 }
 890
 891 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 892 {
 893         struct nfs_inode *nfsi = NFS_I(dir);
 894
 895         spin_lock(&dir->i_lock);
 896         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 897         if (!cinfo->atomic || cinfo->before != dir->i_version)
 898                 nfs_force_lookup_revalidate(dir);
 899         dir->i_version = cinfo->after;
 900         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 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 u32
 941 nfs4_map_atomic_open_share(struct nfs_server *server,
 942                 fmode_t fmode, int openflags)
 943 {
 944         u32 res = 0;
 945
 946         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
 947         case FMODE_READ:
 948                 res = NFS4_SHARE_ACCESS_READ;
 949                 break;
 950         case FMODE_WRITE:
 951                 res = NFS4_SHARE_ACCESS_WRITE;
 952                 break;
 953         case FMODE_READ|FMODE_WRITE:
 954                 res = NFS4_SHARE_ACCESS_BOTH;
 955         }
 956         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 957                 goto out;
 958         /* Want no delegation if we're using O_DIRECT */
 959         if (openflags & O_DIRECT)
 960                 res |= NFS4_SHARE_WANT_NO_DELEG;
 961 out:
 962         return res;
 963 }
 964
 965 static enum open_claim_type4
 966 nfs4_map_atomic_open_claim(struct nfs_server *server,
 967                 enum open_claim_type4 claim)
 968 {
 969         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 970                 return claim;
 971         switch (claim) {
 972         default:
 973                 return claim;
 974         case NFS4_OPEN_CLAIM_FH:
 975                 return NFS4_OPEN_CLAIM_NULL;
 976         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 977                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 978         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 979                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 980         }
 981 }
 982
 983 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 984 {
 985         p->o_res.f_attr = &p->f_attr;
 986         p->o_res.f_label = p->f_label;
 987         p->o_res.seqid = p->o_arg.seqid;
 988         p->c_res.seqid = p->c_arg.seqid;
 989         p->o_res.server = p->o_arg.server;
 990         p->o_res.access_request = p->o_arg.access;
 991         nfs_fattr_init(&p->f_attr);
 992         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 993 }
 994
 995 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 996                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 997                 const struct iattr *attrs,
 998                 struct nfs4_label *label,
 999                 enum open_claim_type4 claim,
1000                 gfp_t gfp_mask)
1001 {
1002         struct dentry *parent = dget_parent(dentry);
1003         struct inode *dir = d_inode(parent);
1004         struct nfs_server *server = NFS_SERVER(dir);
1005         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1006         struct nfs4_opendata *p;
1007
1008         p = kzalloc(sizeof(*p), gfp_mask);
1009         if (p == NULL)
1010                 goto err;
1011
1012         p->f_label = nfs4_label_alloc(server, gfp_mask);
1013         if (IS_ERR(p->f_label))
1014                 goto err_free_p;
1015
1016         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1017         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1018         if (IS_ERR(p->o_arg.seqid))
1019                 goto err_free_label;
1020         nfs_sb_active(dentry->d_sb);
1021         p->dentry = dget(dentry);
1022         p->dir = parent;
1023         p->owner = sp;
1024         atomic_inc(&sp->so_count);
1025         p->o_arg.open_flags = flags;
1026         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1027         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1028                         fmode, flags);
1029         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1030          * will return permission denied for all bits until close */
1031         if (!(flags & O_EXCL)) {
1032                 /* ask server to check for all possible rights as results
1033                  * are cached */
1034                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1035                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1036         }
1037         p->o_arg.clientid = server->nfs_client->cl_clientid;
1038         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1039         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1040         p->o_arg.name = &dentry->d_name;
1041         p->o_arg.server = server;
1042         p->o_arg.bitmask = nfs4_bitmask(server, label);
1043         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1044         p->o_arg.label = label;
1045         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1046         switch (p->o_arg.claim) {
1047         case NFS4_OPEN_CLAIM_NULL:
1048         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1049         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1050                 p->o_arg.fh = NFS_FH(dir);
1051                 break;
1052         case NFS4_OPEN_CLAIM_PREVIOUS:
1053         case NFS4_OPEN_CLAIM_FH:
1054         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1055         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1056                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1057         }
1058         if (attrs != NULL && attrs->ia_valid != 0) {
1059                 __u32 verf[2];
1060
1061                 p->o_arg.u.attrs = &p->attrs;
1062                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1063
1064                 verf[0] = jiffies;
1065                 verf[1] = current->pid;
1066                 memcpy(p->o_arg.u.verifier.data, verf,
1067                                 sizeof(p->o_arg.u.verifier.data));
1068         }
1069         p->c_arg.fh = &p->o_res.fh;
1070         p->c_arg.stateid = &p->o_res.stateid;
1071         p->c_arg.seqid = p->o_arg.seqid;
1072         nfs4_init_opendata_res(p);
1073         kref_init(&p->kref);
1074         return p;
1075
1076 err_free_label:
1077         nfs4_label_free(p->f_label);
1078 err_free_p:
1079         kfree(p);
1080 err:
1081         dput(parent);
1082         return NULL;
1083 }
1084
1085 static void nfs4_opendata_free(struct kref *kref)
1086 {
1087         struct nfs4_opendata *p = container_of(kref,
1088                         struct nfs4_opendata, kref);
1089         struct super_block *sb = p->dentry->d_sb;
1090
1091         nfs_free_seqid(p->o_arg.seqid);
1092         if (p->state != NULL)
1093                 nfs4_put_open_state(p->state);
1094         nfs4_put_state_owner(p->owner);
1095
1096         nfs4_label_free(p->f_label);
1097
1098         dput(p->dir);
1099         dput(p->dentry);
1100         nfs_sb_deactive(sb);
1101         nfs_fattr_free_names(&p->f_attr);
1102         kfree(p->f_attr.mdsthreshold);
1103         kfree(p);
1104 }
1105
1106 static void nfs4_opendata_put(struct nfs4_opendata *p)
1107 {
1108         if (p != NULL)
1109                 kref_put(&p->kref, nfs4_opendata_free);
1110 }
1111
1112 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1113 {
1114         int ret;
1115
1116         ret = rpc_wait_for_completion_task(task);
1117         return ret;
1118 }
1119
1120 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1121 {
1122         int ret = 0;
1123
1124         if (open_mode & (O_EXCL|O_TRUNC))
1125                 goto out;
1126         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1127                 case FMODE_READ:
1128                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1129                                 && state->n_rdonly != 0;
1130                         break;
1131                 case FMODE_WRITE:
1132                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1133                                 && state->n_wronly != 0;
1134                         break;
1135                 case FMODE_READ|FMODE_WRITE:
1136                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1137                                 && state->n_rdwr != 0;
1138         }
1139 out:
1140         return ret;
1141 }
1142
1143 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1144 {
1145         if (delegation == NULL)
1146                 return 0;
1147         if ((delegation->type & fmode) != fmode)
1148                 return 0;
1149         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1150                 return 0;
1151         nfs_mark_delegation_referenced(delegation);
1152         return 1;
1153 }
1154
1155 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1156 {
1157         switch (fmode) {
1158                 case FMODE_WRITE:
1159                         state->n_wronly++;
1160                         break;
1161                 case FMODE_READ:
1162                         state->n_rdonly++;
1163                         break;
1164                 case FMODE_READ|FMODE_WRITE:
1165                         state->n_rdwr++;
1166         }
1167         nfs4_state_set_mode_locked(state, state->state | fmode);
1168 }
1169
1170 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1171 {
1172         struct nfs_client *clp = state->owner->so_server->nfs_client;
1173         bool need_recover = false;
1174
1175         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1176                 need_recover = true;
1177         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1178                 need_recover = true;
1179         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1180                 need_recover = true;
1181         if (need_recover)
1182                 nfs4_state_mark_reclaim_nograce(clp, state);
1183 }
1184
1185 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1186                 nfs4_stateid *stateid)
1187 {
1188         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1189                 return true;
1190         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1191                 nfs_test_and_clear_all_open_stateid(state);
1192                 return true;
1193         }
1194         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1195                 return true;
1196         return false;
1197 }
1198
1199 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1200 {
1201         if (state->n_wronly)
1202                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1203         if (state->n_rdonly)
1204                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1205         if (state->n_rdwr)
1206                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1207 }
1208
1209 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1210                 nfs4_stateid *stateid, fmode_t fmode)
1211 {
1212         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1213         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1214         case FMODE_WRITE:
1215                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1216                 break;
1217         case FMODE_READ:
1218                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1219                 break;
1220         case 0:
1221                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1222                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1223                 clear_bit(NFS_OPEN_STATE, &state->flags);
1224         }
1225         if (stateid == NULL)
1226                 return;
1227         /* Handle races with OPEN */
1228         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1229             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1230                 nfs_resync_open_stateid_locked(state);
1231                 return;
1232         }
1233         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1234                 nfs4_stateid_copy(&state->stateid, stateid);
1235         nfs4_stateid_copy(&state->open_stateid, stateid);
1236 }
1237
1238 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1239 {
1240         write_seqlock(&state->seqlock);
1241         nfs_clear_open_stateid_locked(state, stateid, fmode);
1242         write_sequnlock(&state->seqlock);
1243         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1244                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1245 }
1246
1247 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1248 {
1249         switch (fmode) {
1250                 case FMODE_READ:
1251                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1252                         break;
1253                 case FMODE_WRITE:
1254                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1255                         break;
1256                 case FMODE_READ|FMODE_WRITE:
1257                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1258         }
1259         if (!nfs_need_update_open_stateid(state, stateid))
1260                 return;
1261         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1262                 nfs4_stateid_copy(&state->stateid, stateid);
1263         nfs4_stateid_copy(&state->open_stateid, stateid);
1264 }
1265
1266 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1267 {
1268         /*
1269          * Protect the call to nfs4_state_set_mode_locked and
1270          * serialise the stateid update
1271          */
1272         write_seqlock(&state->seqlock);
1273         if (deleg_stateid != NULL) {
1274                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1275                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1276         }
1277         if (open_stateid != NULL)
1278                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1279         write_sequnlock(&state->seqlock);
1280         spin_lock(&state->owner->so_lock);
1281         update_open_stateflags(state, fmode);
1282         spin_unlock(&state->owner->so_lock);
1283 }
1284
1285 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1286 {
1287         struct nfs_inode *nfsi = NFS_I(state->inode);
1288         struct nfs_delegation *deleg_cur;
1289         int ret = 0;
1290
1291         fmode &= (FMODE_READ|FMODE_WRITE);
1292
1293         rcu_read_lock();
1294         deleg_cur = rcu_dereference(nfsi->delegation);
1295         if (deleg_cur == NULL)
1296                 goto no_delegation;
1297
1298         spin_lock(&deleg_cur->lock);
1299         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1300            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1301             (deleg_cur->type & fmode) != fmode)
1302                 goto no_delegation_unlock;
1303
1304         if (delegation == NULL)
1305                 delegation = &deleg_cur->stateid;
1306         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1307                 goto no_delegation_unlock;
1308
1309         nfs_mark_delegation_referenced(deleg_cur);
1310         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1311         ret = 1;
1312 no_delegation_unlock:
1313         spin_unlock(&deleg_cur->lock);
1314 no_delegation:
1315         rcu_read_unlock();
1316
1317         if (!ret && open_stateid != NULL) {
1318                 __update_open_stateid(state, open_stateid, NULL, fmode);
1319                 ret = 1;
1320         }
1321         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1322                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1323
1324         return ret;
1325 }
1326
1327 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1328                 const nfs4_stateid *stateid)
1329 {
1330         struct nfs4_state *state = lsp->ls_state;
1331         bool ret = false;
1332
1333         spin_lock(&state->state_lock);
1334         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1335                 goto out_noupdate;
1336         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1337                 goto out_noupdate;
1338         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1339         ret = true;
1340 out_noupdate:
1341         spin_unlock(&state->state_lock);
1342         return ret;
1343 }
1344
1345 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1346 {
1347         struct nfs_delegation *delegation;
1348
1349         rcu_read_lock();
1350         delegation = rcu_dereference(NFS_I(inode)->delegation);
1351         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1352                 rcu_read_unlock();
1353                 return;
1354         }
1355         rcu_read_unlock();
1356         nfs4_inode_return_delegation(inode);
1357 }
1358
1359 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1360 {
1361         struct nfs4_state *state = opendata->state;
1362         struct nfs_inode *nfsi = NFS_I(state->inode);
1363         struct nfs_delegation *delegation;
1364         int open_mode = opendata->o_arg.open_flags;
1365         fmode_t fmode = opendata->o_arg.fmode;
1366         nfs4_stateid stateid;
1367         int ret = -EAGAIN;
1368
1369         for (;;) {
1370                 spin_lock(&state->owner->so_lock);
1371                 if (can_open_cached(state, fmode, open_mode)) {
1372                         update_open_stateflags(state, fmode);
1373                         spin_unlock(&state->owner->so_lock);
1374                         goto out_return_state;
1375                 }
1376                 spin_unlock(&state->owner->so_lock);
1377                 rcu_read_lock();
1378                 delegation = rcu_dereference(nfsi->delegation);
1379                 if (!can_open_delegated(delegation, fmode)) {
1380                         rcu_read_unlock();
1381                         break;
1382                 }
1383                 /* Save the delegation */
1384                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1385                 rcu_read_unlock();
1386                 nfs_release_seqid(opendata->o_arg.seqid);
1387                 if (!opendata->is_recover) {
1388                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1389                         if (ret != 0)
1390                                 goto out;
1391                 }
1392                 ret = -EAGAIN;
1393
1394                 /* Try to update the stateid using the delegation */
1395                 if (update_open_stateid(state, NULL, &stateid, fmode))
1396                         goto out_return_state;
1397         }
1398 out:
1399         return ERR_PTR(ret);
1400 out_return_state:
1401         atomic_inc(&state->count);
1402         return state;
1403 }
1404
1405 static void
1406 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1407 {
1408         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1409         struct nfs_delegation *delegation;
1410         int delegation_flags = 0;
1411
1412         rcu_read_lock();
1413         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1414         if (delegation)
1415                 delegation_flags = delegation->flags;
1416         rcu_read_unlock();
1417         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1418                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1419                                    "returning a delegation for "
1420                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1421                                    clp->cl_hostname);
1422         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1423                 nfs_inode_set_delegation(state->inode,
1424                                          data->owner->so_cred,
1425                                          &data->o_res);
1426         else
1427                 nfs_inode_reclaim_delegation(state->inode,
1428                                              data->owner->so_cred,
1429                                              &data->o_res);
1430 }
1431
1432 /*
1433  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1434  * and update the nfs4_state.
1435  */
1436 static struct nfs4_state *
1437 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1438 {
1439         struct inode *inode = data->state->inode;
1440         struct nfs4_state *state = data->state;
1441         int ret;
1442
1443         if (!data->rpc_done) {
1444                 if (data->rpc_status) {
1445                         ret = data->rpc_status;
1446                         goto err;
1447                 }
1448                 /* cached opens have already been processed */
1449                 goto update;
1450         }
1451
1452         ret = nfs_refresh_inode(inode, &data->f_attr);
1453         if (ret)
1454                 goto err;
1455
1456         if (data->o_res.delegation_type != 0)
1457                 nfs4_opendata_check_deleg(data, state);
1458 update:
1459         update_open_stateid(state, &data->o_res.stateid, NULL,
1460                             data->o_arg.fmode);
1461         atomic_inc(&state->count);
1462
1463         return state;
1464 err:
1465         return ERR_PTR(ret);
1466
1467 }
1468
1469 static struct nfs4_state *
1470 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1471 {
1472         struct inode *inode;
1473         struct nfs4_state *state = NULL;
1474         int ret;
1475
1476         if (!data->rpc_done) {
1477                 state = nfs4_try_open_cached(data);
1478                 goto out;
1479         }
1480
1481         ret = -EAGAIN;
1482         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1483                 goto err;
1484         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1485         ret = PTR_ERR(inode);
1486         if (IS_ERR(inode))
1487                 goto err;
1488         ret = -ENOMEM;
1489         state = nfs4_get_open_state(inode, data->owner);
1490         if (state == NULL)
1491                 goto err_put_inode;
1492         if (data->o_res.delegation_type != 0)
1493                 nfs4_opendata_check_deleg(data, state);
1494         update_open_stateid(state, &data->o_res.stateid, NULL,
1495                         data->o_arg.fmode);
1496         iput(inode);
1497 out:
1498         nfs_release_seqid(data->o_arg.seqid);
1499         return state;
1500 err_put_inode:
1501         iput(inode);
1502 err:
1503         return ERR_PTR(ret);
1504 }
1505
1506 static struct nfs4_state *
1507 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1508 {
1509         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1510                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1511         return _nfs4_opendata_to_nfs4_state(data);
1512 }
1513
1514 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1515 {
1516         struct nfs_inode *nfsi = NFS_I(state->inode);
1517         struct nfs_open_context *ctx;
1518
1519         spin_lock(&state->inode->i_lock);
1520         list_for_each_entry(ctx, &nfsi->open_files, list) {
1521                 if (ctx->state != state)
1522                         continue;
1523                 get_nfs_open_context(ctx);
1524                 spin_unlock(&state->inode->i_lock);
1525                 return ctx;
1526         }
1527         spin_unlock(&state->inode->i_lock);
1528         return ERR_PTR(-ENOENT);
1529 }
1530
1531 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1532                 struct nfs4_state *state, enum open_claim_type4 claim)
1533 {
1534         struct nfs4_opendata *opendata;
1535
1536         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1537                         NULL, NULL, claim, GFP_NOFS);
1538         if (opendata == NULL)
1539                 return ERR_PTR(-ENOMEM);
1540         opendata->state = state;
1541         atomic_inc(&state->count);
1542         return opendata;
1543 }
1544
1545 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1546 {
1547         struct nfs4_state *newstate;
1548         int ret;
1549
1550         opendata->o_arg.open_flags = 0;
1551         opendata->o_arg.fmode = fmode;
1552         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1553                         NFS_SB(opendata->dentry->d_sb),
1554                         fmode, 0);
1555         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1556         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1557         nfs4_init_opendata_res(opendata);
1558         ret = _nfs4_recover_proc_open(opendata);
1559         if (ret != 0)
1560                 return ret;
1561         newstate = nfs4_opendata_to_nfs4_state(opendata);
1562         if (IS_ERR(newstate))
1563                 return PTR_ERR(newstate);
1564         nfs4_close_state(newstate, fmode);
1565         *res = newstate;
1566         return 0;
1567 }
1568
1569 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1570 {
1571         struct nfs4_state *newstate;
1572         int ret;
1573
1574         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1575         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1576         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1577         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1578         /* memory barrier prior to reading state->n_* */
1579         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1580         clear_bit(NFS_OPEN_STATE, &state->flags);
1581         smp_rmb();
1582         if (state->n_rdwr != 0) {
1583                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1584                 if (ret != 0)
1585                         return ret;
1586                 if (newstate != state)
1587                         return -ESTALE;
1588         }
1589         if (state->n_wronly != 0) {
1590                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1591                 if (ret != 0)
1592                         return ret;
1593                 if (newstate != state)
1594                         return -ESTALE;
1595         }
1596         if (state->n_rdonly != 0) {
1597                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1598                 if (ret != 0)
1599                         return ret;
1600                 if (newstate != state)
1601                         return -ESTALE;
1602         }
1603         /*
1604          * We may have performed cached opens for all three recoveries.
1605          * Check if we need to update the current stateid.
1606          */
1607         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1608             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1609                 write_seqlock(&state->seqlock);
1610                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1611                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1612                 write_sequnlock(&state->seqlock);
1613         }
1614         return 0;
1615 }
1616
1617 /*
1618  * OPEN_RECLAIM:
1619  *      reclaim state on the server after a reboot.
1620  */
1621 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1622 {
1623         struct nfs_delegation *delegation;
1624         struct nfs4_opendata *opendata;
1625         fmode_t delegation_type = 0;
1626         int status;
1627
1628         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1629                         NFS4_OPEN_CLAIM_PREVIOUS);
1630         if (IS_ERR(opendata))
1631                 return PTR_ERR(opendata);
1632         rcu_read_lock();
1633         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1634         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1635                 delegation_type = delegation->type;
1636         rcu_read_unlock();
1637         opendata->o_arg.u.delegation_type = delegation_type;
1638         status = nfs4_open_recover(opendata, state);
1639         nfs4_opendata_put(opendata);
1640         return status;
1641 }
1642
1643 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1644 {
1645         struct nfs_server *server = NFS_SERVER(state->inode);
1646         struct nfs4_exception exception = { };
1647         int err;
1648         do {
1649                 err = _nfs4_do_open_reclaim(ctx, state);
1650                 trace_nfs4_open_reclaim(ctx, 0, err);
1651                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1652                         continue;
1653                 if (err != -NFS4ERR_DELAY)
1654                         break;
1655                 nfs4_handle_exception(server, err, &exception);
1656         } while (exception.retry);
1657         return err;
1658 }
1659
1660 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1661 {
1662         struct nfs_open_context *ctx;
1663         int ret;
1664
1665         ctx = nfs4_state_find_open_context(state);
1666         if (IS_ERR(ctx))
1667                 return -EAGAIN;
1668         ret = nfs4_do_open_reclaim(ctx, state);
1669         put_nfs_open_context(ctx);
1670         return ret;
1671 }
1672
1673 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1674 {
1675         switch (err) {
1676                 default:
1677                         printk(KERN_ERR "NFS: %s: unhandled error "
1678                                         "%d.\n", __func__, err);
1679                 case 0:
1680                 case -ENOENT:
1681                 case -EAGAIN:
1682                 case -ESTALE:
1683                         break;
1684                 case -NFS4ERR_BADSESSION:
1685                 case -NFS4ERR_BADSLOT:
1686                 case -NFS4ERR_BAD_HIGH_SLOT:
1687                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1688                 case -NFS4ERR_DEADSESSION:
1689                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1690                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1691                         return -EAGAIN;
1692                 case -NFS4ERR_STALE_CLIENTID:
1693                 case -NFS4ERR_STALE_STATEID:
1694                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1695                 case -NFS4ERR_EXPIRED:
1696                         /* Don't recall a delegation if it was lost */
1697                         nfs4_schedule_lease_recovery(server->nfs_client);
1698                         return -EAGAIN;
1699                 case -NFS4ERR_MOVED:
1700                         nfs4_schedule_migration_recovery(server);
1701                         return -EAGAIN;
1702                 case -NFS4ERR_LEASE_MOVED:
1703                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1704                         return -EAGAIN;
1705                 case -NFS4ERR_DELEG_REVOKED:
1706                 case -NFS4ERR_ADMIN_REVOKED:
1707                 case -NFS4ERR_BAD_STATEID:
1708                 case -NFS4ERR_OPENMODE:
1709                         nfs_inode_find_state_and_recover(state->inode,
1710                                         stateid);
1711                         nfs4_schedule_stateid_recovery(server, state);
1712                         return -EAGAIN;
1713                 case -NFS4ERR_DELAY:
1714                 case -NFS4ERR_GRACE:
1715                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1716                         ssleep(1);
1717                         return -EAGAIN;
1718                 case -ENOMEM:
1719                 case -NFS4ERR_DENIED:
1720                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1721                         return 0;
1722         }
1723         return err;
1724 }
1725
1726 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1727 {
1728         struct nfs_server *server = NFS_SERVER(state->inode);
1729         struct nfs4_opendata *opendata;
1730         int err;
1731
1732         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1733                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1734         if (IS_ERR(opendata))
1735                 return PTR_ERR(opendata);
1736         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1737         err = nfs4_open_recover(opendata, state);
1738         nfs4_opendata_put(opendata);
1739         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1740 }
1741
1742 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1743 {
1744         struct nfs4_opendata *data = calldata;
1745
1746         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1747                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1748 }
1749
1750 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1751 {
1752         struct nfs4_opendata *data = calldata;
1753
1754         nfs40_sequence_done(task, &data->c_res.seq_res);
1755
1756         data->rpc_status = task->tk_status;
1757         if (data->rpc_status == 0) {
1758                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1759                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1760                 renew_lease(data->o_res.server, data->timestamp);
1761                 data->rpc_done = 1;
1762         }
1763 }
1764
1765 static void nfs4_open_confirm_release(void *calldata)
1766 {
1767         struct nfs4_opendata *data = calldata;
1768         struct nfs4_state *state = NULL;
1769
1770         /* If this request hasn't been cancelled, do nothing */
1771         if (data->cancelled == 0)
1772                 goto out_free;
1773         /* In case of error, no cleanup! */
1774         if (!data->rpc_done)
1775                 goto out_free;
1776         state = nfs4_opendata_to_nfs4_state(data);
1777         if (!IS_ERR(state))
1778                 nfs4_close_state(state, data->o_arg.fmode);
1779 out_free:
1780         nfs4_opendata_put(data);
1781 }
1782
1783 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1784         .rpc_call_prepare = nfs4_open_confirm_prepare,
1785         .rpc_call_done = nfs4_open_confirm_done,
1786         .rpc_release = nfs4_open_confirm_release,
1787 };
1788
1789 /*
1790  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1791  */
1792 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1793 {
1794         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1795         struct rpc_task *task;
1796         struct  rpc_message msg = {
1797                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1798                 .rpc_argp = &data->c_arg,
1799                 .rpc_resp = &data->c_res,
1800                 .rpc_cred = data->owner->so_cred,
1801         };
1802         struct rpc_task_setup task_setup_data = {
1803                 .rpc_client = server->client,
1804                 .rpc_message = &msg,
1805                 .callback_ops = &nfs4_open_confirm_ops,
1806                 .callback_data = data,
1807                 .workqueue = nfsiod_workqueue,
1808                 .flags = RPC_TASK_ASYNC,
1809         };
1810         int status;
1811
1812         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1813         kref_get(&data->kref);
1814         data->rpc_done = 0;
1815         data->rpc_status = 0;
1816         data->timestamp = jiffies;
1817         task = rpc_run_task(&task_setup_data);
1818         if (IS_ERR(task))
1819                 return PTR_ERR(task);
1820         status = nfs4_wait_for_completion_rpc_task(task);
1821         if (status != 0) {
1822                 data->cancelled = 1;
1823                 smp_wmb();
1824         } else
1825                 status = data->rpc_status;
1826         rpc_put_task(task);
1827         return status;
1828 }
1829
1830 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1831 {
1832         struct nfs4_opendata *data = calldata;
1833         struct nfs4_state_owner *sp = data->owner;
1834         struct nfs_client *clp = sp->so_server->nfs_client;
1835
1836         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1837                 goto out_wait;
1838         /*
1839          * Check if we still need to send an OPEN call, or if we can use
1840          * a delegation instead.
