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 <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  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         if ((opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
1551              opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEG_CUR_FH) &&
1552             (opendata->o_arg.u.delegation_type & fmode) != fmode)
1553                 /* This mode can't have been delegated, so we must have
1554                  * a valid open_stateid to cover it - not need to reclaim.
1555                  */
1556                 return 0;
1557         opendata->o_arg.open_flags = 0;
1558         opendata->o_arg.fmode = fmode;
1559         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1560                         NFS_SB(opendata->dentry->d_sb),
1561                         fmode, 0);
1562         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1563         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1564         nfs4_init_opendata_res(opendata);
1565         ret = _nfs4_recover_proc_open(opendata);
1566         if (ret != 0)
1567                 return ret;
1568         newstate = nfs4_opendata_to_nfs4_state(opendata);
1569         if (IS_ERR(newstate))
1570                 return PTR_ERR(newstate);
1571         nfs4_close_state(newstate, fmode);
1572         *res = newstate;
1573         return 0;
1574 }
1575
1576 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1577 {
1578         struct nfs4_state *newstate;
1579         int ret;
1580
1581         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1582         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1583         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1584         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1585         /* memory barrier prior to reading state->n_* */
1586         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1587         clear_bit(NFS_OPEN_STATE, &state->flags);
1588         smp_rmb();
1589         if (state->n_rdwr != 0) {
1590                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1591                 if (ret != 0)
1592                         return ret;
1593                 if (newstate != state)
1594                         return -ESTALE;
1595         }
1596         if (state->n_wronly != 0) {
1597                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1598                 if (ret != 0)
1599                         return ret;
1600                 if (newstate != state)
1601                         return -ESTALE;
1602         }
1603         if (state->n_rdonly != 0) {
1604                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1605                 if (ret != 0)
1606                         return ret;
1607                 if (newstate != state)
1608                         return -ESTALE;
1609         }
1610         /*
1611          * We may have performed cached opens for all three recoveries.
1612          * Check if we need to update the current stateid.
1613          */
1614         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1615             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1616                 write_seqlock(&state->seqlock);
1617                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1618                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1619                 write_sequnlock(&state->seqlock);
1620         }
1621         return 0;
1622 }
1623
1624 /*
1625  * OPEN_RECLAIM:
1626  *      reclaim state on the server after a reboot.
1627  */
1628 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1629 {
1630         struct nfs_delegation *delegation;
1631         struct nfs4_opendata *opendata;
1632         fmode_t delegation_type = 0;
1633         int status;
1634
1635         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1636                         NFS4_OPEN_CLAIM_PREVIOUS);
1637         if (IS_ERR(opendata))
1638                 return PTR_ERR(opendata);
1639         rcu_read_lock();
1640         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1641         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1642                 delegation_type = delegation->type;
1643         rcu_read_unlock();
1644         opendata->o_arg.u.delegation_type = delegation_type;
1645         status = nfs4_open_recover(opendata, state);
1646         nfs4_opendata_put(opendata);
1647         return status;
1648 }
1649
1650 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1651 {
1652         struct nfs_server *server = NFS_SERVER(state->inode);
1653         struct nfs4_exception exception = { };
1654         int err;
1655         do {
1656                 err = _nfs4_do_open_reclaim(ctx, state);
1657                 trace_nfs4_open_reclaim(ctx, 0, err);
1658                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1659                         continue;
1660                 if (err != -NFS4ERR_DELAY)
1661                         break;
1662                 nfs4_handle_exception(server, err, &exception);
1663         } while (exception.retry);
1664         return err;
1665 }
1666
1667 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1668 {
1669         struct nfs_open_context *ctx;
1670         int ret;
1671
1672         ctx = nfs4_state_find_open_context(state);
1673         if (IS_ERR(ctx))
1674                 return -EAGAIN;
1675         ret = nfs4_do_open_reclaim(ctx, state);
1676         put_nfs_open_context(ctx);
1677         return ret;
1678 }
1679
1680 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1681 {
1682         switch (err) {
1683                 default:
1684                         printk(KERN_ERR "NFS: %s: unhandled error "
1685                                         "%d.\n", __func__, err);
1686                 case 0:
1687                 case -ENOENT:
1688                 case -EAGAIN:
1689                 case -ESTALE:
1690                         break;
1691                 case -NFS4ERR_BADSESSION:
1692                 case -NFS4ERR_BADSLOT:
1693                 case -NFS4ERR_BAD_HIGH_SLOT:
1694                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1695                 case -NFS4ERR_DEADSESSION:
1696                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1697                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1698                         return -EAGAIN;
1699                 case -NFS4ERR_STALE_CLIENTID:
1700                 case -NFS4ERR_STALE_STATEID:
1701                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1702                 case -NFS4ERR_EXPIRED:
1703                         /* Don't recall a delegation if it was lost */
1704                         nfs4_schedule_lease_recovery(server->nfs_client);
1705                         return -EAGAIN;
1706                 case -NFS4ERR_MOVED:
1707                         nfs4_schedule_migration_recovery(server);
1708                         return -EAGAIN;
1709                 case -NFS4ERR_LEASE_MOVED:
1710                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1711                         return -EAGAIN;
1712                 case -NFS4ERR_DELEG_REVOKED:
1713                 case -NFS4ERR_ADMIN_REVOKED:
1714                 case -NFS4ERR_BAD_STATEID:
1715                 case -NFS4ERR_OPENMODE:
1716                         nfs_inode_find_state_and_recover(state->inode,
1717                                         stateid);
1718                         nfs4_schedule_stateid_recovery(server, state);
1719                         return -EAGAIN;
1720                 case -NFS4ERR_DELAY:
1721                 case -NFS4ERR_GRACE:
1722                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1723                         ssleep(1);
1724                         return -EAGAIN;
1725                 case -ENOMEM:
1726                 case -NFS4ERR_DENIED:
1727                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1728                         return 0;
1729         }
1730         return err;
1731 }
1732
1733 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1734 {
1735         struct nfs_server *server = NFS_SERVER(state->inode);
1736         struct nfs4_opendata *opendata;
1737         int err;
1738
1739         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1740                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1741         if (IS_ERR(opendata))
1742                 return PTR_ERR(opendata);
1743         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1744         err = nfs4_open_recover(opendata, state);
1745         nfs4_opendata_put(opendata);
1746         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1747 }
1748
1749 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1750 {
1751         struct nfs4_opendata *data = calldata;
1752
1753         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1754                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1755 }
1756
1757 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1758 {
1759         struct nfs4_opendata *data = calldata;
1760
1761         nfs40_sequence_done(task, &data->c_res.seq_res);
1762
1763         data->rpc_status = task->tk_status;
1764         if (data->rpc_status == 0) {
1765                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1766                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1767                 renew_lease(data->o_res.server, data->timestamp);
1768                 data->rpc_done = 1;
1769         }
1770 }
1771
1772 static void nfs4_open_confirm_release(void *calldata)
1773 {
1774         struct nfs4_opendata *data = calldata;
1775         struct nfs4_state *state = NULL;
1776
1777         /* If this request hasn't been cancelled, do nothing */
1778         if (data->cancelled == 0)
1779                 goto out_free;
1780         /* In case of error, no cleanup! */
1781         if (!data->rpc_done)
1782                 goto out_free;
1783         state = nfs4_opendata_to_nfs4_state(data);
1784         if (!IS_ERR(state))
1785                 nfs4_close_state(state, data->o_arg.fmode);
1786 out_free:
1787         nfs4_opendata_put(data);
1788 }
1789
1790 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1791         .rpc_call_prepare = nfs4_open_confirm_prepare,
1792         .rpc_call_done = nfs4_open_confirm_done,
1793         .rpc_release = nfs4_open_confirm_release,
1794 };
1795
1796 /*
1797  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1798  */
1799 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1800 {
1801         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1802         struct rpc_task *task;
1803         struct  rpc_message msg = {
1804                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1805                 .rpc_argp = &data->c_arg,
1806                 .rpc_resp = &data->c_res,
1807                 .rpc_cred = data->owner->so_cred,
1808         };
1809         struct rpc_task_setup task_setup_data = {
1810                 .rpc_client = server->client,
1811                 .rpc_message = &msg,
1812                 .callback_ops = &nfs4_open_confirm_ops,
1813                 .callback_data = data,
1814                 .workqueue = nfsiod_workqueue,
1815                 .flags = RPC_TASK_ASYNC,
1816         };
1817         int status;
1818
1819         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1820         kref_get(&data->kref);
1821         data->rpc_done = 0;
1822         data->rpc_status = 0;
1823         data->timestamp = jiffies;
1824         task = rpc_run_task(&task_setup_data);
1825         if (IS_ERR(task))
1826                 return PTR_ERR(task);
1827         status = nfs4_wait_for_completion_rpc_task(task);
1828         if (status != 0) {
1829                 data->cancelled = 1;
1830                 smp_wmb();
1831         } else
1832                 status = data->rpc_status;
1833         rpc_put_task(task);
1834         return status;
1835 }
1836
1837 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1838 {
1839         struct nfs4_opendata *data = calldata;
1840         struct nfs4_state_owner *sp = data->owner;
1841         struct nfs_client *clp = sp->so_server->nfs_client;
1842
1843         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1844                 goto out_wait;
1845         /*
1846          * Check if we still need to send an OPEN call, or if we can use
1847          * a delegation instead.
1848          */
1849         if (data->state != NULL) {
1850                 struct nfs_delegation *delegation;
1851
1852                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1853                         goto out_no_action;
1854                 rcu_read_lock();
1855                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1856                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1857                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1858                     can_open_delegated(delegation, data->o_arg.fmode))
1859                         goto unlock_no_action;
1860                 rcu_read_unlock();
1861         }
1862         /* Update client id. */
1863         data->o_arg.clientid = clp->cl_clientid;
1864         switch (data->o_arg.claim) {
1865         case NFS4_OPEN_CLAIM_PREVIOUS:
1866         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1867         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1868                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1869         case NFS4_OPEN_CLAIM_FH:
1870                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1871                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1872         }
1873         data->timestamp = jiffies;
1874         if (nfs4_setup_sequence(data->o_arg.server,
1875                                 &data->o_arg.seq_args,
1876                                 &data->o_res.seq_res,
1877                                 task) != 0)
1878                 nfs_release_seqid(data->o_arg.seqid);
1879
1880         /* Set the create mode (note dependency on the session type) */
1881         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1882         if (data->o_arg.open_flags & O_EXCL) {
1883                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1884                 if (nfs4_has_persistent_session(clp))
1885                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1886                 else if (clp->cl_mvops->minor_version > 0)
1887                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1888         }
1889         return;
1890 unlock_no_action:
1891         rcu_read_unlock();
1892 out_no_action:
1893         task->tk_action = NULL;
1894 out_wait:
1895         nfs4_sequence_done(task, &data->o_res.seq_res);
1896 }
1897
1898 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1899 {
1900         struct nfs4_opendata *data = calldata;
1901
1902         data->rpc_status = task->tk_status;
1903
1904         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1905                 return;
1906
1907         if (task->tk_status == 0) {
1908                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1909                         switch (data->o_res.f_attr->mode & S_IFMT) {
1910                         case S_IFREG:
1911                                 break;
1912                         case S_IFLNK:
1913                                 data->rpc_status = -ELOOP;
1914                                 break;
1915                         case S_IFDIR:
1916                                 data->rpc_status = -EISDIR;
1917                                 break;
1918                         default:
1919                                 data->rpc_status = -ENOTDIR;
1920                         }
1921                 }
1922                 renew_lease(data->o_res.server, data->timestamp);
1923                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1924                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1925         }
1926         data->rpc_done = 1;
1927 }
1928
1929 static void nfs4_open_release(void *calldata)
1930 {
1931         struct nfs4_opendata *data = calldata;
1932         struct nfs4_state *state = NULL;
1933
1934         /* If this request hasn't been cancelled, do nothing */
1935         if (data->cancelled == 0)
1936                 goto out_free;
1937         /* In case of error, no cleanup! */
1938         if (data->rpc_status != 0 || !data->rpc_done)
1939                 goto out_free;
1940         /* In case we need an open_confirm, no cleanup! */
1941         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1942                 goto out_free;
1943         state = nfs4_opendata_to_nfs4_state(data);
1944         if (!IS_ERR(state))
1945                 nfs4_close_state(state, data->o_arg.fmode);
1946 out_free:
1947         nfs4_opendata_put(data);
1948 }
1949
1950 static const struct rpc_call_ops nfs4_open_ops = {
1951         .rpc_call_prepare = nfs4_open_prepare,
1952         .rpc_call_done = nfs4_open_done,
1953         .rpc_release = nfs4_open_release,
1954 };
1955
1956 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1957 {
1958         struct inode *dir = d_inode(data->dir);
1959         struct nfs_server *server = NFS_SERVER(dir);
1960         struct nfs_openargs *o_arg = &data->o_arg;
1961         struct nfs_openres *o_res = &data->o_res;
1962         struct rpc_task *task;
1963         struct rpc_message msg = {
1964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1965                 .rpc_argp = o_arg,
1966                 .rpc_resp = o_res,
1967                 .rpc_cred = data->owner->so_cred,
1968         };
1969         struct rpc_task_setup task_setup_data = {
1970                 .rpc_client = server->client,
1971                 .rpc_message = &msg,
1972                 .callback_ops = &nfs4_open_ops,
1973                 .callback_data = data,
1974                 .workqueue = nfsiod_workqueue,
1975                 .flags = RPC_TASK_ASYNC,
1976         };
1977         int status;
1978
1979         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1980         kref_get(&data->kref);
1981         data->rpc_done = 0;
1982         data->rpc_status = 0;
1983         data->cancelled = 0;
1984         data->is_recover = 0;
1985         if (isrecover) {
1986                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1987                 data->is_recover = 1;
1988         }
1989         task = rpc_run_task(&task_setup_data);
1990         if (IS_ERR(task))
1991                 return PTR_ERR(task);
1992         status = nfs4_wait_for_completion_rpc_task(task);
1993         if (status != 0) {
1994                 data->cancelled = 1;
1995                 smp_wmb();
1996         } else
1997                 status = data->rpc_status;
1998         rpc_put_task(task);
1999
2000         return status;
2001 }
2002
2003 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2004 {
2005         struct inode *dir = d_inode(data->dir);
2006         struct nfs_openres *o_res = &data->o_res;
2007         int status;
2008
2009         status = nfs4_run_open_task(data, 1);
2010         if (status != 0 || !data->rpc_done)
2011                 return status;
2012
2013         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2014
2015         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2016                 status = _nfs4_proc_open_confirm(data);
2017                 if (status != 0)
2018                         return status;
2019         }
2020
2021         return status;
2022 }
2023
2024 /*
2025  * Additional permission checks in order to distinguish between an
2026  * open for read, and an open for execute. This works around the
2027  * fact that NFSv4 OPEN treats read and execute permissions as being
2028  * the same.
2029  * Note that in the non-execute case, we want to turn off permission
2030  * checking if we just created a new file (POSIX open() semantics).
2031  */
2032 static int nfs4_opendata_access(struct rpc_cred *cred,
2033                                 struct nfs4_opendata *opendata,
2034                                 struct nfs4_state *state, fmode_t fmode,
2035                                 int openflags)
2036 {
2037         struct nfs_access_entry cache;
2038         u32 mask;
2039
2040         /* access call failed or for some reason the server doesn't
2041          * support any access modes -- defer access call until later */
2042         if (opendata->o_res.access_supported == 0)
2043                 return 0;
2044
2045         mask = 0;
2046         /*
2047          * Use openflags to check for exec, because fmode won't
2048          * always have FMODE_EXEC set when file open for exec.
2049          */
2050         if (openflags & __FMODE_EXEC) {
2051                 /* ONLY check for exec rights */
2052                 mask = MAY_EXEC;
2053         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2054                 mask = MAY_READ;
2055
2056         cache.cred = cred;
2057         cache.jiffies = jiffies;
2058         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2059         nfs_access_add_cache(state->inode, &cache);
2060
2061         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2062                 return 0;
2063
2064         /* even though OPEN succeeded, access is denied. Close the file */
2065         nfs4_close_state(state, fmode);
2066         return -EACCES;
2067 }
2068
2069 /*
2070  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2071  */
2072 static int _nfs4_proc_open(struct nfs4_opendata *data)
2073 {
2074         struct inode *dir = d_inode(data->dir);
2075         struct nfs_server *server = NFS_SERVER(dir);
2076         struct nfs_openargs *o_arg = &data->o_arg;
2077         struct nfs_openres *o_res = &data->o_res;
2078         int status;
2079
2080         status = nfs4_run_open_task(data, 0);
2081         if (!data->rpc_done)
2082                 return status;
2083         if (status != 0) {
2084                 if (status == -NFS4ERR_BADNAME &&
2085                                 !(o_arg->open_flags & O_CREAT))
2086                         return -ENOENT;
2087                 return status;
2088         }
2089
2090         nfs_fattr_map_and_free_names(server, &data->f_attr);
2091
2092         if (o_arg->open_flags & O_CREAT) {
2093                 update_changeattr(dir, &o_res->cinfo);
2094                 if (o_arg->open_flags & O_EXCL)
2095                         data->file_created = 1;
2096                 else if (o_res->cinfo.before != o_res->cinfo.after)
2097                         data->file_created = 1;
2098         }
2099         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2100                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2101         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2102                 status = _nfs4_proc_open_confirm(data);
2103                 if (status != 0)
2104                         return status;
2105         }
2106         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2107                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2108         return 0;
2109 }
2110
2111 static int nfs4_recover_expired_lease(struct nfs_server *server)
2112 {
2113         return nfs4_client_recover_expired_lease(server->nfs_client);
2114 }
2115
2116 /*
2117  * OPEN_EXPIRED:
2118  *      reclaim state on the server after a network partition.
2119  *      Assumes caller holds the appropriate lock
2120  */
2121 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2122 {
2123         struct nfs4_opendata *opendata;
2124         int ret;
2125
2126         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2127                         NFS4_OPEN_CLAIM_FH);
2128         if (IS_ERR(opendata))
2129                 return PTR_ERR(opendata);
2130         ret = nfs4_open_recover(opendata, state);
2131         if (ret == -ESTALE)
2132                 d_drop(ctx->dentry);
2133         nfs4_opendata_put(opendata);
2134         return ret;
2135 }
2136
2137 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2138 {
2139         struct nfs_server *server = NFS_SERVER(state->inode);
2140         struct nfs4_exception exception = { };
2141         int err;
2142
2143         do {
2144                 err = _nfs4_open_expired(ctx, state);
2145                 trace_nfs4_open_expired(ctx, 0, err);
2146                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2147                         continue;
2148                 switch (err) {
2149                 default:
2150                         goto out;
2151                 case -NFS4ERR_GRACE:
2152                 case -NFS4ERR_DELAY:
2153                         nfs4_handle_exception(server, err, &exception);
2154                         err = 0;
2155                 }
2156         } while (exception.retry);
2157 out:
2158         return err;
2159 }
2160
2161 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2162 {
2163         struct nfs_open_context *ctx;
2164         int ret;
2165
2166         ctx = nfs4_state_find_open_context(state);
2167         if (IS_ERR(ctx))
2168                 return -EAGAIN;
2169         ret = nfs4_do_open_expired(ctx, state);
2170         put_nfs_open_context(ctx);
2171         return ret;
2172 }
2173
2174 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2175 {
2176         nfs_remove_bad_delegation(state->inode);
2177         write_seqlock(&state->seqlock);
2178         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2179         write_sequnlock(&state->seqlock);
2180         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2181 }
2182
2183 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2184 {
2185         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2186                 nfs_finish_clear_delegation_stateid(state);
2187 }
2188
2189 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2190 {
2191         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2192         nfs40_clear_delegation_stateid(state);
2193         return nfs4_open_expired(sp, state);
2194 }
2195
2196 #if defined(CONFIG_NFS_V4_1)
2197 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2198 {
2199         struct nfs_server *server = NFS_SERVER(state->inode);
2200         nfs4_stateid stateid;
2201         struct nfs_delegation *delegation;
2202         struct rpc_cred *cred;
2203         int status;
2204
2205         /* Get the delegation credential for use by test/free_stateid */
2206         rcu_read_lock();
2207         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2208         if (delegation == NULL) {
2209                 rcu_read_unlock();
2210                 return;
2211         }
2212
2213         nfs4_stateid_copy(&stateid, &delegation->stateid);
2214         cred = get_rpccred(delegation->cred);
2215         rcu_read_unlock();
2216         status = nfs41_test_stateid(server, &stateid, cred);
2217         trace_nfs4_test_delegation_stateid(state, NULL, status);
2218
2219         if (status != NFS_OK) {
2220                 /* Free the stateid unless the server explicitly
2221                  * informs us the stateid is unrecognized. */
2222                 if (status != -NFS4ERR_BAD_STATEID)
2223                         nfs41_free_stateid(server, &stateid, cred);
2224                 nfs_finish_clear_delegation_stateid(state);
2225         }
2226
2227         put_rpccred(cred);
2228 }
2229
2230 /**
2231  * nfs41_check_open_stateid - possibly free an open stateid
2232  *
2233  * @state: NFSv4 state for an inode
2234  *
2235  * Returns NFS_OK if recovery for this stateid is now finished.
2236  * Otherwise a negative NFS4ERR value is returned.
2237  */
2238 static int nfs41_check_open_stateid(struct nfs4_state *state)
2239 {
2240         struct nfs_server *server = NFS_SERVER(state->inode);
2241         nfs4_stateid *stateid = &state->open_stateid;
2242         struct rpc_cred *cred = state->owner->so_cred;
2243         int status;
2244
2245         /* If a state reset has been done, test_stateid is unneeded */
2246         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2247             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2248             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2249                 return -NFS4ERR_BAD_STATEID;
2250
2251         status = nfs41_test_stateid(server, stateid, cred);
2252         trace_nfs4_test_open_stateid(state, NULL, status);
2253         if (status != NFS_OK) {
2254                 /* Free the stateid unless the server explicitly
2255                  * informs us the stateid is unrecognized. */
2256                 if (status != -NFS4ERR_BAD_STATEID)
2257                         nfs41_free_stateid(server, stateid, cred);
2258
2259                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2260                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2261                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2262                 clear_bit(NFS_OPEN_STATE, &state->flags);
2263         }
2264         return status;
2265 }
2266
2267 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2268 {
2269         int status;
2270
2271         nfs41_check_delegation_stateid(state);
2272         status = nfs41_check_open_stateid(state);
2273         if (status != NFS_OK)
2274                 status = nfs4_open_expired(sp, state);
2275         return status;
2276 }
2277 #endif
2278
2279 /*
2280  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2281  * fields corresponding to attributes that were used to store the verifier.
2282  * Make sure we clobber those fields in the later setattr call
2283  */
2284 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2285 {
2286         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2287             !(sattr->ia_valid & ATTR_ATIME_SET))
2288                 sattr->ia_valid |= ATTR_ATIME;
2289
2290         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2291             !(sattr->ia_valid & ATTR_MTIME_SET))
2292                 sattr->ia_valid |= ATTR_MTIME;
2293 }
2294
2295 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2296                 fmode_t fmode,
2297                 int flags,
2298                 struct nfs_open_context *ctx)
2299 {
2300         struct nfs4_state_owner *sp = opendata->owner;
2301         struct nfs_server *server = sp->so_server;
2302         struct dentry *dentry;
2303         struct nfs4_state *state;
2304         unsigned int seq;
2305         int ret;
2306
2307         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2308
2309         ret = _nfs4_proc_open(opendata);
2310         if (ret != 0)
2311                 goto out;
2312
2313         state = nfs4_opendata_to_nfs4_state(opendata);
2314         ret = PTR_ERR(state);
2315         if (IS_ERR(state))
2316                 goto out;
2317         if (server->caps & NFS_CAP_POSIX_LOCK)
2318                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2319
2320         dentry = opendata->dentry;
2321         if (d_really_is_negative(dentry)) {
2322                 /* FIXME: Is this d_drop() ever needed? */
2323                 d_drop(dentry);
2324                 dentry = d_add_unique(dentry, igrab(state->inode));
2325                 if (dentry == NULL) {
2326                         dentry = opendata->dentry;
2327                 } else if (dentry != ctx->dentry) {
2328                         dput(ctx->dentry);
2329                         ctx->dentry = dget(dentry);
2330                 }
2331                 nfs_set_verifier(dentry,
2332                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2333         }
2334
2335         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2336         if (ret != 0)
2337                 goto out;
2338
2339         ctx->state = state;
2340         if (d_inode(dentry) == state->inode) {
2341                 nfs_inode_attach_open_context(ctx);
2342                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2343                         nfs4_schedule_stateid_recovery(server, state);
2344         }
2345 out:
2346         return ret;
2347 }
2348
2349 /*
2350  * Returns a referenced nfs4_state
2351  */
2352 static int _nfs4_do_open(struct inode *dir,
2353                         struct nfs_open_context *ctx,
2354                         int flags,
2355                         struct iattr *sattr,
2356                         struct nfs4_label *label,
2357                         int *opened)
2358 {
2359         struct nfs4_state_owner  *sp;
2360         struct nfs4_state     *state = NULL;
2361         struct nfs_server       *server = NFS_SERVER(dir);
2362         struct nfs4_opendata *opendata;
2363         struct dentry *dentry = ctx->dentry;
2364         struct rpc_cred *cred = ctx->cred;
2365         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2366         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2367         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2368         struct nfs4_label *olabel = NULL;
2369         int status;
2370
2371         /* Protect against reboot recovery conflicts */
2372         status = -ENOMEM;
2373         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2374         if (sp == NULL) {
2375                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2376                 goto out_err;
2377         }
2378         status = nfs4_recover_expired_lease(server);
2379         if (status != 0)
2380                 goto err_put_state_owner;
2381         if (d_really_is_positive(dentry))
2382                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2383         status = -ENOMEM;
2384         if (d_really_is_positive(dentry))
2385                 claim = NFS4_OPEN_CLAIM_FH;
2386         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2387                         label, claim, GFP_KERNEL);
2388         if (opendata == NULL)
2389                 goto err_put_state_owner;
2390
2391         if (label) {
2392                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2393                 if (IS_ERR(olabel)) {
2394                         status = PTR_ERR(olabel);
2395                         goto err_opendata_put;
2396                 }
2397         }
2398
2399         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2400                 if (!opendata->f_attr.mdsthreshold) {
2401                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2402                         if (!opendata->f_attr.mdsthreshold)
2403                                 goto err_free_label;
2404                 }
2405                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2406         }
2407         if (d_really_is_positive(dentry))
2408                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2409
2410         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2411         if (status != 0)
2412                 goto err_free_label;
2413         state = ctx->state;
2414
2415         if ((opendata->o_arg.open_flags & O_EXCL) &&
2416             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2417                 nfs4_exclusive_attrset(opendata, sattr);
2418
2419                 nfs_fattr_init(opendata->o_res.f_attr);
2420                 status = nfs4_do_setattr(state->inode, cred,
2421                                 opendata->o_res.f_attr, sattr,
2422                                 state, label, olabel);
2423                 if (status == 0) {
2424                         nfs_setattr_update_inode(state->inode, sattr,
2425                                         opendata->o_res.f_attr);
2426                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2427                 }
2428         }
2429         if (opendata->file_created)
2430                 *opened |= FILE_CREATED;
2431
2432         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2433                 *ctx_th = opendata->f_attr.mdsthreshold;
2434                 opendata->f_attr.mdsthreshold = NULL;
2435         }
2436
2437         nfs4_label_free(olabel);
2438
2439         nfs4_opendata_put(opendata);
2440         nfs4_put_state_owner(sp);
2441         return 0;
2442 err_free_label:
2443         nfs4_label_free(olabel);
2444 err_opendata_put:
2445         nfs4_opendata_put(opendata);
2446 err_put_state_owner:
2447         nfs4_put_state_owner(sp);
2448 out_err:
2449         return status;
2450 }
2451
2452
2453 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2454                                         struct nfs_open_context *ctx,
2455                                         int flags,
2456                                         struct iattr *sattr,
2457                                         struct nfs4_label *label,
2458                                         int *opened)
2459 {
2460         struct nfs_server *server = NFS_SERVER(dir);
2461         struct nfs4_exception exception = { };
2462         struct nfs4_state *res;
2463         int status;
2464
2465         do {
2466                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2467                 res = ctx->state;
2468                 trace_nfs4_open_file(ctx, flags, status);
2469                 if (status == 0)
2470                         break;
2471                 /* NOTE: BAD_SEQID means the server and client disagree about the
2472                  * book-keeping w.r.t. state-changing operations
2473                  * (OPEN/CLOSE/LOCK/LOCKU...)