1841          */
1842         if (data->state != NULL) {
1843                 struct nfs_delegation *delegation;
1844
1845                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1846                         goto out_no_action;
1847                 rcu_read_lock();
1848                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1849                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1850                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1851                     can_open_delegated(delegation, data->o_arg.fmode))
1852                         goto unlock_no_action;
1853                 rcu_read_unlock();
1854         }
1855         /* Update client id. */
1856         data->o_arg.clientid = clp->cl_clientid;
1857         switch (data->o_arg.claim) {
1858         case NFS4_OPEN_CLAIM_PREVIOUS:
1859         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1860         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1861                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1862         case NFS4_OPEN_CLAIM_FH:
1863                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1864                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1865         }
1866         data->timestamp = jiffies;
1867         if (nfs4_setup_sequence(data->o_arg.server,
1868                                 &data->o_arg.seq_args,
1869                                 &data->o_res.seq_res,
1870                                 task) != 0)
1871                 nfs_release_seqid(data->o_arg.seqid);
1872
1873         /* Set the create mode (note dependency on the session type) */
1874         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1875         if (data->o_arg.open_flags & O_EXCL) {
1876                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1877                 if (nfs4_has_persistent_session(clp))
1878                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1879                 else if (clp->cl_mvops->minor_version > 0)
1880                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1881         }
1882         return;
1883 unlock_no_action:
1884         rcu_read_unlock();
1885 out_no_action:
1886         task->tk_action = NULL;
1887 out_wait:
1888         nfs4_sequence_done(task, &data->o_res.seq_res);
1889 }
1890
1891 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1892 {
1893         struct nfs4_opendata *data = calldata;
1894
1895         data->rpc_status = task->tk_status;
1896
1897         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1898                 return;
1899
1900         if (task->tk_status == 0) {
1901                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1902                         switch (data->o_res.f_attr->mode & S_IFMT) {
1903                         case S_IFREG:
1904                                 break;
1905                         case S_IFLNK:
1906                                 data->rpc_status = -ELOOP;
1907                                 break;
1908                         case S_IFDIR:
1909                                 data->rpc_status = -EISDIR;
1910                                 break;
1911                         default:
1912                                 data->rpc_status = -ENOTDIR;
1913                         }
1914                 }
1915                 renew_lease(data->o_res.server, data->timestamp);
1916                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1917                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1918         }
1919         data->rpc_done = 1;
1920 }
1921
1922 static void nfs4_open_release(void *calldata)
1923 {
1924         struct nfs4_opendata *data = calldata;
1925         struct nfs4_state *state = NULL;
1926
1927         /* If this request hasn't been cancelled, do nothing */
1928         if (data->cancelled == 0)
1929                 goto out_free;
1930         /* In case of error, no cleanup! */
1931         if (data->rpc_status != 0 || !data->rpc_done)
1932                 goto out_free;
1933         /* In case we need an open_confirm, no cleanup! */
1934         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1935                 goto out_free;
1936         state = nfs4_opendata_to_nfs4_state(data);
1937         if (!IS_ERR(state))
1938                 nfs4_close_state(state, data->o_arg.fmode);
1939 out_free:
1940         nfs4_opendata_put(data);
1941 }
1942
1943 static const struct rpc_call_ops nfs4_open_ops = {
1944         .rpc_call_prepare = nfs4_open_prepare,
1945         .rpc_call_done = nfs4_open_done,
1946         .rpc_release = nfs4_open_release,
1947 };
1948
1949 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1950 {
1951         struct inode *dir = d_inode(data->dir);
1952         struct nfs_server *server = NFS_SERVER(dir);
1953         struct nfs_openargs *o_arg = &data->o_arg;
1954         struct nfs_openres *o_res = &data->o_res;
1955         struct rpc_task *task;
1956         struct rpc_message msg = {
1957                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1958                 .rpc_argp = o_arg,
1959                 .rpc_resp = o_res,
1960                 .rpc_cred = data->owner->so_cred,
1961         };
1962         struct rpc_task_setup task_setup_data = {
1963                 .rpc_client = server->client,
1964                 .rpc_message = &msg,
1965                 .callback_ops = &nfs4_open_ops,
1966                 .callback_data = data,
1967                 .workqueue = nfsiod_workqueue,
1968                 .flags = RPC_TASK_ASYNC,
1969         };
1970         int status;
1971
1972         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1973         kref_get(&data->kref);
1974         data->rpc_done = 0;
1975         data->rpc_status = 0;
1976         data->cancelled = 0;
1977         data->is_recover = 0;
1978         if (isrecover) {
1979                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1980                 data->is_recover = 1;
1981         }
1982         task = rpc_run_task(&task_setup_data);
1983         if (IS_ERR(task))
1984                 return PTR_ERR(task);
1985         status = nfs4_wait_for_completion_rpc_task(task);
1986         if (status != 0) {
1987                 data->cancelled = 1;
1988                 smp_wmb();
1989         } else
1990                 status = data->rpc_status;
1991         rpc_put_task(task);
1992
1993         return status;
1994 }
1995
1996 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1997 {
1998         struct inode *dir = d_inode(data->dir);
1999         struct nfs_openres *o_res = &data->o_res;
2000         int status;
2001
2002         status = nfs4_run_open_task(data, 1);
2003         if (status != 0 || !data->rpc_done)
2004                 return status;
2005
2006         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2007
2008         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2009                 status = _nfs4_proc_open_confirm(data);
2010                 if (status != 0)
2011                         return status;
2012         }
2013
2014         return status;
2015 }
2016
2017 /*
2018  * Additional permission checks in order to distinguish between an
2019  * open for read, and an open for execute. This works around the
2020  * fact that NFSv4 OPEN treats read and execute permissions as being
2021  * the same.
2022  * Note that in the non-execute case, we want to turn off permission
2023  * checking if we just created a new file (POSIX open() semantics).
2024  */
2025 static int nfs4_opendata_access(struct rpc_cred *cred,
2026                                 struct nfs4_opendata *opendata,
2027                                 struct nfs4_state *state, fmode_t fmode,
2028                                 int openflags)
2029 {
2030         struct nfs_access_entry cache;
2031         u32 mask;
2032
2033         /* access call failed or for some reason the server doesn't
2034          * support any access modes -- defer access call until later */
2035         if (opendata->o_res.access_supported == 0)
2036                 return 0;
2037
2038         mask = 0;
2039         /*
2040          * Use openflags to check for exec, because fmode won't
2041          * always have FMODE_EXEC set when file open for exec.
2042          */
2043         if (openflags & __FMODE_EXEC) {
2044                 /* ONLY check for exec rights */
2045                 mask = MAY_EXEC;
2046         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2047                 mask = MAY_READ;
2048
2049         cache.cred = cred;
2050         cache.jiffies = jiffies;
2051         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2052         nfs_access_add_cache(state->inode, &cache);
2053
2054         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2055                 return 0;
2056
2057         /* even though OPEN succeeded, access is denied. Close the file */
2058         nfs4_close_state(state, fmode);
2059         return -EACCES;
2060 }
2061
2062 /*
2063  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2064  */
2065 static int _nfs4_proc_open(struct nfs4_opendata *data)
2066 {
2067         struct inode *dir = d_inode(data->dir);
2068         struct nfs_server *server = NFS_SERVER(dir);
2069         struct nfs_openargs *o_arg = &data->o_arg;
2070         struct nfs_openres *o_res = &data->o_res;
2071         int status;
2072
2073         status = nfs4_run_open_task(data, 0);
2074         if (!data->rpc_done)
2075                 return status;
2076         if (status != 0) {
2077                 if (status == -NFS4ERR_BADNAME &&
2078                                 !(o_arg->open_flags & O_CREAT))
2079                         return -ENOENT;
2080                 return status;
2081         }
2082
2083         nfs_fattr_map_and_free_names(server, &data->f_attr);
2084
2085         if (o_arg->open_flags & O_CREAT) {
2086                 update_changeattr(dir, &o_res->cinfo);
2087                 if (o_arg->open_flags & O_EXCL)
2088                         data->file_created = 1;
2089                 else if (o_res->cinfo.before != o_res->cinfo.after)
2090                         data->file_created = 1;
2091         }
2092         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2093                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2094         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2095                 status = _nfs4_proc_open_confirm(data);
2096                 if (status != 0)
2097                         return status;
2098         }
2099         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2100                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2101         return 0;
2102 }
2103
2104 static int nfs4_recover_expired_lease(struct nfs_server *server)
2105 {
2106         return nfs4_client_recover_expired_lease(server->nfs_client);
2107 }
2108
2109 /*
2110  * OPEN_EXPIRED:
2111  *      reclaim state on the server after a network partition.
2112  *      Assumes caller holds the appropriate lock
2113  */
2114 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2115 {
2116         struct nfs4_opendata *opendata;
2117         int ret;
2118
2119         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2120                         NFS4_OPEN_CLAIM_FH);
2121         if (IS_ERR(opendata))
2122                 return PTR_ERR(opendata);
2123         ret = nfs4_open_recover(opendata, state);
2124         if (ret == -ESTALE)
2125                 d_drop(ctx->dentry);
2126         nfs4_opendata_put(opendata);
2127         return ret;
2128 }
2129
2130 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2131 {
2132         struct nfs_server *server = NFS_SERVER(state->inode);
2133         struct nfs4_exception exception = { };
2134         int err;
2135
2136         do {
2137                 err = _nfs4_open_expired(ctx, state);
2138                 trace_nfs4_open_expired(ctx, 0, err);
2139                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2140                         continue;
2141                 switch (err) {
2142                 default:
2143                         goto out;
2144                 case -NFS4ERR_GRACE:
2145                 case -NFS4ERR_DELAY:
2146                         nfs4_handle_exception(server, err, &exception);
2147                         err = 0;
2148                 }
2149         } while (exception.retry);
2150 out:
2151         return err;
2152 }
2153
2154 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2155 {
2156         struct nfs_open_context *ctx;
2157         int ret;
2158
2159         ctx = nfs4_state_find_open_context(state);
2160         if (IS_ERR(ctx))
2161                 return -EAGAIN;
2162         ret = nfs4_do_open_expired(ctx, state);
2163         put_nfs_open_context(ctx);
2164         return ret;
2165 }
2166
2167 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2168 {
2169         nfs_remove_bad_delegation(state->inode);
2170         write_seqlock(&state->seqlock);
2171         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2172         write_sequnlock(&state->seqlock);
2173         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2174 }
2175
2176 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2177 {
2178         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2179                 nfs_finish_clear_delegation_stateid(state);
2180 }
2181
2182 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2183 {
2184         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2185         nfs40_clear_delegation_stateid(state);
2186         return nfs4_open_expired(sp, state);
2187 }
2188
2189 #if defined(CONFIG_NFS_V4_1)
2190 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2191 {
2192         struct nfs_server *server = NFS_SERVER(state->inode);
2193         nfs4_stateid stateid;
2194         struct nfs_delegation *delegation;
2195         struct rpc_cred *cred;
2196         int status;
2197
2198         /* Get the delegation credential for use by test/free_stateid */
2199         rcu_read_lock();
2200         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2201         if (delegation == NULL) {
2202                 rcu_read_unlock();
2203                 return;
2204         }
2205
2206         nfs4_stateid_copy(&stateid, &delegation->stateid);
2207         cred = get_rpccred(delegation->cred);
2208         rcu_read_unlock();
2209         status = nfs41_test_stateid(server, &stateid, cred);
2210         trace_nfs4_test_delegation_stateid(state, NULL, status);
2211
2212         if (status != NFS_OK) {
2213                 /* Free the stateid unless the server explicitly
2214                  * informs us the stateid is unrecognized. */
2215                 if (status != -NFS4ERR_BAD_STATEID)
2216                         nfs41_free_stateid(server, &stateid, cred);
2217                 nfs_finish_clear_delegation_stateid(state);
2218         }
2219
2220         put_rpccred(cred);
2221 }
2222
2223 /**
2224  * nfs41_check_open_stateid - possibly free an open stateid
2225  *
2226  * @state: NFSv4 state for an inode
2227  *
2228  * Returns NFS_OK if recovery for this stateid is now finished.
2229  * Otherwise a negative NFS4ERR value is returned.
2230  */
2231 static int nfs41_check_open_stateid(struct nfs4_state *state)
2232 {
2233         struct nfs_server *server = NFS_SERVER(state->inode);
2234         nfs4_stateid *stateid = &state->open_stateid;
2235         struct rpc_cred *cred = state->owner->so_cred;
2236         int status;
2237
2238         /* If a state reset has been done, test_stateid is unneeded */
2239         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2240             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2241             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2242                 return -NFS4ERR_BAD_STATEID;
2243
2244         status = nfs41_test_stateid(server, stateid, cred);
2245         trace_nfs4_test_open_stateid(state, NULL, status);
2246         if (status != NFS_OK) {
2247                 /* Free the stateid unless the server explicitly
2248                  * informs us the stateid is unrecognized. */
2249                 if (status != -NFS4ERR_BAD_STATEID)
2250                         nfs41_free_stateid(server, stateid, cred);
2251
2252                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2253                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2254                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2255                 clear_bit(NFS_OPEN_STATE, &state->flags);
2256         }
2257         return status;
2258 }
2259
2260 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2261 {
2262         int status;
2263
2264         nfs41_check_delegation_stateid(state);
2265         status = nfs41_check_open_stateid(state);
2266         if (status != NFS_OK)
2267                 status = nfs4_open_expired(sp, state);
2268         return status;
2269 }
2270 #endif
2271
2272 /*
2273  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2274  * fields corresponding to attributes that were used to store the verifier.
2275  * Make sure we clobber those fields in the later setattr call
2276  */
2277 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2278 {
2279         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2280             !(sattr->ia_valid & ATTR_ATIME_SET))
2281                 sattr->ia_valid |= ATTR_ATIME;
2282
2283         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2284             !(sattr->ia_valid & ATTR_MTIME_SET))
2285                 sattr->ia_valid |= ATTR_MTIME;
2286 }
2287
2288 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2289                 fmode_t fmode,
2290                 int flags,
2291                 struct nfs_open_context *ctx)
2292 {
2293         struct nfs4_state_owner *sp = opendata->owner;
2294         struct nfs_server *server = sp->so_server;
2295         struct dentry *dentry;
2296         struct nfs4_state *state;
2297         unsigned int seq;
2298         int ret;
2299
2300         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2301
2302         ret = _nfs4_proc_open(opendata);
2303         if (ret != 0)
2304                 goto out;
2305
2306         state = nfs4_opendata_to_nfs4_state(opendata);
2307         ret = PTR_ERR(state);
2308         if (IS_ERR(state))
2309                 goto out;
2310         if (server->caps & NFS_CAP_POSIX_LOCK)
2311                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2312
2313         dentry = opendata->dentry;
2314         if (d_really_is_negative(dentry)) {
2315                 /* FIXME: Is this d_drop() ever needed? */
2316                 d_drop(dentry);
2317                 dentry = d_add_unique(dentry, igrab(state->inode));
2318                 if (dentry == NULL) {
2319                         dentry = opendata->dentry;
2320                 } else if (dentry != ctx->dentry) {
2321                         dput(ctx->dentry);
2322                         ctx->dentry = dget(dentry);
2323                 }
2324                 nfs_set_verifier(dentry,
2325                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2326         }
2327
2328         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2329         if (ret != 0)
2330                 goto out;
2331
2332         ctx->state = state;
2333         if (d_inode(dentry) == state->inode) {
2334                 nfs_inode_attach_open_context(ctx);
2335                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2336                         nfs4_schedule_stateid_recovery(server, state);
2337         }
2338 out:
2339         return ret;
2340 }
2341
2342 /*
2343  * Returns a referenced nfs4_state
2344  */
2345 static int _nfs4_do_open(struct inode *dir,
2346                         struct nfs_open_context *ctx,
2347                         int flags,
2348                         struct iattr *sattr,
2349                         struct nfs4_label *label,
2350                         int *opened)
2351 {
2352         struct nfs4_state_owner  *sp;
2353         struct nfs4_state     *state = NULL;
2354         struct nfs_server       *server = NFS_SERVER(dir);
2355         struct nfs4_opendata *opendata;
2356         struct dentry *dentry = ctx->dentry;
2357         struct rpc_cred *cred = ctx->cred;
2358         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2359         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2360         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2361         struct nfs4_label *olabel = NULL;
2362         int status;
2363
2364         /* Protect against reboot recovery conflicts */
2365         status = -ENOMEM;
2366         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2367         if (sp == NULL) {
2368                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2369                 goto out_err;
2370         }
2371         status = nfs4_recover_expired_lease(server);
2372         if (status != 0)
2373                 goto err_put_state_owner;
2374         if (d_really_is_positive(dentry))
2375                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2376         status = -ENOMEM;
2377         if (d_really_is_positive(dentry))
2378                 claim = NFS4_OPEN_CLAIM_FH;
2379         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2380                         label, claim, GFP_KERNEL);
2381         if (opendata == NULL)
2382                 goto err_put_state_owner;
2383
2384         if (label) {
2385                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2386                 if (IS_ERR(olabel)) {
2387                         status = PTR_ERR(olabel);
2388                         goto err_opendata_put;
2389                 }
2390         }
2391
2392         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2393                 if (!opendata->f_attr.mdsthreshold) {
2394                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2395                         if (!opendata->f_attr.mdsthreshold)
2396                                 goto err_free_label;
2397                 }
2398                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2399         }
2400         if (d_really_is_positive(dentry))
2401                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2402
2403         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2404         if (status != 0)
2405                 goto err_free_label;
2406         state = ctx->state;
2407
2408         if ((opendata->o_arg.open_flags & O_EXCL) &&
2409             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2410                 nfs4_exclusive_attrset(opendata, sattr);
2411
2412                 nfs_fattr_init(opendata->o_res.f_attr);
2413                 status = nfs4_do_setattr(state->inode, cred,
2414                                 opendata->o_res.f_attr, sattr,
2415                                 state, label, olabel);
2416                 if (status == 0) {
2417                         nfs_setattr_update_inode(state->inode, sattr,
2418                                         opendata->o_res.f_attr);
2419                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2420                 }
2421         }
2422         if (opendata->file_created)
2423                 *opened |= FILE_CREATED;
2424
2425         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2426                 *ctx_th = opendata->f_attr.mdsthreshold;
2427                 opendata->f_attr.mdsthreshold = NULL;
2428         }
2429
2430         nfs4_label_free(olabel);
2431
2432         nfs4_opendata_put(opendata);
2433         nfs4_put_state_owner(sp);
2434         return 0;
2435 err_free_label:
2436         nfs4_label_free(olabel);
2437 err_opendata_put:
2438         nfs4_opendata_put(opendata);
2439 err_put_state_owner:
2440         nfs4_put_state_owner(sp);
2441 out_err:
2442         return status;
2443 }
2444
2445
2446 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2447                                         struct nfs_open_context *ctx,
2448                                         int flags,
2449                                         struct iattr *sattr,
2450                                         struct nfs4_label *label,
2451                                         int *opened)
2452 {
2453         struct nfs_server *server = NFS_SERVER(dir);
2454         struct nfs4_exception exception = { };
2455         struct nfs4_state *res;
2456         int status;
2457
2458         do {
2459                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2460                 res = ctx->state;
2461                 trace_nfs4_open_file(ctx, flags, status);
2462                 if (status == 0)
2463                         break;
2464                 /* NOTE: BAD_SEQID means the server and client disagree about the
2465                  * book-keeping w.r.t. state-changing operations
2466                  * (OPEN/CLOSE/LOCK/LOCKU...)
2467                  * It is actually a sign of a bug on the client or on the server.
2468                  *
2469                  * If we receive a BAD_SEQID error in the particular case of
2470                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2471                  * have unhashed the old state_owner for us, and that we can
2472                  * therefore safely retry using a new one. We should still warn
2473                  * the user though...
2474                  */
2475                 if (status == -NFS4ERR_BAD_SEQID) {
2476                         pr_warn_ratelimited("NFS: v4 server %s "
2477                                         " returned a bad sequence-id error!\n",
2478                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2479                         exception.retry = 1;
2480                         continue;
2481                 }
2482                 /*
2483                  * BAD_STATEID on OPEN means that the server cancelled our
2484                  * state before it received the OPEN_CONFIRM.
2485                  * Recover by retrying the request as per the discussion
2486                  * on Page 181 of RFC3530.
2487                  */
2488                 if (status == -NFS4ERR_BAD_STATEID) {
2489                         exception.retry = 1;
2490                         continue;
2491                 }
2492                 if (status == -EAGAIN) {
2493                         /* We must have found a delegation */
2494                         exception.retry = 1;
2495                         continue;
2496                 }
2497                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2498                         continue;
2499                 res = ERR_PTR(nfs4_handle_exception(server,
2500                                         status, &exception));
2501         } while (exception.retry);
2502         return res;
2503 }
2504
2505 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2506                             struct nfs_fattr *fattr, struct iattr *sattr,
2507                             struct nfs4_state *state, struct nfs4_label *ilabel,
2508                             struct nfs4_label *olabel)
2509 {
2510         struct nfs_server *server = NFS_SERVER(inode);
2511         struct nfs_setattrargs  arg = {
2512                 .fh             = NFS_FH(inode),
2513                 .iap            = sattr,
2514                 .server         = server,
2515                 .bitmask = server->attr_bitmask,
2516                 .label          = ilabel,
2517         };
2518         struct nfs_setattrres  res = {
2519                 .fattr          = fattr,
2520                 .label          = olabel,
2521                 .server         = server,
2522         };
2523         struct rpc_message msg = {
2524                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2525                 .rpc_argp       = &arg,
2526                 .rpc_resp       = &res,
2527                 .rpc_cred       = cred,
2528         };
2529         unsigned long timestamp = jiffies;
2530         fmode_t fmode;
2531         bool truncate;
2532         int status;
2533
2534         arg.bitmask = nfs4_bitmask(server, ilabel);
2535         if (ilabel)
2536                 arg.bitmask = nfs4_bitmask(server, olabel);
2537
2538         nfs_fattr_init(fattr);
2539
2540         /* Servers should only apply open mode checks for file size changes */
2541         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2542         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2543
2544         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2545                 /* Use that stateid */
2546         } else if (truncate && state != NULL) {
2547                 struct nfs_lockowner lockowner = {
2548                         .l_owner = current->files,
2549                         .l_pid = current->tgid,
2550                 };
2551                 if (!nfs4_valid_open_stateid(state))
2552                         return -EBADF;
2553                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2554                                 &lockowner) == -EIO)
2555                         return -EBADF;
2556         } else
2557                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2558
2559         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2560         if (status == 0 && state != NULL)
2561                 renew_lease(server, timestamp);
2562         return status;
2563 }
2564
2565 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2566                            struct nfs_fattr *fattr, struct iattr *sattr,
2567                            struct nfs4_state *state, struct nfs4_label *ilabel,
2568                            struct nfs4_label *olabel)
2569 {
2570         struct nfs_server *server = NFS_SERVER(inode);
2571         struct nfs4_exception exception = {
2572                 .state = state,
2573                 .inode = inode,
2574         };
2575         int err;
2576         do {
2577                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2578                 trace_nfs4_setattr(inode, err);
2579                 switch (err) {
2580                 case -NFS4ERR_OPENMODE:
2581                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2582                                 pr_warn_once("NFSv4: server %s is incorrectly "
2583                                                 "applying open mode checks to "
2584                                                 "a SETATTR that is not "
2585                                                 "changing file size.\n",
2586                                                 server->nfs_client->cl_hostname);
2587                         }
2588                         if (state && !(state->state & FMODE_WRITE)) {
2589                                 err = -EBADF;
2590                                 if (sattr->ia_valid & ATTR_OPEN)
2591                                         err = -EACCES;
2592                                 goto out;
2593                         }
2594                 }
2595                 err = nfs4_handle_exception(server, err, &exception);
2596         } while (exception.retry);
2597 out:
2598         return err;
2599 }
2600
2601 struct nfs4_closedata {
2602         struct inode *inode;
2603         struct nfs4_state *state;
2604         struct nfs_closeargs arg;
2605         struct nfs_closeres res;
2606         struct nfs_fattr fattr;
2607         unsigned long timestamp;
2608         bool roc;
2609         u32 roc_barrier;
2610 };
2611
2612 static void nfs4_free_closedata(void *data)
2613 {
2614         struct nfs4_closedata *calldata = data;
2615         struct nfs4_state_owner *sp = calldata->state->owner;
2616         struct super_block *sb = calldata->state->inode->i_sb;
2617
2618         if (calldata->roc)
2619                 pnfs_roc_release(calldata->state->inode);
2620         nfs4_put_open_state(calldata->state);
2621         nfs_free_seqid(calldata->arg.seqid);
2622         nfs4_put_state_owner(sp);
2623         nfs_sb_deactive(sb);
2624         kfree(calldata);
2625 }
2626
2627 static void nfs4_close_done(struct rpc_task *task, void *data)
2628 {
2629         struct nfs4_closedata *calldata = data;
2630         struct nfs4_state *state = calldata->state;
2631         struct nfs_server *server = NFS_SERVER(calldata->inode);
2632         nfs4_stateid *res_stateid = NULL;
2633
2634         dprintk("%s: begin!\n", __func__);
2635         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2636                 return;
2637         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2638         /* hmm. we are done with the inode, and in the process of freeing
2639          * the state_owner. we keep this around to process errors
2640          */
2641         switch (task->tk_status) {
2642                 case 0:
2643                         res_stateid = &calldata->res.stateid;
2644                         if (calldata->arg.fmode == 0 && calldata->roc)
2645                                 pnfs_roc_set_barrier(state->inode,
2646                                                      calldata->roc_barrier);
2647                         renew_lease(server, calldata->timestamp);
2648                         break;
2649                 case -NFS4ERR_ADMIN_REVOKED:
2650                 case -NFS4ERR_STALE_STATEID:
2651                 case -NFS4ERR_OLD_STATEID:
2652                 case -NFS4ERR_BAD_STATEID:
2653                 case -NFS4ERR_EXPIRED:
2654                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2655                                                 &state->open_stateid)) {
2656                                 rpc_restart_call_prepare(task);
2657                                 goto out_release;
2658                         }
2659                         if (calldata->arg.fmode == 0)
2660                                 break;
2661                 default:
2662                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2663                                 rpc_restart_call_prepare(task);
2664                                 goto out_release;
2665                         }
2666         }
2667         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2668 out_release:
2669         nfs_release_seqid(calldata->arg.seqid);
2670         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2671         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2672 }
2673
2674 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2675 {
2676         struct nfs4_closedata *calldata = data;
2677         struct nfs4_state *state = calldata->state;
2678         struct inode *inode = calldata->inode;
2679         bool is_rdonly, is_wronly, is_rdwr;
2680         int call_close = 0;
2681
2682         dprintk("%s: begin!\n", __func__);
2683         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2684                 goto out_wait;
2685
2686         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2687         spin_lock(&state->owner->so_lock);
2688         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2689         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2690         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2691         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2692         /* Calculate the change in open mode */
2693         calldata->arg.fmode = 0;
2694         if (state->n_rdwr == 0) {
2695                 if (state->n_rdonly == 0)
2696                         call_close |= is_rdonly;
2697                 else if (is_rdonly)
2698                         calldata->arg.fmode |= FMODE_READ;
2699                 if (state->n_wronly == 0)
2700                         call_close |= is_wronly;
2701                 else if (is_wronly)
2702                         calldata->arg.fmode |= FMODE_WRITE;
2703         } else if (is_rdwr)
2704                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2705
2706         if (calldata->arg.fmode == 0)
2707                 call_close |= is_rdwr;
2708
2709         if (!nfs4_valid_open_stateid(state))
2710                 call_close = 0;
2711         spin_unlock(&state->owner->so_lock);
2712
2713         if (!call_close) {
2714                 /* Note: exit _without_ calling nfs4_close_done */
2715                 goto out_no_action;
2716         }
2717
2718         if (calldata->arg.fmode == 0) {
2719                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2720                 if (calldata->roc &&
2721                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2722                         nfs_release_seqid(calldata->arg.seqid);
2723                         goto out_wait;
2724                     }
2725         }
2726         calldata->arg.share_access =
2727                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2728                                 calldata->arg.fmode, 0);
2729
2730         nfs_fattr_init(calldata->res.fattr);
2731         calldata->timestamp = jiffies;
2732         if (nfs4_setup_sequence(NFS_SERVER(inode),
2733                                 &calldata->arg.seq_args,
2734                                 &calldata->res.seq_res,
2735                                 task) != 0)
2736                 nfs_release_seqid(calldata->arg.seqid);
2737         dprintk("%s: done!\n", __func__);
2738         return;
2739 out_no_action:
2740         task->tk_action = NULL;
2741 out_wait:
2742         nfs4_sequence_done(task, &calldata->res.seq_res);
2743 }
2744
2745 static const struct rpc_call_ops nfs4_close_ops = {
2746         .rpc_call_prepare = nfs4_close_prepare,
2747         .rpc_call_done = nfs4_close_done,
2748         .rpc_release = nfs4_free_closedata,
2749 };
2750
2751 static bool nfs4_roc(struct inode *inode)
2752 {
2753         if (!nfs_have_layout(inode))
2754                 return false;
2755         return pnfs_roc(inode);
2756 }
2757
2758 /*
2759  * It is possible for data to be read/written from a mem-mapped file
2760  * after the sys_close call (which hits the vfs layer as a flush).
2761  * This means that we can't safely call nfsv4 close on a file until
2762  * the inode is cleared. This in turn means that we are not good
2763  * NFSv4 citizens - we do not indicate to the server to update the file's
2764  * share state even when we are done with one of the three share
2765  * stateid's in the inode.
2766  *
2767  * NOTE: Caller must be holding the sp->so_owner semaphore!
2768  */
2769 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2770 {
2771         struct nfs_server *server = NFS_SERVER(state->inode);
2772         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2773         struct nfs4_closedata *calldata;
2774         struct nfs4_state_owner *sp = state->owner;
2775         struct rpc_task *task;
2776         struct rpc_message msg = {
2777                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2778                 .rpc_cred = state->owner->so_cred,
2779         };
2780         struct rpc_task_setup task_setup_data = {
2781                 .rpc_client = server->client,
2782                 .rpc_message = &msg,
2783                 .callback_ops = &nfs4_close_ops,
2784                 .workqueue = nfsiod_workqueue,
2785                 .flags = RPC_TASK_ASYNC,
2786         };
2787         int status = -ENOMEM;
2788
2789         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2790                 &task_setup_data.rpc_client, &msg);
2791
2792         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2793         if (calldata == NULL)
2794                 goto out;
2795         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2796         calldata->inode = state->inode;
2797         calldata->state = state;
2798         calldata->arg.fh = NFS_FH(state->inode);
2799         /* Serialization for the sequence id */
2800         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2801         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2802         if (IS_ERR(calldata->arg.seqid))
2803                 goto out_free_calldata;
2804         calldata->arg.fmode = 0;
2805         calldata->arg.bitmask = server->cache_consistency_bitmask;
2806         calldata->res.fattr = &calldata->fattr;
2807         calldata->res.seqid = calldata->arg.seqid;
2808         calldata->res.server = server;
2809         calldata->roc = nfs4_roc(state->inode);
2810         nfs_sb_active(calldata->inode->i_sb);
2811
2812         msg.rpc_argp = &calldata->arg;
2813         msg.rpc_resp = &calldata->res;
2814         task_setup_data.callback_data = calldata;
2815         task = rpc_run_task(&task_setup_data);
2816         if (IS_ERR(task))
2817                 return PTR_ERR(task);
2818         status = 0;
2819         if (wait)
2820                 status = rpc_wait_for_completion_task(task);
2821         rpc_put_task(task);
2822         return status;
2823 out_free_calldata:
2824         kfree(calldata);
2825 out:
2826         nfs4_put_open_state(state);
2827         nfs4_put_state_owner(sp);
2828         return status;
2829 }
2830
2831 static struct inode *
2832 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2833                 int open_flags, struct iattr *attr, int *opened)
2834 {
2835         struct nfs4_state *state;
2836         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2837
2838         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2839
2840         /* Protect against concurrent sillydeletes */
2841         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2842
2843         nfs4_label_release_security(label);
2844
2845         if (IS_ERR(state))
2846                 return ERR_CAST(state);
2847         return state->inode;
2848 }
2849
2850 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2851 {
2852         if (ctx->state == NULL)
2853                 return;
2854         if (is_sync)
2855                 nfs4_close_sync(ctx->state, ctx->mode);
2856         else
2857                 nfs4_close_state(ctx->state, ctx->mode);
2858 }
2859
2860 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2861 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2862 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2863
2864 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2865 {
2866         struct nfs4_server_caps_arg args = {
2867                 .fhandle = fhandle,
2868         };
2869         struct nfs4_server_caps_res res = {};
2870         struct rpc_message msg = {
2871                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2872                 .rpc_argp = &args,
2873                 .rpc_resp = &res,
2874         };
2875         int status;
2876
2877         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2878         if (status == 0) {
2879                 /* Sanity check the server answers */
2880                 switch (server->nfs_client->cl_minorversion) {
2881                 case 0:
2882                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2883                         res.attr_bitmask[2] = 0;
2884                         break;
2885                 case 1:
2886                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2887                         break;
2888                 case 2:
2889                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2890                 }
2891                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2892                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2893                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2894                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2895                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2896                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2897                                 NFS_CAP_SECURITY_LABEL);
2898                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2899                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2900                         server->caps |= NFS_CAP_ACLS;
2901                 if (res.has_links != 0)
2902                         server->caps |= NFS_CAP_HARDLINKS;
2903                 if (res.has_symlinks != 0)
2904                         server->caps |= NFS_CAP_SYMLINKS;
2905                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2906                         server->caps |= NFS_CAP_FILEID;
2907                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2908                         server->caps |= NFS_CAP_MODE;
2909                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2910                         server->caps |= NFS_CAP_NLINK;
2911                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2912                         server->caps |= NFS_CAP_OWNER;
2913                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2914                         server->caps |= NFS_CAP_OWNER_GROUP;
2915                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2916                         server->caps |= NFS_CAP_ATIME;
2917                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2918                         server->caps |= NFS_CAP_CTIME;
2919                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2920                         server->caps |= NFS_CAP_MTIME;
2921 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2922                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2923                         server->caps |= NFS_CAP_SECURITY_LABEL;
2924 #endif
2925                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2926                                 sizeof(server->attr_bitmask));
2927                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2928
2929                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2930                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2931                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2932                 server->cache_consistency_bitmask[2] = 0;
2933                 server->acl_bitmask = res.acl_bitmask;
2934                 server->fh_expire_type = res.fh_expire_type;
2935         }
2936
2937         return status;
2938 }
2939
2940 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2941 {
2942         struct nfs4_exception exception = { };
2943         int err;
2944         do {
2945                 err = nfs4_handle_exception(server,
2946                                 _nfs4_server_capabilities(server, fhandle),
2947                                 &exception);
2948         } while (exception.retry);
2949         return err;
2950 }
2951
2952 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2953                 struct nfs_fsinfo *info)
2954 {
2955         u32 bitmask[3];
2956         struct nfs4_lookup_root_arg args = {
2957                 .bitmask = bitmask,
2958         };
2959         struct nfs4_lookup_res res = {
2960                 .server = server,
2961                 .fattr = info->fattr,
2962                 .fh = fhandle,
2963         };
2964         struct rpc_message msg = {
2965                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2966                 .rpc_argp = &args,
2967                 .rpc_resp = &res,
2968         };
2969
2970         bitmask[0] = nfs4_fattr_bitmap[0];
2971         bitmask[1] = nfs4_fattr_bitmap[1];
2972         /*
2973          * Process the label in the upcoming getfattr
2974          */
2975         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2976
2977         nfs_fattr_init(info->fattr);
2978         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2979 }
2980
2981 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2982                 struct nfs_fsinfo *info)
2983 {
2984         struct nfs4_exception exception = { };
2985         int err;
2986         do {
2987                 err = _nfs4_lookup_root(server, fhandle, info);
2988                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2989                 switch (err) {
2990                 case 0:
2991                 case -NFS4ERR_WRONGSEC:
2992                         goto out;
2993                 default:
2994                         err = nfs4_handle_exception(server, err, &exception);
2995                 }
2996         } while (exception.retry);
2997 out:
2998         return err;
2999 }
3000
3001 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3002                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3003 {
3004         struct rpc_auth_create_args auth_args = {
3005                 .pseudoflavor = flavor,
3006         };
3007         struct rpc_auth *auth;
3008         int ret;
3009
3010         auth = rpcauth_create(&auth_args, server->client);
3011         if (IS_ERR(auth)) {
3012                 ret = -EACCES;
3013                 goto out;
3014         }
3015         ret = nfs4_lookup_root(server, fhandle, info);
3016 out:
3017         return ret;
3018 }
3019
3020 /*
3021  * Retry pseudoroot lookup with various security flavors.  We do this when:
3022  *
3023  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3024  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3025  *
3026  * Returns zero on success, or a negative NFS4ERR value, or a
3027  * negative errno value.
3028  */
3029 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3030                               struct nfs_fsinfo *info)
3031 {
3032         /* Per 3530bis 15.33.5 */
3033         static const rpc_authflavor_t flav_array[] = {
3034                 RPC_AUTH_GSS_KRB5P,
3035                 RPC_AUTH_GSS_KRB5I,
3036                 RPC_AUTH_GSS_KRB5,
3037                 RPC_AUTH_UNIX,                  /* courtesy */
3038                 RPC_AUTH_NULL,
3039         };
3040         int status = -EPERM;
3041         size_t i;
3042
3043         if (server->auth_info.flavor_len > 0) {
3044                 /* try each flavor specified by user */
3045                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3046                         status = nfs4_lookup_root_sec(server, fhandle, info,
3047                                                 server->auth_info.flavors[i]);
3048                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3049                                 continue;
3050                         break;
3051                 }
3052         } else {
3053                 /* no flavors specified by user, try default list */
3054                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3055                         status = nfs4_lookup_root_sec(server, fhandle, info,
3056                                                       flav_array[i]);
3057                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3058                                 continue;
3059                         break;
3060                 }
3061         }
3062
3063         /*
3064          * -EACCESS could mean that the user doesn't have correct permissions
3065          * to access the mount.  It could also mean that we tried to mount
3066          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3067          * existing mount programs don't handle -EACCES very well so it should
3068          * be mapped to -EPERM instead.
3069          */
3070         if (status == -EACCES)
3071                 status = -EPERM;
3072         return status;
3073 }
3074
3075 static int nfs4_do_find_root_sec(struct nfs_server *server,
3076                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3077 {
3078         int mv = server->nfs_client->cl_minorversion;
3079         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3080 }
3081
3082 /**
3083  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3084  * @server: initialized nfs_server handle
3085  * @fhandle: we fill in the pseudo-fs root file handle
3086  * @info: we fill in an FSINFO struct
3087  * @auth_probe: probe the auth flavours
3088  *
3089  * Returns zero on success, or a negative errno.
3090  */
3091 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3092                          struct nfs_fsinfo *info,
3093                          bool auth_probe)
3094 {
3095         int status = 0;
3096
3097         if (!auth_probe)
3098                 status = nfs4_lookup_root(server, fhandle, info);
3099
3100         if (auth_probe || status == NFS4ERR_WRONGSEC)
3101                 status = nfs4_do_find_root_sec(server, fhandle, info);
3102
3103         if (status == 0)
3104                 status = nfs4_server_capabilities(server, fhandle);
3105         if (status == 0)
3106                 status = nfs4_do_fsinfo(server, fhandle, info);
3107
3108         return nfs4_map_errors(status);
3109 }
3110
3111 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3112                               struct nfs_fsinfo *info)
3113 {
3114         int error;
3115         struct nfs_fattr *fattr = info->fattr;
3116         struct nfs4_label *label = NULL;
3117
3118         error = nfs4_server_capabilities(server, mntfh);
3119         if (error < 0) {
3120                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3121                 return error;
3122         }
3123
3124         label = nfs4_label_alloc(server, GFP_KERNEL);
3125         if (IS_ERR(label))
3126                 return PTR_ERR(label);
3127
3128         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3129         if (error < 0) {
3130                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3131                 goto err_free_label;
3132         }
3133
3134         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3135             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3136                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3137
3138 err_free_label:
3139         nfs4_label_free(label);
3140
3141         return error;
3142 }
3143
3144 /*
3145  * Get locations and (maybe) other attributes of a referral.
3146  * Note that we'll actually follow the referral later when
3147  * we detect fsid mismatch in inode revalidation
3148  */
3149 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3150                              const struct qstr *name, struct nfs_fattr *fattr,
3151                              struct nfs_fh *fhandle)
3152 {
3153         int status = -ENOMEM;
3154         struct page *page = NULL;
3155         struct nfs4_fs_locations *locations = NULL;
3156
3157         page = alloc_page(GFP_KERNEL);
3158         if (page == NULL)
3159                 goto out;
3160         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3161         if (locations == NULL)
3162                 goto out;
3163
3164         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3165         if (status != 0)
3166                 goto out;
3167
3168         /*
3169          * If the fsid didn't change, this is a migration event, not a
3170          * referral.  Cause us to drop into the exception handler, which
3171          * will kick off migration recovery.
3172          */
3173         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3174                 dprintk("%s: server did not return a different fsid for"
3175                         " a referral at %s\n", __func__, name->name);
3176                 status = -NFS4ERR_MOVED;
3177                 goto out;
3178         }
3179         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3180         nfs_fixup_referral_attributes(&locations->fattr);
3181
3182         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3183         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3184         memset(fhandle, 0, sizeof(struct nfs_fh));
3185 out:
3186         if (page)
3187                 __free_page(page);
3188         kfree(locations);
3189         return status;
3190 }
3191
3192 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3193                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3194 {
3195         struct nfs4_getattr_arg args = {
3196                 .fh = fhandle,
3197                 .bitmask = server->attr_bitmask,
3198         };
3199         struct nfs4_getattr_res res = {
3200                 .fattr = fattr,
3201                 .label = label,
3202                 .server = server,
3203         };
3204         struct rpc_message msg = {
3205                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3206                 .rpc_argp = &args,
3207                 .rpc_resp = &res,
3208         };
3209
3210         args.bitmask = nfs4_bitmask(server, label);
3211
3212         nfs_fattr_init(fattr);
3213         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3214 }
3215
3216 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3217                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3218 {
3219         struct nfs4_exception exception = { };
3220         int err;
3221         do {
3222                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3223                 trace_nfs4_getattr(server, fhandle, fattr, err);
3224                 err = nfs4_handle_exception(server, err,
3225                                 &exception);
3226         } while (exception.retry);
3227         return err;
3228 }
3229
3230 /*
3231  * The file is not closed if it is opened due to the a request to change
3232  * the size of the file. The open call will not be needed once the
3233  * VFS layer lookup-intents are implemented.
3234  *
3235  * Close is called when the inode is destroyed.
3236  * If we haven't opened the file for O_WRONLY, we
3237  * need to in the size_change case to obtain a stateid.
3238  *
3239  * Got race?
3240  * Because OPEN is always done by name in nfsv4, it is
3241  * possible that we opened a different file by the same
3242  * name.  We can recognize this race condition, but we
3243  * can't do anything about it besides returning an error.
3244  *
3245  * This will be fixed with VFS changes (lookup-intent).
3246  */
3247 static int
3248 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3249                   struct iattr *sattr)
3250 {
3251         struct inode *inode = d_inode(dentry);
3252         struct rpc_cred *cred = NULL;
3253         struct nfs4_state *state = NULL;
3254         struct nfs4_label *label = NULL;
3255         int status;
3256
3257         if (pnfs_ld_layoutret_on_setattr(inode) &&
3258             sattr->ia_valid & ATTR_SIZE &&
3259             sattr->ia_size < i_size_read(inode))
3260                 pnfs_commit_and_return_layout(inode);
3261
3262         nfs_fattr_init(fattr);
3263
3264         /* Deal with open(O_TRUNC) */
3265         if (sattr->ia_valid & ATTR_OPEN)
3266                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3267
3268         /* Optimization: if the end result is no change, don't RPC */
3269         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3270                 return 0;
3271
3272         /* Search for an existing open(O_WRITE) file */
3273         if (sattr->ia_valid & ATTR_FILE) {
3274                 struct nfs_open_context *ctx;
3275
3276                 ctx = nfs_file_open_context(sattr->ia_file);
3277                 if (ctx) {
3278                         cred = ctx->cred;
3279                         state = ctx->state;
3280                 }
3281         }
3282
3283         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3284         if (IS_ERR(label))
3285                 return PTR_ERR(label);
3286
3287         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3288         if (status == 0) {
3289                 nfs_setattr_update_inode(inode, sattr, fattr);
3290                 nfs_setsecurity(inode, fattr, label);
3291         }
3292         nfs4_label_free(label);
3293         return status;
3294 }
3295
3296 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3297                 const struct qstr *name, struct nfs_fh *fhandle,
3298                 struct nfs_fattr *fattr, struct nfs4_label *label)
3299 {
3300         struct nfs_server *server = NFS_SERVER(dir);
3301         int                    status;
3302         struct nfs4_lookup_arg args = {
3303                 .bitmask = server->attr_bitmask,
3304                 .dir_fh = NFS_FH(dir),
3305                 .name = name,
3306         };
3307         struct nfs4_lookup_res res = {
3308                 .server = server,
3309                 .fattr = fattr,
3310                 .label = label,
3311                 .fh = fhandle,
3312         };
3313         struct rpc_message msg = {
3314                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3315                 .rpc_argp = &args,
3316                 .rpc_resp = &res,
3317         };
3318
3319         args.bitmask = nfs4_bitmask(server, label);
3320
3321         nfs_fattr_init(fattr);
3322
3323         dprintk("NFS call  lookup %s\n", name->name);
3324         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3325         dprintk("NFS reply lookup: %d\n", status);
3326         return status;
3327 }
3328
3329 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3330 {
3331         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3332                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3333         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3334         fattr->nlink = 2;
3335 }
3336
3337 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3338                                    struct qstr *name, struct nfs_fh *fhandle,
3339                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3340 {
3341         struct nfs4_exception exception = { };
3342         struct rpc_clnt *client = *clnt;
3343         int err;
3344         do {
3345                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3346                 trace_nfs4_lookup(dir, name, err);
3347                 switch (err) {
3348                 case -NFS4ERR_BADNAME:
3349                         err = -ENOENT;
3350                         goto out;
3351                 case -NFS4ERR_MOVED:
3352                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3353                         if (err == -NFS4ERR_MOVED)
3354                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3355                         goto out;
3356                 case -NFS4ERR_WRONGSEC:
3357                         err = -EPERM;
3358                         if (client != *clnt)
3359                                 goto out;
3360                         client = nfs4_negotiate_security(client, dir, name);
3361                         if (IS_ERR(client))
3362                                 return PTR_ERR(client);
3363
3364                         exception.retry = 1;
3365                         break;
3366                 default:
3367                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3368                 }
3369         } while (exception.retry);
3370
3371 out:
3372         if (err == 0)
3373                 *clnt = client;
3374         else if (client != *clnt)
3375                 rpc_shutdown_client(client);
3376
3377         return err;
3378 }
3379
3380 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3381                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3382                             struct nfs4_label *label)
3383 {
3384         int status;
3385         struct rpc_clnt *client = NFS_CLIENT(dir);
3386
3387         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3388         if (client != NFS_CLIENT(dir)) {
3389                 rpc_shutdown_client(client);
3390                 nfs_fixup_secinfo_attributes(fattr);
3391         }
3392         return status;
3393 }
3394
3395 struct rpc_clnt *
3396 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3397                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3398 {
3399         struct rpc_clnt *client = NFS_CLIENT(dir);
3400         int status;
3401
3402         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3403         if (status < 0)
3404                 return ERR_PTR(status);
3405         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3406 }
3407
3408 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3409 {
3410         struct nfs_server *server = NFS_SERVER(inode);
3411         struct nfs4_accessargs args = {
3412                 .fh = NFS_FH(inode),
3413                 .bitmask = server->cache_consistency_bitmask,
3414         };
3415         struct nfs4_accessres res = {
3416                 .server = server,
3417         };
3418         struct rpc_message msg = {
3419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3420                 .rpc_argp = &args,
3421                 .rpc_resp = &res,
3422                 .rpc_cred = entry->cred,
3423         };
3424         int mode = entry->mask;
3425         int status = 0;
3426
3427         /*
3428          * Determine which access bits we want to ask for...
3429          */
3430         if (mode & MAY_READ)
3431                 args.access |= NFS4_ACCESS_READ;
3432         if (S_ISDIR(inode->i_mode)) {
3433                 if (mode & MAY_WRITE)
3434                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3435                 if (mode & MAY_EXEC)
3436                         args.access |= NFS4_ACCESS_LOOKUP;
3437         } else {
3438                 if (mode & MAY_WRITE)
3439                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3440                 if (mode & MAY_EXEC)
3441                         args.access |= NFS4_ACCESS_EXECUTE;
3442         }
3443
3444         res.fattr = nfs_alloc_fattr();
3445         if (res.fattr == NULL)
3446                 return -ENOMEM;
3447
3448         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3449         if (!status) {
3450                 nfs_access_set_mask(entry, res.access);
3451                 nfs_refresh_inode(inode, res.fattr);
3452         }
3453         nfs_free_fattr(res.fattr);
3454         return status;
3455 }
3456
3457 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3458 {
3459         struct nfs4_exception exception = { };
3460         int err;
3461         do {
3462                 err = _nfs4_proc_access(inode, entry);
3463                 trace_nfs4_access(inode, err);
3464                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3465                                 &exception);
3466         } while (exception.retry);
3467         return err;
3468 }
3469
3470 /*
3471  * TODO: For the time being, we don't try to get any attributes
3472  * along with any of the zero-copy operations READ, READDIR,
3473  * READLINK, WRITE.
3474  *
3475  * In the case of the first three, we want to put the GETATTR
3476  * after the read-type operation -- this is because it is hard
3477  * to predict the length of a GETATTR response in v4, and thus
3478  * align the READ data correctly.  This means that the GETATTR
3479  * may end up partially falling into the page cache, and we should
3480  * shift it into the 'tail' of the xdr_buf before processing.
3481  * To do this efficiently, we need to know the total length
3482  * of data received, which doesn't seem to be available outside
3483  * of the RPC layer.
3484  *
3485  * In the case of WRITE, we also want to put the GETATTR after
3486  * the operation -- in this case because we want to make sure
3487  * we get the post-operation mtime and size.
3488  *
3489  * Both of these changes to the XDR layer would in fact be quite
3490  * minor, but I decided to leave them for a subsequent patch.
3491  */
3492 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3493                 unsigned int pgbase, unsigned int pglen)
3494 {
3495         struct nfs4_readlink args = {
3496                 .fh       = NFS_FH(inode),
3497                 .pgbase   = pgbase,
3498                 .pglen    = pglen,
3499                 .pages    = &page,
3500         };
3501         struct nfs4_readlink_res res;
3502         struct rpc_message msg = {
3503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3504                 .rpc_argp = &args,
3505                 .rpc_resp = &res,
3506         };
3507
3508         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3509 }
3510
3511 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3512                 unsigned int pgbase, unsigned int pglen)
3513 {
3514         struct nfs4_exception exception = { };
3515         int err;
3516         do {
3517                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3518                 trace_nfs4_readlink(inode, err);
3519                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3520                                 &exception);
3521         } while (exception.retry);
3522         return err;
3523 }
3524
3525 /*
3526  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3527  */
3528 static int
3529 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3530                  int flags)
3531 {
3532         struct nfs4_label l, *ilabel = NULL;
3533         struct nfs_open_context *ctx;
3534         struct nfs4_state *state;
3535         int opened = 0;
3536         int status = 0;
3537
3538         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3539         if (IS_ERR(ctx))
3540                 return PTR_ERR(ctx);
3541
3542         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3543
3544         sattr->ia_mode &= ~current_umask();
3545         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3546         if (IS_ERR(state)) {
3547                 status = PTR_ERR(state);
3548                 goto out;
3549         }
3550 out:
3551         nfs4_label_release_security(ilabel);
3552         put_nfs_open_context(ctx);
3553         return status;
3554 }
3555
3556 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3557 {
3558         struct nfs_server *server = NFS_SERVER(dir);
3559         struct nfs_removeargs args = {
3560                 .fh = NFS_FH(dir),
3561                 .name = *name,
3562         };
3563         struct nfs_removeres res = {
3564                 .server = server,
3565         };
3566         struct rpc_message msg = {
3567                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3568                 .rpc_argp = &args,
3569                 .rpc_resp = &res,
3570         };
3571         int status;
3572
3573         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3574         if (status == 0)
3575                 update_changeattr(dir, &res.cinfo);
3576         return status;
3577 }
3578
3579 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3580 {
3581         struct nfs4_exception exception = { };
3582         int err;
3583         do {
3584                 err = _nfs4_proc_remove(dir, name);
3585                 trace_nfs4_remove(dir, name, err);
3586                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3587                                 &exception);
3588         } while (exception.retry);
3589         return err;
3590 }
3591
3592 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3593 {
3594         struct nfs_server *server = NFS_SERVER(dir);
3595         struct nfs_removeargs *args = msg->rpc_argp;
3596         struct nfs_removeres *res = msg->rpc_resp;
3597
3598         res->server = server;
3599         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3600         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3601
3602         nfs_fattr_init(res->dir_attr);
3603 }
3604
3605 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3606 {
3607         nfs4_setup_sequence(NFS_SERVER(data->dir),
3608                         &data->args.seq_args,
3609                         &data->res.seq_res,
3610                         task);
3611 }
3612
3613 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3614 {
3615         struct nfs_unlinkdata *data = task->tk_calldata;
3616         struct nfs_removeres *res = &data->res;
3617
3618         if (!nfs4_sequence_done(task, &res->seq_res))
3619                 return 0;
3620         if (nfs4_async_handle_error(task, res->server, NULL,
3621                                     &data->timeout) == -EAGAIN)
3622                 return 0;
3623         update_changeattr(dir, &res->cinfo);
3624         return 1;
3625 }
3626
3627 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3628 {
3629         struct nfs_server *server = NFS_SERVER(dir);
3630         struct nfs_renameargs *arg = msg->rpc_argp;
3631         struct nfs_renameres *res = msg->rpc_resp;
3632
3633         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3634         res->server = server;
3635         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3636 }
3637
3638 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3639 {
3640         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3641                         &data->args.seq_args,
3642                         &data->res.seq_res,
3643                         task);
3644 }
3645
3646 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3647                                  struct inode *new_dir)
3648 {
3649         struct nfs_renamedata *data = task->tk_calldata;
3650         struct nfs_renameres *res = &data->res;
3651
3652         if (!nfs4_sequence_done(task, &res->seq_res))
3653                 return 0;
3654         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3655                 return 0;
3656
3657         update_changeattr(old_dir, &res->old_cinfo);
3658         update_changeattr(new_dir, &res->new_cinfo);
3659         return 1;
3660 }
3661
3662 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3663 {
3664         struct nfs_server *server = NFS_SERVER(inode);
3665         struct nfs4_link_arg arg = {
3666                 .fh     = NFS_FH(inode),
3667                 .dir_fh = NFS_FH(dir),
3668                 .name   = name,
3669                 .bitmask = server->attr_bitmask,
3670         };
3671         struct nfs4_link_res res = {
3672                 .server = server,
3673                 .label = NULL,
3674         };
3675         struct rpc_message msg = {
3676                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3677                 .rpc_argp = &arg,
3678                 .rpc_resp = &res,
3679         };
3680         int status = -ENOMEM;
3681
3682         res.fattr = nfs_alloc_fattr();
3683         if (res.fattr == NULL)
3684                 goto out;
3685
3686         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3687         if (IS_ERR(res.label)) {
3688                 status = PTR_ERR(res.label);
3689                 goto out;
3690         }
3691         arg.bitmask = nfs4_bitmask(server, res.label);
3692
3693         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3694         if (!status) {
3695                 update_changeattr(dir, &res.cinfo);
3696                 status = nfs_post_op_update_inode(inode, res.fattr);
3697                 if (!status)
3698                         nfs_setsecurity(inode, res.fattr, res.label);
3699         }
3700
3701
3702         nfs4_label_free(res.label);
3703
3704 out:
3705         nfs_free_fattr(res.fattr);
3706         return status;
3707 }
3708
3709 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3710 {
3711         struct nfs4_exception exception = { };
3712         int err;
3713         do {
3714                 err = nfs4_handle_exception(NFS_SERVER(inode),
3715                                 _nfs4_proc_link(inode, dir, name),
3716                                 &exception);
3717         } while (exception.retry);
3718         return err;
3719 }
3720
3721 struct nfs4_createdata {
3722         struct rpc_message msg;
3723         struct nfs4_create_arg arg;
3724         struct nfs4_create_res res;
3725         struct nfs_fh fh;
3726         struct nfs_fattr fattr;
3727         struct nfs4_label *label;
3728 };
3729
3730 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3731                 struct qstr *name, struct iattr *sattr, u32 ftype)
3732 {
3733         struct nfs4_createdata *data;
3734
3735         data = kzalloc(sizeof(*data), GFP_KERNEL);
3736         if (data != NULL) {
3737                 struct nfs_server *server = NFS_SERVER(dir);
3738
3739                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3740                 if (IS_ERR(data->label))
3741                         goto out_free;
3742
3743                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3744                 data->msg.rpc_argp = &data->arg;
3745                 data->msg.rpc_resp = &data->res;
3746                 data->arg.dir_fh = NFS_FH(dir);
3747                 data->arg.server = server;
3748                 data->arg.name = name;
3749                 data->arg.attrs = sattr;
3750                 data->arg.ftype = ftype;
3751                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3752                 data->res.server = server;
3753                 data->res.fh = &data->fh;
3754                 data->res.fattr = &data->fattr;
3755                 data->res.label = data->label;
3756                 nfs_fattr_init(data->res.fattr);
3757         }
3758         return data;
3759 out_free:
3760         kfree(data);
3761         return NULL;
3762 }
3763
3764 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3765 {
3766         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3767                                     &data->arg.seq_args, &data->res.seq_res, 1);
3768         if (status == 0) {
3769                 update_changeattr(dir, &data->res.dir_cinfo);
3770                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3771         }
3772         return status;
3773 }
3774
3775 static void nfs4_free_createdata(struct nfs4_createdata *data)
3776 {
3777         nfs4_label_free(data->label);
3778         kfree(data);
3779 }
3780
3781 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3782                 struct page *page, unsigned int len, struct iattr *sattr,
3783                 struct nfs4_label *label)
3784 {
3785         struct nfs4_createdata *data;
3786         int status = -ENAMETOOLONG;
3787
3788         if (len > NFS4_MAXPATHLEN)
3789                 goto out;
3790
3791         status = -ENOMEM;
3792         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3793         if (data == NULL)
3794                 goto out;
3795
3796         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3797         data->arg.u.symlink.pages = &page;
3798         data->arg.u.symlink.len = len;
3799         data->arg.label = label;
3800
3801         status = nfs4_do_create(dir, dentry, data);
3802
3803         nfs4_free_createdata(data);
3804 out:
3805         return status;
3806 }
3807
3808 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3809                 struct page *page, unsigned int len, struct iattr *sattr)
3810 {
3811         struct nfs4_exception exception = { };
3812         struct nfs4_label l, *label = NULL;
3813         int err;
3814
3815         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3816
3817         do {
3818                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3819                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3820                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3821                                 &exception);
3822         } while (exception.retry);
3823
3824         nfs4_label_release_security(label);
3825         return err;
3826 }
3827
3828 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3829                 struct iattr *sattr, struct nfs4_label *label)
3830 {
3831         struct nfs4_createdata *data;
3832         int status = -ENOMEM;
3833
3834         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3835         if (data == NULL)
3836                 goto out;
3837
3838         data->arg.label = label;
3839         status = nfs4_do_create(dir, dentry, data);
3840
3841         nfs4_free_createdata(data);
3842 out:
3843         return status;
3844 }
3845
3846 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3847                 struct iattr *sattr)
3848 {
3849         struct nfs4_exception exception = { };
3850         struct nfs4_label l, *label = NULL;
3851         int err;
3852
3853         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3854
3855         sattr->ia_mode &= ~current_umask();
3856         do {