2474                  * It is actually a sign of a bug on the client or on the server.
2475                  *
2476                  * If we receive a BAD_SEQID error in the particular case of
2477                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2478                  * have unhashed the old state_owner for us, and that we can
2479                  * therefore safely retry using a new one. We should still warn
2480                  * the user though...
2481                  */
2482                 if (status == -NFS4ERR_BAD_SEQID) {
2483                         pr_warn_ratelimited("NFS: v4 server %s "
2484                                         " returned a bad sequence-id error!\n",
2485                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2486                         exception.retry = 1;
2487                         continue;
2488                 }
2489                 /*
2490                  * BAD_STATEID on OPEN means that the server cancelled our
2491                  * state before it received the OPEN_CONFIRM.
2492                  * Recover by retrying the request as per the discussion
2493                  * on Page 181 of RFC3530.
2494                  */
2495                 if (status == -NFS4ERR_BAD_STATEID) {
2496                         exception.retry = 1;
2497                         continue;
2498                 }
2499                 if (status == -EAGAIN) {
2500                         /* We must have found a delegation */
2501                         exception.retry = 1;
2502                         continue;
2503                 }
2504                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2505                         continue;
2506                 res = ERR_PTR(nfs4_handle_exception(server,
2507                                         status, &exception));
2508         } while (exception.retry);
2509         return res;
2510 }
2511
2512 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2513                             struct nfs_fattr *fattr, struct iattr *sattr,
2514                             struct nfs4_state *state, struct nfs4_label *ilabel,
2515                             struct nfs4_label *olabel)
2516 {
2517         struct nfs_server *server = NFS_SERVER(inode);
2518         struct nfs_setattrargs  arg = {
2519                 .fh             = NFS_FH(inode),
2520                 .iap            = sattr,
2521                 .server         = server,
2522                 .bitmask = server->attr_bitmask,
2523                 .label          = ilabel,
2524         };
2525         struct nfs_setattrres  res = {
2526                 .fattr          = fattr,
2527                 .label          = olabel,
2528                 .server         = server,
2529         };
2530         struct rpc_message msg = {
2531                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2532                 .rpc_argp       = &arg,
2533                 .rpc_resp       = &res,
2534                 .rpc_cred       = cred,
2535         };
2536         unsigned long timestamp = jiffies;
2537         fmode_t fmode;
2538         bool truncate;
2539         int status;
2540
2541         arg.bitmask = nfs4_bitmask(server, ilabel);
2542         if (ilabel)
2543                 arg.bitmask = nfs4_bitmask(server, olabel);
2544
2545         nfs_fattr_init(fattr);
2546
2547         /* Servers should only apply open mode checks for file size changes */
2548         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2549         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2550
2551         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2552                 /* Use that stateid */
2553         } else if (truncate && state != NULL) {
2554                 struct nfs_lockowner lockowner = {
2555                         .l_owner = current->files,
2556                         .l_pid = current->tgid,
2557                 };
2558                 if (!nfs4_valid_open_stateid(state))
2559                         return -EBADF;
2560                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2561                                 &lockowner) == -EIO)
2562                         return -EBADF;
2563         } else
2564                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2565
2566         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2567         if (status == 0 && state != NULL)
2568                 renew_lease(server, timestamp);
2569         return status;
2570 }
2571
2572 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2573                            struct nfs_fattr *fattr, struct iattr *sattr,
2574                            struct nfs4_state *state, struct nfs4_label *ilabel,
2575                            struct nfs4_label *olabel)
2576 {
2577         struct nfs_server *server = NFS_SERVER(inode);
2578         struct nfs4_exception exception = {
2579                 .state = state,
2580                 .inode = inode,
2581         };
2582         int err;
2583         do {
2584                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2585                 trace_nfs4_setattr(inode, err);
2586                 switch (err) {
2587                 case -NFS4ERR_OPENMODE:
2588                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2589                                 pr_warn_once("NFSv4: server %s is incorrectly "
2590                                                 "applying open mode checks to "
2591                                                 "a SETATTR that is not "
2592                                                 "changing file size.\n",
2593                                                 server->nfs_client->cl_hostname);
2594                         }
2595                         if (state && !(state->state & FMODE_WRITE)) {
2596                                 err = -EBADF;
2597                                 if (sattr->ia_valid & ATTR_OPEN)
2598                                         err = -EACCES;
2599                                 goto out;
2600                         }
2601                 }
2602                 err = nfs4_handle_exception(server, err, &exception);
2603         } while (exception.retry);
2604 out:
2605         return err;
2606 }
2607
2608 struct nfs4_closedata {
2609         struct inode *inode;
2610         struct nfs4_state *state;
2611         struct nfs_closeargs arg;
2612         struct nfs_closeres res;
2613         struct nfs_fattr fattr;
2614         unsigned long timestamp;
2615         bool roc;
2616         u32 roc_barrier;
2617 };
2618
2619 static void nfs4_free_closedata(void *data)
2620 {
2621         struct nfs4_closedata *calldata = data;
2622         struct nfs4_state_owner *sp = calldata->state->owner;
2623         struct super_block *sb = calldata->state->inode->i_sb;
2624
2625         if (calldata->roc)
2626                 pnfs_roc_release(calldata->state->inode);
2627         nfs4_put_open_state(calldata->state);
2628         nfs_free_seqid(calldata->arg.seqid);
2629         nfs4_put_state_owner(sp);
2630         nfs_sb_deactive(sb);
2631         kfree(calldata);
2632 }
2633
2634 static void nfs4_close_done(struct rpc_task *task, void *data)
2635 {
2636         struct nfs4_closedata *calldata = data;
2637         struct nfs4_state *state = calldata->state;
2638         struct nfs_server *server = NFS_SERVER(calldata->inode);
2639         nfs4_stateid *res_stateid = NULL;
2640
2641         dprintk("%s: begin!\n", __func__);
2642         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2643                 return;
2644         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2645         /* hmm. we are done with the inode, and in the process of freeing
2646          * the state_owner. we keep this around to process errors
2647          */
2648         switch (task->tk_status) {
2649                 case 0:
2650                         res_stateid = &calldata->res.stateid;
2651                         if (calldata->arg.fmode == 0 && calldata->roc)
2652                                 pnfs_roc_set_barrier(state->inode,
2653                                                      calldata->roc_barrier);
2654                         renew_lease(server, calldata->timestamp);
2655                         break;
2656                 case -NFS4ERR_ADMIN_REVOKED:
2657                 case -NFS4ERR_STALE_STATEID:
2658                 case -NFS4ERR_OLD_STATEID:
2659                 case -NFS4ERR_BAD_STATEID:
2660                 case -NFS4ERR_EXPIRED:
2661                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2662                                                 &state->open_stateid)) {
2663                                 rpc_restart_call_prepare(task);
2664                                 goto out_release;
2665                         }
2666                         if (calldata->arg.fmode == 0)
2667                                 break;
2668                 default:
2669                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2670                                 rpc_restart_call_prepare(task);
2671                                 goto out_release;
2672                         }
2673         }
2674         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2675 out_release:
2676         nfs_release_seqid(calldata->arg.seqid);
2677         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2678         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2679 }
2680
2681 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2682 {
2683         struct nfs4_closedata *calldata = data;
2684         struct nfs4_state *state = calldata->state;
2685         struct inode *inode = calldata->inode;
2686         bool is_rdonly, is_wronly, is_rdwr;
2687         int call_close = 0;
2688
2689         dprintk("%s: begin!\n", __func__);
2690         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2691                 goto out_wait;
2692
2693         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2694         spin_lock(&state->owner->so_lock);
2695         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2696         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2697         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2698         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2699         /* Calculate the change in open mode */
2700         calldata->arg.fmode = 0;
2701         if (state->n_rdwr == 0) {
2702                 if (state->n_rdonly == 0)
2703                         call_close |= is_rdonly;
2704                 else if (is_rdonly)
2705                         calldata->arg.fmode |= FMODE_READ;
2706                 if (state->n_wronly == 0)
2707                         call_close |= is_wronly;
2708                 else if (is_wronly)
2709                         calldata->arg.fmode |= FMODE_WRITE;
2710         } else if (is_rdwr)
2711                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2712
2713         if (calldata->arg.fmode == 0)
2714                 call_close |= is_rdwr;
2715
2716         if (!nfs4_valid_open_stateid(state))
2717                 call_close = 0;
2718         spin_unlock(&state->owner->so_lock);
2719
2720         if (!call_close) {
2721                 /* Note: exit _without_ calling nfs4_close_done */
2722                 goto out_no_action;
2723         }
2724
2725         if (calldata->arg.fmode == 0) {
2726                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2727                 if (calldata->roc &&
2728                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2729                         nfs_release_seqid(calldata->arg.seqid);
2730                         goto out_wait;
2731                     }
2732         }
2733         calldata->arg.share_access =
2734                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2735                                 calldata->arg.fmode, 0);
2736
2737         nfs_fattr_init(calldata->res.fattr);
2738         calldata->timestamp = jiffies;
2739         if (nfs4_setup_sequence(NFS_SERVER(inode),
2740                                 &calldata->arg.seq_args,
2741                                 &calldata->res.seq_res,
2742                                 task) != 0)
2743                 nfs_release_seqid(calldata->arg.seqid);
2744         dprintk("%s: done!\n", __func__);
2745         return;
2746 out_no_action:
2747         task->tk_action = NULL;
2748 out_wait:
2749         nfs4_sequence_done(task, &calldata->res.seq_res);
2750 }
2751
2752 static const struct rpc_call_ops nfs4_close_ops = {
2753         .rpc_call_prepare = nfs4_close_prepare,
2754         .rpc_call_done = nfs4_close_done,
2755         .rpc_release = nfs4_free_closedata,
2756 };
2757
2758 static bool nfs4_roc(struct inode *inode)
2759 {
2760         if (!nfs_have_layout(inode))
2761                 return false;
2762         return pnfs_roc(inode);
2763 }
2764
2765 /*
2766  * It is possible for data to be read/written from a mem-mapped file
2767  * after the sys_close call (which hits the vfs layer as a flush).
2768  * This means that we can't safely call nfsv4 close on a file until
2769  * the inode is cleared. This in turn means that we are not good
2770  * NFSv4 citizens - we do not indicate to the server to update the file's
2771  * share state even when we are done with one of the three share
2772  * stateid's in the inode.
2773  *
2774  * NOTE: Caller must be holding the sp->so_owner semaphore!
2775  */
2776 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2777 {
2778         struct nfs_server *server = NFS_SERVER(state->inode);
2779         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2780         struct nfs4_closedata *calldata;
2781         struct nfs4_state_owner *sp = state->owner;
2782         struct rpc_task *task;
2783         struct rpc_message msg = {
2784                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2785                 .rpc_cred = state->owner->so_cred,
2786         };
2787         struct rpc_task_setup task_setup_data = {
2788                 .rpc_client = server->client,
2789                 .rpc_message = &msg,
2790                 .callback_ops = &nfs4_close_ops,
2791                 .workqueue = nfsiod_workqueue,
2792                 .flags = RPC_TASK_ASYNC,
2793         };
2794         int status = -ENOMEM;
2795
2796         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2797                 &task_setup_data.rpc_client, &msg);
2798
2799         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2800         if (calldata == NULL)
2801                 goto out;
2802         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2803         calldata->inode = state->inode;
2804         calldata->state = state;
2805         calldata->arg.fh = NFS_FH(state->inode);
2806         /* Serialization for the sequence id */
2807         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2808         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2809         if (IS_ERR(calldata->arg.seqid))
2810                 goto out_free_calldata;
2811         calldata->arg.fmode = 0;
2812         calldata->arg.bitmask = server->cache_consistency_bitmask;
2813         calldata->res.fattr = &calldata->fattr;
2814         calldata->res.seqid = calldata->arg.seqid;
2815         calldata->res.server = server;
2816         calldata->roc = nfs4_roc(state->inode);
2817         nfs_sb_active(calldata->inode->i_sb);
2818
2819         msg.rpc_argp = &calldata->arg;
2820         msg.rpc_resp = &calldata->res;
2821         task_setup_data.callback_data = calldata;
2822         task = rpc_run_task(&task_setup_data);
2823         if (IS_ERR(task))
2824                 return PTR_ERR(task);
2825         status = 0;
2826         if (wait)
2827                 status = rpc_wait_for_completion_task(task);
2828         rpc_put_task(task);
2829         return status;
2830 out_free_calldata:
2831         kfree(calldata);
2832 out:
2833         nfs4_put_open_state(state);
2834         nfs4_put_state_owner(sp);
2835         return status;
2836 }
2837
2838 static struct inode *
2839 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2840                 int open_flags, struct iattr *attr, int *opened)
2841 {
2842         struct nfs4_state *state;
2843         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2844
2845         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2846
2847         /* Protect against concurrent sillydeletes */
2848         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2849
2850         nfs4_label_release_security(label);
2851
2852         if (IS_ERR(state))
2853                 return ERR_CAST(state);
2854         return state->inode;
2855 }
2856
2857 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2858 {
2859         if (ctx->state == NULL)
2860                 return;
2861         if (is_sync)
2862                 nfs4_close_sync(ctx->state, ctx->mode);
2863         else
2864                 nfs4_close_state(ctx->state, ctx->mode);
2865 }
2866
2867 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2868 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2869 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2870
2871 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2872 {
2873         struct nfs4_server_caps_arg args = {
2874                 .fhandle = fhandle,
2875         };
2876         struct nfs4_server_caps_res res = {};
2877         struct rpc_message msg = {
2878                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2879                 .rpc_argp = &args,
2880                 .rpc_resp = &res,
2881         };
2882         int status;
2883
2884         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2885         if (status == 0) {
2886                 /* Sanity check the server answers */
2887                 switch (server->nfs_client->cl_minorversion) {
2888                 case 0:
2889                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2890                         res.attr_bitmask[2] = 0;
2891                         break;
2892                 case 1:
2893                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2894                         break;
2895                 case 2:
2896                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2897                 }
2898                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2899                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2900                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2901                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2902                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2903                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2904                                 NFS_CAP_SECURITY_LABEL);
2905                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2906                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2907                         server->caps |= NFS_CAP_ACLS;
2908                 if (res.has_links != 0)
2909                         server->caps |= NFS_CAP_HARDLINKS;
2910                 if (res.has_symlinks != 0)
2911                         server->caps |= NFS_CAP_SYMLINKS;
2912                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2913                         server->caps |= NFS_CAP_FILEID;
2914                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2915                         server->caps |= NFS_CAP_MODE;
2916                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2917                         server->caps |= NFS_CAP_NLINK;
2918                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2919                         server->caps |= NFS_CAP_OWNER;
2920                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2921                         server->caps |= NFS_CAP_OWNER_GROUP;
2922                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2923                         server->caps |= NFS_CAP_ATIME;
2924                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2925                         server->caps |= NFS_CAP_CTIME;
2926                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2927                         server->caps |= NFS_CAP_MTIME;
2928 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2929                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2930                         server->caps |= NFS_CAP_SECURITY_LABEL;
2931 #endif
2932                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2933                                 sizeof(server->attr_bitmask));
2934                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2935
2936                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2937                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2938                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2939                 server->cache_consistency_bitmask[2] = 0;
2940                 server->acl_bitmask = res.acl_bitmask;
2941                 server->fh_expire_type = res.fh_expire_type;
2942         }
2943
2944         return status;
2945 }
2946
2947 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2948 {
2949         struct nfs4_exception exception = { };
2950         int err;
2951         do {
2952                 err = nfs4_handle_exception(server,
2953                                 _nfs4_server_capabilities(server, fhandle),
2954                                 &exception);
2955         } while (exception.retry);
2956         return err;
2957 }
2958
2959 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2960                 struct nfs_fsinfo *info)
2961 {
2962         u32 bitmask[3];
2963         struct nfs4_lookup_root_arg args = {
2964                 .bitmask = bitmask,
2965         };
2966         struct nfs4_lookup_res res = {
2967                 .server = server,
2968                 .fattr = info->fattr,
2969                 .fh = fhandle,
2970         };
2971         struct rpc_message msg = {
2972                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2973                 .rpc_argp = &args,
2974                 .rpc_resp = &res,
2975         };
2976
2977         bitmask[0] = nfs4_fattr_bitmap[0];
2978         bitmask[1] = nfs4_fattr_bitmap[1];
2979         /*
2980          * Process the label in the upcoming getfattr
2981          */
2982         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2983
2984         nfs_fattr_init(info->fattr);
2985         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2986 }
2987
2988 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2989                 struct nfs_fsinfo *info)
2990 {
2991         struct nfs4_exception exception = { };
2992         int err;
2993         do {
2994                 err = _nfs4_lookup_root(server, fhandle, info);
2995                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2996                 switch (err) {
2997                 case 0:
2998                 case -NFS4ERR_WRONGSEC:
2999                         goto out;
3000                 default:
3001                         err = nfs4_handle_exception(server, err, &exception);
3002                 }
3003         } while (exception.retry);
3004 out:
3005         return err;
3006 }
3007
3008 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3009                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3010 {
3011         struct rpc_auth_create_args auth_args = {
3012                 .pseudoflavor = flavor,
3013         };
3014         struct rpc_auth *auth;
3015         int ret;
3016
3017         auth = rpcauth_create(&auth_args, server->client);
3018         if (IS_ERR(auth)) {
3019                 ret = -EACCES;
3020                 goto out;
3021         }
3022         ret = nfs4_lookup_root(server, fhandle, info);
3023 out:
3024         return ret;
3025 }
3026
3027 /*
3028  * Retry pseudoroot lookup with various security flavors.  We do this when:
3029  *
3030  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3031  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3032  *
3033  * Returns zero on success, or a negative NFS4ERR value, or a
3034  * negative errno value.
3035  */
3036 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3037                               struct nfs_fsinfo *info)
3038 {
3039         /* Per 3530bis 15.33.5 */
3040         static const rpc_authflavor_t flav_array[] = {
3041                 RPC_AUTH_GSS_KRB5P,
3042                 RPC_AUTH_GSS_KRB5I,
3043                 RPC_AUTH_GSS_KRB5,
3044                 RPC_AUTH_UNIX,                  /* courtesy */
3045                 RPC_AUTH_NULL,
3046         };
3047         int status = -EPERM;
3048         size_t i;
3049
3050         if (server->auth_info.flavor_len > 0) {
3051                 /* try each flavor specified by user */
3052                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3053                         status = nfs4_lookup_root_sec(server, fhandle, info,
3054                                                 server->auth_info.flavors[i]);
3055                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3056                                 continue;
3057                         break;
3058                 }
3059         } else {
3060                 /* no flavors specified by user, try default list */
3061                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3062                         status = nfs4_lookup_root_sec(server, fhandle, info,
3063                                                       flav_array[i]);
3064                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3065                                 continue;
3066                         break;
3067                 }
3068         }
3069
3070         /*
3071          * -EACCESS could mean that the user doesn't have correct permissions
3072          * to access the mount.  It could also mean that we tried to mount
3073          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3074          * existing mount programs don't handle -EACCES very well so it should
3075          * be mapped to -EPERM instead.
3076          */
3077         if (status == -EACCES)
3078                 status = -EPERM;
3079         return status;
3080 }
3081
3082 static int nfs4_do_find_root_sec(struct nfs_server *server,
3083                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3084 {
3085         int mv = server->nfs_client->cl_minorversion;
3086         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3087 }
3088
3089 /**
3090  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3091  * @server: initialized nfs_server handle
3092  * @fhandle: we fill in the pseudo-fs root file handle
3093  * @info: we fill in an FSINFO struct
3094  * @auth_probe: probe the auth flavours
3095  *
3096  * Returns zero on success, or a negative errno.
3097  */
3098 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3099                          struct nfs_fsinfo *info,
3100                          bool auth_probe)
3101 {
3102         int status = 0;
3103
3104         if (!auth_probe)
3105                 status = nfs4_lookup_root(server, fhandle, info);
3106
3107         if (auth_probe || status == NFS4ERR_WRONGSEC)
3108                 status = nfs4_do_find_root_sec(server, fhandle, info);
3109
3110         if (status == 0)
3111                 status = nfs4_server_capabilities(server, fhandle);
3112         if (status == 0)
3113                 status = nfs4_do_fsinfo(server, fhandle, info);
3114
3115         return nfs4_map_errors(status);
3116 }
3117
3118 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3119                               struct nfs_fsinfo *info)
3120 {
3121         int error;
3122         struct nfs_fattr *fattr = info->fattr;
3123         struct nfs4_label *label = NULL;
3124
3125         error = nfs4_server_capabilities(server, mntfh);
3126         if (error < 0) {
3127                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3128                 return error;
3129         }
3130
3131         label = nfs4_label_alloc(server, GFP_KERNEL);
3132         if (IS_ERR(label))
3133                 return PTR_ERR(label);
3134
3135         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3136         if (error < 0) {
3137                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3138                 goto err_free_label;
3139         }
3140
3141         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3142             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3143                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3144
3145 err_free_label:
3146         nfs4_label_free(label);
3147
3148         return error;
3149 }
3150
3151 /*
3152  * Get locations and (maybe) other attributes of a referral.
3153  * Note that we'll actually follow the referral later when
3154  * we detect fsid mismatch in inode revalidation
3155  */
3156 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3157                              const struct qstr *name, struct nfs_fattr *fattr,
3158                              struct nfs_fh *fhandle)
3159 {
3160         int status = -ENOMEM;
3161         struct page *page = NULL;
3162         struct nfs4_fs_locations *locations = NULL;
3163
3164         page = alloc_page(GFP_KERNEL);
3165         if (page == NULL)
3166                 goto out;
3167         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3168         if (locations == NULL)
3169                 goto out;
3170
3171         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3172         if (status != 0)
3173                 goto out;
3174
3175         /*
3176          * If the fsid didn't change, this is a migration event, not a
3177          * referral.  Cause us to drop into the exception handler, which
3178          * will kick off migration recovery.
3179          */
3180         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3181                 dprintk("%s: server did not return a different fsid for"
3182                         " a referral at %s\n", __func__, name->name);
3183                 status = -NFS4ERR_MOVED;
3184                 goto out;
3185         }
3186         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3187         nfs_fixup_referral_attributes(&locations->fattr);
3188
3189         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3190         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3191         memset(fhandle, 0, sizeof(struct nfs_fh));
3192 out:
3193         if (page)
3194                 __free_page(page);
3195         kfree(locations);
3196         return status;
3197 }
3198
3199 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3200                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3201 {
3202         struct nfs4_getattr_arg args = {
3203                 .fh = fhandle,
3204                 .bitmask = server->attr_bitmask,
3205         };
3206         struct nfs4_getattr_res res = {
3207                 .fattr = fattr,
3208                 .label = label,
3209                 .server = server,
3210         };
3211         struct rpc_message msg = {
3212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3213                 .rpc_argp = &args,
3214                 .rpc_resp = &res,
3215         };
3216
3217         args.bitmask = nfs4_bitmask(server, label);
3218
3219         nfs_fattr_init(fattr);
3220         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3221 }
3222
3223 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3224                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3225 {
3226         struct nfs4_exception exception = { };
3227         int err;
3228         do {
3229                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3230                 trace_nfs4_getattr(server, fhandle, fattr, err);
3231                 err = nfs4_handle_exception(server, err,
3232                                 &exception);
3233         } while (exception.retry);
3234         return err;
3235 }
3236
3237 /*
3238  * The file is not closed if it is opened due to the a request to change
3239  * the size of the file. The open call will not be needed once the
3240  * VFS layer lookup-intents are implemented.
3241  *
3242  * Close is called when the inode is destroyed.
3243  * If we haven't opened the file for O_WRONLY, we
3244  * need to in the size_change case to obtain a stateid.
3245  *
3246  * Got race?
3247  * Because OPEN is always done by name in nfsv4, it is
3248  * possible that we opened a different file by the same
3249  * name.  We can recognize this race condition, but we
3250  * can't do anything about it besides returning an error.
3251  *
3252  * This will be fixed with VFS changes (lookup-intent).