3857                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3858                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3859                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3860                                 &exception);
3861         } while (exception.retry);
3862         nfs4_label_release_security(label);
3863
3864         return err;
3865 }
3866
3867 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3868                 u64 cookie, struct page **pages, unsigned int count, int plus)
3869 {
3870         struct inode            *dir = d_inode(dentry);
3871         struct nfs4_readdir_arg args = {
3872                 .fh = NFS_FH(dir),
3873                 .pages = pages,
3874                 .pgbase = 0,
3875                 .count = count,
3876                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3877                 .plus = plus,
3878         };
3879         struct nfs4_readdir_res res;
3880         struct rpc_message msg = {
3881                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3882                 .rpc_argp = &args,
3883                 .rpc_resp = &res,
3884                 .rpc_cred = cred,
3885         };
3886         int                     status;
3887
3888         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3889                         dentry,
3890                         (unsigned long long)cookie);
3891         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3892         res.pgbase = args.pgbase;
3893         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3894         if (status >= 0) {
3895                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3896                 status += args.pgbase;
3897         }
3898
3899         nfs_invalidate_atime(dir);
3900
3901         dprintk("%s: returns %d\n", __func__, status);
3902         return status;
3903 }
3904
3905 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3906                 u64 cookie, struct page **pages, unsigned int count, int plus)
3907 {
3908         struct nfs4_exception exception = { };
3909         int err;
3910         do {
3911                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3912                                 pages, count, plus);
3913                 trace_nfs4_readdir(d_inode(dentry), err);
3914                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3915                                 &exception);
3916         } while (exception.retry);
3917         return err;
3918 }
3919
3920 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3921                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3922 {
3923         struct nfs4_createdata *data;
3924         int mode = sattr->ia_mode;
3925         int status = -ENOMEM;
3926
3927         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3928         if (data == NULL)
3929                 goto out;
3930
3931         if (S_ISFIFO(mode))
3932                 data->arg.ftype = NF4FIFO;
3933         else if (S_ISBLK(mode)) {
3934                 data->arg.ftype = NF4BLK;
3935                 data->arg.u.device.specdata1 = MAJOR(rdev);
3936                 data->arg.u.device.specdata2 = MINOR(rdev);
3937         }
3938         else if (S_ISCHR(mode)) {
3939                 data->arg.ftype = NF4CHR;
3940                 data->arg.u.device.specdata1 = MAJOR(rdev);
3941                 data->arg.u.device.specdata2 = MINOR(rdev);
3942         } else if (!S_ISSOCK(mode)) {
3943                 status = -EINVAL;
3944                 goto out_free;
3945         }
3946
3947         data->arg.label = label;
3948         status = nfs4_do_create(dir, dentry, data);
3949 out_free:
3950         nfs4_free_createdata(data);
3951 out:
3952         return status;
3953 }
3954
3955 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3956                 struct iattr *sattr, dev_t rdev)
3957 {
3958         struct nfs4_exception exception = { };
3959         struct nfs4_label l, *label = NULL;
3960         int err;
3961
3962         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3963
3964         sattr->ia_mode &= ~current_umask();
3965         do {
3966                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3967                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3968                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3969                                 &exception);
3970         } while (exception.retry);
3971
3972         nfs4_label_release_security(label);
3973
3974         return err;
3975 }
3976
3977 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3978                  struct nfs_fsstat *fsstat)
3979 {
3980         struct nfs4_statfs_arg args = {
3981                 .fh = fhandle,
3982                 .bitmask = server->attr_bitmask,
3983         };
3984         struct nfs4_statfs_res res = {
3985                 .fsstat = fsstat,
3986         };
3987         struct rpc_message msg = {
3988                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3989                 .rpc_argp = &args,
3990                 .rpc_resp = &res,
3991         };
3992
3993         nfs_fattr_init(fsstat->fattr);
3994         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3995 }
3996
3997 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3998 {
3999         struct nfs4_exception exception = { };
4000         int err;
4001         do {
4002                 err = nfs4_handle_exception(server,
4003                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4004                                 &exception);
4005         } while (exception.retry);
4006         return err;
4007 }
4008
4009 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4010                 struct nfs_fsinfo *fsinfo)
4011 {
4012         struct nfs4_fsinfo_arg args = {
4013                 .fh = fhandle,
4014                 .bitmask = server->attr_bitmask,
4015         };
4016         struct nfs4_fsinfo_res res = {
4017                 .fsinfo = fsinfo,
4018         };
4019         struct rpc_message msg = {
4020                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4021                 .rpc_argp = &args,
4022                 .rpc_resp = &res,
4023         };
4024
4025         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4026 }
4027
4028 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4029 {
4030         struct nfs4_exception exception = { };
4031         unsigned long now = jiffies;
4032         int err;
4033
4034         do {
4035                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4036                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4037                 if (err == 0) {
4038                         struct nfs_client *clp = server->nfs_client;
4039
4040                         spin_lock(&clp->cl_lock);
4041                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4042                         clp->cl_last_renewal = now;
4043                         spin_unlock(&clp->cl_lock);
4044                         break;
4045                 }
4046                 err = nfs4_handle_exception(server, err, &exception);
4047         } while (exception.retry);
4048         return err;
4049 }
4050
4051 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4052 {
4053         int error;
4054
4055         nfs_fattr_init(fsinfo->fattr);
4056         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4057         if (error == 0) {
4058                 /* block layout checks this! */
4059                 server->pnfs_blksize = fsinfo->blksize;
4060                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4061         }
4062
4063         return error;
4064 }
4065
4066 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4067                 struct nfs_pathconf *pathconf)
4068 {
4069         struct nfs4_pathconf_arg args = {
4070                 .fh = fhandle,
4071                 .bitmask = server->attr_bitmask,
4072         };
4073         struct nfs4_pathconf_res res = {
4074                 .pathconf = pathconf,
4075         };
4076         struct rpc_message msg = {
4077                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4078                 .rpc_argp = &args,
4079                 .rpc_resp = &res,
4080         };
4081
4082         /* None of the pathconf attributes are mandatory to implement */
4083         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4084                 memset(pathconf, 0, sizeof(*pathconf));
4085                 return 0;
4086         }
4087
4088         nfs_fattr_init(pathconf->fattr);
4089         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4090 }
4091
4092 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4093                 struct nfs_pathconf *pathconf)
4094 {
4095         struct nfs4_exception exception = { };
4096         int err;
4097
4098         do {
4099                 err = nfs4_handle_exception(server,
4100                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4101                                 &exception);
4102         } while (exception.retry);
4103         return err;
4104 }
4105
4106 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4107                 const struct nfs_open_context *ctx,
4108                 const struct nfs_lock_context *l_ctx,
4109                 fmode_t fmode)
4110 {
4111         const struct nfs_lockowner *lockowner = NULL;
4112
4113         if (l_ctx != NULL)
4114                 lockowner = &l_ctx->lockowner;
4115         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4116 }
4117 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4118
4119 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4120                 const struct nfs_open_context *ctx,
4121                 const struct nfs_lock_context *l_ctx,
4122                 fmode_t fmode)
4123 {
4124         nfs4_stateid current_stateid;
4125
4126         /* If the current stateid represents a lost lock, then exit */
4127         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4128                 return true;
4129         return nfs4_stateid_match(stateid, &current_stateid);
4130 }
4131
4132 static bool nfs4_error_stateid_expired(int err)
4133 {
4134         switch (err) {
4135         case -NFS4ERR_DELEG_REVOKED:
4136         case -NFS4ERR_ADMIN_REVOKED:
4137         case -NFS4ERR_BAD_STATEID:
4138         case -NFS4ERR_STALE_STATEID:
4139         case -NFS4ERR_OLD_STATEID:
4140         case -NFS4ERR_OPENMODE:
4141         case -NFS4ERR_EXPIRED:
4142                 return true;
4143         }
4144         return false;
4145 }
4146
4147 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4148 {
4149         nfs_invalidate_atime(hdr->inode);
4150 }
4151
4152 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4153 {
4154         struct nfs_server *server = NFS_SERVER(hdr->inode);
4155
4156         trace_nfs4_read(hdr, task->tk_status);
4157         if (nfs4_async_handle_error(task, server,
4158                                     hdr->args.context->state,
4159                                     NULL) == -EAGAIN) {
4160                 rpc_restart_call_prepare(task);
4161                 return -EAGAIN;
4162         }
4163
4164         __nfs4_read_done_cb(hdr);
4165         if (task->tk_status > 0)
4166                 renew_lease(server, hdr->timestamp);
4167         return 0;
4168 }
4169
4170 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4171                 struct nfs_pgio_args *args)
4172 {
4173
4174         if (!nfs4_error_stateid_expired(task->tk_status) ||
4175                 nfs4_stateid_is_current(&args->stateid,
4176                                 args->context,
4177                                 args->lock_context,
4178                                 FMODE_READ))
4179                 return false;
4180         rpc_restart_call_prepare(task);
4181         return true;
4182 }
4183
4184 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4185 {
4186
4187         dprintk("--> %s\n", __func__);
4188
4189         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4190                 return -EAGAIN;
4191         if (nfs4_read_stateid_changed(task, &hdr->args))
4192                 return -EAGAIN;
4193         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4194                                     nfs4_read_done_cb(task, hdr);
4195 }
4196
4197 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4198                                  struct rpc_message *msg)
4199 {
4200         hdr->timestamp   = jiffies;
4201         hdr->pgio_done_cb = nfs4_read_done_cb;
4202         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4203         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4204 }
4205
4206 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4207                                       struct nfs_pgio_header *hdr)
4208 {
4209         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4210                         &hdr->args.seq_args,
4211                         &hdr->res.seq_res,
4212                         task))
4213                 return 0;
4214         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4215                                 hdr->args.lock_context,
4216                                 hdr->rw_ops->rw_mode) == -EIO)
4217                 return -EIO;
4218         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4219                 return -EIO;
4220         return 0;
4221 }
4222
4223 static int nfs4_write_done_cb(struct rpc_task *task,
4224                               struct nfs_pgio_header *hdr)
4225 {
4226         struct inode *inode = hdr->inode;
4227
4228         trace_nfs4_write(hdr, task->tk_status);
4229         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4230                                     hdr->args.context->state,
4231                                     NULL) == -EAGAIN) {
4232                 rpc_restart_call_prepare(task);
4233                 return -EAGAIN;
4234         }
4235         if (task->tk_status >= 0) {
4236                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4237                 nfs_writeback_update_inode(hdr);
4238         }
4239         return 0;
4240 }
4241
4242 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4243                 struct nfs_pgio_args *args)
4244 {
4245
4246         if (!nfs4_error_stateid_expired(task->tk_status) ||
4247                 nfs4_stateid_is_current(&args->stateid,
4248                                 args->context,
4249                                 args->lock_context,
4250                                 FMODE_WRITE))
4251                 return false;
4252         rpc_restart_call_prepare(task);
4253         return true;
4254 }
4255
4256 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4257 {
4258         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4259                 return -EAGAIN;
4260         if (nfs4_write_stateid_changed(task, &hdr->args))
4261                 return -EAGAIN;
4262         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4263                 nfs4_write_done_cb(task, hdr);
4264 }
4265
4266 static
4267 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4268 {
4269         /* Don't request attributes for pNFS or O_DIRECT writes */
4270         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4271                 return false;
4272         /* Otherwise, request attributes if and only if we don't hold
4273          * a delegation
4274          */
4275         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4276 }
4277
4278 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4279                                   struct rpc_message *msg)
4280 {
4281         struct nfs_server *server = NFS_SERVER(hdr->inode);
4282
4283         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4284                 hdr->args.bitmask = NULL;
4285                 hdr->res.fattr = NULL;
4286         } else
4287                 hdr->args.bitmask = server->cache_consistency_bitmask;
4288
4289         if (!hdr->pgio_done_cb)
4290                 hdr->pgio_done_cb = nfs4_write_done_cb;
4291         hdr->res.server = server;
4292         hdr->timestamp   = jiffies;
4293
4294         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4295         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4296 }
4297
4298 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4299 {
4300         nfs4_setup_sequence(NFS_SERVER(data->inode),
4301                         &data->args.seq_args,
4302                         &data->res.seq_res,
4303                         task);
4304 }
4305
4306 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4307 {
4308         struct inode *inode = data->inode;
4309
4310         trace_nfs4_commit(data, task->tk_status);
4311         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4312                                     NULL, NULL) == -EAGAIN) {
4313                 rpc_restart_call_prepare(task);
4314                 return -EAGAIN;
4315         }
4316         return 0;
4317 }
4318
4319 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4320 {
4321         if (!nfs4_sequence_done(task, &data->res.seq_res))
4322                 return -EAGAIN;
4323         return data->commit_done_cb(task, data);
4324 }
4325
4326 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4327 {
4328         struct nfs_server *server = NFS_SERVER(data->inode);
4329
4330         if (data->commit_done_cb == NULL)
4331                 data->commit_done_cb = nfs4_commit_done_cb;
4332         data->res.server = server;
4333         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4334         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4335 }
4336
4337 struct nfs4_renewdata {
4338         struct nfs_client       *client;
4339         unsigned long           timestamp;
4340 };
4341
4342 /*
4343  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4344  * standalone procedure for queueing an asynchronous RENEW.
4345  */
4346 static void nfs4_renew_release(void *calldata)
4347 {
4348         struct nfs4_renewdata *data = calldata;
4349         struct nfs_client *clp = data->client;
4350
4351         if (atomic_read(&clp->cl_count) > 1)
4352                 nfs4_schedule_state_renewal(clp);
4353         nfs_put_client(clp);
4354         kfree(data);
4355 }
4356
4357 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4358 {
4359         struct nfs4_renewdata *data = calldata;
4360         struct nfs_client *clp = data->client;
4361         unsigned long timestamp = data->timestamp;
4362
4363         trace_nfs4_renew_async(clp, task->tk_status);
4364         switch (task->tk_status) {
4365         case 0:
4366                 break;
4367         case -NFS4ERR_LEASE_MOVED:
4368                 nfs4_schedule_lease_moved_recovery(clp);
4369                 break;
4370         default:
4371                 /* Unless we're shutting down, schedule state recovery! */
4372                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4373                         return;
4374                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4375                         nfs4_schedule_lease_recovery(clp);
4376                         return;
4377                 }
4378                 nfs4_schedule_path_down_recovery(clp);
4379         }
4380         do_renew_lease(clp, timestamp);
4381 }
4382
4383 static const struct rpc_call_ops nfs4_renew_ops = {
4384         .rpc_call_done = nfs4_renew_done,
4385         .rpc_release = nfs4_renew_release,
4386 };
4387
4388 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4389 {
4390         struct rpc_message msg = {
4391                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4392                 .rpc_argp       = clp,
4393                 .rpc_cred       = cred,
4394         };
4395         struct nfs4_renewdata *data;
4396
4397         if (renew_flags == 0)
4398                 return 0;
4399         if (!atomic_inc_not_zero(&clp->cl_count))
4400                 return -EIO;
4401         data = kmalloc(sizeof(*data), GFP_NOFS);
4402         if (data == NULL)
4403                 return -ENOMEM;
4404         data->client = clp;
4405         data->timestamp = jiffies;
4406         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4407                         &nfs4_renew_ops, data);
4408 }
4409
4410 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4411 {
4412         struct rpc_message msg = {
4413                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4414                 .rpc_argp       = clp,
4415                 .rpc_cred       = cred,
4416         };
4417         unsigned long now = jiffies;
4418         int status;
4419
4420         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4421         if (status < 0)
4422                 return status;
4423         do_renew_lease(clp, now);
4424         return 0;
4425 }
4426
4427 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4428 {
4429         return server->caps & NFS_CAP_ACLS;
4430 }
4431
4432 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4433  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4434  * the stack.
4435  */
4436 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4437
4438 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4439                 struct page **pages, unsigned int *pgbase)
4440 {
4441         struct page *newpage, **spages;
4442         int rc = 0;
4443         size_t len;
4444         spages = pages;
4445
4446         do {
4447                 len = min_t(size_t, PAGE_SIZE, buflen);
4448                 newpage = alloc_page(GFP_KERNEL);
4449
4450                 if (newpage == NULL)
4451                         goto unwind;
4452                 memcpy(page_address(newpage), buf, len);
4453                 buf += len;
4454                 buflen -= len;
4455                 *pages++ = newpage;
4456                 rc++;
4457         } while (buflen != 0);
4458
4459         return rc;
4460
4461 unwind:
4462         for(; rc > 0; rc--)
4463                 __free_page(spages[rc-1]);
4464         return -ENOMEM;
4465 }
4466
4467 struct nfs4_cached_acl {
4468         int cached;
4469         size_t len;
4470         char data[0];
4471 };
4472
4473 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4474 {
4475         struct nfs_inode *nfsi = NFS_I(inode);
4476
4477         spin_lock(&inode->i_lock);
4478         kfree(nfsi->nfs4_acl);
4479         nfsi->nfs4_acl = acl;
4480         spin_unlock(&inode->i_lock);
4481 }
4482
4483 static void nfs4_zap_acl_attr(struct inode *inode)
4484 {
4485         nfs4_set_cached_acl(inode, NULL);
4486 }
4487
4488 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4489 {
4490         struct nfs_inode *nfsi = NFS_I(inode);
4491         struct nfs4_cached_acl *acl;
4492         int ret = -ENOENT;
4493
4494         spin_lock(&inode->i_lock);
4495         acl = nfsi->nfs4_acl;
4496         if (acl == NULL)
4497                 goto out;
4498         if (buf == NULL) /* user is just asking for length */
4499                 goto out_len;
4500         if (acl->cached == 0)
4501                 goto out;
4502         ret = -ERANGE; /* see getxattr(2) man page */
4503         if (acl->len > buflen)
4504                 goto out;
4505         memcpy(buf, acl->data, acl->len);
4506 out_len:
4507         ret = acl->len;
4508 out:
4509         spin_unlock(&inode->i_lock);
4510         return ret;
4511 }
4512
4513 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4514 {
4515         struct nfs4_cached_acl *acl;
4516         size_t buflen = sizeof(*acl) + acl_len;
4517
4518         if (buflen <= PAGE_SIZE) {
4519                 acl = kmalloc(buflen, GFP_KERNEL);
4520                 if (acl == NULL)
4521                         goto out;
4522                 acl->cached = 1;
4523                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4524         } else {
4525                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4526                 if (acl == NULL)
4527                         goto out;
4528                 acl->cached = 0;
4529         }
4530         acl->len = acl_len;
4531 out:
4532         nfs4_set_cached_acl(inode, acl);
4533 }
4534
4535 /*
4536  * The getxattr API returns the required buffer length when called with a
4537  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4538  * the required buf.  On a NULL buf, we send a page of data to the server
4539  * guessing that the ACL request can be serviced by a page. If so, we cache
4540  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4541  * the cache. If not so, we throw away the page, and cache the required
4542  * length. The next getxattr call will then produce another round trip to
4543  * the server, this time with the input buf of the required size.
4544  */
4545 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4546 {
4547         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4548         struct nfs_getaclargs args = {
4549                 .fh = NFS_FH(inode),
4550                 .acl_pages = pages,
4551                 .acl_len = buflen,
4552         };
4553         struct nfs_getaclres res = {
4554                 .acl_len = buflen,
4555         };
4556         struct rpc_message msg = {
4557                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4558                 .rpc_argp = &args,
4559                 .rpc_resp = &res,
4560         };
4561         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4562         int ret = -ENOMEM, i;
4563
4564         /* As long as we're doing a round trip to the server anyway,
4565          * let's be prepared for a page of acl data. */
4566         if (npages == 0)
4567                 npages = 1;
4568         if (npages > ARRAY_SIZE(pages))
4569                 return -ERANGE;
4570
4571         for (i = 0; i < npages; i++) {
4572                 pages[i] = alloc_page(GFP_KERNEL);
4573                 if (!pages[i])
4574                         goto out_free;
4575         }
4576
4577         /* for decoding across pages */
4578         res.acl_scratch = alloc_page(GFP_KERNEL);
4579         if (!res.acl_scratch)
4580                 goto out_free;
4581
4582         args.acl_len = npages * PAGE_SIZE;
4583         args.acl_pgbase = 0;
4584
4585         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4586                 __func__, buf, buflen, npages, args.acl_len);
4587         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4588                              &msg, &args.seq_args, &res.seq_res, 0);
4589         if (ret)
4590                 goto out_free;
4591
4592         /* Handle the case where the passed-in buffer is too short */
4593         if (res.acl_flags & NFS4_ACL_TRUNC) {
4594                 /* Did the user only issue a request for the acl length? */
4595                 if (buf == NULL)
4596                         goto out_ok;
4597                 ret = -ERANGE;
4598                 goto out_free;
4599         }
4600         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4601         if (buf) {
4602                 if (res.acl_len > buflen) {
4603                         ret = -ERANGE;
4604                         goto out_free;
4605                 }
4606                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4607         }
4608 out_ok:
4609         ret = res.acl_len;
4610 out_free:
4611         for (i = 0; i < npages; i++)
4612                 if (pages[i])
4613                         __free_page(pages[i]);
4614         if (res.acl_scratch)
4615                 __free_page(res.acl_scratch);
4616         return ret;
4617 }
4618
4619 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4620 {
4621         struct nfs4_exception exception = { };
4622         ssize_t ret;
4623         do {
4624                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4625                 trace_nfs4_get_acl(inode, ret);
4626                 if (ret >= 0)
4627                         break;
4628                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4629         } while (exception.retry);
4630         return ret;
4631 }
4632
4633 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4634 {
4635         struct nfs_server *server = NFS_SERVER(inode);
4636         int ret;
4637
4638         if (!nfs4_server_supports_acls(server))
4639                 return -EOPNOTSUPP;
4640         ret = nfs_revalidate_inode(server, inode);
4641         if (ret < 0)
4642                 return ret;
4643         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4644                 nfs_zap_acl_cache(inode);
4645         ret = nfs4_read_cached_acl(inode, buf, buflen);
4646         if (ret != -ENOENT)
4647                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4648                  * but no cached acl data, just the acl length */
4649                 return ret;
4650         return nfs4_get_acl_uncached(inode, buf, buflen);
4651 }
4652
4653 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4654 {
4655         struct nfs_server *server = NFS_SERVER(inode);
4656         struct page *pages[NFS4ACL_MAXPAGES];
4657         struct nfs_setaclargs arg = {
4658                 .fh             = NFS_FH(inode),
4659                 .acl_pages      = pages,
4660                 .acl_len        = buflen,
4661         };
4662         struct nfs_setaclres res;
4663         struct rpc_message msg = {
4664                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4665                 .rpc_argp       = &arg,
4666                 .rpc_resp       = &res,
4667         };
4668         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4669         int ret, i;
4670
4671         if (!nfs4_server_supports_acls(server))
4672                 return -EOPNOTSUPP;
4673         if (npages > ARRAY_SIZE(pages))
4674                 return -ERANGE;
4675         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4676         if (i < 0)
4677                 return i;
4678         nfs4_inode_return_delegation(inode);
4679         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4680
4681         /*
4682          * Free each page after tx, so the only ref left is
4683          * held by the network stack
4684          */
4685         for (; i > 0; i--)
4686                 put_page(pages[i-1]);
4687
4688         /*
4689          * Acl update can result in inode attribute update.
4690          * so mark the attribute cache invalid.
4691          */
4692         spin_lock(&inode->i_lock);
4693         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4694         spin_unlock(&inode->i_lock);
4695         nfs_access_zap_cache(inode);
4696         nfs_zap_acl_cache(inode);
4697         return ret;
4698 }
4699
4700 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4701 {
4702         struct nfs4_exception exception = { };
4703         int err;
4704         do {
4705                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4706                 trace_nfs4_set_acl(inode, err);
4707                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4708                                 &exception);
4709         } while (exception.retry);
4710         return err;
4711 }
4712
4713 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4714 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4715                                         size_t buflen)
4716 {
4717         struct nfs_server *server = NFS_SERVER(inode);
4718         struct nfs_fattr fattr;
4719         struct nfs4_label label = {0, 0, buflen, buf};
4720
4721         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4722         struct nfs4_getattr_arg arg = {
4723                 .fh             = NFS_FH(inode),
4724                 .bitmask        = bitmask,
4725         };
4726         struct nfs4_getattr_res res = {
4727                 .fattr          = &fattr,
4728                 .label          = &label,
4729                 .server         = server,
4730         };
4731         struct rpc_message msg = {
4732                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4733                 .rpc_argp       = &arg,
4734                 .rpc_resp       = &res,
4735         };
4736         int ret;
4737
4738         nfs_fattr_init(&fattr);
4739
4740         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4741         if (ret)
4742                 return ret;
4743         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4744                 return -ENOENT;
4745         if (buflen < label.len)
4746                 return -ERANGE;
4747         return 0;
4748 }
4749
4750 static int nfs4_get_security_label(struct inode *inode, void *buf,
4751                                         size_t buflen)
4752 {
4753         struct nfs4_exception exception = { };
4754         int err;
4755
4756         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4757                 return -EOPNOTSUPP;
4758
4759         do {
4760                 err = _nfs4_get_security_label(inode, buf, buflen);
4761                 trace_nfs4_get_security_label(inode, err);
4762                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4763                                 &exception);
4764         } while (exception.retry);
4765         return err;
4766 }
4767
4768 static int _nfs4_do_set_security_label(struct inode *inode,
4769                 struct nfs4_label *ilabel,
4770                 struct nfs_fattr *fattr,
4771                 struct nfs4_label *olabel)
4772 {
4773
4774         struct iattr sattr = {0};
4775         struct nfs_server *server = NFS_SERVER(inode);
4776         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4777         struct nfs_setattrargs arg = {
4778                 .fh             = NFS_FH(inode),
4779                 .iap            = &sattr,
4780                 .server         = server,
4781                 .bitmask        = bitmask,
4782                 .label          = ilabel,
4783         };
4784         struct nfs_setattrres res = {
4785                 .fattr          = fattr,
4786                 .label          = olabel,
4787                 .server         = server,
4788         };
4789         struct rpc_message msg = {
4790                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4791                 .rpc_argp       = &arg,
4792                 .rpc_resp       = &res,
4793         };
4794         int status;
4795
4796         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4797
4798         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4799         if (status)
4800                 dprintk("%s failed: %d\n", __func__, status);
4801
4802         return status;
4803 }
4804
4805 static int nfs4_do_set_security_label(struct inode *inode,
4806                 struct nfs4_label *ilabel,
4807                 struct nfs_fattr *fattr,
4808                 struct nfs4_label *olabel)
4809 {
4810         struct nfs4_exception exception = { };
4811         int err;
4812
4813         do {
4814                 err = _nfs4_do_set_security_label(inode, ilabel,
4815                                 fattr, olabel);
4816                 trace_nfs4_set_security_label(inode, err);
4817                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4818                                 &exception);
4819         } while (exception.retry);
4820         return err;
4821 }
4822
4823 static int
4824 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4825 {
4826         struct nfs4_label ilabel, *olabel = NULL;
4827         struct nfs_fattr fattr;
4828         struct rpc_cred *cred;
4829         struct inode *inode = d_inode(dentry);
4830         int status;
4831
4832         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4833                 return -EOPNOTSUPP;
4834
4835         nfs_fattr_init(&fattr);
4836
4837         ilabel.pi = 0;
4838         ilabel.lfs = 0;
4839         ilabel.label = (char *)buf;
4840         ilabel.len = buflen;
4841
4842         cred = rpc_lookup_cred();
4843         if (IS_ERR(cred))
4844                 return PTR_ERR(cred);
4845
4846         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4847         if (IS_ERR(olabel)) {
4848                 status = -PTR_ERR(olabel);
4849                 goto out;
4850         }
4851
4852         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4853         if (status == 0)
4854                 nfs_setsecurity(inode, &fattr, olabel);
4855
4856         nfs4_label_free(olabel);
4857 out:
4858         put_rpccred(cred);
4859         return status;
4860 }
4861 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4862
4863
4864 static int
4865 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4866                         struct nfs4_state *state, long *timeout)
4867 {
4868         struct nfs_client *clp = server->nfs_client;
4869
4870         if (task->tk_status >= 0)
4871                 return 0;
4872         switch(task->tk_status) {
4873                 case -NFS4ERR_DELEG_REVOKED:
4874                 case -NFS4ERR_ADMIN_REVOKED:
4875                 case -NFS4ERR_BAD_STATEID:
4876                 case -NFS4ERR_OPENMODE:
4877                         if (state == NULL)
4878                                 break;
4879                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4880                                 goto recovery_failed;
4881                         goto wait_on_recovery;
4882                 case -NFS4ERR_EXPIRED:
4883                         if (state != NULL) {
4884                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4885                                         goto recovery_failed;
4886                         }
4887                 case -NFS4ERR_STALE_STATEID:
4888                 case -NFS4ERR_STALE_CLIENTID:
4889                         nfs4_schedule_lease_recovery(clp);
4890                         goto wait_on_recovery;
4891                 case -NFS4ERR_MOVED:
4892                         if (nfs4_schedule_migration_recovery(server) < 0)
4893                                 goto recovery_failed;
4894                         goto wait_on_recovery;
4895                 case -NFS4ERR_LEASE_MOVED:
4896                         nfs4_schedule_lease_moved_recovery(clp);
4897                         goto wait_on_recovery;
4898 #if defined(CONFIG_NFS_V4_1)
4899                 case -NFS4ERR_BADSESSION:
4900                 case -NFS4ERR_BADSLOT:
4901                 case -NFS4ERR_BAD_HIGH_SLOT:
4902                 case -NFS4ERR_DEADSESSION:
4903                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4904                 case -NFS4ERR_SEQ_FALSE_RETRY:
4905                 case -NFS4ERR_SEQ_MISORDERED:
4906                         dprintk("%s ERROR %d, Reset session\n", __func__,
4907                                 task->tk_status);
4908                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4909                         goto wait_on_recovery;
4910 #endif /* CONFIG_NFS_V4_1 */
4911                 case -NFS4ERR_DELAY:
4912                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4913                         rpc_delay(task, nfs4_update_delay(timeout));
4914                         goto restart_call;
4915                 case -NFS4ERR_GRACE:
4916                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4917                 case -NFS4ERR_RETRY_UNCACHED_REP:
4918                 case -NFS4ERR_OLD_STATEID:
4919                         goto restart_call;
4920         }
4921         task->tk_status = nfs4_map_errors(task->tk_status);
4922         return 0;
4923 recovery_failed:
4924         task->tk_status = -EIO;
4925         return 0;
4926 wait_on_recovery:
4927         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4928         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4929                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4930         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4931                 goto recovery_failed;
4932 restart_call:
4933         task->tk_status = 0;
4934         return -EAGAIN;
4935 }
4936
4937 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4938                                     nfs4_verifier *bootverf)
4939 {
4940         __be32 verf[2];
4941
4942         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4943                 /* An impossible timestamp guarantees this value
4944                  * will never match a generated boot time. */
4945                 verf[0] = 0;
4946                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4947         } else {
4948                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4949                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4950                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4951         }
4952         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4953 }
4954
4955 static int
4956 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
4957 {
4958         int result;
4959         size_t len;
4960         char *str;
4961         bool retried = false;
4962
4963         if (clp->cl_owner_id != NULL)
4964                 return 0;
4965 retry:
4966         rcu_read_lock();
4967         len = 10 + strlen(clp->cl_ipaddr) + 1 +
4968                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
4969                 1 +
4970                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
4971                 1;
4972         rcu_read_unlock();
4973
4974         if (len > NFS4_OPAQUE_LIMIT + 1)
4975                 return -EINVAL;
4976
4977         /*
4978          * Since this string is allocated at mount time, and held until the
4979          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
4980          * about a memory-reclaim deadlock.
4981          */
4982         str = kmalloc(len, GFP_KERNEL);
4983         if (!str)
4984                 return -ENOMEM;
4985
4986         rcu_read_lock();
4987         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
4988                         clp->cl_ipaddr,
4989                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
4990                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
4991         rcu_read_unlock();
4992
4993         /* Did something change? */
4994         if (result >= len) {
4995                 kfree(str);
4996                 if (retried)
4997                         return -EINVAL;
4998                 retried = true;
4999                 goto retry;
5000         }
5001         clp->cl_owner_id = str;
5002         return 0;
5003 }
5004
5005 static int
5006 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5007 {
5008         int result;
5009         size_t len;
5010         char *str;
5011
5012         len = 10 + 10 + 1 + 10 + 1 +
5013                 strlen(nfs4_client_id_uniquifier) + 1 +
5014                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5015
5016         if (len > NFS4_OPAQUE_LIMIT + 1)
5017                 return -EINVAL;
5018
5019         /*
5020          * Since this string is allocated at mount time, and held until the
5021          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5022          * about a memory-reclaim deadlock.
5023          */
5024         str = kmalloc(len, GFP_KERNEL);
5025         if (!str)
5026                 return -ENOMEM;
5027
5028         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5029                         clp->rpc_ops->version, clp->cl_minorversion,
5030                         nfs4_client_id_uniquifier,
5031                         clp->cl_rpcclient->cl_nodename);
5032         if (result >= len) {
5033                 kfree(str);
5034                 return -EINVAL;
5035         }
5036         clp->cl_owner_id = str;
5037         return 0;
5038 }
5039
5040 static int
5041 nfs4_init_uniform_client_string(struct nfs_client *clp)
5042 {
5043         int result;
5044         size_t len;
5045         char *str;
5046
5047         if (clp->cl_owner_id != NULL)
5048                 return 0;
5049
5050         if (nfs4_client_id_uniquifier[0] != '\0')
5051                 return nfs4_init_uniquifier_client_string(clp);
5052
5053         len = 10 + 10 + 1 + 10 + 1 +
5054                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5055
5056         if (len > NFS4_OPAQUE_LIMIT + 1)
5057                 return -EINVAL;
5058
5059         /*
5060          * Since this string is allocated at mount time, and held until the
5061          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5062          * about a memory-reclaim deadlock.
5063          */
5064         str = kmalloc(len, GFP_KERNEL);
5065         if (!str)
5066                 return -ENOMEM;
5067
5068         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5069                         clp->rpc_ops->version, clp->cl_minorversion,
5070                         clp->cl_rpcclient->cl_nodename);
5071         if (result >= len) {
5072                 kfree(str);
5073                 return -EINVAL;
5074         }
5075         clp->cl_owner_id = str;
5076         return 0;
5077 }
5078
5079 /*
5080  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5081  * services.  Advertise one based on the address family of the
5082  * clientaddr.
5083  */
5084 static unsigned int
5085 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5086 {
5087         if (strchr(clp->cl_ipaddr, ':') != NULL)
5088                 return scnprintf(buf, len, "tcp6");
5089         else
5090                 return scnprintf(buf, len, "tcp");
5091 }
5092
5093 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5094 {
5095         struct nfs4_setclientid *sc = calldata;
5096
5097         if (task->tk_status == 0)
5098                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5099 }
5100
5101 static const struct rpc_call_ops nfs4_setclientid_ops = {
5102         .rpc_call_done = nfs4_setclientid_done,
5103 };
5104
5105 /**
5106  * nfs4_proc_setclientid - Negotiate client ID
5107  * @clp: state data structure
5108  * @program: RPC program for NFSv4 callback service
5109  * @port: IP port number for NFS4 callback service
5110  * @cred: RPC credential to use for this call
5111  * @res: where to place the result
5112  *
5113  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5114  */
5115 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5116                 unsigned short port, struct rpc_cred *cred,
5117                 struct nfs4_setclientid_res *res)
5118 {
5119         nfs4_verifier sc_verifier;
5120         struct nfs4_setclientid setclientid = {
5121                 .sc_verifier = &sc_verifier,
5122                 .sc_prog = program,
5123                 .sc_clnt = clp,
5124         };
5125         struct rpc_message msg = {
5126                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5127                 .rpc_argp = &setclientid,
5128                 .rpc_resp = res,
5129                 .rpc_cred = cred,
5130         };
5131         struct rpc_task *task;
5132         struct rpc_task_setup task_setup_data = {
5133                 .rpc_client = clp->cl_rpcclient,
5134                 .rpc_message = &msg,
5135                 .callback_ops = &nfs4_setclientid_ops,
5136                 .callback_data = &setclientid,
5137                 .flags = RPC_TASK_TIMEOUT,
5138         };
5139         int status;
5140
5141         /* nfs_client_id4 */
5142         nfs4_init_boot_verifier(clp, &sc_verifier);
5143
5144         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5145                 status = nfs4_init_uniform_client_string(clp);
5146         else
5147                 status = nfs4_init_nonuniform_client_string(clp);
5148
5149         if (status)
5150                 goto out;
5151
5152         /* cb_client4 */
5153         setclientid.sc_netid_len =
5154                                 nfs4_init_callback_netid(clp,
5155                                                 setclientid.sc_netid,
5156                                                 sizeof(setclientid.sc_netid));
5157         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5158                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5159                                 clp->cl_ipaddr, port >> 8, port & 255);
5160
5161         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5162                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5163                 clp->cl_owner_id);
5164         task = rpc_run_task(&task_setup_data);
5165         if (IS_ERR(task)) {
5166                 status = PTR_ERR(task);
5167                 goto out;
5168         }
5169         status = task->tk_status;
5170         if (setclientid.sc_cred) {
5171                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5172                 put_rpccred(setclientid.sc_cred);
5173         }
5174         rpc_put_task(task);
5175 out:
5176         trace_nfs4_setclientid(clp, status);
5177         dprintk("NFS reply setclientid: %d\n", status);
5178         return status;
5179 }
5180
5181 /**
5182  * nfs4_proc_setclientid_confirm - Confirm client ID
5183  * @clp: state data structure
5184  * @res: result of a previous SETCLIENTID
5185  * @cred: RPC credential to use for this call
5186  *
5187  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5188  */
5189 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5190                 struct nfs4_setclientid_res *arg,
5191                 struct rpc_cred *cred)
5192 {
5193         struct rpc_message msg = {
5194                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5195                 .rpc_argp = arg,
5196                 .rpc_cred = cred,
5197         };
5198         int status;
5199
5200         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5201                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5202                 clp->cl_clientid);
5203         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5204         trace_nfs4_setclientid_confirm(clp, status);
5205         dprintk("NFS reply setclientid_confirm: %d\n", status);
5206         return status;
5207 }
5208
5209 struct nfs4_delegreturndata {
5210         struct nfs4_delegreturnargs args;
5211         struct nfs4_delegreturnres res;
5212         struct nfs_fh fh;
5213         nfs4_stateid stateid;
5214         unsigned long timestamp;
5215         struct nfs_fattr fattr;
5216         int rpc_status;
5217         struct inode *inode;
5218         bool roc;
5219         u32 roc_barrier;
5220 };
5221
5222 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5223 {
5224         struct nfs4_delegreturndata *data = calldata;
5225
5226         if (!nfs4_sequence_done(task, &data->res.seq_res))
5227                 return;
5228
5229         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5230         switch (task->tk_status) {
5231         case 0:
5232                 renew_lease(data->res.server, data->timestamp);
5233         case -NFS4ERR_ADMIN_REVOKED:
5234         case -NFS4ERR_DELEG_REVOKED:
5235         case -NFS4ERR_BAD_STATEID:
5236         case -NFS4ERR_OLD_STATEID:
5237         case -NFS4ERR_STALE_STATEID:
5238         case -NFS4ERR_EXPIRED:
5239                 task->tk_status = 0;
5240                 if (data->roc)
5241                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5242                 break;
5243         default:
5244                 if (nfs4_async_handle_error(task, data->res.server,
5245                                             NULL, NULL) == -EAGAIN) {
5246                         rpc_restart_call_prepare(task);
5247                         return;
5248                 }
5249         }
5250         data->rpc_status = task->tk_status;
5251 }
5252
5253 static void nfs4_delegreturn_release(void *calldata)
5254 {
5255         struct nfs4_delegreturndata *data = calldata;
5256         struct inode *inode = data->inode;
5257
5258         if (inode) {
5259                 if (data->roc)
5260                         pnfs_roc_release(inode);
5261                 nfs_iput_and_deactive(inode);
5262         }
5263         kfree(calldata);
5264 }
5265
5266 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5267 {
5268         struct nfs4_delegreturndata *d_data;
5269
5270         d_data = (struct nfs4_delegreturndata *)data;
5271
5272         if (d_data->roc &&
5273             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5274                 return;
5275
5276         nfs4_setup_sequence(d_data->res.server,
5277                         &d_data->args.seq_args,
5278                         &d_data->res.seq_res,
5279                         task);
5280 }
5281
5282 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5283         .rpc_call_prepare = nfs4_delegreturn_prepare,
5284         .rpc_call_done = nfs4_delegreturn_done,
5285         .rpc_release = nfs4_delegreturn_release,
5286 };
5287
5288 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5289 {
5290         struct nfs4_delegreturndata *data;
5291         struct nfs_server *server = NFS_SERVER(inode);
5292         struct rpc_task *task;
5293         struct rpc_message msg = {
5294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5295                 .rpc_cred = cred,
5296         };
5297         struct rpc_task_setup task_setup_data = {
5298                 .rpc_client = server->client,
5299                 .rpc_message = &msg,
5300                 .callback_ops = &nfs4_delegreturn_ops,
5301                 .flags = RPC_TASK_ASYNC,
5302         };
5303         int status = 0;
5304
5305         data = kzalloc(sizeof(*data), GFP_NOFS);
5306         if (data == NULL)
5307                 return -ENOMEM;
5308         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5309         data->args.fhandle = &data->fh;
5310         data->args.stateid = &data->stateid;
5311         data->args.bitmask = server->cache_consistency_bitmask;
5312         nfs_copy_fh(&data->fh, NFS_FH(inode));
5313         nfs4_stateid_copy(&data->stateid, stateid);
5314         data->res.fattr = &data->fattr;
5315         data->res.server = server;
5316         nfs_fattr_init(data->res.fattr);
5317         data->timestamp = jiffies;
5318         data->rpc_status = 0;
5319         data->inode = nfs_igrab_and_active(inode);
5320         if (data->inode)
5321                 data->roc = nfs4_roc(inode);
5322
5323         task_setup_data.callback_data = data;
5324         msg.rpc_argp = &data->args;
5325         msg.rpc_resp = &data->res;
5326         task = rpc_run_task(&task_setup_data);
5327         if (IS_ERR(task))
5328                 return PTR_ERR(task);
5329         if (!issync)
5330                 goto out;
5331         status = nfs4_wait_for_completion_rpc_task(task);
5332         if (status != 0)
5333                 goto out;
5334         status = data->rpc_status;
5335         if (status == 0)
5336                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5337         else
5338                 nfs_refresh_inode(inode, &data->fattr);
5339 out:
5340         rpc_put_task(task);
5341         return status;
5342 }
5343
5344 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5345 {
5346         struct nfs_server *server = NFS_SERVER(inode);
5347         struct nfs4_exception exception = { };
5348         int err;
5349         do {
5350                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5351                 trace_nfs4_delegreturn(inode, err);
5352                 switch (err) {
5353                         case -NFS4ERR_STALE_STATEID:
5354                         case -NFS4ERR_EXPIRED:
5355                         case 0:
5356                                 return 0;
5357                 }
5358                 err = nfs4_handle_exception(server, err, &exception);
5359         } while (exception.retry);
5360         return err;
5361 }
5362
5363 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5364 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5365
5366 /*
5367  * sleep, with exponential backoff, and retry the LOCK operation.
5368  */
5369 static unsigned long
5370 nfs4_set_lock_task_retry(unsigned long timeout)
5371 {
5372         freezable_schedule_timeout_killable_unsafe(timeout);
5373         timeout <<= 1;
5374         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5375                 return NFS4_LOCK_MAXTIMEOUT;
5376         return timeout;
5377 }
5378
5379 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5380 {
5381         struct inode *inode = state->inode;
5382         struct nfs_server *server = NFS_SERVER(inode);
5383         struct nfs_client *clp = server->nfs_client;
5384         struct nfs_lockt_args arg = {
5385                 .fh = NFS_FH(inode),
5386                 .fl = request,
5387         };
5388         struct nfs_lockt_res res = {
5389                 .denied = request,
5390         };
5391         struct rpc_message msg = {
5392                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5393                 .rpc_argp       = &arg,
5394                 .rpc_resp       = &res,
5395                 .rpc_cred       = state->owner->so_cred,
5396         };
5397         struct nfs4_lock_state *lsp;
5398         int status;
5399
5400         arg.lock_owner.clientid = clp->cl_clientid;
5401         status = nfs4_set_lock_state(state, request);
5402         if (status != 0)
5403                 goto out;
5404         lsp = request->fl_u.nfs4_fl.owner;
5405         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5406         arg.lock_owner.s_dev = server->s_dev;
5407         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5408         switch (status) {
5409                 case 0:
5410                         request->fl_type = F_UNLCK;
5411                         break;
5412                 case -NFS4ERR_DENIED:
5413                         status = 0;
5414         }
5415         request->fl_ops->fl_release_private(request);
5416         request->fl_ops = NULL;
5417 out:
5418         return status;
5419 }
5420
5421 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5422 {
5423         struct nfs4_exception exception = { };
5424         int err;
5425
5426         do {
5427                 err = _nfs4_proc_getlk(state, cmd, request);
5428                 trace_nfs4_get_lock(request, state, cmd, err);
5429                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5430                                 &exception);
5431         } while (exception.retry);
5432         return err;
5433 }
5434
5435 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5436 {
5437         int res = 0;
5438         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5439                 case FL_POSIX:
5440                         res = posix_lock_file_wait(file, fl);
5441                         break;
5442                 case FL_FLOCK:
5443                         res = flock_lock_file_wait(file, fl);
5444                         break;
5445                 default:
5446                         BUG();
5447         }
5448         return res;
5449 }
5450
5451 struct nfs4_unlockdata {
5452         struct nfs_locku_args arg;
5453         struct nfs_locku_res res;
5454         struct nfs4_lock_state *lsp;
5455         struct nfs_open_context *ctx;
5456         struct file_lock fl;
5457         const struct nfs_server *server;
5458         unsigned long timestamp;
5459 };
5460
5461 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5462                 struct nfs_open_context *ctx,
5463                 struct nfs4_lock_state *lsp,
5464                 struct nfs_seqid *seqid)
5465 {
5466         struct nfs4_unlockdata *p;
5467         struct inode *inode = lsp->ls_state->inode;
5468
5469         p = kzalloc(sizeof(*p), GFP_NOFS);
5470         if (p == NULL)
5471                 return NULL;
5472         p->arg.fh = NFS_FH(inode);
5473         p->arg.fl = &p->fl;
5474         p->arg.seqid = seqid;
5475         p->res.seqid = seqid;
5476         p->lsp = lsp;
5477         atomic_inc(&lsp->ls_count);
5478         /* Ensure we don't close file until we're done freeing locks! */
5479         p->ctx = get_nfs_open_context(ctx);
5480         memcpy(&p->fl, fl, sizeof(p->fl));
5481         p->server = NFS_SERVER(inode);
5482         return p;
5483 }
5484
5485 static void nfs4_locku_release_calldata(void *data)
5486 {
5487         struct nfs4_unlockdata *calldata = data;
5488         nfs_free_seqid(calldata->arg.seqid);
5489         nfs4_put_lock_state(calldata->lsp);
5490         put_nfs_open_context(calldata->ctx);
5491         kfree(calldata);
5492 }
5493
5494 static void nfs4_locku_done(struct rpc_task *task, void *data)
5495 {
5496         struct nfs4_unlockdata *calldata = data;
5497
5498         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5499                 return;
5500         switch (task->tk_status) {
5501                 case 0:
5502                         renew_lease(calldata->server, calldata->timestamp);
5503                         do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5504                         if (nfs4_update_lock_stateid(calldata->lsp,
5505                                         &calldata->res.stateid))
5506                                 break;
5507                 case -NFS4ERR_BAD_STATEID:
5508                 case -NFS4ERR_OLD_STATEID:
5509                 case -NFS4ERR_STALE_STATEID:
5510                 case -NFS4ERR_EXPIRED:
5511                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5512                                                 &calldata->lsp->ls_stateid))
5513                                 rpc_restart_call_prepare(task);
5514                         break;
5515                 default:
5516                         if (nfs4_async_handle_error(task, calldata->server,
5517                                                     NULL, NULL) == -EAGAIN)
5518                                 rpc_restart_call_prepare(task);
5519         }
5520         nfs_release_seqid(calldata->arg.seqid);
5521 }
5522
5523 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5524 {
5525         struct nfs4_unlockdata *calldata = data;
5526
5527         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5528                 goto out_wait;
5529         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5530         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5531                 /* Note: exit _without_ running nfs4_locku_done */
5532                 goto out_no_action;
5533         }
5534         calldata->timestamp = jiffies;
5535         if (nfs4_setup_sequence(calldata->server,
5536                                 &calldata->arg.seq_args,
5537                                 &calldata->res.seq_res,
5538                                 task) != 0)
5539                 nfs_release_seqid(calldata->arg.seqid);
5540         return;
5541 out_no_action:
5542         task->tk_action = NULL;
5543 out_wait:
5544         nfs4_sequence_done(task, &calldata->res.seq_res);
5545 }
5546
5547 static const struct rpc_call_ops nfs4_locku_ops = {
5548         .rpc_call_prepare = nfs4_locku_prepare,
5549         .rpc_call_done = nfs4_locku_done,
5550         .rpc_release = nfs4_locku_release_calldata,
5551 };
5552
5553 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5554                 struct nfs_open_context *ctx,
5555                 struct nfs4_lock_state *lsp,
5556                 struct nfs_seqid *seqid)
5557 {
5558         struct nfs4_unlockdata *data;
5559         struct rpc_message msg = {
5560                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5561                 .rpc_cred = ctx->cred,
5562         };
5563         struct rpc_task_setup task_setup_data = {
5564                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5565                 .rpc_message = &msg,
5566                 .callback_ops = &nfs4_locku_ops,
5567                 .workqueue = nfsiod_workqueue,
5568                 .flags = RPC_TASK_ASYNC,
5569         };
5570
5571         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5572                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5573
5574         /* Ensure this is an unlock - when canceling a lock, the
5575          * canceled lock is passed in, and it won't be an unlock.
5576          */
5577         fl->fl_type = F_UNLCK;
5578
5579         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5580         if (data == NULL) {
5581                 nfs_free_seqid(seqid);
5582                 return ERR_PTR(-ENOMEM);
5583         }
5584
5585         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5586         msg.rpc_argp = &data->arg;
5587         msg.rpc_resp = &data->res;
5588         task_setup_data.callback_data = data;
5589         return rpc_run_task(&task_setup_data);
5590 }
5591
5592 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5593 {
5594         struct inode *inode = state->inode;
5595         struct nfs4_state_owner *sp = state->owner;
5596         struct nfs_inode *nfsi = NFS_I(inode);
5597         struct nfs_seqid *seqid;
5598         struct nfs4_lock_state *lsp;
5599         struct rpc_task *task;
5600         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5601         int status = 0;
5602         unsigned char fl_flags = request->fl_flags;
5603
5604         status = nfs4_set_lock_state(state, request);
5605         /* Unlock _before_ we do the RPC call */
5606         request->fl_flags |= FL_EXISTS;
5607         /* Exclude nfs_delegation_claim_locks() */
5608         mutex_lock(&sp->so_delegreturn_mutex);
5609         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5610         down_read(&nfsi->rwsem);
5611         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5612                 up_read(&nfsi->rwsem);
5613                 mutex_unlock(&sp->so_delegreturn_mutex);
5614                 goto out;
5615         }
5616         up_read(&nfsi->rwsem);
5617         mutex_unlock(&sp->so_delegreturn_mutex);
5618         if (status != 0)
5619                 goto out;
5620         /* Is this a delegated lock? */
5621         lsp = request->fl_u.nfs4_fl.owner;
5622         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5623                 goto out;
5624         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5625         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5626         status = -ENOMEM;
5627         if (IS_ERR(seqid))
5628                 goto out;
5629         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5630         status = PTR_ERR(task);
5631         if (IS_ERR(task))
5632                 goto out;
5633         status = nfs4_wait_for_completion_rpc_task(task);
5634         rpc_put_task(task);
5635 out:
5636         request->fl_flags = fl_flags;
5637         trace_nfs4_unlock(request, state, F_SETLK, status);
5638         return status;
5639 }
5640
5641 struct nfs4_lockdata {
5642         struct nfs_lock_args arg;
5643         struct nfs_lock_res res;
5644         struct nfs4_lock_state *lsp;
5645         struct nfs_open_context *ctx;
5646         struct file_lock fl;
5647         unsigned long timestamp;
5648         int rpc_status;
5649         int cancelled;
5650         struct nfs_server *server;
5651 };
5652
5653 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5654                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5655                 gfp_t gfp_mask)
5656 {
5657         struct nfs4_lockdata *p;
5658         struct inode *inode = lsp->ls_state->inode;
5659         struct nfs_server *server = NFS_SERVER(inode);
5660         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5661
5662         p = kzalloc(sizeof(*p), gfp_mask);
5663         if (p == NULL)
5664                 return NULL;
5665
5666         p->arg.fh = NFS_FH(inode);
5667         p->arg.fl = &p->fl;
5668         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5669         if (IS_ERR(p->arg.open_seqid))
5670                 goto out_free;
5671         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5672         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5673         if (IS_ERR(p->arg.lock_seqid))
5674                 goto out_free_seqid;
5675         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5676         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5677         p->arg.lock_owner.s_dev = server->s_dev;
5678         p->res.lock_seqid = p->arg.lock_seqid;
5679         p->lsp = lsp;
5680         p->server = server;
5681         atomic_inc(&lsp->ls_count);
5682         p->ctx = get_nfs_open_context(ctx);
5683         get_file(fl->fl_file);
5684         memcpy(&p->fl, fl, sizeof(p->fl));
5685         return p;
5686 out_free_seqid:
5687         nfs_free_seqid(p->arg.open_seqid);
5688 out_free:
5689         kfree(p);
5690         return NULL;
5691 }
5692
5693 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5694 {
5695         struct nfs4_lockdata *data = calldata;
5696         struct nfs4_state *state = data->lsp->ls_state;
5697
5698         dprintk("%s: begin!\n", __func__);
5699         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5700                 goto out_wait;
5701         /* Do we need to do an open_to_lock_owner? */
5702         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5703                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5704                         goto out_release_lock_seqid;
5705                 }
5706                 nfs4_stateid_copy(&data->arg.open_stateid,
5707                                 &state->open_stateid);
5708                 data->arg.new_lock_owner = 1;
5709                 data->res.open_seqid = data->arg.open_seqid;
5710         } else {
5711                 data->arg.new_lock_owner = 0;
5712                 nfs4_stateid_copy(&data->arg.lock_stateid,
5713                                 &data->lsp->ls_stateid);
5714         }
5715         if (!nfs4_valid_open_stateid(state)) {
5716                 data->rpc_status = -EBADF;
5717                 task->tk_action = NULL;
5718                 goto out_release_open_seqid;
5719         }
5720         data->timestamp = jiffies;
5721         if (nfs4_setup_sequence(data->server,
5722                                 &data->arg.seq_args,
5723                                 &data->res.seq_res,
5724                                 task) == 0)
5725                 return;
5726 out_release_open_seqid:
5727         nfs_release_seqid(data->arg.open_seqid);
5728 out_release_lock_seqid:
5729         nfs_release_seqid(data->arg.lock_seqid);
5730 out_wait:
5731         nfs4_sequence_done(task, &data->res.seq_res);
5732         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5733 }
5734
5735 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5736 {
5737         struct nfs4_lockdata *data = calldata;
5738         struct nfs4_lock_state *lsp = data->lsp;
5739
5740         dprintk("%s: begin!\n", __func__);
5741
5742         if (!nfs4_sequence_done(task, &data->res.seq_res))
5743                 return;
5744
5745         data->rpc_status = task->tk_status;
5746         switch (task->tk_status) {
5747         case 0:
5748                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5749                                 data->timestamp);
5750                 if (data->arg.new_lock) {
5751                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5752                         if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5753                                 rpc_restart_call_prepare(task);
5754                                 break;
5755                         }
5756                 }
5757                 if (data->arg.new_lock_owner != 0) {
5758                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5759                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5760                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5761                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5762                         rpc_restart_call_prepare(task);
5763                 break;
5764         case -NFS4ERR_BAD_STATEID:
5765         case -NFS4ERR_OLD_STATEID:
5766         case -NFS4ERR_STALE_STATEID:
5767         case -NFS4ERR_EXPIRED:
5768                 if (data->arg.new_lock_owner != 0) {
5769                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5770                                                 &lsp->ls_state->open_stateid))
5771                                 rpc_restart_call_prepare(task);
5772                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5773                                                 &lsp->ls_stateid))
5774                                 rpc_restart_call_prepare(task);
5775         }
5776         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5777 }
5778
5779 static void nfs4_lock_release(void *calldata)
5780 {
5781         struct nfs4_lockdata *data = calldata;
5782
5783         dprintk("%s: begin!\n", __func__);
5784         nfs_free_seqid(data->arg.open_seqid);
5785         if (data->cancelled != 0) {
5786                 struct rpc_task *task;
5787                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5788                                 data->arg.lock_seqid);
5789                 if (!IS_ERR(task))
5790                         rpc_put_task_async(task);
5791                 dprintk("%s: cancelling lock!\n", __func__);
5792         } else
5793                 nfs_free_seqid(data->arg.lock_seqid);
5794         nfs4_put_lock_state(data->lsp);
5795         put_nfs_open_context(data->ctx);
5796         fput(data->fl.fl_file);
5797         kfree(data);
5798         dprintk("%s: done!\n", __func__);
5799 }
5800
5801 static const struct rpc_call_ops nfs4_lock_ops = {
5802         .rpc_call_prepare = nfs4_lock_prepare,
5803         .rpc_call_done = nfs4_lock_done,
5804         .rpc_release = nfs4_lock_release,
5805 };
5806
5807 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5808 {
5809         switch (error) {
5810         case -NFS4ERR_ADMIN_REVOKED:
5811         case -NFS4ERR_BAD_STATEID:
5812                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5813                 if (new_lock_owner != 0 ||
5814                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5815                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5816                 break;
5817         case -NFS4ERR_STALE_STATEID:
5818                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5819         case -NFS4ERR_EXPIRED:
5820                 nfs4_schedule_lease_recovery(server->nfs_client);
5821         };
5822 }
5823
5824 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5825 {
5826         struct nfs4_lockdata *data;
5827         struct rpc_task *task;
5828         struct rpc_message msg = {
5829                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5830                 .rpc_cred = state->owner->so_cred,
5831         };
5832         struct rpc_task_setup task_setup_data = {
5833                 .rpc_client = NFS_CLIENT(state->inode),
5834                 .rpc_message = &msg,
5835                 .callback_ops = &nfs4_lock_ops,
5836                 .workqueue = nfsiod_workqueue,
5837                 .flags = RPC_TASK_ASYNC,
5838         };
5839         int ret;
5840
5841         dprintk("%s: begin!\n", __func__);
5842         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5843                         fl->fl_u.nfs4_fl.owner,
5844                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5845         if (data == NULL)
5846                 return -ENOMEM;
5847         if (IS_SETLKW(cmd))
5848                 data->arg.block = 1;
5849         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5850         msg.rpc_argp = &data->arg;
5851         msg.rpc_resp = &data->res;
5852         task_setup_data.callback_data = data;
5853         if (recovery_type > NFS_LOCK_NEW) {
5854                 if (recovery_type == NFS_LOCK_RECLAIM)
5855                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5856                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5857         } else
5858                 data->arg.new_lock = 1;
5859         task = rpc_run_task(&task_setup_data);
5860         if (IS_ERR(task))
5861                 return PTR_ERR(task);
5862         ret = nfs4_wait_for_completion_rpc_task(task);
5863         if (ret == 0) {
5864                 ret = data->rpc_status;
5865                 if (ret)
5866                         nfs4_handle_setlk_error(data->server, data->lsp,
5867                                         data->arg.new_lock_owner, ret);
5868         } else
5869                 data->cancelled = 1;
5870         rpc_put_task(task);
5871         dprintk("%s: done, ret = %d!\n", __func__, ret);
5872         return ret;
5873 }
5874
5875 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5876 {
5877         struct nfs_server *server = NFS_SERVER(state->inode);
5878         struct nfs4_exception exception = {
5879                 .inode = state->inode,
5880         };
5881         int err;
5882
5883         do {
5884                 /* Cache the lock if possible... */
5885                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5886                         return 0;
5887                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5888                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5889                 if (err != -NFS4ERR_DELAY)
5890                         break;
5891                 nfs4_handle_exception(server, err, &exception);
5892         } while (exception.retry);
5893         return err;
5894 }
5895
5896 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5897 {
5898         struct nfs_server *server = NFS_SERVER(state->inode);
5899         struct nfs4_exception exception = {
5900                 .inode = state->inode,
5901         };
5902         int err;
5903
5904         err = nfs4_set_lock_state(state, request);
5905         if (err != 0)
5906                 return err;
5907         if (!recover_lost_locks) {
5908                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5909                 return 0;
5910         }
5911         do {
5912                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5913                         return 0;
5914                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5915                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5916                 switch (err) {
5917                 default:
5918                         goto out;
5919                 case -NFS4ERR_GRACE:
5920                 case -NFS4ERR_DELAY:
5921                         nfs4_handle_exception(server, err, &exception);
5922                         err = 0;
5923                 }
5924         } while (exception.retry);
5925 out:
5926         return err;
5927 }
5928
5929 #if defined(CONFIG_NFS_V4_1)
5930 /**
5931  * nfs41_check_expired_locks - possibly free a lock stateid
5932  *
5933  * @state: NFSv4 state for an inode
5934  *
5935  * Returns NFS_OK if recovery for this stateid is now finished.