3253  */
3254 static int
3255 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3256                   struct iattr *sattr)
3257 {
3258         struct inode *inode = d_inode(dentry);
3259         struct rpc_cred *cred = NULL;
3260         struct nfs4_state *state = NULL;
3261         struct nfs4_label *label = NULL;
3262         int status;
3263
3264         if (pnfs_ld_layoutret_on_setattr(inode) &&
3265             sattr->ia_valid & ATTR_SIZE &&
3266             sattr->ia_size < i_size_read(inode))
3267                 pnfs_commit_and_return_layout(inode);
3268
3269         nfs_fattr_init(fattr);
3270
3271         /* Deal with open(O_TRUNC) */
3272         if (sattr->ia_valid & ATTR_OPEN)
3273                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3274
3275         /* Optimization: if the end result is no change, don't RPC */
3276         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3277                 return 0;
3278
3279         /* Search for an existing open(O_WRITE) file */
3280         if (sattr->ia_valid & ATTR_FILE) {
3281                 struct nfs_open_context *ctx;
3282
3283                 ctx = nfs_file_open_context(sattr->ia_file);
3284                 if (ctx) {
3285                         cred = ctx->cred;
3286                         state = ctx->state;
3287                 }
3288         }
3289
3290         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3291         if (IS_ERR(label))
3292                 return PTR_ERR(label);
3293
3294         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3295         if (status == 0) {
3296                 nfs_setattr_update_inode(inode, sattr, fattr);
3297                 nfs_setsecurity(inode, fattr, label);
3298         }
3299         nfs4_label_free(label);
3300         return status;
3301 }
3302
3303 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3304                 const struct qstr *name, struct nfs_fh *fhandle,
3305                 struct nfs_fattr *fattr, struct nfs4_label *label)
3306 {
3307         struct nfs_server *server = NFS_SERVER(dir);
3308         int                    status;
3309         struct nfs4_lookup_arg args = {
3310                 .bitmask = server->attr_bitmask,
3311                 .dir_fh = NFS_FH(dir),
3312                 .name = name,
3313         };
3314         struct nfs4_lookup_res res = {
3315                 .server = server,
3316                 .fattr = fattr,
3317                 .label = label,
3318                 .fh = fhandle,
3319         };
3320         struct rpc_message msg = {
3321                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3322                 .rpc_argp = &args,
3323                 .rpc_resp = &res,
3324         };
3325
3326         args.bitmask = nfs4_bitmask(server, label);
3327
3328         nfs_fattr_init(fattr);
3329
3330         dprintk("NFS call  lookup %s\n", name->name);
3331         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3332         dprintk("NFS reply lookup: %d\n", status);
3333         return status;
3334 }
3335
3336 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3337 {
3338         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3339                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3340         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3341         fattr->nlink = 2;
3342 }
3343
3344 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3345                                    struct qstr *name, struct nfs_fh *fhandle,
3346                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3347 {
3348         struct nfs4_exception exception = { };
3349         struct rpc_clnt *client = *clnt;
3350         int err;
3351         do {
3352                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3353                 trace_nfs4_lookup(dir, name, err);
3354                 switch (err) {
3355                 case -NFS4ERR_BADNAME:
3356                         err = -ENOENT;
3357                         goto out;
3358                 case -NFS4ERR_MOVED:
3359                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3360                         if (err == -NFS4ERR_MOVED)
3361                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3362                         goto out;
3363                 case -NFS4ERR_WRONGSEC:
3364                         err = -EPERM;
3365                         if (client != *clnt)
3366                                 goto out;
3367                         client = nfs4_negotiate_security(client, dir, name);
3368                         if (IS_ERR(client))
3369                                 return PTR_ERR(client);
3370
3371                         exception.retry = 1;
3372                         break;
3373                 default:
3374                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3375                 }
3376         } while (exception.retry);
3377
3378 out:
3379         if (err == 0)
3380                 *clnt = client;
3381         else if (client != *clnt)
3382                 rpc_shutdown_client(client);
3383
3384         return err;
3385 }
3386
3387 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3388                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3389                             struct nfs4_label *label)
3390 {
3391         int status;
3392         struct rpc_clnt *client = NFS_CLIENT(dir);
3393
3394         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3395         if (client != NFS_CLIENT(dir)) {
3396                 rpc_shutdown_client(client);
3397                 nfs_fixup_secinfo_attributes(fattr);
3398         }
3399         return status;
3400 }
3401
3402 struct rpc_clnt *
3403 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3404                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3405 {
3406         struct rpc_clnt *client = NFS_CLIENT(dir);
3407         int status;
3408
3409         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3410         if (status < 0)
3411                 return ERR_PTR(status);
3412         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3413 }
3414
3415 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3416 {
3417         struct nfs_server *server = NFS_SERVER(inode);
3418         struct nfs4_accessargs args = {
3419                 .fh = NFS_FH(inode),
3420                 .bitmask = server->cache_consistency_bitmask,
3421         };
3422         struct nfs4_accessres res = {
3423                 .server = server,
3424         };
3425         struct rpc_message msg = {
3426                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3427                 .rpc_argp = &args,
3428                 .rpc_resp = &res,
3429                 .rpc_cred = entry->cred,
3430         };
3431         int mode = entry->mask;
3432         int status = 0;
3433
3434         /*
3435          * Determine which access bits we want to ask for...
3436          */
3437         if (mode & MAY_READ)
3438                 args.access |= NFS4_ACCESS_READ;
3439         if (S_ISDIR(inode->i_mode)) {
3440                 if (mode & MAY_WRITE)
3441                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3442                 if (mode & MAY_EXEC)
3443                         args.access |= NFS4_ACCESS_LOOKUP;
3444         } else {
3445                 if (mode & MAY_WRITE)
3446                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3447                 if (mode & MAY_EXEC)
3448                         args.access |= NFS4_ACCESS_EXECUTE;
3449         }
3450
3451         res.fattr = nfs_alloc_fattr();
3452         if (res.fattr == NULL)
3453                 return -ENOMEM;
3454
3455         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3456         if (!status) {
3457                 nfs_access_set_mask(entry, res.access);
3458                 nfs_refresh_inode(inode, res.fattr);
3459         }
3460         nfs_free_fattr(res.fattr);
3461         return status;
3462 }
3463
3464 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3465 {
3466         struct nfs4_exception exception = { };
3467         int err;
3468         do {
3469                 err = _nfs4_proc_access(inode, entry);
3470                 trace_nfs4_access(inode, err);
3471                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3472                                 &exception);
3473         } while (exception.retry);
3474         return err;
3475 }
3476
3477 /*
3478  * TODO: For the time being, we don't try to get any attributes
3479  * along with any of the zero-copy operations READ, READDIR,
3480  * READLINK, WRITE.
3481  *
3482  * In the case of the first three, we want to put the GETATTR
3483  * after the read-type operation -- this is because it is hard
3484  * to predict the length of a GETATTR response in v4, and thus
3485  * align the READ data correctly.  This means that the GETATTR
3486  * may end up partially falling into the page cache, and we should
3487  * shift it into the 'tail' of the xdr_buf before processing.
3488  * To do this efficiently, we need to know the total length
3489  * of data received, which doesn't seem to be available outside
3490  * of the RPC layer.
3491  *
3492  * In the case of WRITE, we also want to put the GETATTR after
3493  * the operation -- in this case because we want to make sure
3494  * we get the post-operation mtime and size.
3495  *
3496  * Both of these changes to the XDR layer would in fact be quite
3497  * minor, but I decided to leave them for a subsequent patch.
3498  */
3499 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3500                 unsigned int pgbase, unsigned int pglen)
3501 {
3502         struct nfs4_readlink args = {
3503                 .fh       = NFS_FH(inode),
3504                 .pgbase   = pgbase,
3505                 .pglen    = pglen,
3506                 .pages    = &page,
3507         };
3508         struct nfs4_readlink_res res;
3509         struct rpc_message msg = {
3510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3511                 .rpc_argp = &args,
3512                 .rpc_resp = &res,
3513         };
3514
3515         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3516 }
3517
3518 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3519                 unsigned int pgbase, unsigned int pglen)
3520 {
3521         struct nfs4_exception exception = { };
3522         int err;
3523         do {
3524                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3525                 trace_nfs4_readlink(inode, err);
3526                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3527                                 &exception);
3528         } while (exception.retry);
3529         return err;
3530 }
3531
3532 /*
3533  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3534  */
3535 static int
3536 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3537                  int flags)
3538 {
3539         struct nfs4_label l, *ilabel = NULL;
3540         struct nfs_open_context *ctx;
3541         struct nfs4_state *state;
3542         int opened = 0;
3543         int status = 0;
3544
3545         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3546         if (IS_ERR(ctx))
3547                 return PTR_ERR(ctx);
3548
3549         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3550
3551         sattr->ia_mode &= ~current_umask();
3552         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3553         if (IS_ERR(state)) {
3554                 status = PTR_ERR(state);
3555                 goto out;
3556         }
3557 out:
3558         nfs4_label_release_security(ilabel);
3559         put_nfs_open_context(ctx);
3560         return status;
3561 }
3562
3563 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3564 {
3565         struct nfs_server *server = NFS_SERVER(dir);
3566         struct nfs_removeargs args = {
3567                 .fh = NFS_FH(dir),
3568                 .name = *name,
3569         };
3570         struct nfs_removeres res = {
3571                 .server = server,
3572         };
3573         struct rpc_message msg = {
3574                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3575                 .rpc_argp = &args,
3576                 .rpc_resp = &res,
3577         };
3578         int status;
3579
3580         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3581         if (status == 0)
3582                 update_changeattr(dir, &res.cinfo);
3583         return status;
3584 }
3585
3586 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3587 {
3588         struct nfs4_exception exception = { };
3589         int err;
3590         do {
3591                 err = _nfs4_proc_remove(dir, name);
3592                 trace_nfs4_remove(dir, name, err);
3593                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3594                                 &exception);
3595         } while (exception.retry);
3596         return err;
3597 }
3598
3599 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3600 {
3601         struct nfs_server *server = NFS_SERVER(dir);
3602         struct nfs_removeargs *args = msg->rpc_argp;
3603         struct nfs_removeres *res = msg->rpc_resp;
3604
3605         res->server = server;
3606         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3607         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3608
3609         nfs_fattr_init(res->dir_attr);
3610 }
3611
3612 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3613 {
3614         nfs4_setup_sequence(NFS_SERVER(data->dir),
3615                         &data->args.seq_args,
3616                         &data->res.seq_res,
3617                         task);
3618 }
3619
3620 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3621 {
3622         struct nfs_unlinkdata *data = task->tk_calldata;
3623         struct nfs_removeres *res = &data->res;
3624
3625         if (!nfs4_sequence_done(task, &res->seq_res))
3626                 return 0;
3627         if (nfs4_async_handle_error(task, res->server, NULL,
3628                                     &data->timeout) == -EAGAIN)
3629                 return 0;
3630         update_changeattr(dir, &res->cinfo);
3631         return 1;
3632 }
3633
3634 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3635 {
3636         struct nfs_server *server = NFS_SERVER(dir);
3637         struct nfs_renameargs *arg = msg->rpc_argp;
3638         struct nfs_renameres *res = msg->rpc_resp;
3639
3640         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3641         res->server = server;
3642         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3643 }
3644
3645 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3646 {
3647         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3648                         &data->args.seq_args,
3649                         &data->res.seq_res,
3650                         task);
3651 }
3652
3653 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3654                                  struct inode *new_dir)
3655 {
3656         struct nfs_renamedata *data = task->tk_calldata;
3657         struct nfs_renameres *res = &data->res;
3658
3659         if (!nfs4_sequence_done(task, &res->seq_res))
3660                 return 0;
3661         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3662                 return 0;
3663
3664         update_changeattr(old_dir, &res->old_cinfo);
3665         update_changeattr(new_dir, &res->new_cinfo);
3666         return 1;
3667 }
3668
3669 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3670 {
3671         struct nfs_server *server = NFS_SERVER(inode);
3672         struct nfs4_link_arg arg = {
3673                 .fh     = NFS_FH(inode),
3674                 .dir_fh = NFS_FH(dir),
3675                 .name   = name,
3676                 .bitmask = server->attr_bitmask,
3677         };
3678         struct nfs4_link_res res = {
3679                 .server = server,
3680                 .label = NULL,
3681         };
3682         struct rpc_message msg = {
3683                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3684                 .rpc_argp = &arg,
3685                 .rpc_resp = &res,
3686         };
3687         int status = -ENOMEM;
3688
3689         res.fattr = nfs_alloc_fattr();
3690         if (res.fattr == NULL)
3691                 goto out;
3692
3693         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3694         if (IS_ERR(res.label)) {
3695                 status = PTR_ERR(res.label);
3696                 goto out;
3697         }
3698         arg.bitmask = nfs4_bitmask(server, res.label);
3699
3700         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3701         if (!status) {
3702                 update_changeattr(dir, &res.cinfo);
3703                 status = nfs_post_op_update_inode(inode, res.fattr);
3704                 if (!status)
3705                         nfs_setsecurity(inode, res.fattr, res.label);
3706         }
3707
3708
3709         nfs4_label_free(res.label);
3710
3711 out:
3712         nfs_free_fattr(res.fattr);
3713         return status;
3714 }
3715
3716 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3717 {
3718         struct nfs4_exception exception = { };
3719         int err;
3720         do {
3721                 err = nfs4_handle_exception(NFS_SERVER(inode),
3722                                 _nfs4_proc_link(inode, dir, name),
3723                                 &exception);
3724         } while (exception.retry);
3725         return err;
3726 }
3727
3728 struct nfs4_createdata {
3729         struct rpc_message msg;
3730         struct nfs4_create_arg arg;
3731         struct nfs4_create_res res;
3732         struct nfs_fh fh;
3733         struct nfs_fattr fattr;
3734         struct nfs4_label *label;
3735 };
3736
3737 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3738                 struct qstr *name, struct iattr *sattr, u32 ftype)
3739 {
3740         struct nfs4_createdata *data;
3741
3742         data = kzalloc(sizeof(*data), GFP_KERNEL);
3743         if (data != NULL) {
3744                 struct nfs_server *server = NFS_SERVER(dir);
3745
3746                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3747                 if (IS_ERR(data->label))
3748                         goto out_free;
3749
3750                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3751                 data->msg.rpc_argp = &data->arg;
3752                 data->msg.rpc_resp = &data->res;
3753                 data->arg.dir_fh = NFS_FH(dir);
3754                 data->arg.server = server;
3755                 data->arg.name = name;
3756                 data->arg.attrs = sattr;
3757                 data->arg.ftype = ftype;
3758                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3759                 data->res.server = server;
3760                 data->res.fh = &data->fh;
3761                 data->res.fattr = &data->fattr;
3762                 data->res.label = data->label;
3763                 nfs_fattr_init(data->res.fattr);
3764         }
3765         return data;
3766 out_free:
3767         kfree(data);
3768         return NULL;
3769 }
3770
3771 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3772 {
3773         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3774                                     &data->arg.seq_args, &data->res.seq_res, 1);
3775         if (status == 0) {
3776                 update_changeattr(dir, &data->res.dir_cinfo);
3777                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3778         }
3779         return status;
3780 }
3781
3782 static void nfs4_free_createdata(struct nfs4_createdata *data)
3783 {
3784         nfs4_label_free(data->label);
3785         kfree(data);
3786 }
3787
3788 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3789                 struct page *page, unsigned int len, struct iattr *sattr,
3790                 struct nfs4_label *label)
3791 {
3792         struct nfs4_createdata *data;
3793         int status = -ENAMETOOLONG;
3794
3795         if (len > NFS4_MAXPATHLEN)
3796                 goto out;
3797
3798         status = -ENOMEM;
3799         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3800         if (data == NULL)
3801                 goto out;
3802
3803         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3804         data->arg.u.symlink.pages = &page;
3805         data->arg.u.symlink.len = len;
3806         data->arg.label = label;
3807
3808         status = nfs4_do_create(dir, dentry, data);
3809
3810         nfs4_free_createdata(data);
3811 out:
3812         return status;
3813 }
3814
3815 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3816                 struct page *page, unsigned int len, struct iattr *sattr)
3817 {
3818         struct nfs4_exception exception = { };
3819         struct nfs4_label l, *label = NULL;
3820         int err;
3821
3822         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3823
3824         do {
3825                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3826                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3827                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3828                                 &exception);
3829         } while (exception.retry);
3830
3831         nfs4_label_release_security(label);
3832         return err;
3833 }
3834
3835 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3836                 struct iattr *sattr, struct nfs4_label *label)
3837 {
3838         struct nfs4_createdata *data;
3839         int status = -ENOMEM;
3840
3841         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3842         if (data == NULL)
3843                 goto out;
3844
3845         data->arg.label = label;
3846         status = nfs4_do_create(dir, dentry, data);
3847
3848         nfs4_free_createdata(data);
3849 out:
3850         return status;
3851 }
3852
3853 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3854                 struct iattr *sattr)
3855 {
3856         struct nfs4_exception exception = { };
3857         struct nfs4_label l, *label = NULL;
3858         int err;
3859
3860         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3861
3862         sattr->ia_mode &= ~current_umask();
3863         do {
3864                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3865                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3866                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3867                                 &exception);
3868         } while (exception.retry);
3869         nfs4_label_release_security(label);
3870
3871         return err;
3872 }
3873
3874 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3875                 u64 cookie, struct page **pages, unsigned int count, int plus)
3876 {
3877         struct inode            *dir = d_inode(dentry);
3878         struct nfs4_readdir_arg args = {
3879                 .fh = NFS_FH(dir),
3880                 .pages = pages,
3881                 .pgbase = 0,
3882                 .count = count,
3883                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3884                 .plus = plus,
3885         };
3886         struct nfs4_readdir_res res;
3887         struct rpc_message msg = {
3888                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3889                 .rpc_argp = &args,
3890                 .rpc_resp = &res,
3891                 .rpc_cred = cred,
3892         };
3893         int                     status;
3894
3895         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3896                         dentry,
3897                         (unsigned long long)cookie);
3898         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3899         res.pgbase = args.pgbase;
3900         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3901         if (status >= 0) {
3902                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3903                 status += args.pgbase;
3904         }
3905
3906         nfs_invalidate_atime(dir);
3907
3908         dprintk("%s: returns %d\n", __func__, status);
3909         return status;
3910 }
3911
3912 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3913                 u64 cookie, struct page **pages, unsigned int count, int plus)
3914 {
3915         struct nfs4_exception exception = { };
3916         int err;
3917         do {
3918                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3919                                 pages, count, plus);
3920                 trace_nfs4_readdir(d_inode(dentry), err);
3921                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3922                                 &exception);
3923         } while (exception.retry);
3924         return err;
3925 }
3926
3927 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3928                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3929 {
3930         struct nfs4_createdata *data;
3931         int mode = sattr->ia_mode;
3932         int status = -ENOMEM;
3933
3934         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3935         if (data == NULL)
3936                 goto out;
3937
3938         if (S_ISFIFO(mode))
3939                 data->arg.ftype = NF4FIFO;
3940         else if (S_ISBLK(mode)) {
3941                 data->arg.ftype = NF4BLK;
3942                 data->arg.u.device.specdata1 = MAJOR(rdev);
3943                 data->arg.u.device.specdata2 = MINOR(rdev);
3944         }
3945         else if (S_ISCHR(mode)) {
3946                 data->arg.ftype = NF4CHR;
3947                 data->arg.u.device.specdata1 = MAJOR(rdev);
3948                 data->arg.u.device.specdata2 = MINOR(rdev);
3949         } else if (!S_ISSOCK(mode)) {
3950                 status = -EINVAL;
3951                 goto out_free;
3952         }
3953
3954         data->arg.label = label;
3955         status = nfs4_do_create(dir, dentry, data);
3956 out_free:
3957         nfs4_free_createdata(data);
3958 out:
3959         return status;
3960 }
3961
3962 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3963                 struct iattr *sattr, dev_t rdev)
3964 {
3965         struct nfs4_exception exception = { };
3966         struct nfs4_label l, *label = NULL;
3967         int err;
3968
3969         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3970
3971         sattr->ia_mode &= ~current_umask();
3972         do {
3973                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3974                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3975                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3976                                 &exception);
3977         } while (exception.retry);
3978
3979         nfs4_label_release_security(label);
3980
3981         return err;
3982 }
3983
3984 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3985                  struct nfs_fsstat *fsstat)
3986 {
3987         struct nfs4_statfs_arg args = {
3988                 .fh = fhandle,
3989                 .bitmask = server->attr_bitmask,
3990         };
3991         struct nfs4_statfs_res res = {
3992                 .fsstat = fsstat,
3993         };
3994         struct rpc_message msg = {
3995                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3996                 .rpc_argp = &args,
3997                 .rpc_resp = &res,
3998         };
3999
4000         nfs_fattr_init(fsstat->fattr);
4001         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4002 }
4003
4004 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4005 {
4006         struct nfs4_exception exception = { };
4007         int err;
4008         do {
4009                 err = nfs4_handle_exception(server,
4010                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4011                                 &exception);
4012         } while (exception.retry);
4013         return err;
4014 }
4015
4016 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4017                 struct nfs_fsinfo *fsinfo)
4018 {
4019         struct nfs4_fsinfo_arg args = {
4020                 .fh = fhandle,
4021                 .bitmask = server->attr_bitmask,
4022         };
4023         struct nfs4_fsinfo_res res = {
4024                 .fsinfo = fsinfo,
4025         };
4026         struct rpc_message msg = {
4027                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4028                 .rpc_argp = &args,
4029                 .rpc_resp = &res,
4030         };
4031
4032         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4033 }
4034
4035 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4036 {
4037         struct nfs4_exception exception = { };
4038         unsigned long now = jiffies;
4039         int err;
4040
4041         do {
4042                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4043                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4044                 if (err == 0) {
4045                         struct nfs_client *clp = server->nfs_client;
4046
4047                         spin_lock(&clp->cl_lock);
4048                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4049                         clp->cl_last_renewal = now;
4050                         spin_unlock(&clp->cl_lock);
4051                         break;
4052                 }
4053                 err = nfs4_handle_exception(server, err, &exception);
4054         } while (exception.retry);
4055         return err;
4056 }
4057
4058 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4059 {
4060         int error;
4061
4062         nfs_fattr_init(fsinfo->fattr);
4063         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4064         if (error == 0) {
4065                 /* block layout checks this! */
4066                 server->pnfs_blksize = fsinfo->blksize;
4067                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4068         }
4069
4070         return error;
4071 }
4072
4073 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4074                 struct nfs_pathconf *pathconf)
4075 {
4076         struct nfs4_pathconf_arg args = {
4077                 .fh = fhandle,
4078                 .bitmask = server->attr_bitmask,
4079         };
4080         struct nfs4_pathconf_res res = {
4081                 .pathconf = pathconf,
4082         };
4083         struct rpc_message msg = {
4084                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4085                 .rpc_argp = &args,
4086                 .rpc_resp = &res,
4087         };
4088
4089         /* None of the pathconf attributes are mandatory to implement */
4090         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4091                 memset(pathconf, 0, sizeof(*pathconf));
4092                 return 0;
4093         }
4094
4095         nfs_fattr_init(pathconf->fattr);
4096         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4097 }
4098
4099 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4100                 struct nfs_pathconf *pathconf)
4101 {
4102         struct nfs4_exception exception = { };
4103         int err;
4104
4105         do {
4106                 err = nfs4_handle_exception(server,
4107                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4108                                 &exception);
4109         } while (exception.retry);
4110         return err;
4111 }
4112
4113 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4114                 const struct nfs_open_context *ctx,
4115                 const struct nfs_lock_context *l_ctx,
4116                 fmode_t fmode)
4117 {
4118         const struct nfs_lockowner *lockowner = NULL;
4119
4120         if (l_ctx != NULL)
4121                 lockowner = &l_ctx->lockowner;
4122         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4123 }
4124 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4125
4126 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4127                 const struct nfs_open_context *ctx,
4128                 const struct nfs_lock_context *l_ctx,
4129                 fmode_t fmode)
4130 {
4131         nfs4_stateid current_stateid;
4132
4133         /* If the current stateid represents a lost lock, then exit */
4134         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4135                 return true;
4136         return nfs4_stateid_match(stateid, &current_stateid);
4137 }
4138
4139 static bool nfs4_error_stateid_expired(int err)
4140 {
4141         switch (err) {
4142         case -NFS4ERR_DELEG_REVOKED:
4143         case -NFS4ERR_ADMIN_REVOKED:
4144         case -NFS4ERR_BAD_STATEID:
4145         case -NFS4ERR_STALE_STATEID:
4146         case -NFS4ERR_OLD_STATEID:
4147         case -NFS4ERR_OPENMODE:
4148         case -NFS4ERR_EXPIRED:
4149                 return true;
4150         }
4151         return false;
4152 }
4153
4154 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4155 {
4156         nfs_invalidate_atime(hdr->inode);
4157 }
4158
4159 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4160 {
4161         struct nfs_server *server = NFS_SERVER(hdr->inode);
4162
4163         trace_nfs4_read(hdr, task->tk_status);
4164         if (nfs4_async_handle_error(task, server,
4165                                     hdr->args.context->state,
4166                                     NULL) == -EAGAIN) {
4167                 rpc_restart_call_prepare(task);
4168                 return -EAGAIN;
4169         }
4170
4171         __nfs4_read_done_cb(hdr);
4172         if (task->tk_status > 0)
4173                 renew_lease(server, hdr->timestamp);
4174         return 0;
4175 }
4176
4177 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4178                 struct nfs_pgio_args *args)
4179 {
4180
4181         if (!nfs4_error_stateid_expired(task->tk_status) ||
4182                 nfs4_stateid_is_current(&args->stateid,
4183                                 args->context,
4184                                 args->lock_context,
4185                                 FMODE_READ))
4186                 return false;
4187         rpc_restart_call_prepare(task);
4188         return true;
4189 }
4190
4191 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4192 {
4193
4194         dprintk("--> %s\n", __func__);
4195
4196         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4197                 return -EAGAIN;
4198         if (nfs4_read_stateid_changed(task, &hdr->args))
4199                 return -EAGAIN;
4200         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4201                                     nfs4_read_done_cb(task, hdr);
4202 }
4203
4204 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4205                                  struct rpc_message *msg)
4206 {
4207         hdr->timestamp   = jiffies;
4208         hdr->pgio_done_cb = nfs4_read_done_cb;
4209         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4210         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4211 }
4212
4213 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4214                                       struct nfs_pgio_header *hdr)
4215 {
4216         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4217                         &hdr->args.seq_args,
4218                         &hdr->res.seq_res,
4219                         task))
4220                 return 0;
4221         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4222                                 hdr->args.lock_context,
4223                                 hdr->rw_ops->rw_mode) == -EIO)
4224                 return -EIO;
4225         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4226                 return -EIO;
4227         return 0;
4228 }
4229
4230 static int nfs4_write_done_cb(struct rpc_task *task,
4231                               struct nfs_pgio_header *hdr)
4232 {
4233         struct inode *inode = hdr->inode;
4234
4235         trace_nfs4_write(hdr, task->tk_status);
4236         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4237                                     hdr->args.context->state,
4238                                     NULL) == -EAGAIN) {
4239                 rpc_restart_call_prepare(task);
4240                 return -EAGAIN;
4241         }
4242         if (task->tk_status >= 0) {
4243                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4244                 nfs_writeback_update_inode(hdr);
4245         }
4246         return 0;
4247 }
4248
4249 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4250                 struct nfs_pgio_args *args)
4251 {
4252
4253         if (!nfs4_error_stateid_expired(task->tk_status) ||
4254                 nfs4_stateid_is_current(&args->stateid,
4255                                 args->context,
4256                                 args->lock_context,
4257                                 FMODE_WRITE))
4258                 return false;
4259         rpc_restart_call_prepare(task);
4260         return true;
4261 }
4262
4263 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4264 {
4265         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4266                 return -EAGAIN;
4267         if (nfs4_write_stateid_changed(task, &hdr->args))
4268                 return -EAGAIN;
4269         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4270                 nfs4_write_done_cb(task, hdr);
4271 }
4272
4273 static
4274 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4275 {
4276         /* Don't request attributes for pNFS or O_DIRECT writes */
4277         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4278                 return false;
4279         /* Otherwise, request attributes if and only if we don't hold
4280          * a delegation
4281          */
4282         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4283 }
4284
4285 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4286                                   struct rpc_message *msg)
4287 {
4288         struct nfs_server *server = NFS_SERVER(hdr->inode);
4289
4290         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4291                 hdr->args.bitmask = NULL;
4292                 hdr->res.fattr = NULL;
4293         } else
4294                 hdr->args.bitmask = server->cache_consistency_bitmask;
4295
4296         if (!hdr->pgio_done_cb)
4297                 hdr->pgio_done_cb = nfs4_write_done_cb;
4298         hdr->res.server = server;
4299         hdr->timestamp   = jiffies;
4300
4301         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4302         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4303 }
4304
4305 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4306 {
4307         nfs4_setup_sequence(NFS_SERVER(data->inode),
4308                         &data->args.seq_args,
4309                         &data->res.seq_res,
4310                         task);
4311 }
4312
4313 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4314 {
4315         struct inode *inode = data->inode;
4316
4317         trace_nfs4_commit(data, task->tk_status);
4318         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4319                                     NULL, NULL) == -EAGAIN) {
4320                 rpc_restart_call_prepare(task);
4321                 return -EAGAIN;
4322         }
4323         return 0;
4324 }
4325
4326 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4327 {
4328         if (!nfs4_sequence_done(task, &data->res.seq_res))
4329                 return -EAGAIN;
4330         return data->commit_done_cb(task, data);
4331 }
4332
4333 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4334 {
4335         struct nfs_server *server = NFS_SERVER(data->inode);
4336
4337         if (data->commit_done_cb == NULL)
4338                 data->commit_done_cb = nfs4_commit_done_cb;
4339         data->res.server = server;
4340         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4341         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4342 }
4343
4344 struct nfs4_renewdata {
4345         struct nfs_client       *client;
4346         unsigned long           timestamp;
4347 };
4348
4349 /*
4350  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4351  * standalone procedure for queueing an asynchronous RENEW.