5936  * Otherwise a negative NFS4ERR value is returned.
5937  */
5938 static int nfs41_check_expired_locks(struct nfs4_state *state)
5939 {
5940         int status, ret = -NFS4ERR_BAD_STATEID;
5941         struct nfs4_lock_state *lsp;
5942         struct nfs_server *server = NFS_SERVER(state->inode);
5943
5944         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5945                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5946                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5947
5948                         status = nfs41_test_stateid(server,
5949                                         &lsp->ls_stateid,
5950                                         cred);
5951                         trace_nfs4_test_lock_stateid(state, lsp, status);
5952                         if (status != NFS_OK) {
5953                                 /* Free the stateid unless the server
5954                                  * informs us the stateid is unrecognized. */
5955                                 if (status != -NFS4ERR_BAD_STATEID)
5956                                         nfs41_free_stateid(server,
5957                                                         &lsp->ls_stateid,
5958                                                         cred);
5959                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5960                                 ret = status;
5961                         }
5962                 }
5963         };
5964
5965         return ret;
5966 }
5967
5968 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5969 {
5970         int status = NFS_OK;
5971
5972         if (test_bit(LK_STATE_IN_USE, &state->flags))
5973                 status = nfs41_check_expired_locks(state);
5974         if (status != NFS_OK)
5975                 status = nfs4_lock_expired(state, request);
5976         return status;
5977 }
5978 #endif
5979
5980 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5981 {
5982         struct nfs_inode *nfsi = NFS_I(state->inode);
5983         unsigned char fl_flags = request->fl_flags;
5984         int status = -ENOLCK;
5985
5986         if ((fl_flags & FL_POSIX) &&
5987                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5988                 goto out;
5989         /* Is this a delegated open? */
5990         status = nfs4_set_lock_state(state, request);
5991         if (status != 0)
5992                 goto out;
5993         request->fl_flags |= FL_ACCESS;
5994         status = do_vfs_lock(request->fl_file, request);
5995         if (status < 0)
5996                 goto out;
5997         down_read(&nfsi->rwsem);
5998         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5999                 /* Yes: cache locks! */
6000                 /* ...but avoid races with delegation recall... */
6001                 request->fl_flags = fl_flags & ~FL_SLEEP;
6002                 status = do_vfs_lock(request->fl_file, request);
6003                 up_read(&nfsi->rwsem);
6004                 goto out;
6005         }
6006         up_read(&nfsi->rwsem);
6007         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6008 out:
6009         request->fl_flags = fl_flags;
6010         return status;
6011 }
6012
6013 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6014 {
6015         struct nfs4_exception exception = {
6016                 .state = state,
6017                 .inode = state->inode,
6018         };
6019         int err;
6020
6021         do {
6022                 err = _nfs4_proc_setlk(state, cmd, request);
6023                 trace_nfs4_set_lock(request, state, cmd, err);
6024                 if (err == -NFS4ERR_DENIED)
6025                         err = -EAGAIN;
6026                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6027                                 err, &exception);
6028         } while (exception.retry);
6029         return err;
6030 }
6031
6032 static int
6033 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6034 {
6035         struct nfs_open_context *ctx;
6036         struct nfs4_state *state;
6037         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6038         int status;
6039
6040         /* verify open state */
6041         ctx = nfs_file_open_context(filp);
6042         state = ctx->state;
6043
6044         if (request->fl_start < 0 || request->fl_end < 0)
6045                 return -EINVAL;
6046
6047         if (IS_GETLK(cmd)) {
6048                 if (state != NULL)
6049                         return nfs4_proc_getlk(state, F_GETLK, request);
6050                 return 0;
6051         }
6052
6053         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6054                 return -EINVAL;
6055
6056         if (request->fl_type == F_UNLCK) {
6057                 if (state != NULL)
6058                         return nfs4_proc_unlck(state, cmd, request);
6059                 return 0;
6060         }
6061
6062         if (state == NULL)
6063                 return -ENOLCK;
6064         /*
6065          * Don't rely on the VFS having checked the file open mode,
6066          * since it won't do this for flock() locks.
6067          */
6068         switch (request->fl_type) {
6069         case F_RDLCK:
6070                 if (!(filp->f_mode & FMODE_READ))
6071                         return -EBADF;
6072                 break;
6073         case F_WRLCK:
6074                 if (!(filp->f_mode & FMODE_WRITE))
6075                         return -EBADF;
6076         }
6077
6078         do {
6079                 status = nfs4_proc_setlk(state, cmd, request);
6080                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6081                         break;
6082                 timeout = nfs4_set_lock_task_retry(timeout);
6083                 status = -ERESTARTSYS;
6084                 if (signalled())
6085                         break;
6086         } while(status < 0);
6087         return status;
6088 }
6089
6090 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6091 {
6092         struct nfs_server *server = NFS_SERVER(state->inode);
6093         int err;
6094
6095         err = nfs4_set_lock_state(state, fl);
6096         if (err != 0)
6097                 return err;
6098         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6099         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6100 }
6101
6102 struct nfs_release_lockowner_data {
6103         struct nfs4_lock_state *lsp;
6104         struct nfs_server *server;
6105         struct nfs_release_lockowner_args args;
6106         struct nfs_release_lockowner_res res;
6107         unsigned long timestamp;
6108 };
6109
6110 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6111 {
6112         struct nfs_release_lockowner_data *data = calldata;
6113         struct nfs_server *server = data->server;
6114         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6115                              &data->args.seq_args, &data->res.seq_res, task);
6116         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6117         data->timestamp = jiffies;
6118 }
6119
6120 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6121 {
6122         struct nfs_release_lockowner_data *data = calldata;
6123         struct nfs_server *server = data->server;
6124
6125         nfs40_sequence_done(task, &data->res.seq_res);
6126
6127         switch (task->tk_status) {
6128         case 0:
6129                 renew_lease(server, data->timestamp);
6130                 break;
6131         case -NFS4ERR_STALE_CLIENTID:
6132         case -NFS4ERR_EXPIRED:
6133                 nfs4_schedule_lease_recovery(server->nfs_client);
6134                 break;
6135         case -NFS4ERR_LEASE_MOVED:
6136         case -NFS4ERR_DELAY:
6137                 if (nfs4_async_handle_error(task, server,
6138                                             NULL, NULL) == -EAGAIN)
6139                         rpc_restart_call_prepare(task);
6140         }
6141 }
6142
6143 static void nfs4_release_lockowner_release(void *calldata)
6144 {
6145         struct nfs_release_lockowner_data *data = calldata;
6146         nfs4_free_lock_state(data->server, data->lsp);
6147         kfree(calldata);
6148 }
6149
6150 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6151         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6152         .rpc_call_done = nfs4_release_lockowner_done,
6153         .rpc_release = nfs4_release_lockowner_release,
6154 };
6155
6156 static void
6157 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6158 {
6159         struct nfs_release_lockowner_data *data;
6160         struct rpc_message msg = {
6161                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6162         };
6163
6164         if (server->nfs_client->cl_mvops->minor_version != 0)
6165                 return;
6166
6167         data = kmalloc(sizeof(*data), GFP_NOFS);
6168         if (!data)
6169                 return;
6170         data->lsp = lsp;
6171         data->server = server;
6172         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6173         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6174         data->args.lock_owner.s_dev = server->s_dev;
6175
6176         msg.rpc_argp = &data->args;
6177         msg.rpc_resp = &data->res;
6178         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6179         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6180 }
6181
6182 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6183
6184 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6185                                    const void *buf, size_t buflen,
6186                                    int flags, int type)
6187 {
6188         if (strcmp(key, "") != 0)
6189                 return -EINVAL;
6190
6191         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6192 }
6193
6194 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6195                                    void *buf, size_t buflen, int type)
6196 {
6197         if (strcmp(key, "") != 0)
6198                 return -EINVAL;
6199
6200         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6201 }
6202
6203 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6204                                        size_t list_len, const char *name,
6205                                        size_t name_len, int type)
6206 {
6207         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6208
6209         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6210                 return 0;
6211
6212         if (list && len <= list_len)
6213                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6214         return len;
6215 }
6216
6217 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6218 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6219 {
6220         return server->caps & NFS_CAP_SECURITY_LABEL;
6221 }
6222
6223 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6224                                    const void *buf, size_t buflen,
6225                                    int flags, int type)
6226 {
6227         if (security_ismaclabel(key))
6228                 return nfs4_set_security_label(dentry, buf, buflen);
6229
6230         return -EOPNOTSUPP;
6231 }
6232
6233 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6234                                    void *buf, size_t buflen, int type)
6235 {
6236         if (security_ismaclabel(key))
6237                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6238         return -EOPNOTSUPP;
6239 }
6240
6241 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6242                                        size_t list_len, const char *name,
6243                                        size_t name_len, int type)
6244 {
6245         size_t len = 0;
6246
6247         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6248                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6249                 if (list && len <= list_len)
6250                         security_inode_listsecurity(d_inode(dentry), list, len);
6251         }
6252         return len;
6253 }
6254
6255 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6256         .prefix = XATTR_SECURITY_PREFIX,
6257         .list   = nfs4_xattr_list_nfs4_label,
6258         .get    = nfs4_xattr_get_nfs4_label,
6259         .set    = nfs4_xattr_set_nfs4_label,
6260 };
6261 #endif
6262
6263
6264 /*
6265  * nfs_fhget will use either the mounted_on_fileid or the fileid
6266  */
6267 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6268 {
6269         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6270                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6271               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6272               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6273                 return;
6274
6275         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6276                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6277         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6278         fattr->nlink = 2;
6279 }
6280
6281 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6282                                    const struct qstr *name,
6283                                    struct nfs4_fs_locations *fs_locations,
6284                                    struct page *page)
6285 {
6286         struct nfs_server *server = NFS_SERVER(dir);
6287         u32 bitmask[3] = {
6288                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6289         };
6290         struct nfs4_fs_locations_arg args = {
6291                 .dir_fh = NFS_FH(dir),
6292                 .name = name,
6293                 .page = page,
6294                 .bitmask = bitmask,
6295         };
6296         struct nfs4_fs_locations_res res = {
6297                 .fs_locations = fs_locations,
6298         };
6299         struct rpc_message msg = {
6300                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6301                 .rpc_argp = &args,
6302                 .rpc_resp = &res,
6303         };
6304         int status;
6305
6306         dprintk("%s: start\n", __func__);
6307
6308         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6309          * is not supported */
6310         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6311                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6312         else
6313                 bitmask[0] |= FATTR4_WORD0_FILEID;
6314
6315         nfs_fattr_init(&fs_locations->fattr);
6316         fs_locations->server = server;
6317         fs_locations->nlocations = 0;
6318         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6319         dprintk("%s: returned status = %d\n", __func__, status);
6320         return status;
6321 }
6322
6323 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6324                            const struct qstr *name,
6325                            struct nfs4_fs_locations *fs_locations,
6326                            struct page *page)
6327 {
6328         struct nfs4_exception exception = { };
6329         int err;
6330         do {
6331                 err = _nfs4_proc_fs_locations(client, dir, name,
6332                                 fs_locations, page);
6333                 trace_nfs4_get_fs_locations(dir, name, err);
6334                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6335                                 &exception);
6336         } while (exception.retry);
6337         return err;
6338 }
6339
6340 /*
6341  * This operation also signals the server that this client is
6342  * performing migration recovery.  The server can stop returning
6343  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6344  * appended to this compound to identify the client ID which is
6345  * performing recovery.
6346  */
6347 static int _nfs40_proc_get_locations(struct inode *inode,
6348                                      struct nfs4_fs_locations *locations,
6349                                      struct page *page, struct rpc_cred *cred)
6350 {
6351         struct nfs_server *server = NFS_SERVER(inode);
6352         struct rpc_clnt *clnt = server->client;
6353         u32 bitmask[2] = {
6354                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6355         };
6356         struct nfs4_fs_locations_arg args = {
6357                 .clientid       = server->nfs_client->cl_clientid,
6358                 .fh             = NFS_FH(inode),
6359                 .page           = page,
6360                 .bitmask        = bitmask,
6361                 .migration      = 1,            /* skip LOOKUP */
6362                 .renew          = 1,            /* append RENEW */
6363         };
6364         struct nfs4_fs_locations_res res = {
6365                 .fs_locations   = locations,
6366                 .migration      = 1,
6367                 .renew          = 1,
6368         };
6369         struct rpc_message msg = {
6370                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6371                 .rpc_argp       = &args,
6372                 .rpc_resp       = &res,
6373                 .rpc_cred       = cred,
6374         };
6375         unsigned long now = jiffies;
6376         int status;
6377
6378         nfs_fattr_init(&locations->fattr);
6379         locations->server = server;
6380         locations->nlocations = 0;
6381
6382         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6383         nfs4_set_sequence_privileged(&args.seq_args);
6384         status = nfs4_call_sync_sequence(clnt, server, &msg,
6385                                         &args.seq_args, &res.seq_res);
6386         if (status)
6387                 return status;
6388
6389         renew_lease(server, now);
6390         return 0;
6391 }
6392
6393 #ifdef CONFIG_NFS_V4_1
6394
6395 /*
6396  * This operation also signals the server that this client is
6397  * performing migration recovery.  The server can stop asserting
6398  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6399  * performing this operation is identified in the SEQUENCE
6400  * operation in this compound.
6401  *
6402  * When the client supports GETATTR(fs_locations_info), it can
6403  * be plumbed in here.
6404  */
6405 static int _nfs41_proc_get_locations(struct inode *inode,
6406                                      struct nfs4_fs_locations *locations,
6407                                      struct page *page, struct rpc_cred *cred)
6408 {
6409         struct nfs_server *server = NFS_SERVER(inode);
6410         struct rpc_clnt *clnt = server->client;
6411         u32 bitmask[2] = {
6412                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6413         };
6414         struct nfs4_fs_locations_arg args = {
6415                 .fh             = NFS_FH(inode),
6416                 .page           = page,
6417                 .bitmask        = bitmask,
6418                 .migration      = 1,            /* skip LOOKUP */
6419         };
6420         struct nfs4_fs_locations_res res = {
6421                 .fs_locations   = locations,
6422                 .migration      = 1,
6423         };
6424         struct rpc_message msg = {
6425                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6426                 .rpc_argp       = &args,
6427                 .rpc_resp       = &res,
6428                 .rpc_cred       = cred,
6429         };
6430         int status;
6431
6432         nfs_fattr_init(&locations->fattr);
6433         locations->server = server;
6434         locations->nlocations = 0;
6435
6436         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6437         nfs4_set_sequence_privileged(&args.seq_args);
6438         status = nfs4_call_sync_sequence(clnt, server, &msg,
6439                                         &args.seq_args, &res.seq_res);
6440         if (status == NFS4_OK &&
6441             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6442                 status = -NFS4ERR_LEASE_MOVED;
6443         return status;
6444 }
6445
6446 #endif  /* CONFIG_NFS_V4_1 */
6447
6448 /**
6449  * nfs4_proc_get_locations - discover locations for a migrated FSID
6450  * @inode: inode on FSID that is migrating
6451  * @locations: result of query
6452  * @page: buffer
6453  * @cred: credential to use for this operation
6454  *
6455  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6456  * operation failed, or a negative errno if a local error occurred.
6457  *
6458  * On success, "locations" is filled in, but if the server has
6459  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6460  * asserted.
6461  *
6462  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6463  * from this client that require migration recovery.
6464  */
6465 int nfs4_proc_get_locations(struct inode *inode,
6466                             struct nfs4_fs_locations *locations,
6467                             struct page *page, struct rpc_cred *cred)
6468 {
6469         struct nfs_server *server = NFS_SERVER(inode);
6470         struct nfs_client *clp = server->nfs_client;
6471         const struct nfs4_mig_recovery_ops *ops =
6472                                         clp->cl_mvops->mig_recovery_ops;
6473         struct nfs4_exception exception = { };
6474         int status;
6475
6476         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6477                 (unsigned long long)server->fsid.major,
6478                 (unsigned long long)server->fsid.minor,
6479                 clp->cl_hostname);
6480         nfs_display_fhandle(NFS_FH(inode), __func__);
6481
6482         do {
6483                 status = ops->get_locations(inode, locations, page, cred);
6484                 if (status != -NFS4ERR_DELAY)
6485                         break;
6486                 nfs4_handle_exception(server, status, &exception);
6487         } while (exception.retry);
6488         return status;
6489 }
6490
6491 /*
6492  * This operation also signals the server that this client is
6493  * performing "lease moved" recovery.  The server can stop
6494  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6495  * is appended to this compound to identify the client ID which is
6496  * performing recovery.
6497  */
6498 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6499 {
6500         struct nfs_server *server = NFS_SERVER(inode);
6501         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6502         struct rpc_clnt *clnt = server->client;
6503         struct nfs4_fsid_present_arg args = {
6504                 .fh             = NFS_FH(inode),
6505                 .clientid       = clp->cl_clientid,
6506                 .renew          = 1,            /* append RENEW */
6507         };
6508         struct nfs4_fsid_present_res res = {
6509                 .renew          = 1,
6510         };
6511         struct rpc_message msg = {
6512                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6513                 .rpc_argp       = &args,
6514                 .rpc_resp       = &res,
6515                 .rpc_cred       = cred,
6516         };
6517         unsigned long now = jiffies;
6518         int status;
6519
6520         res.fh = nfs_alloc_fhandle();
6521         if (res.fh == NULL)
6522                 return -ENOMEM;
6523
6524         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6525         nfs4_set_sequence_privileged(&args.seq_args);
6526         status = nfs4_call_sync_sequence(clnt, server, &msg,
6527                                                 &args.seq_args, &res.seq_res);
6528         nfs_free_fhandle(res.fh);
6529         if (status)
6530                 return status;
6531
6532         do_renew_lease(clp, now);
6533         return 0;
6534 }
6535
6536 #ifdef CONFIG_NFS_V4_1
6537
6538 /*
6539  * This operation also signals the server that this client is
6540  * performing "lease moved" recovery.  The server can stop asserting
6541  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6542  * this operation is identified in the SEQUENCE operation in this
6543  * compound.
6544  */
6545 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6546 {
6547         struct nfs_server *server = NFS_SERVER(inode);
6548         struct rpc_clnt *clnt = server->client;
6549         struct nfs4_fsid_present_arg args = {
6550                 .fh             = NFS_FH(inode),
6551         };
6552         struct nfs4_fsid_present_res res = {
6553         };
6554         struct rpc_message msg = {
6555                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6556                 .rpc_argp       = &args,
6557                 .rpc_resp       = &res,
6558                 .rpc_cred       = cred,
6559         };
6560         int status;
6561
6562         res.fh = nfs_alloc_fhandle();
6563         if (res.fh == NULL)
6564                 return -ENOMEM;
6565
6566         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6567         nfs4_set_sequence_privileged(&args.seq_args);
6568         status = nfs4_call_sync_sequence(clnt, server, &msg,
6569                                                 &args.seq_args, &res.seq_res);
6570         nfs_free_fhandle(res.fh);
6571         if (status == NFS4_OK &&
6572             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6573                 status = -NFS4ERR_LEASE_MOVED;
6574         return status;
6575 }
6576
6577 #endif  /* CONFIG_NFS_V4_1 */
6578
6579 /**
6580  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6581  * @inode: inode on FSID to check
6582  * @cred: credential to use for this operation
6583  *
6584  * Server indicates whether the FSID is present, moved, or not
6585  * recognized.  This operation is necessary to clear a LEASE_MOVED
6586  * condition for this client ID.
6587  *
6588  * Returns NFS4_OK if the FSID is present on this server,
6589  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6590  *  NFS4ERR code if some error occurred on the server, or a
6591  *  negative errno if a local failure occurred.
6592  */
6593 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6594 {
6595         struct nfs_server *server = NFS_SERVER(inode);
6596         struct nfs_client *clp = server->nfs_client;
6597         const struct nfs4_mig_recovery_ops *ops =
6598                                         clp->cl_mvops->mig_recovery_ops;
6599         struct nfs4_exception exception = { };
6600         int status;
6601
6602         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6603                 (unsigned long long)server->fsid.major,
6604                 (unsigned long long)server->fsid.minor,
6605                 clp->cl_hostname);
6606         nfs_display_fhandle(NFS_FH(inode), __func__);
6607
6608         do {
6609                 status = ops->fsid_present(inode, cred);
6610                 if (status != -NFS4ERR_DELAY)
6611                         break;
6612                 nfs4_handle_exception(server, status, &exception);
6613         } while (exception.retry);
6614         return status;
6615 }
6616
6617 /**
6618  * If 'use_integrity' is true and the state managment nfs_client
6619  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6620  * and the machine credential as per RFC3530bis and RFC5661 Security
6621  * Considerations sections. Otherwise, just use the user cred with the
6622  * filesystem's rpc_client.
6623  */
6624 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6625 {
6626         int status;
6627         struct nfs4_secinfo_arg args = {
6628                 .dir_fh = NFS_FH(dir),
6629                 .name   = name,
6630         };
6631         struct nfs4_secinfo_res res = {
6632                 .flavors     = flavors,
6633         };
6634         struct rpc_message msg = {
6635                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6636                 .rpc_argp = &args,
6637                 .rpc_resp = &res,
6638         };
6639         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6640         struct rpc_cred *cred = NULL;
6641
6642         if (use_integrity) {
6643                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6644                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6645                 msg.rpc_cred = cred;
6646         }
6647
6648         dprintk("NFS call  secinfo %s\n", name->name);
6649
6650         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6651                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6652
6653         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6654                                 &res.seq_res, 0);
6655         dprintk("NFS reply  secinfo: %d\n", status);
6656
6657         if (cred)
6658                 put_rpccred(cred);
6659
6660         return status;
6661 }
6662
6663 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6664                       struct nfs4_secinfo_flavors *flavors)
6665 {
6666         struct nfs4_exception exception = { };
6667         int err;
6668         do {
6669                 err = -NFS4ERR_WRONGSEC;
6670
6671                 /* try to use integrity protection with machine cred */
6672                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6673                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6674
6675                 /*
6676                  * if unable to use integrity protection, or SECINFO with
6677                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6678                  * disallowed by spec, but exists in deployed servers) use
6679                  * the current filesystem's rpc_client and the user cred.
6680                  */
6681                 if (err == -NFS4ERR_WRONGSEC)
6682                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6683
6684                 trace_nfs4_secinfo(dir, name, err);
6685                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6686                                 &exception);
6687         } while (exception.retry);
6688         return err;
6689 }
6690
6691 #ifdef CONFIG_NFS_V4_1
6692 /*
6693  * Check the exchange flags returned by the server for invalid flags, having
6694  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6695  * DS flags set.
6696  */
6697 static int nfs4_check_cl_exchange_flags(u32 flags)
6698 {
6699         if (flags & ~EXCHGID4_FLAG_MASK_R)
6700                 goto out_inval;
6701         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6702             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6703                 goto out_inval;
6704         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6705                 goto out_inval;
6706         return NFS_OK;
6707 out_inval:
6708         return -NFS4ERR_INVAL;
6709 }
6710
6711 static bool
6712 nfs41_same_server_scope(struct nfs41_server_scope *a,
6713                         struct nfs41_server_scope *b)
6714 {
6715         if (a->server_scope_sz == b->server_scope_sz &&
6716             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6717                 return true;
6718
6719         return false;
6720 }
6721
6722 /*
6723  * nfs4_proc_bind_conn_to_session()
6724  *
6725  * The 4.1 client currently uses the same TCP connection for the
6726  * fore and backchannel.
6727  */
6728 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6729 {
6730         int status;
6731         struct nfs41_bind_conn_to_session_args args = {
6732                 .client = clp,
6733                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6734         };
6735         struct nfs41_bind_conn_to_session_res res;
6736         struct rpc_message msg = {
6737                 .rpc_proc =
6738                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6739                 .rpc_argp = &args,
6740                 .rpc_resp = &res,
6741                 .rpc_cred = cred,
6742         };
6743
6744         dprintk("--> %s\n", __func__);
6745
6746         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6747         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6748                 args.dir = NFS4_CDFC4_FORE;
6749
6750         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6751         trace_nfs4_bind_conn_to_session(clp, status);
6752         if (status == 0) {
6753                 if (memcmp(res.sessionid.data,
6754                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6755                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6756                         status = -EIO;
6757                         goto out;
6758                 }
6759                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6760                         dprintk("NFS: %s: Unexpected direction from server\n",
6761                                 __func__);
6762                         status = -EIO;
6763                         goto out;
6764                 }
6765                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6766                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6767                                 __func__);
6768                         status = -EIO;
6769                         goto out;
6770                 }
6771         }
6772 out:
6773         dprintk("<-- %s status= %d\n", __func__, status);
6774         return status;
6775 }
6776
6777 /*
6778  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6779  * and operations we'd like to see to enable certain features in the allow map
6780  */
6781 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6782         .how = SP4_MACH_CRED,
6783         .enforce.u.words = {
6784                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6785                       1 << (OP_EXCHANGE_ID - 32) |
6786                       1 << (OP_CREATE_SESSION - 32) |
6787                       1 << (OP_DESTROY_SESSION - 32) |
6788                       1 << (OP_DESTROY_CLIENTID - 32)
6789         },
6790         .allow.u.words = {
6791                 [0] = 1 << (OP_CLOSE) |
6792                       1 << (OP_LOCKU) |
6793                       1 << (OP_COMMIT),
6794                 [1] = 1 << (OP_SECINFO - 32) |
6795                       1 << (OP_SECINFO_NO_NAME - 32) |
6796                       1 << (OP_TEST_STATEID - 32) |
6797                       1 << (OP_FREE_STATEID - 32) |
6798                       1 << (OP_WRITE - 32)
6799         }
6800 };
6801
6802 /*
6803  * Select the state protection mode for client `clp' given the server results
6804  * from exchange_id in `sp'.
6805  *
6806  * Returns 0 on success, negative errno otherwise.
6807  */
6808 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6809                                  struct nfs41_state_protection *sp)
6810 {
6811         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6812                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6813                       1 << (OP_EXCHANGE_ID - 32) |
6814                       1 << (OP_CREATE_SESSION - 32) |
6815                       1 << (OP_DESTROY_SESSION - 32) |
6816                       1 << (OP_DESTROY_CLIENTID - 32)
6817         };
6818         unsigned int i;
6819
6820         if (sp->how == SP4_MACH_CRED) {
6821                 /* Print state protect result */
6822                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6823                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6824                         if (test_bit(i, sp->enforce.u.longs))
6825                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6826                         if (test_bit(i, sp->allow.u.longs))
6827                                 dfprintk(MOUNT, "  allow op %d\n", i);
6828                 }
6829
6830                 /* make sure nothing is on enforce list that isn't supported */
6831                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6832                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6833                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6834                                 return -EINVAL;
6835                         }
6836                 }
6837
6838                 /*
6839                  * Minimal mode - state operations are allowed to use machine
6840                  * credential.  Note this already happens by default, so the
6841                  * client doesn't have to do anything more than the negotiation.