4352  */
4353 static void nfs4_renew_release(void *calldata)
4354 {
4355         struct nfs4_renewdata *data = calldata;
4356         struct nfs_client *clp = data->client;
4357
4358         if (atomic_read(&clp->cl_count) > 1)
4359                 nfs4_schedule_state_renewal(clp);
4360         nfs_put_client(clp);
4361         kfree(data);
4362 }
4363
4364 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4365 {
4366         struct nfs4_renewdata *data = calldata;
4367         struct nfs_client *clp = data->client;
4368         unsigned long timestamp = data->timestamp;
4369
4370         trace_nfs4_renew_async(clp, task->tk_status);
4371         switch (task->tk_status) {
4372         case 0:
4373                 break;
4374         case -NFS4ERR_LEASE_MOVED:
4375                 nfs4_schedule_lease_moved_recovery(clp);
4376                 break;
4377         default:
4378                 /* Unless we're shutting down, schedule state recovery! */
4379                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4380                         return;
4381                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4382                         nfs4_schedule_lease_recovery(clp);
4383                         return;
4384                 }
4385                 nfs4_schedule_path_down_recovery(clp);
4386         }
4387         do_renew_lease(clp, timestamp);
4388 }
4389
4390 static const struct rpc_call_ops nfs4_renew_ops = {
4391         .rpc_call_done = nfs4_renew_done,
4392         .rpc_release = nfs4_renew_release,
4393 };
4394
4395 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4396 {
4397         struct rpc_message msg = {
4398                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4399                 .rpc_argp       = clp,
4400                 .rpc_cred       = cred,
4401         };
4402         struct nfs4_renewdata *data;
4403
4404         if (renew_flags == 0)
4405                 return 0;
4406         if (!atomic_inc_not_zero(&clp->cl_count))
4407                 return -EIO;
4408         data = kmalloc(sizeof(*data), GFP_NOFS);
4409         if (data == NULL)
4410                 return -ENOMEM;
4411         data->client = clp;
4412         data->timestamp = jiffies;
4413         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4414                         &nfs4_renew_ops, data);
4415 }
4416
4417 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4418 {
4419         struct rpc_message msg = {
4420                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4421                 .rpc_argp       = clp,
4422                 .rpc_cred       = cred,
4423         };
4424         unsigned long now = jiffies;
4425         int status;
4426
4427         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4428         if (status < 0)
4429                 return status;
4430         do_renew_lease(clp, now);
4431         return 0;
4432 }
4433
4434 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4435 {
4436         return server->caps & NFS_CAP_ACLS;
4437 }
4438
4439 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4440  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4441  * the stack.
4442  */
4443 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4444
4445 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4446                 struct page **pages, unsigned int *pgbase)
4447 {
4448         struct page *newpage, **spages;
4449         int rc = 0;
4450         size_t len;
4451         spages = pages;
4452
4453         do {
4454                 len = min_t(size_t, PAGE_SIZE, buflen);
4455                 newpage = alloc_page(GFP_KERNEL);
4456
4457                 if (newpage == NULL)
4458                         goto unwind;
4459                 memcpy(page_address(newpage), buf, len);
4460                 buf += len;
4461                 buflen -= len;
4462                 *pages++ = newpage;
4463                 rc++;
4464         } while (buflen != 0);
4465
4466         return rc;
4467
4468 unwind:
4469         for(; rc > 0; rc--)
4470                 __free_page(spages[rc-1]);
4471         return -ENOMEM;
4472 }
4473
4474 struct nfs4_cached_acl {
4475         int cached;
4476         size_t len;
4477         char data[0];
4478 };
4479
4480 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4481 {
4482         struct nfs_inode *nfsi = NFS_I(inode);
4483
4484         spin_lock(&inode->i_lock);
4485         kfree(nfsi->nfs4_acl);
4486         nfsi->nfs4_acl = acl;
4487         spin_unlock(&inode->i_lock);
4488 }
4489
4490 static void nfs4_zap_acl_attr(struct inode *inode)
4491 {
4492         nfs4_set_cached_acl(inode, NULL);
4493 }
4494
4495 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4496 {
4497         struct nfs_inode *nfsi = NFS_I(inode);
4498         struct nfs4_cached_acl *acl;
4499         int ret = -ENOENT;
4500
4501         spin_lock(&inode->i_lock);
4502         acl = nfsi->nfs4_acl;
4503         if (acl == NULL)
4504                 goto out;
4505         if (buf == NULL) /* user is just asking for length */
4506                 goto out_len;
4507         if (acl->cached == 0)
4508                 goto out;
4509         ret = -ERANGE; /* see getxattr(2) man page */
4510         if (acl->len > buflen)
4511                 goto out;
4512         memcpy(buf, acl->data, acl->len);
4513 out_len:
4514         ret = acl->len;
4515 out:
4516         spin_unlock(&inode->i_lock);
4517         return ret;
4518 }
4519
4520 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4521 {
4522         struct nfs4_cached_acl *acl;
4523         size_t buflen = sizeof(*acl) + acl_len;
4524
4525         if (buflen <= PAGE_SIZE) {
4526                 acl = kmalloc(buflen, GFP_KERNEL);
4527                 if (acl == NULL)
4528                         goto out;
4529                 acl->cached = 1;
4530                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4531         } else {
4532                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4533                 if (acl == NULL)
4534                         goto out;
4535                 acl->cached = 0;
4536         }
4537         acl->len = acl_len;
4538 out:
4539         nfs4_set_cached_acl(inode, acl);
4540 }
4541
4542 /*
4543  * The getxattr API returns the required buffer length when called with a
4544  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4545  * the required buf.  On a NULL buf, we send a page of data to the server
4546  * guessing that the ACL request can be serviced by a page. If so, we cache
4547  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4548  * the cache. If not so, we throw away the page, and cache the required
4549  * length. The next getxattr call will then produce another round trip to
4550  * the server, this time with the input buf of the required size.
4551  */
4552 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4553 {
4554         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4555         struct nfs_getaclargs args = {
4556                 .fh = NFS_FH(inode),
4557                 .acl_pages = pages,
4558                 .acl_len = buflen,
4559         };
4560         struct nfs_getaclres res = {
4561                 .acl_len = buflen,
4562         };
4563         struct rpc_message msg = {
4564                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4565                 .rpc_argp = &args,
4566                 .rpc_resp = &res,
4567         };
4568         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4569         int ret = -ENOMEM, i;
4570
4571         /* As long as we're doing a round trip to the server anyway,
4572          * let's be prepared for a page of acl data. */
4573         if (npages == 0)
4574                 npages = 1;
4575         if (npages > ARRAY_SIZE(pages))
4576                 return -ERANGE;
4577
4578         for (i = 0; i < npages; i++) {
4579                 pages[i] = alloc_page(GFP_KERNEL);
4580                 if (!pages[i])
4581                         goto out_free;
4582         }
4583
4584         /* for decoding across pages */
4585         res.acl_scratch = alloc_page(GFP_KERNEL);
4586         if (!res.acl_scratch)
4587                 goto out_free;
4588
4589         args.acl_len = npages * PAGE_SIZE;
4590         args.acl_pgbase = 0;
4591
4592         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4593                 __func__, buf, buflen, npages, args.acl_len);
4594         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4595                              &msg, &args.seq_args, &res.seq_res, 0);
4596         if (ret)
4597                 goto out_free;
4598
4599         /* Handle the case where the passed-in buffer is too short */
4600         if (res.acl_flags & NFS4_ACL_TRUNC) {
4601                 /* Did the user only issue a request for the acl length? */
4602                 if (buf == NULL)
4603                         goto out_ok;
4604                 ret = -ERANGE;
4605                 goto out_free;
4606         }
4607         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4608         if (buf) {
4609                 if (res.acl_len > buflen) {
4610                         ret = -ERANGE;
4611                         goto out_free;
4612                 }
4613                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4614         }
4615 out_ok:
4616         ret = res.acl_len;
4617 out_free:
4618         for (i = 0; i < npages; i++)
4619                 if (pages[i])
4620                         __free_page(pages[i]);
4621         if (res.acl_scratch)
4622                 __free_page(res.acl_scratch);
4623         return ret;
4624 }
4625
4626 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4627 {
4628         struct nfs4_exception exception = { };
4629         ssize_t ret;
4630         do {
4631                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4632                 trace_nfs4_get_acl(inode, ret);
4633                 if (ret >= 0)
4634                         break;
4635                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4636         } while (exception.retry);
4637         return ret;
4638 }
4639
4640 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4641 {
4642         struct nfs_server *server = NFS_SERVER(inode);
4643         int ret;
4644
4645         if (!nfs4_server_supports_acls(server))
4646                 return -EOPNOTSUPP;
4647         ret = nfs_revalidate_inode(server, inode);
4648         if (ret < 0)
4649                 return ret;
4650         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4651                 nfs_zap_acl_cache(inode);
4652         ret = nfs4_read_cached_acl(inode, buf, buflen);
4653         if (ret != -ENOENT)
4654                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4655                  * but no cached acl data, just the acl length */
4656                 return ret;
4657         return nfs4_get_acl_uncached(inode, buf, buflen);
4658 }
4659
4660 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4661 {
4662         struct nfs_server *server = NFS_SERVER(inode);
4663         struct page *pages[NFS4ACL_MAXPAGES];
4664         struct nfs_setaclargs arg = {
4665                 .fh             = NFS_FH(inode),
4666                 .acl_pages      = pages,
4667                 .acl_len        = buflen,
4668         };
4669         struct nfs_setaclres res;
4670         struct rpc_message msg = {
4671                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4672                 .rpc_argp       = &arg,
4673                 .rpc_resp       = &res,
4674         };
4675         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4676         int ret, i;
4677
4678         if (!nfs4_server_supports_acls(server))
4679                 return -EOPNOTSUPP;
4680         if (npages > ARRAY_SIZE(pages))
4681                 return -ERANGE;
4682         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4683         if (i < 0)
4684                 return i;
4685         nfs4_inode_return_delegation(inode);
4686         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4687
4688         /*
4689          * Free each page after tx, so the only ref left is
4690          * held by the network stack
4691          */
4692         for (; i > 0; i--)
4693                 put_page(pages[i-1]);
4694
4695         /*
4696          * Acl update can result in inode attribute update.
4697          * so mark the attribute cache invalid.
4698          */
4699         spin_lock(&inode->i_lock);
4700         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4701         spin_unlock(&inode->i_lock);
4702         nfs_access_zap_cache(inode);
4703         nfs_zap_acl_cache(inode);
4704         return ret;
4705 }
4706
4707 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4708 {
4709         struct nfs4_exception exception = { };
4710         int err;
4711         do {
4712                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4713                 trace_nfs4_set_acl(inode, err);
4714                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4715                                 &exception);
4716         } while (exception.retry);
4717         return err;
4718 }
4719
4720 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4721 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4722                                         size_t buflen)
4723 {
4724         struct nfs_server *server = NFS_SERVER(inode);
4725         struct nfs_fattr fattr;
4726         struct nfs4_label label = {0, 0, buflen, buf};
4727
4728         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4729         struct nfs4_getattr_arg arg = {
4730                 .fh             = NFS_FH(inode),
4731                 .bitmask        = bitmask,
4732         };
4733         struct nfs4_getattr_res res = {
4734                 .fattr          = &fattr,
4735                 .label          = &label,
4736                 .server         = server,
4737         };
4738         struct rpc_message msg = {
4739                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4740                 .rpc_argp       = &arg,
4741                 .rpc_resp       = &res,
4742         };
4743         int ret;
4744
4745         nfs_fattr_init(&fattr);
4746
4747         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4748         if (ret)
4749                 return ret;
4750         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4751                 return -ENOENT;
4752         if (buflen < label.len)
4753                 return -ERANGE;
4754         return 0;
4755 }
4756
4757 static int nfs4_get_security_label(struct inode *inode, void *buf,
4758                                         size_t buflen)
4759 {
4760         struct nfs4_exception exception = { };
4761         int err;
4762
4763         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4764                 return -EOPNOTSUPP;
4765
4766         do {
4767                 err = _nfs4_get_security_label(inode, buf, buflen);
4768                 trace_nfs4_get_security_label(inode, err);
4769                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4770                                 &exception);
4771         } while (exception.retry);
4772         return err;
4773 }
4774
4775 static int _nfs4_do_set_security_label(struct inode *inode,
4776                 struct nfs4_label *ilabel,
4777                 struct nfs_fattr *fattr,
4778                 struct nfs4_label *olabel)
4779 {
4780
4781         struct iattr sattr = {0};
4782         struct nfs_server *server = NFS_SERVER(inode);
4783         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4784         struct nfs_setattrargs arg = {
4785                 .fh             = NFS_FH(inode),
4786                 .iap            = &sattr,
4787                 .server         = server,
4788                 .bitmask        = bitmask,
4789                 .label          = ilabel,
4790         };
4791         struct nfs_setattrres res = {
4792                 .fattr          = fattr,
4793                 .label          = olabel,
4794                 .server         = server,
4795         };
4796         struct rpc_message msg = {
4797                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4798                 .rpc_argp       = &arg,
4799                 .rpc_resp       = &res,
4800         };
4801         int status;
4802
4803         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4804
4805         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4806         if (status)
4807                 dprintk("%s failed: %d\n", __func__, status);
4808
4809         return status;
4810 }
4811
4812 static int nfs4_do_set_security_label(struct inode *inode,
4813                 struct nfs4_label *ilabel,
4814                 struct nfs_fattr *fattr,
4815                 struct nfs4_label *olabel)
4816 {
4817         struct nfs4_exception exception = { };
4818         int err;
4819
4820         do {
4821                 err = _nfs4_do_set_security_label(inode, ilabel,
4822                                 fattr, olabel);
4823                 trace_nfs4_set_security_label(inode, err);
4824                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4825                                 &exception);
4826         } while (exception.retry);
4827         return err;
4828 }
4829
4830 static int
4831 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4832 {
4833         struct nfs4_label ilabel, *olabel = NULL;
4834         struct nfs_fattr fattr;
4835         struct rpc_cred *cred;
4836         struct inode *inode = d_inode(dentry);
4837         int status;
4838
4839         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4840                 return -EOPNOTSUPP;
4841
4842         nfs_fattr_init(&fattr);
4843
4844         ilabel.pi = 0;
4845         ilabel.lfs = 0;
4846         ilabel.label = (char *)buf;
4847         ilabel.len = buflen;
4848
4849         cred = rpc_lookup_cred();
4850         if (IS_ERR(cred))
4851                 return PTR_ERR(cred);
4852
4853         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4854         if (IS_ERR(olabel)) {
4855                 status = -PTR_ERR(olabel);
4856                 goto out;
4857         }
4858
4859         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4860         if (status == 0)
4861                 nfs_setsecurity(inode, &fattr, olabel);
4862
4863         nfs4_label_free(olabel);
4864 out:
4865         put_rpccred(cred);
4866         return status;
4867 }
4868 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4869
4870
4871 static int
4872 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4873                         struct nfs4_state *state, long *timeout)
4874 {
4875         struct nfs_client *clp = server->nfs_client;
4876
4877         if (task->tk_status >= 0)
4878                 return 0;
4879         switch(task->tk_status) {
4880                 case -NFS4ERR_DELEG_REVOKED:
4881                 case -NFS4ERR_ADMIN_REVOKED:
4882                 case -NFS4ERR_BAD_STATEID:
4883                 case -NFS4ERR_OPENMODE:
4884                         if (state == NULL)
4885                                 break;
4886                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4887                                 goto recovery_failed;
4888                         goto wait_on_recovery;
4889                 case -NFS4ERR_EXPIRED:
4890                         if (state != NULL) {
4891                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4892                                         goto recovery_failed;
4893                         }
4894                 case -NFS4ERR_STALE_STATEID:
4895                 case -NFS4ERR_STALE_CLIENTID:
4896                         nfs4_schedule_lease_recovery(clp);
4897                         goto wait_on_recovery;
4898                 case -NFS4ERR_MOVED:
4899                         if (nfs4_schedule_migration_recovery(server) < 0)
4900                                 goto recovery_failed;
4901                         goto wait_on_recovery;
4902                 case -NFS4ERR_LEASE_MOVED:
4903                         nfs4_schedule_lease_moved_recovery(clp);
4904                         goto wait_on_recovery;
4905 #if defined(CONFIG_NFS_V4_1)
4906                 case -NFS4ERR_BADSESSION:
4907                 case -NFS4ERR_BADSLOT:
4908                 case -NFS4ERR_BAD_HIGH_SLOT:
4909                 case -NFS4ERR_DEADSESSION:
4910                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4911                 case -NFS4ERR_SEQ_FALSE_RETRY:
4912                 case -NFS4ERR_SEQ_MISORDERED:
4913                         dprintk("%s ERROR %d, Reset session\n", __func__,
4914                                 task->tk_status);
4915                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4916                         goto wait_on_recovery;
4917 #endif /* CONFIG_NFS_V4_1 */
4918                 case -NFS4ERR_DELAY:
4919                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4920                         rpc_delay(task, nfs4_update_delay(timeout));
4921                         goto restart_call;
4922                 case -NFS4ERR_GRACE:
4923                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4924                 case -NFS4ERR_RETRY_UNCACHED_REP:
4925                 case -NFS4ERR_OLD_STATEID:
4926                         goto restart_call;
4927         }
4928         task->tk_status = nfs4_map_errors(task->tk_status);
4929         return 0;
4930 recovery_failed:
4931         task->tk_status = -EIO;
4932         return 0;
4933 wait_on_recovery:
4934         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4935         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4936                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4937         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4938                 goto recovery_failed;
4939 restart_call:
4940         task->tk_status = 0;
4941         return -EAGAIN;
4942 }
4943
4944 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4945                                     nfs4_verifier *bootverf)
4946 {
4947         __be32 verf[2];
4948
4949         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4950                 /* An impossible timestamp guarantees this value
4951                  * will never match a generated boot time. */
4952                 verf[0] = 0;
4953                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4954         } else {
4955                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4956                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4957                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4958         }
4959         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4960 }
4961
4962 static int
4963 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
4964 {
4965         int result;
4966         size_t len;
4967         char *str;
4968         bool retried = false;
4969
4970         if (clp->cl_owner_id != NULL)
4971                 return 0;
4972 retry:
4973         rcu_read_lock();
4974         len = 10 + strlen(clp->cl_ipaddr) + 1 +
4975                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
4976                 1 +
4977                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
4978                 1;
4979         rcu_read_unlock();
4980
4981         if (len > NFS4_OPAQUE_LIMIT + 1)
4982                 return -EINVAL;
4983
4984         /*
4985          * Since this string is allocated at mount time, and held until the
4986          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
4987          * about a memory-reclaim deadlock.
4988          */
4989         str = kmalloc(len, GFP_KERNEL);
4990         if (!str)
4991                 return -ENOMEM;
4992
4993         rcu_read_lock();
4994         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
4995                         clp->cl_ipaddr,
4996                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
4997                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
4998         rcu_read_unlock();
4999
5000         /* Did something change? */
5001         if (result >= len) {
5002                 kfree(str);
5003                 if (retried)
5004                         return -EINVAL;
5005                 retried = true;
5006                 goto retry;
5007         }
5008         clp->cl_owner_id = str;
5009         return 0;
5010 }
5011
5012 static int
5013 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5014 {
5015         int result;
5016         size_t len;
5017         char *str;
5018
5019         len = 10 + 10 + 1 + 10 + 1 +
5020                 strlen(nfs4_client_id_uniquifier) + 1 +
5021                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5022
5023         if (len > NFS4_OPAQUE_LIMIT + 1)
5024                 return -EINVAL;
5025
5026         /*
5027          * Since this string is allocated at mount time, and held until the
5028          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5029          * about a memory-reclaim deadlock.
5030          */
5031         str = kmalloc(len, GFP_KERNEL);
5032         if (!str)
5033                 return -ENOMEM;
5034
5035         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5036                         clp->rpc_ops->version, clp->cl_minorversion,
5037                         nfs4_client_id_uniquifier,
5038                         clp->cl_rpcclient->cl_nodename);
5039         if (result >= len) {
5040                 kfree(str);
5041                 return -EINVAL;
5042         }
5043         clp->cl_owner_id = str;
5044         return 0;
5045 }
5046
5047 static int
5048 nfs4_init_uniform_client_string(struct nfs_client *clp)
5049 {
5050         int result;
5051         size_t len;
5052         char *str;
5053
5054         if (clp->cl_owner_id != NULL)
5055                 return 0;
5056
5057         if (nfs4_client_id_uniquifier[0] != '\0')
5058                 return nfs4_init_uniquifier_client_string(clp);
5059
5060         len = 10 + 10 + 1 + 10 + 1 +
5061                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5062
5063         if (len > NFS4_OPAQUE_LIMIT + 1)
5064                 return -EINVAL;
5065
5066         /*
5067          * Since this string is allocated at mount time, and held until the
5068          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5069          * about a memory-reclaim deadlock.
5070          */
5071         str = kmalloc(len, GFP_KERNEL);
5072         if (!str)
5073                 return -ENOMEM;
5074
5075         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5076                         clp->rpc_ops->version, clp->cl_minorversion,
5077                         clp->cl_rpcclient->cl_nodename);
5078         if (result >= len) {
5079                 kfree(str);
5080                 return -EINVAL;
5081         }
5082         clp->cl_owner_id = str;
5083         return 0;
5084 }
5085
5086 /*
5087  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5088  * services.  Advertise one based on the address family of the
5089  * clientaddr.
5090  */
5091 static unsigned int
5092 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5093 {
5094         if (strchr(clp->cl_ipaddr, ':') != NULL)
5095                 return scnprintf(buf, len, "tcp6");
5096         else
5097                 return scnprintf(buf, len, "tcp");
5098 }
5099
5100 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5101 {
5102         struct nfs4_setclientid *sc = calldata;
5103
5104         if (task->tk_status == 0)
5105                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5106 }
5107
5108 static const struct rpc_call_ops nfs4_setclientid_ops = {
5109         .rpc_call_done = nfs4_setclientid_done,
5110 };
5111
5112 /**
5113  * nfs4_proc_setclientid - Negotiate client ID
5114  * @clp: state data structure
5115  * @program: RPC program for NFSv4 callback service
5116  * @port: IP port number for NFS4 callback service
5117  * @cred: RPC credential to use for this call
5118  * @res: where to place the result
5119  *
5120  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5121  */
5122 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5123                 unsigned short port, struct rpc_cred *cred,
5124                 struct nfs4_setclientid_res *res)
5125 {
5126         nfs4_verifier sc_verifier;
5127         struct nfs4_setclientid setclientid = {
5128                 .sc_verifier = &sc_verifier,
5129                 .sc_prog = program,
5130                 .sc_clnt = clp,
5131         };
5132         struct rpc_message msg = {
5133                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5134                 .rpc_argp = &setclientid,
5135                 .rpc_resp = res,
5136                 .rpc_cred = cred,
5137         };
5138         struct rpc_task *task;
5139         struct rpc_task_setup task_setup_data = {
5140                 .rpc_client = clp->cl_rpcclient,
5141                 .rpc_message = &msg,
5142                 .callback_ops = &nfs4_setclientid_ops,
5143                 .callback_data = &setclientid,
5144                 .flags = RPC_TASK_TIMEOUT,
5145         };
5146         int status;
5147
5148         /* nfs_client_id4 */
5149         nfs4_init_boot_verifier(clp, &sc_verifier);
5150
5151         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5152                 status = nfs4_init_uniform_client_string(clp);
5153         else
5154                 status = nfs4_init_nonuniform_client_string(clp);
5155
5156         if (status)
5157                 goto out;
5158
5159         /* cb_client4 */
5160         setclientid.sc_netid_len =
5161                                 nfs4_init_callback_netid(clp,
5162                                                 setclientid.sc_netid,
5163                                                 sizeof(setclientid.sc_netid));
5164         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5165                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5166                                 clp->cl_ipaddr, port >> 8, port & 255);
5167
5168         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5169                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5170                 clp->cl_owner_id);
5171         task = rpc_run_task(&task_setup_data);
5172         if (IS_ERR(task)) {
5173                 status = PTR_ERR(task);
5174                 goto out;
5175         }
5176         status = task->tk_status;
5177         if (setclientid.sc_cred) {
5178                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5179                 put_rpccred(setclientid.sc_cred);
5180         }
5181         rpc_put_task(task);
5182 out:
5183         trace_nfs4_setclientid(clp, status);
5184         dprintk("NFS reply setclientid: %d\n", status);
5185         return status;
5186 }
5187
5188 /**
5189  * nfs4_proc_setclientid_confirm - Confirm client ID
5190  * @clp: state data structure
5191  * @res: result of a previous SETCLIENTID
5192  * @cred: RPC credential to use for this call
5193  *
5194  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5195  */
5196 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5197                 struct nfs4_setclientid_res *arg,
5198                 struct rpc_cred *cred)
5199 {
5200         struct rpc_message msg = {
5201                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5202                 .rpc_argp = arg,
5203                 .rpc_cred = cred,
5204         };
5205         int status;
5206
5207         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5208                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5209                 clp->cl_clientid);
5210         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5211         trace_nfs4_setclientid_confirm(clp, status);
5212         dprintk("NFS reply setclientid_confirm: %d\n", status);
5213         return status;
5214 }
5215
5216 struct nfs4_delegreturndata {
5217         struct nfs4_delegreturnargs args;
5218         struct nfs4_delegreturnres res;
5219         struct nfs_fh fh;
5220         nfs4_stateid stateid;
5221         unsigned long timestamp;
5222         struct nfs_fattr fattr;
5223         int rpc_status;
5224         struct inode *inode;
5225         bool roc;
5226         u32 roc_barrier;
5227 };
5228
5229 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5230 {
5231         struct nfs4_delegreturndata *data = calldata;
5232
5233         if (!nfs4_sequence_done(task, &data->res.seq_res))
5234                 return;
5235
5236         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5237         switch (task->tk_status) {
5238         case 0:
5239                 renew_lease(data->res.server, data->timestamp);
5240         case -NFS4ERR_ADMIN_REVOKED:
5241         case -NFS4ERR_DELEG_REVOKED:
5242         case -NFS4ERR_BAD_STATEID:
5243         case -NFS4ERR_OLD_STATEID:
5244         case -NFS4ERR_STALE_STATEID:
5245         case -NFS4ERR_EXPIRED:
5246                 task->tk_status = 0;
5247                 if (data->roc)
5248                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5249                 break;
5250         default:
5251                 if (nfs4_async_handle_error(task, data->res.server,
5252                                             NULL, NULL) == -EAGAIN) {
5253                         rpc_restart_call_prepare(task);
5254                         return;
5255                 }
5256         }
5257         data->rpc_status = task->tk_status;
5258 }
5259
5260 static void nfs4_delegreturn_release(void *calldata)
5261 {
5262         struct nfs4_delegreturndata *data = calldata;
5263         struct inode *inode = data->inode;
5264
5265         if (inode) {
5266                 if (data->roc)
5267                         pnfs_roc_release(inode);
5268                 nfs_iput_and_deactive(inode);
5269         }
5270         kfree(calldata);
5271 }
5272
5273 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5274 {
5275         struct nfs4_delegreturndata *d_data;
5276
5277         d_data = (struct nfs4_delegreturndata *)data;
5278
5279         if (d_data->roc &&
5280             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5281                 return;
5282
5283         nfs4_setup_sequence(d_data->res.server,
5284                         &d_data->args.seq_args,
5285                         &d_data->res.seq_res,
5286                         task);
5287 }
5288
5289 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5290         .rpc_call_prepare = nfs4_delegreturn_prepare,
5291         .rpc_call_done = nfs4_delegreturn_done,
5292         .rpc_release = nfs4_delegreturn_release,
5293 };
5294
5295 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5296 {
5297         struct nfs4_delegreturndata *data;
5298         struct nfs_server *server = NFS_SERVER(inode);
5299         struct rpc_task *task;
5300         struct rpc_message msg = {
5301                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5302                 .rpc_cred = cred,
5303         };
5304         struct rpc_task_setup task_setup_data = {
5305                 .rpc_client = server->client,
5306                 .rpc_message = &msg,
5307                 .callback_ops = &nfs4_delegreturn_ops,
5308                 .flags = RPC_TASK_ASYNC,
5309         };
5310         int status = 0;
5311
5312         data = kzalloc(sizeof(*data), GFP_NOFS);
5313         if (data == NULL)
5314                 return -ENOMEM;
5315         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5316         data->args.fhandle = &data->fh;
5317         data->args.stateid = &data->stateid;
5318         data->args.bitmask = server->cache_consistency_bitmask;
5319         nfs_copy_fh(&data->fh, NFS_FH(inode));
5320         nfs4_stateid_copy(&data->stateid, stateid);
5321         data->res.fattr = &data->fattr;
5322         data->res.server = server;
5323         nfs_fattr_init(data->res.fattr);
5324         data->timestamp = jiffies;
5325         data->rpc_status = 0;
5326         data->inode = nfs_igrab_and_active(inode);
5327         if (data->inode)
5328                 data->roc = nfs4_roc(inode);
5329
5330         task_setup_data.callback_data = data;
5331         msg.rpc_argp = &data->args;
5332         msg.rpc_resp = &data->res;
5333         task = rpc_run_task(&task_setup_data);
5334         if (IS_ERR(task))
5335                 return PTR_ERR(task);
5336         if (!issync)
5337                 goto out;
5338         status = nfs4_wait_for_completion_rpc_task(task);
5339         if (status != 0)
5340                 goto out;
5341         status = data->rpc_status;
5342         if (status == 0)
5343                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5344         else
5345                 nfs_refresh_inode(inode, &data->fattr);
5346 out:
5347         rpc_put_task(task);
5348         return status;
5349 }
5350
5351 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5352 {
5353         struct nfs_server *server = NFS_SERVER(inode);
5354         struct nfs4_exception exception = { };
5355         int err;
5356         do {
5357                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5358                 trace_nfs4_delegreturn(inode, err);
5359                 switch (err) {
5360                         case -NFS4ERR_STALE_STATEID:
5361                         case -NFS4ERR_EXPIRED:
5362                         case 0:
5363                                 return 0;
5364                 }
5365                 err = nfs4_handle_exception(server, err, &exception);
5366         } while (exception.retry);
5367         return err;
5368 }
5369
5370 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5371 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5372
5373 /*
5374  * sleep, with exponential backoff, and retry the LOCK operation.
5375  */
5376 static unsigned long
5377 nfs4_set_lock_task_retry(unsigned long timeout)
5378 {
5379         freezable_schedule_timeout_killable_unsafe(timeout);
5380         timeout <<= 1;
5381         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5382                 return NFS4_LOCK_MAXTIMEOUT;
5383         return timeout;
5384 }
5385
5386 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5387 {
5388         struct inode *inode = state->inode;
5389         struct nfs_server *server = NFS_SERVER(inode);
5390         struct nfs_client *clp = server->nfs_client;
5391         struct nfs_lockt_args arg = {
5392                 .fh = NFS_FH(inode),
5393                 .fl = request,
5394         };
5395         struct nfs_lockt_res res = {
5396                 .denied = request,
5397         };
5398         struct rpc_message msg = {
5399                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5400                 .rpc_argp       = &arg,
5401                 .rpc_resp       = &res,
5402                 .rpc_cred       = state->owner->so_cred,
5403         };
5404         struct nfs4_lock_state *lsp;
5405         int status;
5406
5407         arg.lock_owner.clientid = clp->cl_clientid;
5408         status = nfs4_set_lock_state(state, request);
5409         if (status != 0)
5410                 goto out;
5411         lsp = request->fl_u.nfs4_fl.owner;
5412         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5413         arg.lock_owner.s_dev = server->s_dev;
5414         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5415         switch (status) {
5416                 case 0:
5417                         request->fl_type = F_UNLCK;
5418                         break;
5419                 case -NFS4ERR_DENIED:
5420                         status = 0;
5421         }
5422         request->fl_ops->fl_release_private(request);
5423         request->fl_ops = NULL;
5424 out:
5425         return status;
5426 }
5427
5428 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5429 {
5430         struct nfs4_exception exception = { };
5431         int err;
5432
5433         do {
5434                 err = _nfs4_proc_getlk(state, cmd, request);
5435                 trace_nfs4_get_lock(request, state, cmd, err);
5436                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5437                                 &exception);
5438         } while (exception.retry);
5439         return err;
5440 }
5441
5442 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5443 {
5444         int res = 0;
5445         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5446                 case FL_POSIX:
5447                         res = posix_lock_file_wait(file, fl);
5448                         break;
5449                 case FL_FLOCK:
5450                         res = flock_lock_file_wait(file, fl);
5451                         break;
5452                 default:
5453                         BUG();
5454         }
5455         return res;
5456 }
5457
5458 struct nfs4_unlockdata {
5459         struct nfs_locku_args arg;
5460         struct nfs_locku_res res;
5461         struct nfs4_lock_state *lsp;
5462         struct nfs_open_context *ctx;
5463         struct file_lock fl;
5464         const struct nfs_server *server;
5465         unsigned long timestamp;
5466 };
5467
5468 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5469                 struct nfs_open_context *ctx,
5470                 struct nfs4_lock_state *lsp,
5471                 struct nfs_seqid *seqid)
5472 {
5473         struct nfs4_unlockdata *p;
5474         struct inode *inode = lsp->ls_state->inode;
5475
5476         p = kzalloc(sizeof(*p), GFP_NOFS);
5477         if (p == NULL)
5478                 return NULL;
5479         p->arg.fh = NFS_FH(inode);
5480         p->arg.fl = &p->fl;
5481         p->arg.seqid = seqid;
5482         p->res.seqid = seqid;
5483         p->lsp = lsp;
5484         atomic_inc(&lsp->ls_count);
5485         /* Ensure we don't close file until we're done freeing locks! */
5486         p->ctx = get_nfs_open_context(ctx);
5487         get_file(fl->fl_file);
5488         memcpy(&p->fl, fl, sizeof(p->fl));
5489         p->server = NFS_SERVER(inode);
5490         return p;
5491 }
5492
5493 static void nfs4_locku_release_calldata(void *data)
5494 {
5495         struct nfs4_unlockdata *calldata = data;
5496         nfs_free_seqid(calldata->arg.seqid);
5497         nfs4_put_lock_state(calldata->lsp);
5498         put_nfs_open_context(calldata->ctx);
5499         fput(calldata->fl.fl_file);
5500         kfree(calldata);
5501 }
5502
5503 static void nfs4_locku_done(struct rpc_task *task, void *data)
5504 {
5505         struct nfs4_unlockdata *calldata = data;
5506
5507         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5508                 return;
5509         switch (task->tk_status) {
5510                 case 0:
5511                         renew_lease(calldata->server, calldata->timestamp);
5512                         do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5513                         if (nfs4_update_lock_stateid(calldata->lsp,
5514                                         &calldata->res.stateid))
5515                                 break;
5516                 case -NFS4ERR_BAD_STATEID:
5517                 case -NFS4ERR_OLD_STATEID:
5518                 case -NFS4ERR_STALE_STATEID:
5519                 case -NFS4ERR_EXPIRED:
5520                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5521                                                 &calldata->lsp->ls_stateid))
5522                                 rpc_restart_call_prepare(task);
5523                         break;
5524                 default:
5525                         if (nfs4_async_handle_error(task, calldata->server,
5526                                                     NULL, NULL) == -EAGAIN)
5527                                 rpc_restart_call_prepare(task);
5528         }
5529         nfs_release_seqid(calldata->arg.seqid);
5530 }
5531
5532 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5533 {
5534         struct nfs4_unlockdata *calldata = data;
5535
5536         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5537                 goto out_wait;
5538         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5539         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5540                 /* Note: exit _without_ running nfs4_locku_done */
5541                 goto out_no_action;
5542         }
5543         calldata->timestamp = jiffies;
5544         if (nfs4_setup_sequence(calldata->server,
5545                                 &calldata->arg.seq_args,
5546                                 &calldata->res.seq_res,
5547                                 task) != 0)
5548                 nfs_release_seqid(calldata->arg.seqid);
5549         return;
5550 out_no_action:
5551         task->tk_action = NULL;
5552 out_wait:
5553         nfs4_sequence_done(task, &calldata->res.seq_res);
5554 }
5555
5556 static const struct rpc_call_ops nfs4_locku_ops = {
5557         .rpc_call_prepare = nfs4_locku_prepare,
5558         .rpc_call_done = nfs4_locku_done,
5559         .rpc_release = nfs4_locku_release_calldata,
5560 };
5561
5562 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5563                 struct nfs_open_context *ctx,
5564                 struct nfs4_lock_state *lsp,
5565                 struct nfs_seqid *seqid)
5566 {
5567         struct nfs4_unlockdata *data;
5568         struct rpc_message msg = {
5569                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5570                 .rpc_cred = ctx->cred,
5571         };
5572         struct rpc_task_setup task_setup_data = {
5573                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5574                 .rpc_message = &msg,
5575                 .callback_ops = &nfs4_locku_ops,
5576                 .workqueue = nfsiod_workqueue,
5577                 .flags = RPC_TASK_ASYNC,
5578         };
5579
5580         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5581                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5582
5583         /* Ensure this is an unlock - when canceling a lock, the
5584          * canceled lock is passed in, and it won't be an unlock.
5585          */
5586         fl->fl_type = F_UNLCK;
5587
5588         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5589         if (data == NULL) {
5590                 nfs_free_seqid(seqid);
5591                 return ERR_PTR(-ENOMEM);
5592         }
5593
5594         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5595         msg.rpc_argp = &data->arg;
5596         msg.rpc_resp = &data->res;
5597         task_setup_data.callback_data = data;
5598         return rpc_run_task(&task_setup_data);
5599 }
5600
5601 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5602 {
5603         struct inode *inode = state->inode;
5604         struct nfs4_state_owner *sp = state->owner;
5605         struct nfs_inode *nfsi = NFS_I(inode);
5606         struct nfs_seqid *seqid;
5607         struct nfs4_lock_state *lsp;
5608         struct rpc_task *task;
5609         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5610         int status = 0;
5611         unsigned char fl_flags = request->fl_flags;
5612
5613         status = nfs4_set_lock_state(state, request);
5614         /* Unlock _before_ we do the RPC call */
5615         request->fl_flags |= FL_EXISTS;
5616         /* Exclude nfs_delegation_claim_locks() */
5617         mutex_lock(&sp->so_delegreturn_mutex);
5618         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5619         down_read(&nfsi->rwsem);
5620         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5621                 up_read(&nfsi->rwsem);
5622                 mutex_unlock(&sp->so_delegreturn_mutex);
5623                 goto out;
5624         }
5625         up_read(&nfsi->rwsem);
5626         mutex_unlock(&sp->so_delegreturn_mutex);
5627         if (status != 0)
5628                 goto out;
5629         /* Is this a delegated lock? */
5630         lsp = request->fl_u.nfs4_fl.owner;
5631         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5632                 goto out;
5633         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5634         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5635         status = -ENOMEM;
5636         if (IS_ERR(seqid))
5637                 goto out;
5638         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5639         status = PTR_ERR(task);
5640         if (IS_ERR(task))
5641                 goto out;
5642         status = nfs4_wait_for_completion_rpc_task(task);
5643         rpc_put_task(task);
5644 out:
5645         request->fl_flags = fl_flags;
5646         trace_nfs4_unlock(request, state, F_SETLK, status);
5647         return status;
5648 }
5649
5650 struct nfs4_lockdata {
5651         struct nfs_lock_args arg;
5652         struct nfs_lock_res res;
5653         struct nfs4_lock_state *lsp;
5654         struct nfs_open_context *ctx;
5655         struct file_lock fl;
5656         unsigned long timestamp;
5657         int rpc_status;
5658         int cancelled;
5659         struct nfs_server *server;
5660 };
5661
5662 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5663                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5664                 gfp_t gfp_mask)
5665 {
5666         struct nfs4_lockdata *p;
5667         struct inode *inode = lsp->ls_state->inode;
5668         struct nfs_server *server = NFS_SERVER(inode);
5669         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5670
5671         p = kzalloc(sizeof(*p), gfp_mask);
5672         if (p == NULL)
5673                 return NULL;
5674
5675         p->arg.fh = NFS_FH(inode);
5676         p->arg.fl = &p->fl;
5677         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5678         if (IS_ERR(p->arg.open_seqid))
5679                 goto out_free;
5680         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5681         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5682         if (IS_ERR(p->arg.lock_seqid))
5683                 goto out_free_seqid;
5684         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5685         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5686         p->arg.lock_owner.s_dev = server->s_dev;
5687         p->res.lock_seqid = p->arg.lock_seqid;
5688         p->lsp = lsp;
5689         p->server = server;
5690         atomic_inc(&lsp->ls_count);
5691         p->ctx = get_nfs_open_context(ctx);
5692         get_file(fl->fl_file);
5693         memcpy(&p->fl, fl, sizeof(p->fl));
5694         return p;
5695 out_free_seqid:
5696         nfs_free_seqid(p->arg.open_seqid);
5697 out_free:
5698         kfree(p);
5699         return NULL;
5700 }
5701
5702 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5703 {
5704         struct nfs4_lockdata *data = calldata;
5705         struct nfs4_state *state = data->lsp->ls_state;
5706
5707         dprintk("%s: begin!\n", __func__);
5708         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5709                 goto out_wait;
5710         /* Do we need to do an open_to_lock_owner? */
5711         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5712                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5713                         goto out_release_lock_seqid;
5714                 }
5715                 nfs4_stateid_copy(&data->arg.open_stateid,
5716                                 &state->open_stateid);
5717                 data->arg.new_lock_owner = 1;
5718                 data->res.open_seqid = data->arg.open_seqid;
5719         } else {
5720                 data->arg.new_lock_owner = 0;
5721                 nfs4_stateid_copy(&data->arg.lock_stateid,
5722                                 &data->lsp->ls_stateid);
5723         }
5724         if (!nfs4_valid_open_stateid(state)) {
5725                 data->rpc_status = -EBADF;
5726                 task->tk_action = NULL;
5727                 goto out_release_open_seqid;
5728         }
5729         data->timestamp = jiffies;
5730         if (nfs4_setup_sequence(data->server,
5731                                 &data->arg.seq_args,
5732                                 &data->res.seq_res,
5733                                 task) == 0)
5734                 return;
5735 out_release_open_seqid:
5736         nfs_release_seqid(data->arg.open_seqid);
5737 out_release_lock_seqid:
5738         nfs_release_seqid(data->arg.lock_seqid);
5739 out_wait:
5740         nfs4_sequence_done(task, &data->res.seq_res);
5741         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5742 }
5743
5744 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5745 {
5746         struct nfs4_lockdata *data = calldata;
5747         struct nfs4_lock_state *lsp = data->lsp;
5748
5749         dprintk("%s: begin!\n", __func__);
5750
5751         if (!nfs4_sequence_done(task, &data->res.seq_res))
5752                 return;
5753
5754         data->rpc_status = task->tk_status;
5755         switch (task->tk_status) {
5756         case 0:
5757                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5758                                 data->timestamp);
5759                 if (data->arg.new_lock) {
5760                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5761                         if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5762                                 rpc_restart_call_prepare(task);
5763                                 break;
5764                         }
5765                 }
5766                 if (data->arg.new_lock_owner != 0) {
5767                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5768                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5769                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5770                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5771                         rpc_restart_call_prepare(task);
5772                 break;
5773         case -NFS4ERR_BAD_STATEID:
5774         case -NFS4ERR_OLD_STATEID:
5775         case -NFS4ERR_STALE_STATEID:
5776         case -NFS4ERR_EXPIRED:
5777                 if (data->arg.new_lock_owner != 0) {
5778                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5779                                                 &lsp->ls_state->open_stateid))
5780                                 rpc_restart_call_prepare(task);
5781                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5782                                                 &lsp->ls_stateid))
5783                                 rpc_restart_call_prepare(task);
5784         }
5785         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5786 }
5787
5788 static void nfs4_lock_release(void *calldata)
5789 {
5790         struct nfs4_lockdata *data = calldata;
5791
5792         dprintk("%s: begin!\n", __func__);
5793         nfs_free_seqid(data->arg.open_seqid);
5794         if (data->cancelled != 0) {
5795                 struct rpc_task *task;
5796                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5797                                 data->arg.lock_seqid);
5798                 if (!IS_ERR(task))
5799                         rpc_put_task_async(task);
5800                 dprintk("%s: cancelling lock!\n", __func__);
5801         } else
5802                 nfs_free_seqid(data->arg.lock_seqid);
5803         nfs4_put_lock_state(data->lsp);
5804         put_nfs_open_context(data->ctx);
5805         fput(data->fl.fl_file);
5806         kfree(data);
5807         dprintk("%s: done!\n", __func__);
5808 }
5809
5810 static const struct rpc_call_ops nfs4_lock_ops = {
5811         .rpc_call_prepare = nfs4_lock_prepare,
5812         .rpc_call_done = nfs4_lock_done,
5813         .rpc_release = nfs4_lock_release,
5814 };
5815
5816 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5817 {
5818         switch (error) {
5819         case -NFS4ERR_ADMIN_REVOKED:
5820         case -NFS4ERR_BAD_STATEID:
5821                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5822                 if (new_lock_owner != 0 ||
5823                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5824                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5825                 break;
5826         case -NFS4ERR_STALE_STATEID:
5827                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5828         case -NFS4ERR_EXPIRED:
5829                 nfs4_schedule_lease_recovery(server->nfs_client);
5830         };
5831 }
5832
5833 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5834 {
5835         struct nfs4_lockdata *data;
5836         struct rpc_task *task;
5837         struct rpc_message msg = {
5838                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5839                 .rpc_cred = state->owner->so_cred,
5840         };
5841         struct rpc_task_setup task_setup_data = {
5842                 .rpc_client = NFS_CLIENT(state->inode),
5843                 .rpc_message = &msg,
5844                 .callback_ops = &nfs4_lock_ops,
5845                 .workqueue = nfsiod_workqueue,
5846                 .flags = RPC_TASK_ASYNC,
5847         };
5848         int ret;
5849
5850         dprintk("%s: begin!\n", __func__);
5851         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5852                         fl->fl_u.nfs4_fl.owner,
5853                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5854         if (data == NULL)
5855                 return -ENOMEM;
5856         if (IS_SETLKW(cmd))
5857                 data->arg.block = 1;
5858         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5859         msg.rpc_argp = &data->arg;
5860         msg.rpc_resp = &data->res;
5861         task_setup_data.callback_data = data;
5862         if (recovery_type > NFS_LOCK_NEW) {
5863                 if (recovery_type == NFS_LOCK_RECLAIM)
5864                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5865                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5866         } else
5867                 data->arg.new_lock = 1;
5868         task = rpc_run_task(&task_setup_data);
5869         if (IS_ERR(task))
5870                 return PTR_ERR(task);
5871         ret = nfs4_wait_for_completion_rpc_task(task);
5872         if (ret == 0) {
5873                 ret = data->rpc_status;
5874                 if (ret)
5875                         nfs4_handle_setlk_error(data->server, data->lsp,
5876                                         data->arg.new_lock_owner, ret);
5877         } else
5878                 data->cancelled = 1;
5879         rpc_put_task(task);
5880         dprintk("%s: done, ret = %d!\n", __func__, ret);
5881         return ret;
5882 }
5883
5884 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5885 {
5886         struct nfs_server *server = NFS_SERVER(state->inode);
5887         struct nfs4_exception exception = {
5888                 .inode = state->inode,
5889         };
5890         int err;
5891
5892         do {
5893                 /* Cache the lock if possible... */
5894                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5895                         return 0;
5896                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5897                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5898                 if (err != -NFS4ERR_DELAY)
5899                         break;
5900                 nfs4_handle_exception(server, err, &exception);
5901         } while (exception.retry);
5902         return err;
5903 }
5904
5905 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5906 {
5907         struct nfs_server *server = NFS_SERVER(state->inode);
5908         struct nfs4_exception exception = {
5909                 .inode = state->inode,
5910         };
5911         int err;
5912
5913         err = nfs4_set_lock_state(state, request);
5914         if (err != 0)
5915                 return err;
5916         if (!recover_lost_locks) {
5917                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5918                 return 0;
5919         }
5920         do {
5921                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5922                         return 0;
5923                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5924                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5925                 switch (err) {
5926                 default:
5927                         goto out;
5928                 case -NFS4ERR_GRACE:
5929                 case -NFS4ERR_DELAY:
5930                         nfs4_handle_exception(server, err, &exception);
5931                         err = 0;
5932                 }
5933         } while (exception.retry);
5934 out:
5935         return err;
5936 }
5937
5938 #if defined(CONFIG_NFS_V4_1)
5939 /**
5940  * nfs41_check_expired_locks - possibly free a lock stateid
5941  *
5942  * @state: NFSv4 state for an inode
5943  *
5944  * Returns NFS_OK if recovery for this stateid is now finished.
5945  * Otherwise a negative NFS4ERR value is returned.
5946  */
5947 static int nfs41_check_expired_locks(struct nfs4_state *state)
5948 {
5949         int status, ret = -NFS4ERR_BAD_STATEID;
5950         struct nfs4_lock_state *lsp;
5951         struct nfs_server *server = NFS_SERVER(state->inode);
5952
5953         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5954                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5955                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5956
5957                         status = nfs41_test_stateid(server,
5958                                         &lsp->ls_stateid,
5959                                         cred);
5960                         trace_nfs4_test_lock_stateid(state, lsp, status);
5961                         if (status != NFS_OK) {
5962                                 /* Free the stateid unless the server
5963                                  * informs us the stateid is unrecognized. */
5964                                 if (status != -NFS4ERR_BAD_STATEID)
5965                                         nfs41_free_stateid(server,
5966                                                         &lsp->ls_stateid,
5967                                                         cred);
5968                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5969                                 ret = status;
5970                         }
5971                 }
5972         };
5973
5974         return ret;
5975 }
5976
5977 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5978 {
5979         int status = NFS_OK;
5980
5981         if (test_bit(LK_STATE_IN_USE, &state->flags))
5982                 status = nfs41_check_expired_locks(state);
5983         if (status != NFS_OK)
5984                 status = nfs4_lock_expired(state, request);
5985         return status;
5986 }
5987 #endif
5988
5989 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5990 {
5991         struct nfs_inode *nfsi = NFS_I(state->inode);
5992         unsigned char fl_flags = request->fl_flags;
5993         int status = -ENOLCK;
5994
5995         if ((fl_flags & FL_POSIX) &&
5996                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5997                 goto out;
5998         /* Is this a delegated open? */
5999         status = nfs4_set_lock_state(state, request);
6000         if (status != 0)
6001                 goto out;
6002         request->fl_flags |= FL_ACCESS;
6003         status = do_vfs_lock(request->fl_file, request);
6004         if (status < 0)
6005                 goto out;
6006         down_read(&nfsi->rwsem);
6007         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6008                 /* Yes: cache locks! */
6009                 /* ...but avoid races with delegation recall... */
6010                 request->fl_flags = fl_flags & ~FL_SLEEP;
6011                 status = do_vfs_lock(request->fl_file, request);
6012                 up_read(&nfsi->rwsem);
6013                 goto out;
6014         }
6015         up_read(&nfsi->rwsem);
6016         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6017 out:
6018         request->fl_flags = fl_flags;
6019         return status;
6020 }
6021
6022 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6023 {
6024         struct nfs4_exception exception = {
6025                 .state = state,
6026                 .inode = state->inode,
6027         };
6028         int err;
6029
6030         do {
6031                 err = _nfs4_proc_setlk(state, cmd, request);
6032                 trace_nfs4_set_lock(request, state, cmd, err);
6033                 if (err == -NFS4ERR_DENIED)
6034                         err = -EAGAIN;
6035                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6036                                 err, &exception);
6037         } while (exception.retry);
6038         return err;
6039 }
6040
6041 static int
6042 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6043 {
6044         struct nfs_open_context *ctx;
6045         struct nfs4_state *state;
6046         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6047         int status;
6048
6049         /* verify open state */
6050         ctx = nfs_file_open_context(filp);
6051         state = ctx->state;
6052
6053         if (request->fl_start < 0 || request->fl_end < 0)
6054                 return -EINVAL;
6055
6056         if (IS_GETLK(cmd)) {
6057                 if (state != NULL)
6058                         return nfs4_proc_getlk(state, F_GETLK, request);
6059                 return 0;
6060         }
6061
6062         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6063                 return -EINVAL;
6064
6065         if (request->fl_type == F_UNLCK) {
6066                 if (state != NULL)
6067                         return nfs4_proc_unlck(state, cmd, request);
6068                 return 0;
6069         }
6070
6071         if (state == NULL)
6072                 return -ENOLCK;
6073         /*
6074          * Don't rely on the VFS having checked the file open mode,
6075          * since it won't do this for flock() locks.