6842                  *
6843                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6844                  *       we're already using the machine cred for exchange_id
6845                  *       and will never use a different cred.
6846                  */
6847                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6848                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6849                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6850                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6851                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6852                         dfprintk(MOUNT, "  minimal mode enabled\n");
6853                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6854                 } else {
6855                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6856                         return -EINVAL;
6857                 }
6858
6859                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6860                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6861                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6862                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6863                 }
6864
6865                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6866                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6867                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6868                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6869                 }
6870
6871                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6872                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6873                         dfprintk(MOUNT, "  stateid mode enabled\n");
6874                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6875                 }
6876
6877                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6878                         dfprintk(MOUNT, "  write mode enabled\n");
6879                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6880                 }
6881
6882                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6883                         dfprintk(MOUNT, "  commit mode enabled\n");
6884                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6885                 }
6886         }
6887
6888         return 0;
6889 }
6890
6891 /*
6892  * _nfs4_proc_exchange_id()
6893  *
6894  * Wrapper for EXCHANGE_ID operation.
6895  */
6896 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6897         u32 sp4_how)
6898 {
6899         nfs4_verifier verifier;
6900         struct nfs41_exchange_id_args args = {
6901                 .verifier = &verifier,
6902                 .client = clp,
6903 #ifdef CONFIG_NFS_V4_1_MIGRATION
6904                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6905                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6906                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6907 #else
6908                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6909                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6910 #endif
6911         };
6912         struct nfs41_exchange_id_res res = {
6913                 0
6914         };
6915         int status;
6916         struct rpc_message msg = {
6917                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6918                 .rpc_argp = &args,
6919                 .rpc_resp = &res,
6920                 .rpc_cred = cred,
6921         };
6922
6923         nfs4_init_boot_verifier(clp, &verifier);
6924
6925         status = nfs4_init_uniform_client_string(clp);
6926         if (status)
6927                 goto out;
6928
6929         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6930                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6931                 clp->cl_owner_id);
6932
6933         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6934                                         GFP_NOFS);
6935         if (unlikely(res.server_owner == NULL)) {
6936                 status = -ENOMEM;
6937                 goto out;
6938         }
6939
6940         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6941                                         GFP_NOFS);
6942         if (unlikely(res.server_scope == NULL)) {
6943                 status = -ENOMEM;
6944                 goto out_server_owner;
6945         }
6946
6947         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6948         if (unlikely(res.impl_id == NULL)) {
6949                 status = -ENOMEM;
6950                 goto out_server_scope;
6951         }
6952
6953         switch (sp4_how) {
6954         case SP4_NONE:
6955                 args.state_protect.how = SP4_NONE;
6956                 break;
6957
6958         case SP4_MACH_CRED:
6959                 args.state_protect = nfs4_sp4_mach_cred_request;
6960                 break;
6961
6962         default:
6963                 /* unsupported! */
6964                 WARN_ON_ONCE(1);
6965                 status = -EINVAL;
6966                 goto out_server_scope;
6967         }
6968
6969         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6970         trace_nfs4_exchange_id(clp, status);
6971         if (status == 0)
6972                 status = nfs4_check_cl_exchange_flags(res.flags);
6973
6974         if (status == 0)
6975                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6976
6977         if (status == 0) {
6978                 clp->cl_clientid = res.clientid;
6979                 clp->cl_exchange_flags = res.flags;
6980                 /* Client ID is not confirmed */
6981                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6982                         clear_bit(NFS4_SESSION_ESTABLISHED,
6983                                         &clp->cl_session->session_state);
6984                         clp->cl_seqid = res.seqid;
6985                 }
6986
6987                 kfree(clp->cl_serverowner);
6988                 clp->cl_serverowner = res.server_owner;
6989                 res.server_owner = NULL;
6990
6991                 /* use the most recent implementation id */
6992                 kfree(clp->cl_implid);
6993                 clp->cl_implid = res.impl_id;
6994
6995                 if (clp->cl_serverscope != NULL &&
6996                     !nfs41_same_server_scope(clp->cl_serverscope,
6997                                              res.server_scope)) {
6998                         dprintk("%s: server_scope mismatch detected\n",
6999                                 __func__);
7000                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7001                         kfree(clp->cl_serverscope);
7002                         clp->cl_serverscope = NULL;
7003                 }
7004
7005                 if (clp->cl_serverscope == NULL) {
7006                         clp->cl_serverscope = res.server_scope;
7007                         goto out;
7008                 }
7009         } else
7010                 kfree(res.impl_id);
7011
7012 out_server_owner:
7013         kfree(res.server_owner);
7014 out_server_scope:
7015         kfree(res.server_scope);
7016 out:
7017         if (clp->cl_implid != NULL)
7018                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7019                         "domain: %s, name: %s, date: %llu,%u\n",
7020                         clp->cl_implid->domain, clp->cl_implid->name,
7021                         clp->cl_implid->date.seconds,
7022                         clp->cl_implid->date.nseconds);
7023         dprintk("NFS reply exchange_id: %d\n", status);
7024         return status;
7025 }
7026
7027 /*
7028  * nfs4_proc_exchange_id()
7029  *
7030  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7031  *
7032  * Since the clientid has expired, all compounds using sessions
7033  * associated with the stale clientid will be returning
7034  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7035  * be in some phase of session reset.
7036  *
7037  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7038  */
7039 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7040 {
7041         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7042         int status;
7043
7044         /* try SP4_MACH_CRED if krb5i/p */
7045         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7046             authflavor == RPC_AUTH_GSS_KRB5P) {
7047                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7048                 if (!status)
7049                         return 0;
7050         }
7051
7052         /* try SP4_NONE */
7053         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7054 }
7055
7056 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7057                 struct rpc_cred *cred)
7058 {
7059         struct rpc_message msg = {
7060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7061                 .rpc_argp = clp,
7062                 .rpc_cred = cred,
7063         };
7064         int status;
7065
7066         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7067         trace_nfs4_destroy_clientid(clp, status);
7068         if (status)
7069                 dprintk("NFS: Got error %d from the server %s on "
7070                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7071         return status;
7072 }
7073
7074 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7075                 struct rpc_cred *cred)
7076 {
7077         unsigned int loop;
7078         int ret;
7079
7080         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7081                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7082                 switch (ret) {
7083                 case -NFS4ERR_DELAY:
7084                 case -NFS4ERR_CLIENTID_BUSY:
7085                         ssleep(1);
7086                         break;
7087                 default:
7088                         return ret;
7089                 }
7090         }
7091         return 0;
7092 }
7093
7094 int nfs4_destroy_clientid(struct nfs_client *clp)
7095 {
7096         struct rpc_cred *cred;
7097         int ret = 0;
7098
7099         if (clp->cl_mvops->minor_version < 1)
7100                 goto out;
7101         if (clp->cl_exchange_flags == 0)
7102                 goto out;
7103         if (clp->cl_preserve_clid)
7104                 goto out;
7105         cred = nfs4_get_clid_cred(clp);
7106         ret = nfs4_proc_destroy_clientid(clp, cred);
7107         if (cred)
7108                 put_rpccred(cred);
7109         switch (ret) {
7110         case 0:
7111         case -NFS4ERR_STALE_CLIENTID:
7112                 clp->cl_exchange_flags = 0;
7113         }
7114 out:
7115         return ret;
7116 }
7117
7118 struct nfs4_get_lease_time_data {
7119         struct nfs4_get_lease_time_args *args;
7120         struct nfs4_get_lease_time_res *res;
7121         struct nfs_client *clp;
7122 };
7123
7124 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7125                                         void *calldata)
7126 {
7127         struct nfs4_get_lease_time_data *data =
7128                         (struct nfs4_get_lease_time_data *)calldata;
7129
7130         dprintk("--> %s\n", __func__);
7131         /* just setup sequence, do not trigger session recovery
7132            since we're invoked within one */
7133         nfs41_setup_sequence(data->clp->cl_session,
7134                         &data->args->la_seq_args,
7135                         &data->res->lr_seq_res,
7136                         task);
7137         dprintk("<-- %s\n", __func__);
7138 }
7139
7140 /*
7141  * Called from nfs4_state_manager thread for session setup, so don't recover
7142  * from sequence operation or clientid errors.
7143  */
7144 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7145 {
7146         struct nfs4_get_lease_time_data *data =
7147                         (struct nfs4_get_lease_time_data *)calldata;
7148
7149         dprintk("--> %s\n", __func__);
7150         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7151                 return;
7152         switch (task->tk_status) {
7153         case -NFS4ERR_DELAY:
7154         case -NFS4ERR_GRACE:
7155                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7156                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7157                 task->tk_status = 0;
7158                 /* fall through */
7159         case -NFS4ERR_RETRY_UNCACHED_REP:
7160                 rpc_restart_call_prepare(task);
7161                 return;
7162         }
7163         dprintk("<-- %s\n", __func__);
7164 }
7165
7166 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7167         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7168         .rpc_call_done = nfs4_get_lease_time_done,
7169 };
7170
7171 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7172 {
7173         struct rpc_task *task;
7174         struct nfs4_get_lease_time_args args;
7175         struct nfs4_get_lease_time_res res = {
7176                 .lr_fsinfo = fsinfo,
7177         };
7178         struct nfs4_get_lease_time_data data = {
7179                 .args = &args,
7180                 .res = &res,
7181                 .clp = clp,
7182         };
7183         struct rpc_message msg = {
7184                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7185                 .rpc_argp = &args,
7186                 .rpc_resp = &res,
7187         };
7188         struct rpc_task_setup task_setup = {
7189                 .rpc_client = clp->cl_rpcclient,
7190                 .rpc_message = &msg,
7191                 .callback_ops = &nfs4_get_lease_time_ops,
7192                 .callback_data = &data,
7193                 .flags = RPC_TASK_TIMEOUT,
7194         };
7195         int status;
7196
7197         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7198         nfs4_set_sequence_privileged(&args.la_seq_args);
7199         dprintk("--> %s\n", __func__);
7200         task = rpc_run_task(&task_setup);
7201
7202         if (IS_ERR(task))
7203                 status = PTR_ERR(task);
7204         else {
7205                 status = task->tk_status;
7206                 rpc_put_task(task);
7207         }
7208         dprintk("<-- %s return %d\n", __func__, status);
7209
7210         return status;
7211 }
7212
7213 /*
7214  * Initialize the values to be used by the client in CREATE_SESSION
7215  * If nfs4_init_session set the fore channel request and response sizes,
7216  * use them.
7217  *
7218  * Set the back channel max_resp_sz_cached to zero to force the client to
7219  * always set csa_cachethis to FALSE because the current implementation
7220  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7221  */
7222 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7223 {
7224         unsigned int max_rqst_sz, max_resp_sz;
7225
7226         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7227         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7228
7229         /* Fore channel attributes */
7230         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7231         args->fc_attrs.max_resp_sz = max_resp_sz;
7232         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7233         args->fc_attrs.max_reqs = max_session_slots;
7234
7235         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7236                 "max_ops=%u max_reqs=%u\n",
7237                 __func__,
7238                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7239                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7240
7241         /* Back channel attributes */
7242         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7243         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7244         args->bc_attrs.max_resp_sz_cached = 0;
7245         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7246         args->bc_attrs.max_reqs = 1;
7247
7248         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7249                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7250                 __func__,
7251                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7252                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7253                 args->bc_attrs.max_reqs);
7254 }
7255
7256 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7257                 struct nfs41_create_session_res *res)
7258 {
7259         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7260         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7261
7262         if (rcvd->max_resp_sz > sent->max_resp_sz)
7263                 return -EINVAL;
7264         /*
7265          * Our requested max_ops is the minimum we need; we're not
7266          * prepared to break up compounds into smaller pieces than that.
7267          * So, no point even trying to continue if the server won't
7268          * cooperate:
7269          */
7270         if (rcvd->max_ops < sent->max_ops)
7271                 return -EINVAL;
7272         if (rcvd->max_reqs == 0)
7273                 return -EINVAL;
7274         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7275                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7276         return 0;
7277 }
7278
7279 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7280                 struct nfs41_create_session_res *res)
7281 {
7282         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7283         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7284
7285         if (!(res->flags & SESSION4_BACK_CHAN))
7286                 goto out;
7287         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7288                 return -EINVAL;
7289         if (rcvd->max_resp_sz < sent->max_resp_sz)
7290                 return -EINVAL;
7291         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7292                 return -EINVAL;
7293         /* These would render the backchannel useless: */
7294         if (rcvd->max_ops != sent->max_ops)
7295                 return -EINVAL;
7296         if (rcvd->max_reqs != sent->max_reqs)
7297                 return -EINVAL;
7298 out:
7299         return 0;
7300 }
7301
7302 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7303                                      struct nfs41_create_session_res *res)
7304 {
7305         int ret;
7306
7307         ret = nfs4_verify_fore_channel_attrs(args, res);
7308         if (ret)
7309                 return ret;
7310         return nfs4_verify_back_channel_attrs(args, res);
7311 }
7312
7313 static void nfs4_update_session(struct nfs4_session *session,
7314                 struct nfs41_create_session_res *res)
7315 {
7316         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7317         /* Mark client id and session as being confirmed */
7318         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7319         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7320         session->flags = res->flags;
7321         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7322         if (res->flags & SESSION4_BACK_CHAN)
7323                 memcpy(&session->bc_attrs, &res->bc_attrs,
7324                                 sizeof(session->bc_attrs));
7325 }
7326
7327 static int _nfs4_proc_create_session(struct nfs_client *clp,
7328                 struct rpc_cred *cred)
7329 {
7330         struct nfs4_session *session = clp->cl_session;
7331         struct nfs41_create_session_args args = {
7332                 .client = clp,
7333                 .clientid = clp->cl_clientid,
7334                 .seqid = clp->cl_seqid,
7335                 .cb_program = NFS4_CALLBACK,
7336         };
7337         struct nfs41_create_session_res res;
7338
7339         struct rpc_message msg = {
7340                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7341                 .rpc_argp = &args,
7342                 .rpc_resp = &res,
7343                 .rpc_cred = cred,
7344         };
7345         int status;
7346
7347         nfs4_init_channel_attrs(&args);
7348         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7349
7350         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7351         trace_nfs4_create_session(clp, status);
7352
7353         if (!status) {
7354                 /* Verify the session's negotiated channel_attrs values */
7355                 status = nfs4_verify_channel_attrs(&args, &res);
7356                 /* Increment the clientid slot sequence id */
7357                 if (clp->cl_seqid == res.seqid)
7358                         clp->cl_seqid++;
7359                 if (status)
7360                         goto out;
7361                 nfs4_update_session(session, &res);
7362         }
7363 out:
7364         return status;
7365 }
7366
7367 /*
7368  * Issues a CREATE_SESSION operation to the server.
7369  * It is the responsibility of the caller to verify the session is
7370  * expired before calling this routine.
7371  */
7372 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7373 {
7374         int status;
7375         unsigned *ptr;
7376         struct nfs4_session *session = clp->cl_session;
7377
7378         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7379
7380         status = _nfs4_proc_create_session(clp, cred);
7381         if (status)
7382                 goto out;
7383
7384         /* Init or reset the session slot tables */
7385         status = nfs4_setup_session_slot_tables(session);
7386         dprintk("slot table setup returned %d\n", status);
7387         if (status)
7388                 goto out;
7389
7390         ptr = (unsigned *)&session->sess_id.data[0];
7391         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7392                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7393 out:
7394         dprintk("<-- %s\n", __func__);
7395         return status;
7396 }
7397
7398 /*
7399  * Issue the over-the-wire RPC DESTROY_SESSION.
7400  * The caller must serialize access to this routine.
7401  */
7402 int nfs4_proc_destroy_session(struct nfs4_session *session,
7403                 struct rpc_cred *cred)
7404 {
7405         struct rpc_message msg = {
7406                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7407                 .rpc_argp = session,
7408                 .rpc_cred = cred,
7409         };
7410         int status = 0;
7411
7412         dprintk("--> nfs4_proc_destroy_session\n");
7413
7414         /* session is still being setup */
7415         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7416                 return 0;
7417
7418         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7419         trace_nfs4_destroy_session(session->clp, status);
7420
7421         if (status)
7422                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7423                         "Session has been destroyed regardless...\n", status);
7424
7425         dprintk("<-- nfs4_proc_destroy_session\n");
7426         return status;
7427 }
7428
7429 /*
7430  * Renew the cl_session lease.
7431  */
7432 struct nfs4_sequence_data {
7433         struct nfs_client *clp;
7434         struct nfs4_sequence_args args;
7435         struct nfs4_sequence_res res;
7436 };
7437
7438 static void nfs41_sequence_release(void *data)
7439 {
7440         struct nfs4_sequence_data *calldata = data;
7441         struct nfs_client *clp = calldata->clp;
7442
7443         if (atomic_read(&clp->cl_count) > 1)
7444                 nfs4_schedule_state_renewal(clp);
7445         nfs_put_client(clp);
7446         kfree(calldata);
7447 }
7448
7449 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7450 {
7451         switch(task->tk_status) {
7452         case -NFS4ERR_DELAY:
7453                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7454                 return -EAGAIN;
7455         default:
7456                 nfs4_schedule_lease_recovery(clp);
7457         }
7458         return 0;
7459 }
7460
7461 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7462 {
7463         struct nfs4_sequence_data *calldata = data;
7464         struct nfs_client *clp = calldata->clp;
7465
7466         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7467                 return;
7468
7469         trace_nfs4_sequence(clp, task->tk_status);
7470         if (task->tk_status < 0) {
7471                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7472                 if (atomic_read(&clp->cl_count) == 1)
7473                         goto out;
7474
7475                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7476                         rpc_restart_call_prepare(task);
7477                         return;
7478                 }
7479         }
7480         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7481 out:
7482         dprintk("<-- %s\n", __func__);
7483 }
7484
7485 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7486 {
7487         struct nfs4_sequence_data *calldata = data;
7488         struct nfs_client *clp = calldata->clp;
7489         struct nfs4_sequence_args *args;
7490         struct nfs4_sequence_res *res;
7491
7492         args = task->tk_msg.rpc_argp;
7493         res = task->tk_msg.rpc_resp;
7494
7495         nfs41_setup_sequence(clp->cl_session, args, res, task);
7496 }
7497
7498 static const struct rpc_call_ops nfs41_sequence_ops = {
7499         .rpc_call_done = nfs41_sequence_call_done,
7500         .rpc_call_prepare = nfs41_sequence_prepare,
7501         .rpc_release = nfs41_sequence_release,
7502 };
7503
7504 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7505                 struct rpc_cred *cred,
7506                 bool is_privileged)
7507 {
7508         struct nfs4_sequence_data *calldata;
7509         struct rpc_message msg = {
7510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7511                 .rpc_cred = cred,
7512         };
7513         struct rpc_task_setup task_setup_data = {
7514                 .rpc_client = clp->cl_rpcclient,
7515                 .rpc_message = &msg,
7516                 .callback_ops = &nfs41_sequence_ops,
7517                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7518         };
7519
7520         if (!atomic_inc_not_zero(&clp->cl_count))
7521                 return ERR_PTR(-EIO);
7522         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7523         if (calldata == NULL) {
7524                 nfs_put_client(clp);
7525                 return ERR_PTR(-ENOMEM);
7526         }
7527         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7528         if (is_privileged)
7529                 nfs4_set_sequence_privileged(&calldata->args);
7530         msg.rpc_argp = &calldata->args;
7531         msg.rpc_resp = &calldata->res;
7532         calldata->clp = clp;
7533         task_setup_data.callback_data = calldata;
7534
7535         return rpc_run_task(&task_setup_data);
7536 }
7537
7538 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7539 {
7540         struct rpc_task *task;
7541         int ret = 0;
7542
7543         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7544                 return -EAGAIN;
7545         task = _nfs41_proc_sequence(clp, cred, false);
7546         if (IS_ERR(task))
7547                 ret = PTR_ERR(task);
7548         else
7549                 rpc_put_task_async(task);
7550         dprintk("<-- %s status=%d\n", __func__, ret);
7551         return ret;
7552 }
7553
7554 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7555 {
7556         struct rpc_task *task;
7557         int ret;
7558
7559         task = _nfs41_proc_sequence(clp, cred, true);
7560         if (IS_ERR(task)) {
7561                 ret = PTR_ERR(task);
7562                 goto out;
7563         }
7564         ret = rpc_wait_for_completion_task(task);
7565         if (!ret) {
7566                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7567
7568                 if (task->tk_status == 0)
7569                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7570                 ret = task->tk_status;
7571         }
7572         rpc_put_task(task);
7573 out:
7574         dprintk("<-- %s status=%d\n", __func__, ret);
7575         return ret;
7576 }
7577
7578 struct nfs4_reclaim_complete_data {
7579         struct nfs_client *clp;
7580         struct nfs41_reclaim_complete_args arg;
7581         struct nfs41_reclaim_complete_res res;
7582 };
7583
7584 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7585 {
7586         struct nfs4_reclaim_complete_data *calldata = data;
7587
7588         nfs41_setup_sequence(calldata->clp->cl_session,
7589                         &calldata->arg.seq_args,
7590                         &calldata->res.seq_res,
7591                         task);
7592 }
7593
7594 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7595 {
7596         switch(task->tk_status) {
7597         case 0:
7598         case -NFS4ERR_COMPLETE_ALREADY:
7599         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7600                 break;
7601         case -NFS4ERR_DELAY:
7602                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7603                 /* fall through */
7604         case -NFS4ERR_RETRY_UNCACHED_REP:
7605                 return -EAGAIN;
7606         default:
7607                 nfs4_schedule_lease_recovery(clp);
7608         }
7609         return 0;
7610 }
7611
7612 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7613 {
7614         struct nfs4_reclaim_complete_data *calldata = data;
7615         struct nfs_client *clp = calldata->clp;
7616         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7617
7618         dprintk("--> %s\n", __func__);
7619         if (!nfs41_sequence_done(task, res))
7620                 return;
7621
7622         trace_nfs4_reclaim_complete(clp, task->tk_status);
7623         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7624                 rpc_restart_call_prepare(task);
7625                 return;
7626         }
7627         dprintk("<-- %s\n", __func__);
7628 }
7629
7630 static void nfs4_free_reclaim_complete_data(void *data)
7631 {
7632         struct nfs4_reclaim_complete_data *calldata = data;
7633
7634         kfree(calldata);
7635 }
7636
7637 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7638         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7639         .rpc_call_done = nfs4_reclaim_complete_done,
7640         .rpc_release = nfs4_free_reclaim_complete_data,
7641 };
7642
7643 /*
7644  * Issue a global reclaim complete.
7645  */
7646 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7647                 struct rpc_cred *cred)
7648 {
7649         struct nfs4_reclaim_complete_data *calldata;
7650         struct rpc_task *task;
7651         struct rpc_message msg = {
7652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7653                 .rpc_cred = cred,
7654         };
7655         struct rpc_task_setup task_setup_data = {
7656                 .rpc_client = clp->cl_rpcclient,
7657                 .rpc_message = &msg,
7658                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7659                 .flags = RPC_TASK_ASYNC,
7660         };
7661         int status = -ENOMEM;
7662
7663         dprintk("--> %s\n", __func__);
7664         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7665         if (calldata == NULL)
7666                 goto out;
7667         calldata->clp = clp;
7668         calldata->arg.one_fs = 0;
7669
7670         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7671         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7672         msg.rpc_argp = &calldata->arg;
7673         msg.rpc_resp = &calldata->res;
7674         task_setup_data.callback_data = calldata;
7675         task = rpc_run_task(&task_setup_data);
7676         if (IS_ERR(task)) {
7677                 status = PTR_ERR(task);
7678                 goto out;
7679         }
7680         status = nfs4_wait_for_completion_rpc_task(task);
7681         if (status == 0)
7682                 status = task->tk_status;
7683         rpc_put_task(task);
7684         return 0;
7685 out:
7686         dprintk("<-- %s status=%d\n", __func__, status);
7687         return status;
7688 }
7689
7690 static void
7691 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7692 {
7693         struct nfs4_layoutget *lgp = calldata;
7694         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7695         struct nfs4_session *session = nfs4_get_session(server);
7696
7697         dprintk("--> %s\n", __func__);
7698         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7699          * right now covering the LAYOUTGET we are about to send.
7700          * However, that is not so catastrophic, and there seems
7701          * to be no way to prevent it completely.
7702          */
7703         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7704                                 &lgp->res.seq_res, task))
7705                 return;
7706         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7707                                           NFS_I(lgp->args.inode)->layout,
7708                                           &lgp->args.range,
7709                                           lgp->args.ctx->state)) {
7710                 rpc_exit(task, NFS4_OK);
7711         }
7712 }
7713
7714 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7715 {
7716         struct nfs4_layoutget *lgp = calldata;
7717         struct inode *inode = lgp->args.inode;
7718         struct nfs_server *server = NFS_SERVER(inode);
7719         struct pnfs_layout_hdr *lo;
7720         struct nfs4_state *state = NULL;
7721         unsigned long timeo, now, giveup;
7722
7723         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7724
7725         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7726                 goto out;
7727
7728         switch (task->tk_status) {
7729         case 0:
7730                 goto out;
7731         /*
7732          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7733          * (or clients) writing to the same RAID stripe
7734          */
7735         case -NFS4ERR_LAYOUTTRYLATER:
7736         /*
7737          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7738          * existing layout before getting a new one).
7739          */
7740         case -NFS4ERR_RECALLCONFLICT:
7741                 timeo = rpc_get_timeout(task->tk_client);
7742                 giveup = lgp->args.timestamp + timeo;
7743                 now = jiffies;
7744                 if (time_after(giveup, now)) {
7745                         unsigned long delay;
7746
7747                         /* Delay for:
7748                          * - Not less then NFS4_POLL_RETRY_MIN.
7749                          * - One last time a jiffie before we give up
7750                          * - exponential backoff (time_now minus start_attempt)
7751                          */
7752                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7753                                     min((giveup - now - 1),
7754                                         now - lgp->args.timestamp));
7755
7756                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7757                                 __func__, delay);
7758                         rpc_delay(task, delay);
7759                         task->tk_status = 0;
7760                         rpc_restart_call_prepare(task);
7761                         goto out; /* Do not call nfs4_async_handle_error() */
7762                 }
7763                 break;
7764         case -NFS4ERR_EXPIRED:
7765         case -NFS4ERR_BAD_STATEID:
7766                 spin_lock(&inode->i_lock);
7767                 lo = NFS_I(inode)->layout;
7768                 if (!lo || list_empty(&lo->plh_segs)) {
7769                         spin_unlock(&inode->i_lock);
7770                         /* If the open stateid was bad, then recover it. */
7771                         state = lgp->args.ctx->state;
7772                 } else {
7773                         LIST_HEAD(head);
7774
7775                         /*
7776                          * Mark the bad layout state as invalid, then retry
7777                          * with the current stateid.