6076          */
6077         switch (request->fl_type) {
6078         case F_RDLCK:
6079                 if (!(filp->f_mode & FMODE_READ))
6080                         return -EBADF;
6081                 break;
6082         case F_WRLCK:
6083                 if (!(filp->f_mode & FMODE_WRITE))
6084                         return -EBADF;
6085         }
6086
6087         do {
6088                 status = nfs4_proc_setlk(state, cmd, request);
6089                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6090                         break;
6091                 timeout = nfs4_set_lock_task_retry(timeout);
6092                 status = -ERESTARTSYS;
6093                 if (signalled())
6094                         break;
6095         } while(status < 0);
6096         return status;
6097 }
6098
6099 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6100 {
6101         struct nfs_server *server = NFS_SERVER(state->inode);
6102         int err;
6103
6104         err = nfs4_set_lock_state(state, fl);
6105         if (err != 0)
6106                 return err;
6107         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6108         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6109 }
6110
6111 struct nfs_release_lockowner_data {
6112         struct nfs4_lock_state *lsp;
6113         struct nfs_server *server;
6114         struct nfs_release_lockowner_args args;
6115         struct nfs_release_lockowner_res res;
6116         unsigned long timestamp;
6117 };
6118
6119 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6120 {
6121         struct nfs_release_lockowner_data *data = calldata;
6122         struct nfs_server *server = data->server;
6123         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6124                              &data->args.seq_args, &data->res.seq_res, task);
6125         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6126         data->timestamp = jiffies;
6127 }
6128
6129 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6130 {
6131         struct nfs_release_lockowner_data *data = calldata;
6132         struct nfs_server *server = data->server;
6133
6134         nfs40_sequence_done(task, &data->res.seq_res);
6135
6136         switch (task->tk_status) {
6137         case 0:
6138                 renew_lease(server, data->timestamp);
6139                 break;
6140         case -NFS4ERR_STALE_CLIENTID:
6141         case -NFS4ERR_EXPIRED:
6142                 nfs4_schedule_lease_recovery(server->nfs_client);
6143                 break;
6144         case -NFS4ERR_LEASE_MOVED:
6145         case -NFS4ERR_DELAY:
6146                 if (nfs4_async_handle_error(task, server,
6147                                             NULL, NULL) == -EAGAIN)
6148                         rpc_restart_call_prepare(task);
6149         }
6150 }
6151
6152 static void nfs4_release_lockowner_release(void *calldata)
6153 {
6154         struct nfs_release_lockowner_data *data = calldata;
6155         nfs4_free_lock_state(data->server, data->lsp);
6156         kfree(calldata);
6157 }
6158
6159 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6160         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6161         .rpc_call_done = nfs4_release_lockowner_done,
6162         .rpc_release = nfs4_release_lockowner_release,
6163 };
6164
6165 static void
6166 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6167 {
6168         struct nfs_release_lockowner_data *data;
6169         struct rpc_message msg = {
6170                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6171         };
6172
6173         if (server->nfs_client->cl_mvops->minor_version != 0)
6174                 return;
6175
6176         data = kmalloc(sizeof(*data), GFP_NOFS);
6177         if (!data)
6178                 return;
6179         data->lsp = lsp;
6180         data->server = server;
6181         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6182         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6183         data->args.lock_owner.s_dev = server->s_dev;
6184
6185         msg.rpc_argp = &data->args;
6186         msg.rpc_resp = &data->res;
6187         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6188         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6189 }
6190
6191 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6192
6193 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6194                                    const void *buf, size_t buflen,
6195                                    int flags, int type)
6196 {
6197         if (strcmp(key, "") != 0)
6198                 return -EINVAL;
6199
6200         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6201 }
6202
6203 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6204                                    void *buf, size_t buflen, int type)
6205 {
6206         if (strcmp(key, "") != 0)
6207                 return -EINVAL;
6208
6209         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6210 }
6211
6212 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6213                                        size_t list_len, const char *name,
6214                                        size_t name_len, int type)
6215 {
6216         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6217
6218         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6219                 return 0;
6220
6221         if (list && len <= list_len)
6222                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6223         return len;
6224 }
6225
6226 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6227 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6228 {
6229         return server->caps & NFS_CAP_SECURITY_LABEL;
6230 }
6231
6232 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6233                                    const void *buf, size_t buflen,
6234                                    int flags, int type)
6235 {
6236         if (security_ismaclabel(key))
6237                 return nfs4_set_security_label(dentry, buf, buflen);
6238
6239         return -EOPNOTSUPP;
6240 }
6241
6242 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6243                                    void *buf, size_t buflen, int type)
6244 {
6245         if (security_ismaclabel(key))
6246                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6247         return -EOPNOTSUPP;
6248 }
6249
6250 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6251                                        size_t list_len, const char *name,
6252                                        size_t name_len, int type)
6253 {
6254         size_t len = 0;
6255
6256         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6257                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6258                 if (list && len <= list_len)
6259                         security_inode_listsecurity(d_inode(dentry), list, len);
6260         }
6261         return len;
6262 }
6263
6264 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6265         .prefix = XATTR_SECURITY_PREFIX,
6266         .list   = nfs4_xattr_list_nfs4_label,
6267         .get    = nfs4_xattr_get_nfs4_label,
6268         .set    = nfs4_xattr_set_nfs4_label,
6269 };
6270 #endif
6271
6272
6273 /*
6274  * nfs_fhget will use either the mounted_on_fileid or the fileid
6275  */
6276 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6277 {
6278         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6279                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6280               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6281               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6282                 return;
6283
6284         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6285                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6286         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6287         fattr->nlink = 2;
6288 }
6289
6290 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6291                                    const struct qstr *name,
6292                                    struct nfs4_fs_locations *fs_locations,
6293                                    struct page *page)
6294 {
6295         struct nfs_server *server = NFS_SERVER(dir);
6296         u32 bitmask[3] = {
6297                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6298         };
6299         struct nfs4_fs_locations_arg args = {
6300                 .dir_fh = NFS_FH(dir),
6301                 .name = name,
6302                 .page = page,
6303                 .bitmask = bitmask,
6304         };
6305         struct nfs4_fs_locations_res res = {
6306                 .fs_locations = fs_locations,
6307         };
6308         struct rpc_message msg = {
6309                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6310                 .rpc_argp = &args,
6311                 .rpc_resp = &res,
6312         };
6313         int status;
6314
6315         dprintk("%s: start\n", __func__);
6316
6317         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6318          * is not supported */
6319         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6320                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6321         else
6322                 bitmask[0] |= FATTR4_WORD0_FILEID;
6323
6324         nfs_fattr_init(&fs_locations->fattr);
6325         fs_locations->server = server;
6326         fs_locations->nlocations = 0;
6327         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6328         dprintk("%s: returned status = %d\n", __func__, status);
6329         return status;
6330 }
6331
6332 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6333                            const struct qstr *name,
6334                            struct nfs4_fs_locations *fs_locations,
6335                            struct page *page)
6336 {
6337         struct nfs4_exception exception = { };
6338         int err;
6339         do {
6340                 err = _nfs4_proc_fs_locations(client, dir, name,
6341                                 fs_locations, page);
6342                 trace_nfs4_get_fs_locations(dir, name, err);
6343                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6344                                 &exception);
6345         } while (exception.retry);
6346         return err;
6347 }
6348
6349 /*
6350  * This operation also signals the server that this client is
6351  * performing migration recovery.  The server can stop returning
6352  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6353  * appended to this compound to identify the client ID which is
6354  * performing recovery.
6355  */
6356 static int _nfs40_proc_get_locations(struct inode *inode,
6357                                      struct nfs4_fs_locations *locations,
6358                                      struct page *page, struct rpc_cred *cred)
6359 {
6360         struct nfs_server *server = NFS_SERVER(inode);
6361         struct rpc_clnt *clnt = server->client;
6362         u32 bitmask[2] = {
6363                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6364         };
6365         struct nfs4_fs_locations_arg args = {
6366                 .clientid       = server->nfs_client->cl_clientid,
6367                 .fh             = NFS_FH(inode),
6368                 .page           = page,
6369                 .bitmask        = bitmask,
6370                 .migration      = 1,            /* skip LOOKUP */
6371                 .renew          = 1,            /* append RENEW */
6372         };
6373         struct nfs4_fs_locations_res res = {
6374                 .fs_locations   = locations,
6375                 .migration      = 1,
6376                 .renew          = 1,
6377         };
6378         struct rpc_message msg = {
6379                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6380                 .rpc_argp       = &args,
6381                 .rpc_resp       = &res,
6382                 .rpc_cred       = cred,
6383         };
6384         unsigned long now = jiffies;
6385         int status;
6386
6387         nfs_fattr_init(&locations->fattr);
6388         locations->server = server;
6389         locations->nlocations = 0;
6390
6391         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6392         nfs4_set_sequence_privileged(&args.seq_args);
6393         status = nfs4_call_sync_sequence(clnt, server, &msg,
6394                                         &args.seq_args, &res.seq_res);
6395         if (status)
6396                 return status;
6397
6398         renew_lease(server, now);
6399         return 0;
6400 }
6401
6402 #ifdef CONFIG_NFS_V4_1
6403
6404 /*
6405  * This operation also signals the server that this client is
6406  * performing migration recovery.  The server can stop asserting
6407  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6408  * performing this operation is identified in the SEQUENCE
6409  * operation in this compound.
6410  *
6411  * When the client supports GETATTR(fs_locations_info), it can
6412  * be plumbed in here.
6413  */
6414 static int _nfs41_proc_get_locations(struct inode *inode,
6415                                      struct nfs4_fs_locations *locations,
6416                                      struct page *page, struct rpc_cred *cred)
6417 {
6418         struct nfs_server *server = NFS_SERVER(inode);
6419         struct rpc_clnt *clnt = server->client;
6420         u32 bitmask[2] = {
6421                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6422         };
6423         struct nfs4_fs_locations_arg args = {
6424                 .fh             = NFS_FH(inode),
6425                 .page           = page,
6426                 .bitmask        = bitmask,
6427                 .migration      = 1,            /* skip LOOKUP */
6428         };
6429         struct nfs4_fs_locations_res res = {
6430                 .fs_locations   = locations,
6431                 .migration      = 1,
6432         };
6433         struct rpc_message msg = {
6434                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6435                 .rpc_argp       = &args,
6436                 .rpc_resp       = &res,
6437                 .rpc_cred       = cred,
6438         };
6439         int status;
6440
6441         nfs_fattr_init(&locations->fattr);
6442         locations->server = server;
6443         locations->nlocations = 0;
6444
6445         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6446         nfs4_set_sequence_privileged(&args.seq_args);
6447         status = nfs4_call_sync_sequence(clnt, server, &msg,
6448                                         &args.seq_args, &res.seq_res);
6449         if (status == NFS4_OK &&
6450             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6451                 status = -NFS4ERR_LEASE_MOVED;
6452         return status;
6453 }
6454
6455 #endif  /* CONFIG_NFS_V4_1 */
6456
6457 /**
6458  * nfs4_proc_get_locations - discover locations for a migrated FSID
6459  * @inode: inode on FSID that is migrating
6460  * @locations: result of query
6461  * @page: buffer
6462  * @cred: credential to use for this operation
6463  *
6464  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6465  * operation failed, or a negative errno if a local error occurred.
6466  *
6467  * On success, "locations" is filled in, but if the server has
6468  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6469  * asserted.
6470  *
6471  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6472  * from this client that require migration recovery.
6473  */
6474 int nfs4_proc_get_locations(struct inode *inode,
6475                             struct nfs4_fs_locations *locations,
6476                             struct page *page, struct rpc_cred *cred)
6477 {
6478         struct nfs_server *server = NFS_SERVER(inode);
6479         struct nfs_client *clp = server->nfs_client;
6480         const struct nfs4_mig_recovery_ops *ops =
6481                                         clp->cl_mvops->mig_recovery_ops;
6482         struct nfs4_exception exception = { };
6483         int status;
6484
6485         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6486                 (unsigned long long)server->fsid.major,
6487                 (unsigned long long)server->fsid.minor,
6488                 clp->cl_hostname);
6489         nfs_display_fhandle(NFS_FH(inode), __func__);
6490
6491         do {
6492                 status = ops->get_locations(inode, locations, page, cred);
6493                 if (status != -NFS4ERR_DELAY)
6494                         break;
6495                 nfs4_handle_exception(server, status, &exception);
6496         } while (exception.retry);
6497         return status;
6498 }
6499
6500 /*
6501  * This operation also signals the server that this client is
6502  * performing "lease moved" recovery.  The server can stop
6503  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6504  * is appended to this compound to identify the client ID which is
6505  * performing recovery.
6506  */
6507 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6508 {
6509         struct nfs_server *server = NFS_SERVER(inode);
6510         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6511         struct rpc_clnt *clnt = server->client;
6512         struct nfs4_fsid_present_arg args = {
6513                 .fh             = NFS_FH(inode),
6514                 .clientid       = clp->cl_clientid,
6515                 .renew          = 1,            /* append RENEW */
6516         };
6517         struct nfs4_fsid_present_res res = {
6518                 .renew          = 1,
6519         };
6520         struct rpc_message msg = {
6521                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6522                 .rpc_argp       = &args,
6523                 .rpc_resp       = &res,
6524                 .rpc_cred       = cred,
6525         };
6526         unsigned long now = jiffies;
6527         int status;
6528
6529         res.fh = nfs_alloc_fhandle();
6530         if (res.fh == NULL)
6531                 return -ENOMEM;
6532
6533         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6534         nfs4_set_sequence_privileged(&args.seq_args);
6535         status = nfs4_call_sync_sequence(clnt, server, &msg,
6536                                                 &args.seq_args, &res.seq_res);
6537         nfs_free_fhandle(res.fh);
6538         if (status)
6539                 return status;
6540
6541         do_renew_lease(clp, now);
6542         return 0;
6543 }
6544
6545 #ifdef CONFIG_NFS_V4_1
6546
6547 /*
6548  * This operation also signals the server that this client is
6549  * performing "lease moved" recovery.  The server can stop asserting
6550  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6551  * this operation is identified in the SEQUENCE operation in this
6552  * compound.
6553  */
6554 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6555 {
6556         struct nfs_server *server = NFS_SERVER(inode);
6557         struct rpc_clnt *clnt = server->client;
6558         struct nfs4_fsid_present_arg args = {
6559                 .fh             = NFS_FH(inode),
6560         };
6561         struct nfs4_fsid_present_res res = {
6562         };
6563         struct rpc_message msg = {
6564                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6565                 .rpc_argp       = &args,
6566                 .rpc_resp       = &res,
6567                 .rpc_cred       = cred,
6568         };
6569         int status;
6570
6571         res.fh = nfs_alloc_fhandle();
6572         if (res.fh == NULL)
6573                 return -ENOMEM;
6574
6575         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6576         nfs4_set_sequence_privileged(&args.seq_args);
6577         status = nfs4_call_sync_sequence(clnt, server, &msg,
6578                                                 &args.seq_args, &res.seq_res);
6579         nfs_free_fhandle(res.fh);
6580         if (status == NFS4_OK &&
6581             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6582                 status = -NFS4ERR_LEASE_MOVED;
6583         return status;
6584 }
6585
6586 #endif  /* CONFIG_NFS_V4_1 */
6587
6588 /**
6589  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6590  * @inode: inode on FSID to check
6591  * @cred: credential to use for this operation
6592  *
6593  * Server indicates whether the FSID is present, moved, or not
6594  * recognized.  This operation is necessary to clear a LEASE_MOVED
6595  * condition for this client ID.
6596  *
6597  * Returns NFS4_OK if the FSID is present on this server,
6598  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6599  *  NFS4ERR code if some error occurred on the server, or a
6600  *  negative errno if a local failure occurred.
6601  */
6602 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6603 {
6604         struct nfs_server *server = NFS_SERVER(inode);
6605         struct nfs_client *clp = server->nfs_client;
6606         const struct nfs4_mig_recovery_ops *ops =
6607                                         clp->cl_mvops->mig_recovery_ops;
6608         struct nfs4_exception exception = { };
6609         int status;
6610
6611         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6612                 (unsigned long long)server->fsid.major,
6613                 (unsigned long long)server->fsid.minor,
6614                 clp->cl_hostname);
6615         nfs_display_fhandle(NFS_FH(inode), __func__);
6616
6617         do {
6618                 status = ops->fsid_present(inode, cred);
6619                 if (status != -NFS4ERR_DELAY)
6620                         break;
6621                 nfs4_handle_exception(server, status, &exception);
6622         } while (exception.retry);
6623         return status;
6624 }
6625
6626 /**
6627  * If 'use_integrity' is true and the state managment nfs_client
6628  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6629  * and the machine credential as per RFC3530bis and RFC5661 Security
6630  * Considerations sections. Otherwise, just use the user cred with the
6631  * filesystem's rpc_client.
6632  */
6633 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6634 {
6635         int status;
6636         struct nfs4_secinfo_arg args = {
6637                 .dir_fh = NFS_FH(dir),
6638                 .name   = name,
6639         };
6640         struct nfs4_secinfo_res res = {
6641                 .flavors     = flavors,
6642         };
6643         struct rpc_message msg = {
6644                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6645                 .rpc_argp = &args,
6646                 .rpc_resp = &res,
6647         };
6648         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6649         struct rpc_cred *cred = NULL;
6650
6651         if (use_integrity) {
6652                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6653                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6654                 msg.rpc_cred = cred;
6655         }
6656
6657         dprintk("NFS call  secinfo %s\n", name->name);
6658
6659         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6660                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6661
6662         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6663                                 &res.seq_res, 0);
6664         dprintk("NFS reply  secinfo: %d\n", status);
6665
6666         if (cred)
6667                 put_rpccred(cred);
6668
6669         return status;
6670 }
6671
6672 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6673                       struct nfs4_secinfo_flavors *flavors)
6674 {
6675         struct nfs4_exception exception = { };
6676         int err;
6677         do {
6678                 err = -NFS4ERR_WRONGSEC;
6679
6680                 /* try to use integrity protection with machine cred */
6681                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6682                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6683
6684                 /*
6685                  * if unable to use integrity protection, or SECINFO with
6686                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6687                  * disallowed by spec, but exists in deployed servers) use
6688                  * the current filesystem's rpc_client and the user cred.
6689                  */
6690                 if (err == -NFS4ERR_WRONGSEC)
6691                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6692
6693                 trace_nfs4_secinfo(dir, name, err);
6694                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6695                                 &exception);
6696         } while (exception.retry);
6697         return err;
6698 }
6699
6700 #ifdef CONFIG_NFS_V4_1
6701 /*
6702  * Check the exchange flags returned by the server for invalid flags, having
6703  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6704  * DS flags set.
6705  */
6706 static int nfs4_check_cl_exchange_flags(u32 flags)
6707 {
6708         if (flags & ~EXCHGID4_FLAG_MASK_R)
6709                 goto out_inval;
6710         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6711             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6712                 goto out_inval;
6713         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6714                 goto out_inval;
6715         return NFS_OK;
6716 out_inval:
6717         return -NFS4ERR_INVAL;
6718 }
6719
6720 static bool
6721 nfs41_same_server_scope(struct nfs41_server_scope *a,
6722                         struct nfs41_server_scope *b)
6723 {
6724         if (a->server_scope_sz == b->server_scope_sz &&
6725             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6726                 return true;
6727
6728         return false;
6729 }
6730
6731 /*
6732  * nfs4_proc_bind_conn_to_session()
6733  *
6734  * The 4.1 client currently uses the same TCP connection for the
6735  * fore and backchannel.
6736  */
6737 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6738 {
6739         int status;
6740         struct nfs41_bind_conn_to_session_args args = {
6741                 .client = clp,
6742                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6743         };
6744         struct nfs41_bind_conn_to_session_res res;
6745         struct rpc_message msg = {
6746                 .rpc_proc =
6747                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6748                 .rpc_argp = &args,
6749                 .rpc_resp = &res,
6750                 .rpc_cred = cred,
6751         };
6752
6753         dprintk("--> %s\n", __func__);
6754
6755         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6756         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6757                 args.dir = NFS4_CDFC4_FORE;
6758
6759         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6760         trace_nfs4_bind_conn_to_session(clp, status);
6761         if (status == 0) {
6762                 if (memcmp(res.sessionid.data,
6763                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6764                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6765                         status = -EIO;
6766                         goto out;
6767                 }
6768                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6769                         dprintk("NFS: %s: Unexpected direction from server\n",
6770                                 __func__);
6771                         status = -EIO;
6772                         goto out;
6773                 }
6774                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6775                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6776                                 __func__);
6777                         status = -EIO;
6778                         goto out;
6779                 }
6780         }
6781 out:
6782         dprintk("<-- %s status= %d\n", __func__, status);
6783         return status;
6784 }
6785
6786 /*
6787  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6788  * and operations we'd like to see to enable certain features in the allow map
6789  */
6790 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6791         .how = SP4_MACH_CRED,
6792         .enforce.u.words = {
6793                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6794                       1 << (OP_EXCHANGE_ID - 32) |
6795                       1 << (OP_CREATE_SESSION - 32) |
6796                       1 << (OP_DESTROY_SESSION - 32) |
6797                       1 << (OP_DESTROY_CLIENTID - 32)
6798         },
6799         .allow.u.words = {
6800                 [0] = 1 << (OP_CLOSE) |
6801                       1 << (OP_LOCKU) |
6802                       1 << (OP_COMMIT),
6803                 [1] = 1 << (OP_SECINFO - 32) |
6804                       1 << (OP_SECINFO_NO_NAME - 32) |
6805                       1 << (OP_TEST_STATEID - 32) |
6806                       1 << (OP_FREE_STATEID - 32) |
6807                       1 << (OP_WRITE - 32)
6808         }
6809 };
6810
6811 /*
6812  * Select the state protection mode for client `clp' given the server results
6813  * from exchange_id in `sp'.
6814  *
6815  * Returns 0 on success, negative errno otherwise.
6816  */
6817 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6818                                  struct nfs41_state_protection *sp)
6819 {
6820         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6821                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6822                       1 << (OP_EXCHANGE_ID - 32) |
6823                       1 << (OP_CREATE_SESSION - 32) |
6824                       1 << (OP_DESTROY_SESSION - 32) |
6825                       1 << (OP_DESTROY_CLIENTID - 32)
6826         };
6827         unsigned int i;
6828
6829         if (sp->how == SP4_MACH_CRED) {
6830                 /* Print state protect result */
6831                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6832                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6833                         if (test_bit(i, sp->enforce.u.longs))
6834                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6835                         if (test_bit(i, sp->allow.u.longs))
6836                                 dfprintk(MOUNT, "  allow op %d\n", i);
6837                 }
6838
6839                 /* make sure nothing is on enforce list that isn't supported */
6840                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6841                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6842                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6843                                 return -EINVAL;
6844                         }
6845                 }
6846
6847                 /*
6848                  * Minimal mode - state operations are allowed to use machine
6849                  * credential.  Note this already happens by default, so the
6850                  * client doesn't have to do anything more than the negotiation.
6851                  *
6852                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6853                  *       we're already using the machine cred for exchange_id
6854                  *       and will never use a different cred.
6855                  */
6856                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6857                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6858                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6859                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6860                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6861                         dfprintk(MOUNT, "  minimal mode enabled\n");
6862                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6863                 } else {
6864                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6865                         return -EINVAL;
6866                 }
6867
6868                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6869                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6870                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6871                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6872                 }
6873
6874                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6875                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6876                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6877                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6878                 }
6879
6880                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6881                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6882                         dfprintk(MOUNT, "  stateid mode enabled\n");
6883                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6884                 }
6885
6886                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6887                         dfprintk(MOUNT, "  write mode enabled\n");
6888                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6889                 }
6890
6891                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6892                         dfprintk(MOUNT, "  commit mode enabled\n");
6893                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6894                 }
6895         }
6896
6897         return 0;
6898 }
6899
6900 /*
6901  * _nfs4_proc_exchange_id()
6902  *
6903  * Wrapper for EXCHANGE_ID operation.
6904  */
6905 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6906         u32 sp4_how)
6907 {
6908         nfs4_verifier verifier;
6909         struct nfs41_exchange_id_args args = {
6910                 .verifier = &verifier,
6911                 .client = clp,
6912 #ifdef CONFIG_NFS_V4_1_MIGRATION
6913                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6914                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6915                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6916 #else
6917                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6918                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6919 #endif
6920         };
6921         struct nfs41_exchange_id_res res = {
6922                 0
6923         };
6924         int status;
6925         struct rpc_message msg = {
6926                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6927                 .rpc_argp = &args,
6928                 .rpc_resp = &res,
6929                 .rpc_cred = cred,
6930         };
6931
6932         nfs4_init_boot_verifier(clp, &verifier);
6933
6934         status = nfs4_init_uniform_client_string(clp);
6935         if (status)
6936                 goto out;
6937
6938         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6939                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6940                 clp->cl_owner_id);
6941
6942         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6943                                         GFP_NOFS);
6944         if (unlikely(res.server_owner == NULL)) {
6945                 status = -ENOMEM;
6946                 goto out;
6947         }
6948
6949         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6950                                         GFP_NOFS);
6951         if (unlikely(res.server_scope == NULL)) {
6952                 status = -ENOMEM;
6953                 goto out_server_owner;
6954         }
6955
6956         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6957         if (unlikely(res.impl_id == NULL)) {
6958                 status = -ENOMEM;
6959                 goto out_server_scope;
6960         }
6961
6962         switch (sp4_how) {
6963         case SP4_NONE:
6964                 args.state_protect.how = SP4_NONE;
6965                 break;
6966
6967         case SP4_MACH_CRED:
6968                 args.state_protect = nfs4_sp4_mach_cred_request;
6969                 break;
6970
6971         default:
6972                 /* unsupported! */
6973                 WARN_ON_ONCE(1);
6974                 status = -EINVAL;
6975                 goto out_impl_id;
6976         }
6977
6978         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6979         trace_nfs4_exchange_id(clp, status);
6980         if (status == 0)
6981                 status = nfs4_check_cl_exchange_flags(res.flags);
6982
6983         if (status == 0)
6984                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6985
6986         if (status == 0) {
6987                 clp->cl_clientid = res.clientid;
6988                 clp->cl_exchange_flags = res.flags;
6989                 /* Client ID is not confirmed */
6990                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6991                         clear_bit(NFS4_SESSION_ESTABLISHED,
6992                                         &clp->cl_session->session_state);
6993                         clp->cl_seqid = res.seqid;
6994                 }
6995
6996                 kfree(clp->cl_serverowner);
6997                 clp->cl_serverowner = res.server_owner;
6998                 res.server_owner = NULL;
6999
7000                 /* use the most recent implementation id */
7001                 kfree(clp->cl_implid);
7002                 clp->cl_implid = res.impl_id;
7003                 res.impl_id = NULL;
7004
7005                 if (clp->cl_serverscope != NULL &&
7006                     !nfs41_same_server_scope(clp->cl_serverscope,
7007                                              res.server_scope)) {
7008                         dprintk("%s: server_scope mismatch detected\n",
7009                                 __func__);
7010                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7011                         kfree(clp->cl_serverscope);
7012                         clp->cl_serverscope = NULL;
7013                 }
7014
7015                 if (clp->cl_serverscope == NULL) {
7016                         clp->cl_serverscope = res.server_scope;
7017                         res.server_scope = NULL;
7018                 }
7019         }
7020
7021 out_impl_id:
7022         kfree(res.impl_id);
7023 out_server_scope:
7024         kfree(res.server_scope);
7025 out_server_owner:
7026         kfree(res.server_owner);
7027 out:
7028         if (clp->cl_implid != NULL)
7029                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7030                         "domain: %s, name: %s, date: %llu,%u\n",
7031                         clp->cl_implid->domain, clp->cl_implid->name,
7032                         clp->cl_implid->date.seconds,
7033                         clp->cl_implid->date.nseconds);
7034         dprintk("NFS reply exchange_id: %d\n", status);
7035         return status;
7036 }
7037
7038 /*
7039  * nfs4_proc_exchange_id()
7040  *
7041  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7042  *
7043  * Since the clientid has expired, all compounds using sessions
7044  * associated with the stale clientid will be returning
7045  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7046  * be in some phase of session reset.
7047  *
7048  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7049  */
7050 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7051 {
7052         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7053         int status;
7054
7055         /* try SP4_MACH_CRED if krb5i/p */
7056         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7057             authflavor == RPC_AUTH_GSS_KRB5P) {
7058                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7059                 if (!status)
7060                         return 0;
7061         }
7062
7063         /* try SP4_NONE */
7064         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7065 }
7066
7067 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7068                 struct rpc_cred *cred)
7069 {
7070         struct rpc_message msg = {
7071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7072                 .rpc_argp = clp,
7073                 .rpc_cred = cred,
7074         };
7075         int status;
7076
7077         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7078         trace_nfs4_destroy_clientid(clp, status);
7079         if (status)
7080                 dprintk("NFS: Got error %d from the server %s on "
7081                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7082         return status;
7083 }
7084
7085 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7086                 struct rpc_cred *cred)
7087 {
7088         unsigned int loop;
7089         int ret;
7090
7091         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7092                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7093                 switch (ret) {
7094                 case -NFS4ERR_DELAY:
7095                 case -NFS4ERR_CLIENTID_BUSY:
7096                         ssleep(1);
7097                         break;
7098                 default:
7099                         return ret;
7100                 }
7101         }
7102         return 0;
7103 }
7104
7105 int nfs4_destroy_clientid(struct nfs_client *clp)
7106 {
7107         struct rpc_cred *cred;
7108         int ret = 0;
7109
7110         if (clp->cl_mvops->minor_version < 1)
7111                 goto out;
7112         if (clp->cl_exchange_flags == 0)
7113                 goto out;
7114         if (clp->cl_preserve_clid)
7115                 goto out;
7116         cred = nfs4_get_clid_cred(clp);
7117         ret = nfs4_proc_destroy_clientid(clp, cred);
7118         if (cred)
7119                 put_rpccred(cred);
7120         switch (ret) {
7121         case 0:
7122         case -NFS4ERR_STALE_CLIENTID:
7123                 clp->cl_exchange_flags = 0;
7124         }
7125 out:
7126         return ret;
7127 }
7128
7129 struct nfs4_get_lease_time_data {
7130         struct nfs4_get_lease_time_args *args;
7131         struct nfs4_get_lease_time_res *res;
7132         struct nfs_client *clp;
7133 };
7134
7135 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7136                                         void *calldata)
7137 {
7138         struct nfs4_get_lease_time_data *data =
7139                         (struct nfs4_get_lease_time_data *)calldata;
7140
7141         dprintk("--> %s\n", __func__);
7142         /* just setup sequence, do not trigger session recovery
7143            since we're invoked within one */
7144         nfs41_setup_sequence(data->clp->cl_session,
7145                         &data->args->la_seq_args,
7146                         &data->res->lr_seq_res,
7147                         task);
7148         dprintk("<-- %s\n", __func__);
7149 }
7150
7151 /*
7152  * Called from nfs4_state_manager thread for session setup, so don't recover
7153  * from sequence operation or clientid errors.
7154  */
7155 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7156 {
7157         struct nfs4_get_lease_time_data *data =
7158                         (struct nfs4_get_lease_time_data *)calldata;
7159
7160         dprintk("--> %s\n", __func__);
7161         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7162                 return;
7163         switch (task->tk_status) {
7164         case -NFS4ERR_DELAY:
7165         case -NFS4ERR_GRACE:
7166                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7167                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7168                 task->tk_status = 0;
7169                 /* fall through */
7170         case -NFS4ERR_RETRY_UNCACHED_REP:
7171                 rpc_restart_call_prepare(task);
7172                 return;
7173         }
7174         dprintk("<-- %s\n", __func__);
7175 }
7176
7177 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7178         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7179         .rpc_call_done = nfs4_get_lease_time_done,
7180 };
7181
7182 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7183 {
7184         struct rpc_task *task;
7185         struct nfs4_get_lease_time_args args;
7186         struct nfs4_get_lease_time_res res = {
7187                 .lr_fsinfo = fsinfo,
7188         };
7189         struct nfs4_get_lease_time_data data = {
7190                 .args = &args,
7191                 .res = &res,
7192                 .clp = clp,
7193         };
7194         struct rpc_message msg = {
7195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7196                 .rpc_argp = &args,
7197                 .rpc_resp = &res,
7198         };
7199         struct rpc_task_setup task_setup = {
7200                 .rpc_client = clp->cl_rpcclient,
7201                 .rpc_message = &msg,
7202                 .callback_ops = &nfs4_get_lease_time_ops,
7203                 .callback_data = &data,
7204                 .flags = RPC_TASK_TIMEOUT,
7205         };
7206         int status;
7207
7208         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7209         nfs4_set_sequence_privileged(&args.la_seq_args);
7210         dprintk("--> %s\n", __func__);
7211         task = rpc_run_task(&task_setup);
7212
7213         if (IS_ERR(task))
7214                 status = PTR_ERR(task);
7215         else {
7216                 status = task->tk_status;
7217                 rpc_put_task(task);
7218         }
7219         dprintk("<-- %s return %d\n", __func__, status);
7220
7221         return status;
7222 }
7223
7224 /*
7225  * Initialize the values to be used by the client in CREATE_SESSION
7226  * If nfs4_init_session set the fore channel request and response sizes,
7227  * use them.
7228  *
7229  * Set the back channel max_resp_sz_cached to zero to force the client to
7230  * always set csa_cachethis to FALSE because the current implementation
7231  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7232  */
7233 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7234 {
7235         unsigned int max_rqst_sz, max_resp_sz;
7236
7237         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7238         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7239
7240         /* Fore channel attributes */
7241         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7242         args->fc_attrs.max_resp_sz = max_resp_sz;
7243         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7244         args->fc_attrs.max_reqs = max_session_slots;
7245
7246         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7247                 "max_ops=%u max_reqs=%u\n",
7248                 __func__,
7249                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7250                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7251
7252         /* Back channel attributes */
7253         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7254         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7255         args->bc_attrs.max_resp_sz_cached = 0;
7256         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7257         args->bc_attrs.max_reqs = 1;
7258
7259         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7260                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7261                 __func__,
7262                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7263                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7264                 args->bc_attrs.max_reqs);
7265 }
7266
7267 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7268                 struct nfs41_create_session_res *res)
7269 {
7270         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7271         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7272
7273         if (rcvd->max_resp_sz > sent->max_resp_sz)
7274                 return -EINVAL;
7275         /*
7276          * Our requested max_ops is the minimum we need; we're not
7277          * prepared to break up compounds into smaller pieces than that.
7278          * So, no point even trying to continue if the server won't
7279          * cooperate:
7280          */
7281         if (rcvd->max_ops < sent->max_ops)
7282                 return -EINVAL;
7283         if (rcvd->max_reqs == 0)
7284                 return -EINVAL;
7285         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7286                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7287         return 0;
7288 }
7289
7290 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7291                 struct nfs41_create_session_res *res)
7292 {
7293         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7294         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7295
7296         if (!(res->flags & SESSION4_BACK_CHAN))
7297                 goto out;
7298         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7299                 return -EINVAL;
7300         if (rcvd->max_resp_sz < sent->max_resp_sz)
7301                 return -EINVAL;
7302         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7303                 return -EINVAL;
7304         /* These would render the backchannel useless: */
7305         if (rcvd->max_ops != sent->max_ops)
7306                 return -EINVAL;
7307         if (rcvd->max_reqs != sent->max_reqs)
7308                 return -EINVAL;
7309 out:
7310         return 0;
7311 }
7312
7313 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7314                                      struct nfs41_create_session_res *res)
7315 {
7316         int ret;
7317
7318         ret = nfs4_verify_fore_channel_attrs(args, res);
7319         if (ret)
7320                 return ret;
7321         return nfs4_verify_back_channel_attrs(args, res);
7322 }
7323
7324 static void nfs4_update_session(struct nfs4_session *session,
7325                 struct nfs41_create_session_res *res)
7326 {
7327         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7328         /* Mark client id and session as being confirmed */
7329         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7330         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7331         session->flags = res->flags;
7332         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7333         if (res->flags & SESSION4_BACK_CHAN)
7334                 memcpy(&session->bc_attrs, &res->bc_attrs,
7335                                 sizeof(session->bc_attrs));
7336 }
7337
7338 static int _nfs4_proc_create_session(struct nfs_client *clp,
7339                 struct rpc_cred *cred)
7340 {
7341         struct nfs4_session *session = clp->cl_session;
7342         struct nfs41_create_session_args args = {
7343                 .client = clp,
7344                 .clientid = clp->cl_clientid,
7345                 .seqid = clp->cl_seqid,
7346                 .cb_program = NFS4_CALLBACK,
7347         };
7348         struct nfs41_create_session_res res;
7349
7350         struct rpc_message msg = {
7351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7352                 .rpc_argp = &args,
7353                 .rpc_resp = &res,
7354                 .rpc_cred = cred,
7355         };
7356         int status;
7357
7358         nfs4_init_channel_attrs(&args);
7359         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7360
7361         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7362         trace_nfs4_create_session(clp, status);
7363
7364         if (!status) {
7365                 /* Verify the session's negotiated channel_attrs values */
7366                 status = nfs4_verify_channel_attrs(&args, &res);
7367                 /* Increment the clientid slot sequence id */
7368                 if (clp->cl_seqid == res.seqid)
7369                         clp->cl_seqid++;
7370                 if (status)
7371                         goto out;
7372                 nfs4_update_session(session, &res);
7373         }
7374 out:
7375         return status;
7376 }
7377
7378 /*
7379  * Issues a CREATE_SESSION operation to the server.
7380  * It is the responsibility of the caller to verify the session is
7381  * expired before calling this routine.
7382  */
7383 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7384 {
7385         int status;
7386         unsigned *ptr;
7387         struct nfs4_session *session = clp->cl_session;
7388
7389         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7390
7391         status = _nfs4_proc_create_session(clp, cred);
7392         if (status)
7393                 goto out;
7394
7395         /* Init or reset the session slot tables */
7396         status = nfs4_setup_session_slot_tables(session);
7397         dprintk("slot table setup returned %d\n", status);
7398         if (status)
7399                 goto out;
7400
7401         ptr = (unsigned *)&session->sess_id.data[0];
7402         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7403                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7404 out:
7405         dprintk("<-- %s\n", __func__);
7406         return status;
7407 }
7408
7409 /*
7410  * Issue the over-the-wire RPC DESTROY_SESSION.
7411  * The caller must serialize access to this routine.
7412  */
7413 int nfs4_proc_destroy_session(struct nfs4_session *session,
7414                 struct rpc_cred *cred)
7415 {
7416         struct rpc_message msg = {
7417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7418                 .rpc_argp = session,
7419                 .rpc_cred = cred,
7420         };
7421         int status = 0;
7422
7423         dprintk("--> nfs4_proc_destroy_session\n");
7424
7425         /* session is still being setup */
7426         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7427                 return 0;
7428
7429         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7430         trace_nfs4_destroy_session(session->clp, status);
7431
7432         if (status)
7433                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7434                         "Session has been destroyed regardless...\n", status);
7435
7436         dprintk("<-- nfs4_proc_destroy_session\n");
7437         return status;
7438 }
7439
7440 /*
7441  * Renew the cl_session lease.
7442  */
7443 struct nfs4_sequence_data {
7444         struct nfs_client *clp;
7445         struct nfs4_sequence_args args;
7446         struct nfs4_sequence_res res;
7447 };
7448
7449 static void nfs41_sequence_release(void *data)
7450 {
7451         struct nfs4_sequence_data *calldata = data;
7452         struct nfs_client *clp = calldata->clp;
7453
7454         if (atomic_read(&clp->cl_count) > 1)
7455                 nfs4_schedule_state_renewal(clp);
7456         nfs_put_client(clp);
7457         kfree(calldata);
7458 }
7459
7460 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7461 {
7462         switch(task->tk_status) {
7463         case -NFS4ERR_DELAY:
7464                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7465                 return -EAGAIN;
7466         default:
7467                 nfs4_schedule_lease_recovery(clp);
7468         }
7469         return 0;
7470 }
7471
7472 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7473 {
7474         struct nfs4_sequence_data *calldata = data;
7475         struct nfs_client *clp = calldata->clp;
7476
7477         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7478                 return;
7479
7480         trace_nfs4_sequence(clp, task->tk_status);
7481         if (task->tk_status < 0) {
7482                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7483                 if (atomic_read(&clp->cl_count) == 1)
7484                         goto out;
7485
7486                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7487                         rpc_restart_call_prepare(task);
7488                         return;
7489                 }
7490         }
7491         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7492 out:
7493         dprintk("<-- %s\n", __func__);
7494 }
7495
7496 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7497 {
7498         struct nfs4_sequence_data *calldata = data;
7499         struct nfs_client *clp = calldata->clp;
7500         struct nfs4_sequence_args *args;
7501         struct nfs4_sequence_res *res;
7502
7503         args = task->tk_msg.rpc_argp;
7504         res = task->tk_msg.rpc_resp;
7505
7506         nfs41_setup_sequence(clp->cl_session, args, res, task);
7507 }
7508
7509 static const struct rpc_call_ops nfs41_sequence_ops = {
7510         .rpc_call_done = nfs41_sequence_call_done,
7511         .rpc_call_prepare = nfs41_sequence_prepare,
7512         .rpc_release = nfs41_sequence_release,
7513 };
7514
7515 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7516                 struct rpc_cred *cred,
7517                 bool is_privileged)
7518 {
7519         struct nfs4_sequence_data *calldata;
7520         struct rpc_message msg = {
7521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7522                 .rpc_cred = cred,
7523         };
7524         struct rpc_task_setup task_setup_data = {
7525                 .rpc_client = clp->cl_rpcclient,
7526                 .rpc_message = &msg,
7527                 .callback_ops = &nfs41_sequence_ops,
7528                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7529         };
7530
7531         if (!atomic_inc_not_zero(&clp->cl_count))
7532                 return ERR_PTR(-EIO);
7533         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7534         if (calldata == NULL) {
7535                 nfs_put_client(clp);
7536                 return ERR_PTR(-ENOMEM);
7537         }
7538         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7539         if (is_privileged)
7540                 nfs4_set_sequence_privileged(&calldata->args);
7541         msg.rpc_argp = &calldata->args;
7542         msg.rpc_resp = &calldata->res;
7543         calldata->clp = clp;
7544         task_setup_data.callback_data = calldata;
7545
7546         return rpc_run_task(&task_setup_data);
7547 }
7548
7549 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7550 {
7551         struct rpc_task *task;
7552         int ret = 0;
7553
7554         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7555                 return -EAGAIN;
7556         task = _nfs41_proc_sequence(clp, cred, false);
7557         if (IS_ERR(task))
7558                 ret = PTR_ERR(task);
7559         else
7560                 rpc_put_task_async(task);
7561         dprintk("<-- %s status=%d\n", __func__, ret);
7562         return ret;
7563 }
7564
7565 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7566 {
7567         struct rpc_task *task;
7568         int ret;
7569
7570         task = _nfs41_proc_sequence(clp, cred, true);
7571         if (IS_ERR(task)) {
7572                 ret = PTR_ERR(task);
7573                 goto out;
7574         }
7575         ret = rpc_wait_for_completion_task(task);
7576         if (!ret) {
7577                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7578
7579                 if (task->tk_status == 0)
7580                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7581                 ret = task->tk_status;
7582         }
7583         rpc_put_task(task);
7584 out:
7585         dprintk("<-- %s status=%d\n", __func__, ret);
7586         return ret;
7587 }
7588
7589 struct nfs4_reclaim_complete_data {
7590         struct nfs_client *clp;
7591         struct nfs41_reclaim_complete_args arg;
7592         struct nfs41_reclaim_complete_res res;
7593 };
7594
7595 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7596 {
7597         struct nfs4_reclaim_complete_data *calldata = data;
7598
7599         nfs41_setup_sequence(calldata->clp->cl_session,
7600                         &calldata->arg.seq_args,
7601                         &calldata->res.seq_res,
7602                         task);
7603 }
7604
7605 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7606 {
7607         switch(task->tk_status) {
7608         case 0:
7609         case -NFS4ERR_COMPLETE_ALREADY:
7610         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7611                 break;
7612         case -NFS4ERR_DELAY:
7613                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7614                 /* fall through */
7615         case -NFS4ERR_RETRY_UNCACHED_REP:
7616                 return -EAGAIN;
7617         default:
7618                 nfs4_schedule_lease_recovery(clp);
7619         }
7620         return 0;
7621 }
7622
7623 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7624 {
7625         struct nfs4_reclaim_complete_data *calldata = data;
7626         struct nfs_client *clp = calldata->clp;
7627         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7628
7629         dprintk("--> %s\n", __func__);
7630         if (!nfs41_sequence_done(task, res))
7631                 return;
7632
7633         trace_nfs4_reclaim_complete(clp, task->tk_status);
7634         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7635                 rpc_restart_call_prepare(task);
7636                 return;
7637         }
7638         dprintk("<-- %s\n", __func__);
7639 }
7640
7641 static void nfs4_free_reclaim_complete_data(void *data)
7642 {
7643         struct nfs4_reclaim_complete_data *calldata = data;
7644
7645         kfree(calldata);
7646 }
7647
7648 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7649         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7650         .rpc_call_done = nfs4_reclaim_complete_done,
7651         .rpc_release = nfs4_free_reclaim_complete_data,
7652 };
7653
7654 /*
7655  * Issue a global reclaim complete.
7656  */
7657 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7658                 struct rpc_cred *cred)
7659 {
7660         struct nfs4_reclaim_complete_data *calldata;
7661         struct rpc_task *task;
7662         struct rpc_message msg = {
7663                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7664                 .rpc_cred = cred,
7665         };
7666         struct rpc_task_setup task_setup_data = {
7667                 .rpc_client = clp->cl_rpcclient,
7668                 .rpc_message = &msg,
7669                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7670                 .flags = RPC_TASK_ASYNC,
7671         };
7672         int status = -ENOMEM;
7673
7674         dprintk("--> %s\n", __func__);
7675         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7676         if (calldata == NULL)
7677                 goto out;
7678         calldata->clp = clp;
7679         calldata->arg.one_fs = 0;
7680
7681         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7682         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7683         msg.rpc_argp = &calldata->arg;
7684         msg.rpc_resp = &calldata->res;
7685         task_setup_data.callback_data = calldata;
7686         task = rpc_run_task(&task_setup_data);
7687         if (IS_ERR(task)) {
7688                 status = PTR_ERR(task);
7689                 goto out;
7690         }
7691         status = nfs4_wait_for_completion_rpc_task(task);
7692         if (status == 0)
7693                 status = task->tk_status;
7694         rpc_put_task(task);
7695         return 0;
7696 out:
7697         dprintk("<-- %s status=%d\n", __func__, status);
7698         return status;
7699 }
7700
7701 static void
7702 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7703 {
7704         struct nfs4_layoutget *lgp = calldata;
7705         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7706         struct nfs4_session *session = nfs4_get_session(server);
7707
7708         dprintk("--> %s\n", __func__);
7709         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7710          * right now covering the LAYOUTGET we are about to send.
7711          * However, that is not so catastrophic, and there seems
7712          * to be no way to prevent it completely.
7713          */
7714         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7715                                 &lgp->res.seq_res, task))
7716                 return;
7717         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7718                                           NFS_I(lgp->args.inode)->layout,
7719                                           &lgp->args.range,
7720                                           lgp->args.ctx->state)) {
7721                 rpc_exit(task, NFS4_OK);
7722         }
7723 }
7724
7725 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7726 {
7727         struct nfs4_layoutget *lgp = calldata;
7728         struct inode *inode = lgp->args.inode;
7729         struct nfs_server *server = NFS_SERVER(inode);
7730         struct pnfs_layout_hdr *lo;
7731         struct nfs4_state *state = NULL;
7732         unsigned long timeo, now, giveup;
7733
7734         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7735
7736         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7737                 goto out;
7738
7739         switch (task->tk_status) {
7740         case 0:
7741                 goto out;
7742         /*
7743          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7744          * (or clients) writing to the same RAID stripe
7745          */
7746         case -NFS4ERR_LAYOUTTRYLATER:
7747         /*
7748          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7749          * existing layout before getting a new one).
7750          */
7751         case -NFS4ERR_RECALLCONFLICT:
7752                 timeo = rpc_get_timeout(task->tk_client);
7753                 giveup = lgp->args.timestamp + timeo;
7754                 now = jiffies;
7755                 if (time_after(giveup, now)) {
7756                         unsigned long delay;
7757
7758                         /* Delay for:
7759                          * - Not less then NFS4_POLL_RETRY_MIN.
7760                          * - One last time a jiffie before we give up
7761                          * - exponential backoff (time_now minus start_attempt)
7762                          */
7763                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7764                                     min((giveup - now - 1),
7765                                         now - lgp->args.timestamp));
7766
7767                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7768                                 __func__, delay);
7769                         rpc_delay(task, delay);
7770                         task->tk_status = 0;
7771                         rpc_restart_call_prepare(task);
7772                         goto out; /* Do not call nfs4_async_handle_error() */
7773                 }
7774                 break;
7775         case -NFS4ERR_EXPIRED:
7776         case -NFS4ERR_BAD_STATEID:
7777                 spin_lock(&inode->i_lock);
7778                 lo = NFS_I(inode)->layout;
7779                 if (!lo || list_empty(&lo->plh_segs)) {
7780                         spin_unlock(&inode->i_lock);
7781                         /* If the open stateid was bad, then recover it. */
7782                         state = lgp->args.ctx->state;
7783                 } else {
7784                         LIST_HEAD(head);
7785
7786                         /*
7787                          * Mark the bad layout state as invalid, then retry
7788                          * with the current stateid.