7778                          */
7779                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7780                         spin_unlock(&inode->i_lock);
7781                         pnfs_free_lseg_list(&head);
7782
7783                         task->tk_status = 0;
7784                         rpc_restart_call_prepare(task);
7785                 }
7786         }
7787         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7788                 rpc_restart_call_prepare(task);
7789 out:
7790         dprintk("<-- %s\n", __func__);
7791 }
7792
7793 static size_t max_response_pages(struct nfs_server *server)
7794 {
7795         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7796         return nfs_page_array_len(0, max_resp_sz);
7797 }
7798
7799 static void nfs4_free_pages(struct page **pages, size_t size)
7800 {
7801         int i;
7802
7803         if (!pages)
7804                 return;
7805
7806         for (i = 0; i < size; i++) {
7807                 if (!pages[i])
7808                         break;
7809                 __free_page(pages[i]);
7810         }
7811         kfree(pages);
7812 }
7813
7814 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7815 {
7816         struct page **pages;
7817         int i;
7818
7819         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7820         if (!pages) {
7821                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7822                 return NULL;
7823         }
7824
7825         for (i = 0; i < size; i++) {
7826                 pages[i] = alloc_page(gfp_flags);
7827                 if (!pages[i]) {
7828                         dprintk("%s: failed to allocate page\n", __func__);
7829                         nfs4_free_pages(pages, size);
7830                         return NULL;
7831                 }
7832         }
7833
7834         return pages;
7835 }
7836
7837 static void nfs4_layoutget_release(void *calldata)
7838 {
7839         struct nfs4_layoutget *lgp = calldata;
7840         struct inode *inode = lgp->args.inode;
7841         struct nfs_server *server = NFS_SERVER(inode);
7842         size_t max_pages = max_response_pages(server);
7843
7844         dprintk("--> %s\n", __func__);
7845         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7846         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7847         put_nfs_open_context(lgp->args.ctx);
7848         kfree(calldata);
7849         dprintk("<-- %s\n", __func__);
7850 }
7851
7852 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7853         .rpc_call_prepare = nfs4_layoutget_prepare,
7854         .rpc_call_done = nfs4_layoutget_done,
7855         .rpc_release = nfs4_layoutget_release,
7856 };
7857
7858 struct pnfs_layout_segment *
7859 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7860 {
7861         struct inode *inode = lgp->args.inode;
7862         struct nfs_server *server = NFS_SERVER(inode);
7863         size_t max_pages = max_response_pages(server);
7864         struct rpc_task *task;
7865         struct rpc_message msg = {
7866                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7867                 .rpc_argp = &lgp->args,
7868                 .rpc_resp = &lgp->res,
7869                 .rpc_cred = lgp->cred,
7870         };
7871         struct rpc_task_setup task_setup_data = {
7872                 .rpc_client = server->client,
7873                 .rpc_message = &msg,
7874                 .callback_ops = &nfs4_layoutget_call_ops,
7875                 .callback_data = lgp,
7876                 .flags = RPC_TASK_ASYNC,
7877         };
7878         struct pnfs_layout_segment *lseg = NULL;
7879         int status = 0;
7880
7881         dprintk("--> %s\n", __func__);
7882
7883         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7884         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7885
7886         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7887         if (!lgp->args.layout.pages) {
7888                 nfs4_layoutget_release(lgp);
7889                 return ERR_PTR(-ENOMEM);
7890         }
7891         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7892         lgp->args.timestamp = jiffies;
7893
7894         lgp->res.layoutp = &lgp->args.layout;
7895         lgp->res.seq_res.sr_slot = NULL;
7896         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7897
7898         task = rpc_run_task(&task_setup_data);
7899         if (IS_ERR(task))
7900                 return ERR_CAST(task);
7901         status = nfs4_wait_for_completion_rpc_task(task);
7902         if (status == 0)
7903                 status = task->tk_status;
7904         trace_nfs4_layoutget(lgp->args.ctx,
7905                         &lgp->args.range,
7906                         &lgp->res.range,
7907                         status);
7908         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7909         if (status == 0 && lgp->res.layoutp->len)
7910                 lseg = pnfs_layout_process(lgp);
7911         rpc_put_task(task);
7912         dprintk("<-- %s status=%d\n", __func__, status);
7913         if (status)
7914                 return ERR_PTR(status);
7915         return lseg;
7916 }
7917
7918 static void
7919 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7920 {
7921         struct nfs4_layoutreturn *lrp = calldata;
7922
7923         dprintk("--> %s\n", __func__);
7924         nfs41_setup_sequence(lrp->clp->cl_session,
7925                         &lrp->args.seq_args,
7926                         &lrp->res.seq_res,
7927                         task);
7928 }
7929
7930 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7931 {
7932         struct nfs4_layoutreturn *lrp = calldata;
7933         struct nfs_server *server;
7934
7935         dprintk("--> %s\n", __func__);
7936
7937         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7938                 return;
7939
7940         server = NFS_SERVER(lrp->args.inode);
7941         switch (task->tk_status) {
7942         default:
7943                 task->tk_status = 0;
7944         case 0:
7945                 break;
7946         case -NFS4ERR_DELAY:
7947                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7948                         break;
7949                 rpc_restart_call_prepare(task);
7950                 return;
7951         }
7952         dprintk("<-- %s\n", __func__);
7953 }
7954
7955 static void nfs4_layoutreturn_release(void *calldata)
7956 {
7957         struct nfs4_layoutreturn *lrp = calldata;
7958         struct pnfs_layout_hdr *lo = lrp->args.layout;
7959
7960         dprintk("--> %s\n", __func__);
7961         spin_lock(&lo->plh_inode->i_lock);
7962         if (lrp->res.lrs_present)
7963                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7964         pnfs_clear_layoutreturn_waitbit(lo);
7965         clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7966         rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7967         lo->plh_block_lgets--;
7968         spin_unlock(&lo->plh_inode->i_lock);
7969         pnfs_put_layout_hdr(lrp->args.layout);
7970         nfs_iput_and_deactive(lrp->inode);
7971         kfree(calldata);
7972         dprintk("<-- %s\n", __func__);
7973 }
7974
7975 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7976         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7977         .rpc_call_done = nfs4_layoutreturn_done,
7978         .rpc_release = nfs4_layoutreturn_release,
7979 };
7980
7981 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7982 {
7983         struct rpc_task *task;
7984         struct rpc_message msg = {
7985                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7986                 .rpc_argp = &lrp->args,
7987                 .rpc_resp = &lrp->res,
7988                 .rpc_cred = lrp->cred,
7989         };
7990         struct rpc_task_setup task_setup_data = {
7991                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
7992                 .rpc_message = &msg,
7993                 .callback_ops = &nfs4_layoutreturn_call_ops,
7994                 .callback_data = lrp,
7995         };
7996         int status = 0;
7997
7998         dprintk("--> %s\n", __func__);
7999         if (!sync) {
8000                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8001                 if (!lrp->inode) {
8002                         nfs4_layoutreturn_release(lrp);
8003                         return -EAGAIN;
8004                 }
8005                 task_setup_data.flags |= RPC_TASK_ASYNC;
8006         }
8007         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8008         task = rpc_run_task(&task_setup_data);
8009         if (IS_ERR(task))
8010                 return PTR_ERR(task);
8011         if (sync)
8012                 status = task->tk_status;
8013         trace_nfs4_layoutreturn(lrp->args.inode, status);
8014         dprintk("<-- %s status=%d\n", __func__, status);
8015         rpc_put_task(task);
8016         return status;
8017 }
8018
8019 static int
8020 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8021                 struct pnfs_device *pdev,
8022                 struct rpc_cred *cred)
8023 {
8024         struct nfs4_getdeviceinfo_args args = {
8025                 .pdev = pdev,
8026                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8027                         NOTIFY_DEVICEID4_DELETE,
8028         };
8029         struct nfs4_getdeviceinfo_res res = {
8030                 .pdev = pdev,
8031         };
8032         struct rpc_message msg = {
8033                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8034                 .rpc_argp = &args,
8035                 .rpc_resp = &res,
8036                 .rpc_cred = cred,
8037         };
8038         int status;
8039
8040         dprintk("--> %s\n", __func__);
8041         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8042         if (res.notification & ~args.notify_types)
8043                 dprintk("%s: unsupported notification\n", __func__);
8044         if (res.notification != args.notify_types)
8045                 pdev->nocache = 1;
8046
8047         dprintk("<-- %s status=%d\n", __func__, status);
8048
8049         return status;
8050 }
8051
8052 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8053                 struct pnfs_device *pdev,
8054                 struct rpc_cred *cred)
8055 {
8056         struct nfs4_exception exception = { };
8057         int err;
8058
8059         do {
8060                 err = nfs4_handle_exception(server,
8061                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8062                                         &exception);
8063         } while (exception.retry);
8064         return err;
8065 }
8066 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8067
8068 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8069 {
8070         struct nfs4_layoutcommit_data *data = calldata;
8071         struct nfs_server *server = NFS_SERVER(data->args.inode);
8072         struct nfs4_session *session = nfs4_get_session(server);
8073
8074         nfs41_setup_sequence(session,
8075                         &data->args.seq_args,
8076                         &data->res.seq_res,
8077                         task);
8078 }
8079
8080 static void
8081 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8082 {
8083         struct nfs4_layoutcommit_data *data = calldata;
8084         struct nfs_server *server = NFS_SERVER(data->args.inode);
8085
8086         if (!nfs41_sequence_done(task, &data->res.seq_res))
8087                 return;
8088
8089         switch (task->tk_status) { /* Just ignore these failures */
8090         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8091         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8092         case -NFS4ERR_BADLAYOUT:     /* no layout */
8093         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8094                 task->tk_status = 0;
8095         case 0:
8096                 break;
8097         default:
8098                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8099                         rpc_restart_call_prepare(task);
8100                         return;
8101                 }
8102         }
8103 }
8104
8105 static void nfs4_layoutcommit_release(void *calldata)
8106 {
8107         struct nfs4_layoutcommit_data *data = calldata;
8108
8109         pnfs_cleanup_layoutcommit(data);
8110         nfs_post_op_update_inode_force_wcc(data->args.inode,
8111                                            data->res.fattr);
8112         put_rpccred(data->cred);
8113         nfs_iput_and_deactive(data->inode);
8114         kfree(data);
8115 }
8116
8117 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8118         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8119         .rpc_call_done = nfs4_layoutcommit_done,
8120         .rpc_release = nfs4_layoutcommit_release,
8121 };
8122
8123 int
8124 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8125 {
8126         struct rpc_message msg = {
8127                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8128                 .rpc_argp = &data->args,
8129                 .rpc_resp = &data->res,
8130                 .rpc_cred = data->cred,
8131         };
8132         struct rpc_task_setup task_setup_data = {
8133                 .task = &data->task,
8134                 .rpc_client = NFS_CLIENT(data->args.inode),
8135                 .rpc_message = &msg,
8136                 .callback_ops = &nfs4_layoutcommit_ops,
8137                 .callback_data = data,
8138         };
8139         struct rpc_task *task;
8140         int status = 0;
8141
8142         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
8143                 "lbw: %llu inode %lu\n",
8144                 data->task.tk_pid, sync,
8145                 data->args.lastbytewritten,
8146                 data->args.inode->i_ino);
8147
8148         if (!sync) {
8149                 data->inode = nfs_igrab_and_active(data->args.inode);
8150                 if (data->inode == NULL) {
8151                         nfs4_layoutcommit_release(data);
8152                         return -EAGAIN;
8153                 }
8154                 task_setup_data.flags = RPC_TASK_ASYNC;
8155         }
8156         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8157         task = rpc_run_task(&task_setup_data);
8158         if (IS_ERR(task))
8159                 return PTR_ERR(task);
8160         if (sync)
8161                 status = task->tk_status;
8162         trace_nfs4_layoutcommit(data->args.inode, status);
8163         dprintk("%s: status %d\n", __func__, status);
8164         rpc_put_task(task);
8165         return status;
8166 }
8167
8168 /**
8169  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8170  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8171  */
8172 static int
8173 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8174                     struct nfs_fsinfo *info,
8175                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8176 {
8177         struct nfs41_secinfo_no_name_args args = {
8178                 .style = SECINFO_STYLE_CURRENT_FH,
8179         };
8180         struct nfs4_secinfo_res res = {
8181                 .flavors = flavors,
8182         };
8183         struct rpc_message msg = {
8184                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8185                 .rpc_argp = &args,
8186                 .rpc_resp = &res,
8187         };
8188         struct rpc_clnt *clnt = server->client;
8189         struct rpc_cred *cred = NULL;
8190         int status;
8191
8192         if (use_integrity) {
8193                 clnt = server->nfs_client->cl_rpcclient;
8194                 cred = nfs4_get_clid_cred(server->nfs_client);
8195                 msg.rpc_cred = cred;
8196         }
8197
8198         dprintk("--> %s\n", __func__);
8199         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8200                                 &res.seq_res, 0);
8201         dprintk("<-- %s status=%d\n", __func__, status);
8202
8203         if (cred)
8204                 put_rpccred(cred);
8205
8206         return status;
8207 }
8208
8209 static int
8210 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8211                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8212 {
8213         struct nfs4_exception exception = { };
8214         int err;
8215         do {
8216                 /* first try using integrity protection */
8217                 err = -NFS4ERR_WRONGSEC;
8218
8219                 /* try to use integrity protection with machine cred */
8220                 if (_nfs4_is_integrity_protected(server->nfs_client))
8221                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8222                                                           flavors, true);
8223
8224                 /*
8225                  * if unable to use integrity protection, or SECINFO with
8226                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8227                  * disallowed by spec, but exists in deployed servers) use
8228                  * the current filesystem's rpc_client and the user cred.
8229                  */
8230                 if (err == -NFS4ERR_WRONGSEC)
8231                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8232                                                           flavors, false);
8233
8234                 switch (err) {
8235                 case 0:
8236                 case -NFS4ERR_WRONGSEC:
8237                 case -ENOTSUPP:
8238                         goto out;
8239                 default:
8240                         err = nfs4_handle_exception(server, err, &exception);
8241                 }
8242         } while (exception.retry);
8243 out:
8244         return err;
8245 }
8246
8247 static int
8248 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8249                     struct nfs_fsinfo *info)
8250 {
8251         int err;
8252         struct page *page;
8253         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8254         struct nfs4_secinfo_flavors *flavors;
8255         struct nfs4_secinfo4 *secinfo;
8256         int i;
8257
8258         page = alloc_page(GFP_KERNEL);
8259         if (!page) {
8260                 err = -ENOMEM;
8261                 goto out;
8262         }
8263
8264         flavors = page_address(page);
8265         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8266
8267         /*
8268          * Fall back on "guess and check" method if
8269          * the server doesn't support SECINFO_NO_NAME
8270          */
8271         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8272                 err = nfs4_find_root_sec(server, fhandle, info);
8273                 goto out_freepage;
8274         }
8275         if (err)
8276                 goto out_freepage;
8277
8278         for (i = 0; i < flavors->num_flavors; i++) {
8279                 secinfo = &flavors->flavors[i];
8280
8281                 switch (secinfo->flavor) {
8282                 case RPC_AUTH_NULL:
8283                 case RPC_AUTH_UNIX:
8284                 case RPC_AUTH_GSS:
8285                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8286                                         &secinfo->flavor_info);
8287                         break;
8288                 default:
8289                         flavor = RPC_AUTH_MAXFLAVOR;
8290                         break;
8291                 }
8292
8293                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8294                         flavor = RPC_AUTH_MAXFLAVOR;
8295
8296                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8297                         err = nfs4_lookup_root_sec(server, fhandle,
8298                                                    info, flavor);
8299                         if (!err)
8300                                 break;
8301                 }
8302         }
8303
8304         if (flavor == RPC_AUTH_MAXFLAVOR)
8305                 err = -EPERM;
8306
8307 out_freepage:
8308         put_page(page);
8309         if (err == -EACCES)
8310                 return -EPERM;
8311 out:
8312         return err;
8313 }
8314
8315 static int _nfs41_test_stateid(struct nfs_server *server,
8316                 nfs4_stateid *stateid,
8317                 struct rpc_cred *cred)
8318 {
8319         int status;
8320         struct nfs41_test_stateid_args args = {
8321                 .stateid = stateid,
8322         };
8323         struct nfs41_test_stateid_res res;
8324         struct rpc_message msg = {
8325                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8326                 .rpc_argp = &args,
8327                 .rpc_resp = &res,
8328                 .rpc_cred = cred,
8329         };
8330         struct rpc_clnt *rpc_client = server->client;
8331
8332         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8333                 &rpc_client, &msg);
8334
8335         dprintk("NFS call  test_stateid %p\n", stateid);
8336         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8337         nfs4_set_sequence_privileged(&args.seq_args);
8338         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8339                         &args.seq_args, &res.seq_res);
8340         if (status != NFS_OK) {
8341                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8342                 return status;
8343         }
8344         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8345         return -res.status;
8346 }
8347
8348 /**
8349  * nfs41_test_stateid - perform a TEST_STATEID operation
8350  *
8351  * @server: server / transport on which to perform the operation
8352  * @stateid: state ID to test
8353  * @cred: credential
8354  *
8355  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8356  * Otherwise a negative NFS4ERR value is returned if the operation
8357  * failed or the state ID is not currently valid.
8358  */
8359 static int nfs41_test_stateid(struct nfs_server *server,
8360                 nfs4_stateid *stateid,
8361                 struct rpc_cred *cred)
8362 {
8363         struct nfs4_exception exception = { };
8364         int err;
8365         do {
8366                 err = _nfs41_test_stateid(server, stateid, cred);
8367                 if (err != -NFS4ERR_DELAY)
8368                         break;
8369                 nfs4_handle_exception(server, err, &exception);
8370         } while (exception.retry);
8371         return err;
8372 }
8373
8374 struct nfs_free_stateid_data {
8375         struct nfs_server *server;
8376         struct nfs41_free_stateid_args args;
8377         struct nfs41_free_stateid_res res;
8378 };
8379
8380 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8381 {
8382         struct nfs_free_stateid_data *data = calldata;
8383         nfs41_setup_sequence(nfs4_get_session(data->server),
8384                         &data->args.seq_args,
8385                         &data->res.seq_res,
8386                         task);
8387 }
8388
8389 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8390 {
8391         struct nfs_free_stateid_data *data = calldata;
8392
8393         nfs41_sequence_done(task, &data->res.seq_res);
8394
8395         switch (task->tk_status) {
8396         case -NFS4ERR_DELAY:
8397                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8398                         rpc_restart_call_prepare(task);
8399         }
8400 }
8401
8402 static void nfs41_free_stateid_release(void *calldata)
8403 {
8404         kfree(calldata);
8405 }
8406
8407 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8408         .rpc_call_prepare = nfs41_free_stateid_prepare,
8409         .rpc_call_done = nfs41_free_stateid_done,
8410         .rpc_release = nfs41_free_stateid_release,
8411 };
8412
8413 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8414                 nfs4_stateid *stateid,
8415                 struct rpc_cred *cred,
8416                 bool privileged)
8417 {
8418         struct rpc_message msg = {
8419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8420                 .rpc_cred = cred,
8421         };
8422         struct rpc_task_setup task_setup = {
8423                 .rpc_client = server->client,
8424                 .rpc_message = &msg,
8425                 .callback_ops = &nfs41_free_stateid_ops,
8426                 .flags = RPC_TASK_ASYNC,
8427         };
8428         struct nfs_free_stateid_data *data;
8429
8430         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8431                 &task_setup.rpc_client, &msg);
8432
8433         dprintk("NFS call  free_stateid %p\n", stateid);
8434         data = kmalloc(sizeof(*data), GFP_NOFS);
8435         if (!data)
8436                 return ERR_PTR(-ENOMEM);
8437         data->server = server;
8438         nfs4_stateid_copy(&data->args.stateid, stateid);
8439
8440         task_setup.callback_data = data;
8441
8442         msg.rpc_argp = &data->args;
8443         msg.rpc_resp = &data->res;
8444         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8445         if (privileged)
8446                 nfs4_set_sequence_privileged(&data->args.seq_args);
8447
8448         return rpc_run_task(&task_setup);
8449 }
8450
8451 /**
8452  * nfs41_free_stateid - perform a FREE_STATEID operation
8453  *
8454  * @server: server / transport on which to perform the operation
8455  * @stateid: state ID to release
8456  * @cred: credential
8457  *
8458  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8459  * negative NFS4ERR value is returned.
8460  */
8461 static int nfs41_free_stateid(struct nfs_server *server,
8462                 nfs4_stateid *stateid,
8463                 struct rpc_cred *cred)
8464 {
8465         struct rpc_task *task;
8466         int ret;
8467
8468         task = _nfs41_free_stateid(server, stateid, cred, true);
8469         if (IS_ERR(task))
8470                 return PTR_ERR(task);
8471         ret = rpc_wait_for_completion_task(task);
8472         if (!ret)
8473                 ret = task->tk_status;
8474         rpc_put_task(task);
8475         return ret;
8476 }
8477
8478 static void
8479 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8480 {
8481         struct rpc_task *task;
8482         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8483
8484         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8485         nfs4_free_lock_state(server, lsp);
8486         if (IS_ERR(task))
8487                 return;
8488         rpc_put_task(task);
8489 }
8490
8491 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8492                 const nfs4_stateid *s2)
8493 {
8494         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8495                 return false;
8496
8497         if (s1->seqid == s2->seqid)
8498                 return true;
8499         if (s1->seqid == 0 || s2->seqid == 0)
8500                 return true;
8501
8502         return false;
8503 }
8504
8505 #endif /* CONFIG_NFS_V4_1 */
8506
8507 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8508                 const nfs4_stateid *s2)
8509 {
8510         return nfs4_stateid_match(s1, s2);
8511 }
8512
8513
8514 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8515         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8516         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8517         .recover_open   = nfs4_open_reclaim,
8518         .recover_lock   = nfs4_lock_reclaim,
8519         .establish_clid = nfs4_init_clientid,
8520         .detect_trunking = nfs40_discover_server_trunking,
8521 };
8522
8523 #if defined(CONFIG_NFS_V4_1)
8524 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8525         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8526         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8527         .recover_open   = nfs4_open_reclaim,
8528         .recover_lock   = nfs4_lock_reclaim,
8529         .establish_clid = nfs41_init_clientid,
8530         .reclaim_complete = nfs41_proc_reclaim_complete,
8531         .detect_trunking = nfs41_discover_server_trunking,
8532 };
8533 #endif /* CONFIG_NFS_V4_1 */
8534
8535 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8536         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8537         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8538         .recover_open   = nfs40_open_expired,
8539         .recover_lock   = nfs4_lock_expired,
8540         .establish_clid = nfs4_init_clientid,
8541 };
8542
8543 #if defined(CONFIG_NFS_V4_1)
8544 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8545         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8546         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8547         .recover_open   = nfs41_open_expired,
8548         .recover_lock   = nfs41_lock_expired,
8549         .establish_clid = nfs41_init_clientid,
8550 };
8551 #endif /* CONFIG_NFS_V4_1 */
8552
8553 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8554         .sched_state_renewal = nfs4_proc_async_renew,
8555         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8556         .renew_lease = nfs4_proc_renew,
8557 };
8558
8559 #if defined(CONFIG_NFS_V4_1)
8560 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8561         .sched_state_renewal = nfs41_proc_async_sequence,
8562         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8563         .renew_lease = nfs4_proc_sequence,
8564 };
8565 #endif
8566
8567 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8568         .get_locations = _nfs40_proc_get_locations,
8569         .fsid_present = _nfs40_proc_fsid_present,
8570 };
8571
8572 #if defined(CONFIG_NFS_V4_1)
8573 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8574         .get_locations = _nfs41_proc_get_locations,
8575         .fsid_present = _nfs41_proc_fsid_present,
8576 };
8577 #endif  /* CONFIG_NFS_V4_1 */
8578
8579 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8580         .minor_version = 0,
8581         .init_caps = NFS_CAP_READDIRPLUS
8582                 | NFS_CAP_ATOMIC_OPEN
8583                 | NFS_CAP_CHANGE_ATTR
8584                 | NFS_CAP_POSIX_LOCK,
8585         .init_client = nfs40_init_client,
8586         .shutdown_client = nfs40_shutdown_client,
8587         .match_stateid = nfs4_match_stateid,
8588         .find_root_sec = nfs4_find_root_sec,
8589         .free_lock_state = nfs4_release_lockowner,
8590         .alloc_seqid = nfs_alloc_seqid,
8591         .call_sync_ops = &nfs40_call_sync_ops,
8592         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8593         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8594         .state_renewal_ops = &nfs40_state_renewal_ops,
8595         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8596 };
8597
8598 #if defined(CONFIG_NFS_V4_1)
8599 static struct nfs_seqid *
8600 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8601 {
8602         return NULL;
8603 }
8604
8605 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8606         .minor_version = 1,
8607         .init_caps = NFS_CAP_READDIRPLUS
8608                 | NFS_CAP_ATOMIC_OPEN
8609                 | NFS_CAP_CHANGE_ATTR
8610                 | NFS_CAP_POSIX_LOCK
8611                 | NFS_CAP_STATEID_NFSV41
8612                 | NFS_CAP_ATOMIC_OPEN_V1,
8613         .init_client = nfs41_init_client,
8614         .shutdown_client = nfs41_shutdown_client,
8615         .match_stateid = nfs41_match_stateid,
8616         .find_root_sec = nfs41_find_root_sec,
8617         .free_lock_state = nfs41_free_lock_state,
8618         .alloc_seqid = nfs_alloc_no_seqid,
8619         .call_sync_ops = &nfs41_call_sync_ops,
8620         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8621         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8622         .state_renewal_ops = &nfs41_state_renewal_ops,
8623         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8624 };
8625 #endif
8626
8627 #if defined(CONFIG_NFS_V4_2)
8628 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8629         .minor_version = 2,
8630         .init_caps = NFS_CAP_READDIRPLUS
8631                 | NFS_CAP_ATOMIC_OPEN
8632                 | NFS_CAP_CHANGE_ATTR
8633                 | NFS_CAP_POSIX_LOCK
8634                 | NFS_CAP_STATEID_NFSV41
8635                 | NFS_CAP_ATOMIC_OPEN_V1
8636                 | NFS_CAP_ALLOCATE
8637                 | NFS_CAP_DEALLOCATE
8638                 | NFS_CAP_SEEK,
8639         .init_client = nfs41_init_client,
8640         .shutdown_client = nfs41_shutdown_client,
8641         .match_stateid = nfs41_match_stateid,
8642         .find_root_sec = nfs41_find_root_sec,
8643         .free_lock_state = nfs41_free_lock_state,
8644         .call_sync_ops = &nfs41_call_sync_ops,
8645         .alloc_seqid = nfs_alloc_no_seqid,
8646         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8647         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8648         .state_renewal_ops = &nfs41_state_renewal_ops,
8649 };
8650 #endif
8651
8652 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8653         [0] = &nfs_v4_0_minor_ops,
8654 #if defined(CONFIG_NFS_V4_1)
8655         [1] = &nfs_v4_1_minor_ops,
8656 #endif
8657 #if defined(CONFIG_NFS_V4_2)
8658         [2] = &nfs_v4_2_minor_ops,
8659 #endif
8660 };
8661
8662 static const struct inode_operations nfs4_dir_inode_operations = {
8663         .create         = nfs_create,
8664         .lookup         = nfs_lookup,
8665         .atomic_open    = nfs_atomic_open,
8666         .link           = nfs_link,
8667         .unlink         = nfs_unlink,
8668         .symlink        = nfs_symlink,
8669         .mkdir          = nfs_mkdir,
8670         .rmdir          = nfs_rmdir,
8671         .mknod          = nfs_mknod,
8672         .rename         = nfs_rename,
8673         .permission     = nfs_permission,
8674         .getattr        = nfs_getattr,
8675         .setattr        = nfs_setattr,
8676         .getxattr       = generic_getxattr,
8677         .setxattr       = generic_setxattr,
8678         .listxattr      = generic_listxattr,
8679         .removexattr    = generic_removexattr,
8680 };
8681
8682 static const struct inode_operations nfs4_file_inode_operations = {
8683         .permission     = nfs_permission,
8684         .getattr        = nfs_getattr,
8685         .setattr        = nfs_setattr,
8686         .getxattr       = generic_getxattr,
8687         .setxattr       = generic_setxattr,
8688         .listxattr      = generic_listxattr,
8689         .removexattr    = generic_removexattr,
8690 };
8691
8692 const struct nfs_rpc_ops nfs_v4_clientops = {
8693         .version        = 4,                    /* protocol version */
8694         .dentry_ops     = &nfs4_dentry_operations,
8695         .dir_inode_ops  = &nfs4_dir_inode_operations,
8696         .file_inode_ops = &nfs4_file_inode_operations,
8697         .file_ops       = &nfs4_file_operations,
8698         .getroot        = nfs4_proc_get_root,
8699         .submount       = nfs4_submount,
8700         .try_mount      = nfs4_try_mount,
8701         .getattr        = nfs4_proc_getattr,
8702         .setattr        = nfs4_proc_setattr,
8703         .lookup         = nfs4_proc_lookup,
8704         .access         = nfs4_proc_access,
8705         .readlink       = nfs4_proc_readlink,
8706         .create         = nfs4_proc_create,
8707         .remove         = nfs4_proc_remove,
8708         .unlink_setup   = nfs4_proc_unlink_setup,
8709         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8710         .unlink_done    = nfs4_proc_unlink_done,
8711         .rename_setup   = nfs4_proc_rename_setup,
8712         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8713         .rename_done    = nfs4_proc_rename_done,
8714         .link           = nfs4_proc_link,
8715         .symlink        = nfs4_proc_symlink,
8716         .mkdir          = nfs4_proc_mkdir,
8717         .rmdir          = nfs4_proc_remove,
8718         .readdir        = nfs4_proc_readdir,
8719         .mknod          = nfs4_proc_mknod,
8720         .statfs         = nfs4_proc_statfs,
8721         .fsinfo         = nfs4_proc_fsinfo,
8722         .pathconf       = nfs4_proc_pathconf,
8723         .set_capabilities = nfs4_server_capabilities,
8724         .decode_dirent  = nfs4_decode_dirent,
8725         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8726         .read_setup     = nfs4_proc_read_setup,
8727         .read_done      = nfs4_read_done,
8728         .write_setup    = nfs4_proc_write_setup,
8729         .write_done     = nfs4_write_done,
8730         .commit_setup   = nfs4_proc_commit_setup,
8731         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8732         .commit_done    = nfs4_commit_done,
8733         .lock           = nfs4_proc_lock,
8734         .clear_acl_cache = nfs4_zap_acl_attr,
8735         .close_context  = nfs4_close_context,
8736         .open_context   = nfs4_atomic_open,
8737         .have_delegation = nfs4_have_delegation,
8738         .return_delegation = nfs4_inode_return_delegation,
8739         .alloc_client   = nfs4_alloc_client,
8740         .init_client    = nfs4_init_client,
8741         .free_client    = nfs4_free_client,
8742         .create_server  = nfs4_create_server,
8743         .clone_server   = nfs_clone_server,
8744 };
8745
8746 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8747         .prefix = XATTR_NAME_NFSV4_ACL,
8748         .list   = nfs4_xattr_list_nfs4_acl,
8749         .get    = nfs4_xattr_get_nfs4_acl,
8750         .set    = nfs4_xattr_set_nfs4_acl,
8751 };
8752
8753 const struct xattr_handler *nfs4_xattr_handlers[] = {
8754         &nfs4_xattr_nfs4_acl_handler,
8755 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8756         &nfs4_xattr_nfs4_label_handler,
8757 #endif
8758         NULL
8759 };
8760
8761 /*
8762  * Local variables:
8763  *  c-basic-offset: 8
8764  * End:
8765  */