7789                          */
7790                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7791                         spin_unlock(&inode->i_lock);
7792                         pnfs_free_lseg_list(&head);
7793
7794                         task->tk_status = 0;
7795                         rpc_restart_call_prepare(task);
7796                 }
7797         }
7798         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7799                 rpc_restart_call_prepare(task);
7800 out:
7801         dprintk("<-- %s\n", __func__);
7802 }
7803
7804 static size_t max_response_pages(struct nfs_server *server)
7805 {
7806         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7807         return nfs_page_array_len(0, max_resp_sz);
7808 }
7809
7810 static void nfs4_free_pages(struct page **pages, size_t size)
7811 {
7812         int i;
7813
7814         if (!pages)
7815                 return;
7816
7817         for (i = 0; i < size; i++) {
7818                 if (!pages[i])
7819                         break;
7820                 __free_page(pages[i]);
7821         }
7822         kfree(pages);
7823 }
7824
7825 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7826 {
7827         struct page **pages;
7828         int i;
7829
7830         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7831         if (!pages) {
7832                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7833                 return NULL;
7834         }
7835
7836         for (i = 0; i < size; i++) {
7837                 pages[i] = alloc_page(gfp_flags);
7838                 if (!pages[i]) {
7839                         dprintk("%s: failed to allocate page\n", __func__);
7840                         nfs4_free_pages(pages, size);
7841                         return NULL;
7842                 }
7843         }
7844
7845         return pages;
7846 }
7847
7848 static void nfs4_layoutget_release(void *calldata)
7849 {
7850         struct nfs4_layoutget *lgp = calldata;
7851         struct inode *inode = lgp->args.inode;
7852         struct nfs_server *server = NFS_SERVER(inode);
7853         size_t max_pages = max_response_pages(server);
7854
7855         dprintk("--> %s\n", __func__);
7856         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7857         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7858         put_nfs_open_context(lgp->args.ctx);
7859         kfree(calldata);
7860         dprintk("<-- %s\n", __func__);
7861 }
7862
7863 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7864         .rpc_call_prepare = nfs4_layoutget_prepare,
7865         .rpc_call_done = nfs4_layoutget_done,
7866         .rpc_release = nfs4_layoutget_release,
7867 };
7868
7869 struct pnfs_layout_segment *
7870 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7871 {
7872         struct inode *inode = lgp->args.inode;
7873         struct nfs_server *server = NFS_SERVER(inode);
7874         size_t max_pages = max_response_pages(server);
7875         struct rpc_task *task;
7876         struct rpc_message msg = {
7877                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7878                 .rpc_argp = &lgp->args,
7879                 .rpc_resp = &lgp->res,
7880                 .rpc_cred = lgp->cred,
7881         };
7882         struct rpc_task_setup task_setup_data = {
7883                 .rpc_client = server->client,
7884                 .rpc_message = &msg,
7885                 .callback_ops = &nfs4_layoutget_call_ops,
7886                 .callback_data = lgp,
7887                 .flags = RPC_TASK_ASYNC,
7888         };
7889         struct pnfs_layout_segment *lseg = NULL;
7890         int status = 0;
7891
7892         dprintk("--> %s\n", __func__);
7893
7894         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7895         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7896
7897         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7898         if (!lgp->args.layout.pages) {
7899                 nfs4_layoutget_release(lgp);
7900                 return ERR_PTR(-ENOMEM);
7901         }
7902         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7903         lgp->args.timestamp = jiffies;
7904
7905         lgp->res.layoutp = &lgp->args.layout;
7906         lgp->res.seq_res.sr_slot = NULL;
7907         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7908
7909         task = rpc_run_task(&task_setup_data);
7910         if (IS_ERR(task))
7911                 return ERR_CAST(task);
7912         status = nfs4_wait_for_completion_rpc_task(task);
7913         if (status == 0)
7914                 status = task->tk_status;
7915         trace_nfs4_layoutget(lgp->args.ctx,
7916                         &lgp->args.range,
7917                         &lgp->res.range,
7918                         status);
7919         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7920         if (status == 0 && lgp->res.layoutp->len)
7921                 lseg = pnfs_layout_process(lgp);
7922         rpc_put_task(task);
7923         dprintk("<-- %s status=%d\n", __func__, status);
7924         if (status)
7925                 return ERR_PTR(status);
7926         return lseg;
7927 }
7928
7929 static void
7930 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7931 {
7932         struct nfs4_layoutreturn *lrp = calldata;
7933
7934         dprintk("--> %s\n", __func__);
7935         nfs41_setup_sequence(lrp->clp->cl_session,
7936                         &lrp->args.seq_args,
7937                         &lrp->res.seq_res,
7938                         task);
7939 }
7940
7941 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7942 {
7943         struct nfs4_layoutreturn *lrp = calldata;
7944         struct nfs_server *server;
7945
7946         dprintk("--> %s\n", __func__);
7947
7948         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7949                 return;
7950
7951         server = NFS_SERVER(lrp->args.inode);
7952         switch (task->tk_status) {
7953         default:
7954                 task->tk_status = 0;
7955         case 0:
7956                 break;
7957         case -NFS4ERR_DELAY:
7958                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7959                         break;
7960                 rpc_restart_call_prepare(task);
7961                 return;
7962         }
7963         dprintk("<-- %s\n", __func__);
7964 }
7965
7966 static void nfs4_layoutreturn_release(void *calldata)
7967 {
7968         struct nfs4_layoutreturn *lrp = calldata;
7969         struct pnfs_layout_hdr *lo = lrp->args.layout;
7970
7971         dprintk("--> %s\n", __func__);
7972         spin_lock(&lo->plh_inode->i_lock);
7973         if (lrp->res.lrs_present)
7974                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7975         pnfs_clear_layoutreturn_waitbit(lo);
7976         clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7977         rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7978         lo->plh_block_lgets--;
7979         spin_unlock(&lo->plh_inode->i_lock);
7980         pnfs_put_layout_hdr(lrp->args.layout);
7981         nfs_iput_and_deactive(lrp->inode);
7982         kfree(calldata);
7983         dprintk("<-- %s\n", __func__);
7984 }
7985
7986 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7987         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7988         .rpc_call_done = nfs4_layoutreturn_done,
7989         .rpc_release = nfs4_layoutreturn_release,
7990 };
7991
7992 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7993 {
7994         struct rpc_task *task;
7995         struct rpc_message msg = {
7996                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7997                 .rpc_argp = &lrp->args,
7998                 .rpc_resp = &lrp->res,
7999                 .rpc_cred = lrp->cred,
8000         };
8001         struct rpc_task_setup task_setup_data = {
8002                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8003                 .rpc_message = &msg,
8004                 .callback_ops = &nfs4_layoutreturn_call_ops,
8005                 .callback_data = lrp,
8006         };
8007         int status = 0;
8008
8009         dprintk("--> %s\n", __func__);
8010         if (!sync) {
8011                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8012                 if (!lrp->inode) {
8013                         nfs4_layoutreturn_release(lrp);
8014                         return -EAGAIN;
8015                 }
8016                 task_setup_data.flags |= RPC_TASK_ASYNC;
8017         }
8018         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8019         task = rpc_run_task(&task_setup_data);
8020         if (IS_ERR(task))
8021                 return PTR_ERR(task);
8022         if (sync)
8023                 status = task->tk_status;
8024         trace_nfs4_layoutreturn(lrp->args.inode, status);
8025         dprintk("<-- %s status=%d\n", __func__, status);
8026         rpc_put_task(task);
8027         return status;
8028 }
8029
8030 static int
8031 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8032                 struct pnfs_device *pdev,
8033                 struct rpc_cred *cred)
8034 {
8035         struct nfs4_getdeviceinfo_args args = {
8036                 .pdev = pdev,
8037                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8038                         NOTIFY_DEVICEID4_DELETE,
8039         };
8040         struct nfs4_getdeviceinfo_res res = {
8041                 .pdev = pdev,
8042         };
8043         struct rpc_message msg = {
8044                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8045                 .rpc_argp = &args,
8046                 .rpc_resp = &res,
8047                 .rpc_cred = cred,
8048         };
8049         int status;
8050
8051         dprintk("--> %s\n", __func__);
8052         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8053         if (res.notification & ~args.notify_types)
8054                 dprintk("%s: unsupported notification\n", __func__);
8055         if (res.notification != args.notify_types)
8056                 pdev->nocache = 1;
8057
8058         dprintk("<-- %s status=%d\n", __func__, status);
8059
8060         return status;
8061 }
8062
8063 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8064                 struct pnfs_device *pdev,
8065                 struct rpc_cred *cred)
8066 {
8067         struct nfs4_exception exception = { };
8068         int err;
8069
8070         do {
8071                 err = nfs4_handle_exception(server,
8072                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8073                                         &exception);
8074         } while (exception.retry);
8075         return err;
8076 }
8077 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8078
8079 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8080 {
8081         struct nfs4_layoutcommit_data *data = calldata;
8082         struct nfs_server *server = NFS_SERVER(data->args.inode);
8083         struct nfs4_session *session = nfs4_get_session(server);
8084
8085         nfs41_setup_sequence(session,
8086                         &data->args.seq_args,
8087                         &data->res.seq_res,
8088                         task);
8089 }
8090
8091 static void
8092 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8093 {
8094         struct nfs4_layoutcommit_data *data = calldata;
8095         struct nfs_server *server = NFS_SERVER(data->args.inode);
8096
8097         if (!nfs41_sequence_done(task, &data->res.seq_res))
8098                 return;
8099
8100         switch (task->tk_status) { /* Just ignore these failures */
8101         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8102         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8103         case -NFS4ERR_BADLAYOUT:     /* no layout */
8104         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8105                 task->tk_status = 0;
8106         case 0:
8107                 break;
8108         default:
8109                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8110                         rpc_restart_call_prepare(task);
8111                         return;
8112                 }
8113         }
8114 }
8115
8116 static void nfs4_layoutcommit_release(void *calldata)
8117 {
8118         struct nfs4_layoutcommit_data *data = calldata;
8119
8120         pnfs_cleanup_layoutcommit(data);
8121         nfs_post_op_update_inode_force_wcc(data->args.inode,
8122                                            data->res.fattr);
8123         put_rpccred(data->cred);
8124         nfs_iput_and_deactive(data->inode);
8125         kfree(data);
8126 }
8127
8128 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8129         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8130         .rpc_call_done = nfs4_layoutcommit_done,
8131         .rpc_release = nfs4_layoutcommit_release,
8132 };
8133
8134 int
8135 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8136 {
8137         struct rpc_message msg = {
8138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8139                 .rpc_argp = &data->args,
8140                 .rpc_resp = &data->res,
8141                 .rpc_cred = data->cred,
8142         };
8143         struct rpc_task_setup task_setup_data = {
8144                 .task = &data->task,
8145                 .rpc_client = NFS_CLIENT(data->args.inode),
8146                 .rpc_message = &msg,
8147                 .callback_ops = &nfs4_layoutcommit_ops,
8148                 .callback_data = data,
8149         };
8150         struct rpc_task *task;
8151         int status = 0;
8152
8153         dprintk("NFS: initiating layoutcommit call. sync %d "
8154                 "lbw: %llu inode %lu\n", sync,
8155                 data->args.lastbytewritten,
8156                 data->args.inode->i_ino);
8157
8158         if (!sync) {
8159                 data->inode = nfs_igrab_and_active(data->args.inode);
8160                 if (data->inode == NULL) {
8161                         nfs4_layoutcommit_release(data);
8162                         return -EAGAIN;
8163                 }
8164                 task_setup_data.flags = RPC_TASK_ASYNC;
8165         }
8166         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8167         task = rpc_run_task(&task_setup_data);
8168         if (IS_ERR(task))
8169                 return PTR_ERR(task);
8170         if (sync)
8171                 status = task->tk_status;
8172         trace_nfs4_layoutcommit(data->args.inode, status);
8173         dprintk("%s: status %d\n", __func__, status);
8174         rpc_put_task(task);
8175         return status;
8176 }
8177
8178 /**
8179  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8180  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8181  */
8182 static int
8183 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8184                     struct nfs_fsinfo *info,
8185                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8186 {
8187         struct nfs41_secinfo_no_name_args args = {
8188                 .style = SECINFO_STYLE_CURRENT_FH,
8189         };
8190         struct nfs4_secinfo_res res = {
8191                 .flavors = flavors,
8192         };
8193         struct rpc_message msg = {
8194                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8195                 .rpc_argp = &args,
8196                 .rpc_resp = &res,
8197         };
8198         struct rpc_clnt *clnt = server->client;
8199         struct rpc_cred *cred = NULL;
8200         int status;
8201
8202         if (use_integrity) {
8203                 clnt = server->nfs_client->cl_rpcclient;
8204                 cred = nfs4_get_clid_cred(server->nfs_client);
8205                 msg.rpc_cred = cred;
8206         }
8207
8208         dprintk("--> %s\n", __func__);
8209         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8210                                 &res.seq_res, 0);
8211         dprintk("<-- %s status=%d\n", __func__, status);
8212
8213         if (cred)
8214                 put_rpccred(cred);
8215
8216         return status;
8217 }
8218
8219 static int
8220 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8221                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8222 {
8223         struct nfs4_exception exception = { };
8224         int err;
8225         do {
8226                 /* first try using integrity protection */
8227                 err = -NFS4ERR_WRONGSEC;
8228
8229                 /* try to use integrity protection with machine cred */
8230                 if (_nfs4_is_integrity_protected(server->nfs_client))
8231                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8232                                                           flavors, true);
8233
8234                 /*
8235                  * if unable to use integrity protection, or SECINFO with
8236                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8237                  * disallowed by spec, but exists in deployed servers) use
8238                  * the current filesystem's rpc_client and the user cred.
8239                  */
8240                 if (err == -NFS4ERR_WRONGSEC)
8241                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8242                                                           flavors, false);
8243
8244                 switch (err) {
8245                 case 0:
8246                 case -NFS4ERR_WRONGSEC:
8247                 case -ENOTSUPP:
8248                         goto out;
8249                 default:
8250                         err = nfs4_handle_exception(server, err, &exception);
8251                 }
8252         } while (exception.retry);
8253 out:
8254         return err;
8255 }
8256
8257 static int
8258 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8259                     struct nfs_fsinfo *info)
8260 {
8261         int err;
8262         struct page *page;
8263         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8264         struct nfs4_secinfo_flavors *flavors;
8265         struct nfs4_secinfo4 *secinfo;
8266         int i;
8267
8268         page = alloc_page(GFP_KERNEL);
8269         if (!page) {
8270                 err = -ENOMEM;
8271                 goto out;
8272         }
8273
8274         flavors = page_address(page);
8275         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8276
8277         /*
8278          * Fall back on "guess and check" method if
8279          * the server doesn't support SECINFO_NO_NAME
8280          */
8281         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8282                 err = nfs4_find_root_sec(server, fhandle, info);
8283                 goto out_freepage;
8284         }
8285         if (err)
8286                 goto out_freepage;
8287
8288         for (i = 0; i < flavors->num_flavors; i++) {
8289                 secinfo = &flavors->flavors[i];
8290
8291                 switch (secinfo->flavor) {
8292                 case RPC_AUTH_NULL:
8293                 case RPC_AUTH_UNIX:
8294                 case RPC_AUTH_GSS:
8295                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8296                                         &secinfo->flavor_info);
8297                         break;
8298                 default:
8299                         flavor = RPC_AUTH_MAXFLAVOR;
8300                         break;
8301                 }
8302
8303                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8304                         flavor = RPC_AUTH_MAXFLAVOR;
8305
8306                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8307                         err = nfs4_lookup_root_sec(server, fhandle,
8308                                                    info, flavor);
8309                         if (!err)
8310                                 break;
8311                 }
8312         }
8313
8314         if (flavor == RPC_AUTH_MAXFLAVOR)
8315                 err = -EPERM;
8316
8317 out_freepage:
8318         put_page(page);
8319         if (err == -EACCES)
8320                 return -EPERM;
8321 out:
8322         return err;
8323 }
8324
8325 static int _nfs41_test_stateid(struct nfs_server *server,
8326                 nfs4_stateid *stateid,
8327                 struct rpc_cred *cred)
8328 {
8329         int status;
8330         struct nfs41_test_stateid_args args = {
8331                 .stateid = stateid,
8332         };
8333         struct nfs41_test_stateid_res res;
8334         struct rpc_message msg = {
8335                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8336                 .rpc_argp = &args,
8337                 .rpc_resp = &res,
8338                 .rpc_cred = cred,
8339         };
8340         struct rpc_clnt *rpc_client = server->client;
8341
8342         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8343                 &rpc_client, &msg);
8344
8345         dprintk("NFS call  test_stateid %p\n", stateid);
8346         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8347         nfs4_set_sequence_privileged(&args.seq_args);
8348         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8349                         &args.seq_args, &res.seq_res);
8350         if (status != NFS_OK) {
8351                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8352                 return status;
8353         }
8354         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8355         return -res.status;
8356 }
8357
8358 /**
8359  * nfs41_test_stateid - perform a TEST_STATEID operation
8360  *
8361  * @server: server / transport on which to perform the operation
8362  * @stateid: state ID to test
8363  * @cred: credential
8364  *
8365  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8366  * Otherwise a negative NFS4ERR value is returned if the operation
8367  * failed or the state ID is not currently valid.
8368  */
8369 static int nfs41_test_stateid(struct nfs_server *server,
8370                 nfs4_stateid *stateid,
8371                 struct rpc_cred *cred)
8372 {
8373         struct nfs4_exception exception = { };
8374         int err;
8375         do {
8376                 err = _nfs41_test_stateid(server, stateid, cred);
8377                 if (err != -NFS4ERR_DELAY)
8378                         break;
8379                 nfs4_handle_exception(server, err, &exception);
8380         } while (exception.retry);
8381         return err;
8382 }
8383
8384 struct nfs_free_stateid_data {
8385         struct nfs_server *server;
8386         struct nfs41_free_stateid_args args;
8387         struct nfs41_free_stateid_res res;
8388 };
8389
8390 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8391 {
8392         struct nfs_free_stateid_data *data = calldata;
8393         nfs41_setup_sequence(nfs4_get_session(data->server),
8394                         &data->args.seq_args,
8395                         &data->res.seq_res,
8396                         task);
8397 }
8398
8399 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8400 {
8401         struct nfs_free_stateid_data *data = calldata;
8402
8403         nfs41_sequence_done(task, &data->res.seq_res);
8404
8405         switch (task->tk_status) {
8406         case -NFS4ERR_DELAY:
8407                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8408                         rpc_restart_call_prepare(task);
8409         }
8410 }
8411
8412 static void nfs41_free_stateid_release(void *calldata)
8413 {
8414         kfree(calldata);
8415 }
8416
8417 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8418         .rpc_call_prepare = nfs41_free_stateid_prepare,
8419         .rpc_call_done = nfs41_free_stateid_done,
8420         .rpc_release = nfs41_free_stateid_release,
8421 };
8422
8423 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8424                 nfs4_stateid *stateid,
8425                 struct rpc_cred *cred,
8426                 bool privileged)
8427 {
8428         struct rpc_message msg = {
8429                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8430                 .rpc_cred = cred,
8431         };
8432         struct rpc_task_setup task_setup = {
8433                 .rpc_client = server->client,
8434                 .rpc_message = &msg,
8435                 .callback_ops = &nfs41_free_stateid_ops,
8436                 .flags = RPC_TASK_ASYNC,
8437         };
8438         struct nfs_free_stateid_data *data;
8439
8440         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8441                 &task_setup.rpc_client, &msg);
8442
8443         dprintk("NFS call  free_stateid %p\n", stateid);
8444         data = kmalloc(sizeof(*data), GFP_NOFS);
8445         if (!data)
8446                 return ERR_PTR(-ENOMEM);
8447         data->server = server;
8448         nfs4_stateid_copy(&data->args.stateid, stateid);
8449
8450         task_setup.callback_data = data;
8451
8452         msg.rpc_argp = &data->args;
8453         msg.rpc_resp = &data->res;
8454         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8455         if (privileged)
8456                 nfs4_set_sequence_privileged(&data->args.seq_args);
8457
8458         return rpc_run_task(&task_setup);
8459 }
8460
8461 /**
8462  * nfs41_free_stateid - perform a FREE_STATEID operation
8463  *
8464  * @server: server / transport on which to perform the operation
8465  * @stateid: state ID to release
8466  * @cred: credential
8467  *
8468  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8469  * negative NFS4ERR value is returned.
8470  */
8471 static int nfs41_free_stateid(struct nfs_server *server,
8472                 nfs4_stateid *stateid,
8473                 struct rpc_cred *cred)
8474 {
8475         struct rpc_task *task;
8476         int ret;
8477
8478         task = _nfs41_free_stateid(server, stateid, cred, true);
8479         if (IS_ERR(task))
8480                 return PTR_ERR(task);
8481         ret = rpc_wait_for_completion_task(task);
8482         if (!ret)
8483                 ret = task->tk_status;
8484         rpc_put_task(task);
8485         return ret;
8486 }
8487
8488 static void
8489 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8490 {
8491         struct rpc_task *task;
8492         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8493
8494         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8495         nfs4_free_lock_state(server, lsp);
8496         if (IS_ERR(task))
8497                 return;
8498         rpc_put_task(task);
8499 }
8500
8501 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8502                 const nfs4_stateid *s2)
8503 {
8504         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8505                 return false;
8506
8507         if (s1->seqid == s2->seqid)
8508                 return true;
8509         if (s1->seqid == 0 || s2->seqid == 0)
8510                 return true;
8511
8512         return false;
8513 }
8514
8515 #endif /* CONFIG_NFS_V4_1 */
8516
8517 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8518                 const nfs4_stateid *s2)
8519 {
8520         return nfs4_stateid_match(s1, s2);
8521 }
8522
8523
8524 static const struct nfs4_state_recovery_ops nfs40_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 = nfs4_init_clientid,
8530         .detect_trunking = nfs40_discover_server_trunking,
8531 };
8532
8533 #if defined(CONFIG_NFS_V4_1)
8534 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8535         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8536         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8537         .recover_open   = nfs4_open_reclaim,
8538         .recover_lock   = nfs4_lock_reclaim,
8539         .establish_clid = nfs41_init_clientid,
8540         .reclaim_complete = nfs41_proc_reclaim_complete,
8541         .detect_trunking = nfs41_discover_server_trunking,
8542 };
8543 #endif /* CONFIG_NFS_V4_1 */
8544
8545 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8546         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8547         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8548         .recover_open   = nfs40_open_expired,
8549         .recover_lock   = nfs4_lock_expired,
8550         .establish_clid = nfs4_init_clientid,
8551 };
8552
8553 #if defined(CONFIG_NFS_V4_1)
8554 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8555         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8556         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8557         .recover_open   = nfs41_open_expired,
8558         .recover_lock   = nfs41_lock_expired,
8559         .establish_clid = nfs41_init_clientid,
8560 };
8561 #endif /* CONFIG_NFS_V4_1 */
8562
8563 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8564         .sched_state_renewal = nfs4_proc_async_renew,
8565         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8566         .renew_lease = nfs4_proc_renew,
8567 };
8568
8569 #if defined(CONFIG_NFS_V4_1)
8570 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8571         .sched_state_renewal = nfs41_proc_async_sequence,
8572         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8573         .renew_lease = nfs4_proc_sequence,
8574 };
8575 #endif
8576
8577 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8578         .get_locations = _nfs40_proc_get_locations,
8579         .fsid_present = _nfs40_proc_fsid_present,
8580 };
8581
8582 #if defined(CONFIG_NFS_V4_1)
8583 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8584         .get_locations = _nfs41_proc_get_locations,
8585         .fsid_present = _nfs41_proc_fsid_present,
8586 };
8587 #endif  /* CONFIG_NFS_V4_1 */
8588
8589 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8590         .minor_version = 0,
8591         .init_caps = NFS_CAP_READDIRPLUS
8592                 | NFS_CAP_ATOMIC_OPEN
8593                 | NFS_CAP_CHANGE_ATTR
8594                 | NFS_CAP_POSIX_LOCK,
8595         .init_client = nfs40_init_client,
8596         .shutdown_client = nfs40_shutdown_client,
8597         .match_stateid = nfs4_match_stateid,
8598         .find_root_sec = nfs4_find_root_sec,
8599         .free_lock_state = nfs4_release_lockowner,
8600         .alloc_seqid = nfs_alloc_seqid,
8601         .call_sync_ops = &nfs40_call_sync_ops,
8602         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8603         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8604         .state_renewal_ops = &nfs40_state_renewal_ops,
8605         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8606 };
8607
8608 #if defined(CONFIG_NFS_V4_1)
8609 static struct nfs_seqid *
8610 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8611 {
8612         return NULL;
8613 }
8614
8615 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8616         .minor_version = 1,
8617         .init_caps = NFS_CAP_READDIRPLUS
8618                 | NFS_CAP_ATOMIC_OPEN
8619                 | NFS_CAP_CHANGE_ATTR
8620                 | NFS_CAP_POSIX_LOCK
8621                 | NFS_CAP_STATEID_NFSV41
8622                 | NFS_CAP_ATOMIC_OPEN_V1,
8623         .init_client = nfs41_init_client,
8624         .shutdown_client = nfs41_shutdown_client,
8625         .match_stateid = nfs41_match_stateid,
8626         .find_root_sec = nfs41_find_root_sec,
8627         .free_lock_state = nfs41_free_lock_state,
8628         .alloc_seqid = nfs_alloc_no_seqid,
8629         .call_sync_ops = &nfs41_call_sync_ops,
8630         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8631         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8632         .state_renewal_ops = &nfs41_state_renewal_ops,
8633         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8634 };
8635 #endif
8636
8637 #if defined(CONFIG_NFS_V4_2)
8638 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8639         .minor_version = 2,
8640         .init_caps = NFS_CAP_READDIRPLUS
8641                 | NFS_CAP_ATOMIC_OPEN
8642                 | NFS_CAP_CHANGE_ATTR
8643                 | NFS_CAP_POSIX_LOCK
8644                 | NFS_CAP_STATEID_NFSV41
8645                 | NFS_CAP_ATOMIC_OPEN_V1
8646                 | NFS_CAP_ALLOCATE
8647                 | NFS_CAP_DEALLOCATE
8648                 | NFS_CAP_SEEK
8649                 | NFS_CAP_LAYOUTSTATS,
8650         .init_client = nfs41_init_client,
8651         .shutdown_client = nfs41_shutdown_client,
8652         .match_stateid = nfs41_match_stateid,
8653         .find_root_sec = nfs41_find_root_sec,
8654         .free_lock_state = nfs41_free_lock_state,
8655         .call_sync_ops = &nfs41_call_sync_ops,
8656         .alloc_seqid = nfs_alloc_no_seqid,
8657         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8658         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8659         .state_renewal_ops = &nfs41_state_renewal_ops,
8660 };
8661 #endif
8662
8663 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8664         [0] = &nfs_v4_0_minor_ops,
8665 #if defined(CONFIG_NFS_V4_1)
8666         [1] = &nfs_v4_1_minor_ops,
8667 #endif
8668 #if defined(CONFIG_NFS_V4_2)
8669         [2] = &nfs_v4_2_minor_ops,
8670 #endif
8671 };
8672
8673 static const struct inode_operations nfs4_dir_inode_operations = {
8674         .create         = nfs_create,
8675         .lookup         = nfs_lookup,
8676         .atomic_open    = nfs_atomic_open,
8677         .link           = nfs_link,
8678         .unlink         = nfs_unlink,
8679         .symlink        = nfs_symlink,
8680         .mkdir          = nfs_mkdir,
8681         .rmdir          = nfs_rmdir,
8682         .mknod          = nfs_mknod,
8683         .rename         = nfs_rename,
8684         .permission     = nfs_permission,
8685         .getattr        = nfs_getattr,
8686         .setattr        = nfs_setattr,
8687         .getxattr       = generic_getxattr,
8688         .setxattr       = generic_setxattr,
8689         .listxattr      = generic_listxattr,
8690         .removexattr    = generic_removexattr,
8691 };
8692
8693 static const struct inode_operations nfs4_file_inode_operations = {
8694         .permission     = nfs_permission,
8695         .getattr        = nfs_getattr,
8696         .setattr        = nfs_setattr,
8697         .getxattr       = generic_getxattr,
8698         .setxattr       = generic_setxattr,
8699         .listxattr      = generic_listxattr,
8700         .removexattr    = generic_removexattr,
8701 };
8702
8703 const struct nfs_rpc_ops nfs_v4_clientops = {
8704         .version        = 4,                    /* protocol version */
8705         .dentry_ops     = &nfs4_dentry_operations,
8706         .dir_inode_ops  = &nfs4_dir_inode_operations,
8707         .file_inode_ops = &nfs4_file_inode_operations,
8708         .file_ops       = &nfs4_file_operations,
8709         .getroot        = nfs4_proc_get_root,
8710         .submount       = nfs4_submount,
8711         .try_mount      = nfs4_try_mount,
8712         .getattr        = nfs4_proc_getattr,
8713         .setattr        = nfs4_proc_setattr,
8714         .lookup         = nfs4_proc_lookup,
8715         .access         = nfs4_proc_access,
8716         .readlink       = nfs4_proc_readlink,
8717         .create         = nfs4_proc_create,
8718         .remove         = nfs4_proc_remove,
8719         .unlink_setup   = nfs4_proc_unlink_setup,
8720         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8721         .unlink_done    = nfs4_proc_unlink_done,
8722         .rename_setup   = nfs4_proc_rename_setup,
8723         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8724         .rename_done    = nfs4_proc_rename_done,
8725         .link           = nfs4_proc_link,
8726         .symlink        = nfs4_proc_symlink,
8727         .mkdir          = nfs4_proc_mkdir,
8728         .rmdir          = nfs4_proc_remove,
8729         .readdir        = nfs4_proc_readdir,
8730         .mknod          = nfs4_proc_mknod,
8731         .statfs         = nfs4_proc_statfs,
8732         .fsinfo         = nfs4_proc_fsinfo,
8733         .pathconf       = nfs4_proc_pathconf,
8734         .set_capabilities = nfs4_server_capabilities,
8735         .decode_dirent  = nfs4_decode_dirent,
8736         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8737         .read_setup     = nfs4_proc_read_setup,
8738         .read_done      = nfs4_read_done,
8739         .write_setup    = nfs4_proc_write_setup,
8740         .write_done     = nfs4_write_done,
8741         .commit_setup   = nfs4_proc_commit_setup,
8742         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8743         .commit_done    = nfs4_commit_done,
8744         .lock           = nfs4_proc_lock,
8745         .clear_acl_cache = nfs4_zap_acl_attr,
8746         .close_context  = nfs4_close_context,
8747         .open_context   = nfs4_atomic_open,
8748         .have_delegation = nfs4_have_delegation,
8749         .return_delegation = nfs4_inode_return_delegation,
8750         .alloc_client   = nfs4_alloc_client,
8751         .init_client    = nfs4_init_client,
8752         .free_client    = nfs4_free_client,
8753         .create_server  = nfs4_create_server,
8754         .clone_server   = nfs_clone_server,
8755 };
8756
8757 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8758         .prefix = XATTR_NAME_NFSV4_ACL,
8759         .list   = nfs4_xattr_list_nfs4_acl,
8760         .get    = nfs4_xattr_get_nfs4_acl,
8761         .set    = nfs4_xattr_set_nfs4_acl,
8762 };
8763
8764 const struct xattr_handler *nfs4_xattr_handlers[] = {
8765         &nfs4_xattr_nfs4_acl_handler,
8766 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8767         &nfs4_xattr_nfs4_label_handler,
8768 #endif
8769         NULL
8770 };
8771
8772 /*
8773  * Local variables:
8774  *  c-basic-offset: 8
8775  * End:
8776  */