fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
  82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  86                             struct nfs_fattr *fattr, struct iattr *sattr,
  87                             struct nfs4_state *state, struct nfs4_label *ilabel,
  88                             struct nfs4_label *olabel);
  89 #ifdef CONFIG_NFS_V4_1
  90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  91                 struct rpc_cred *);
  92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  93                 struct rpc_cred *);
  94 #endif
  95
  96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  97 static inline struct nfs4_label *
  98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  99         struct iattr *sattr, struct nfs4_label *label)
 100 {
 101         int err;
 102
 103         if (label == NULL)
 104                 return NULL;
 105
 106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 107                 return NULL;
 108
 109         err = security_dentry_init_security(dentry, sattr->ia_mode,
 110                                 &dentry->d_name, (void **)&label->label, &label->len);
 111         if (err == 0)
 112                 return label;
 113
 114         return NULL;
 115 }
 116 static inline void
 117 nfs4_label_release_security(struct nfs4_label *label)
 118 {
 119         if (label)
 120                 security_release_secctx(label->label, label->len);
 121 }
 122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 123 {
 124         if (label)
 125                 return server->attr_bitmask;
 126
 127         return server->attr_bitmask_nl;
 128 }
 129 #else
 130 static inline struct nfs4_label *
 131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 132         struct iattr *sattr, struct nfs4_label *l)
 133 { return NULL; }
 134 static inline void
 135 nfs4_label_release_security(struct nfs4_label *label)
 136 { return; }
 137 static inline u32 *
 138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 139 { return server->attr_bitmask; }
 140 #endif
 141
 142 /* Prevent leaks of NFSv4 errors into userland */
 143 static int nfs4_map_errors(int err)
 144 {
 145         if (err >= -1000)
 146                 return err;
 147         switch (err) {
 148         case -NFS4ERR_RESOURCE:
 149         case -NFS4ERR_LAYOUTTRYLATER:
 150         case -NFS4ERR_RECALLCONFLICT:
 151                 return -EREMOTEIO;
 152         case -NFS4ERR_WRONGSEC:
 153         case -NFS4ERR_WRONG_CRED:
 154                 return -EPERM;
 155         case -NFS4ERR_BADOWNER:
 156         case -NFS4ERR_BADNAME:
 157                 return -EINVAL;
 158         case -NFS4ERR_SHARE_DENIED:
 159                 return -EACCES;
 160         case -NFS4ERR_MINOR_VERS_MISMATCH:
 161                 return -EPROTONOSUPPORT;
 162         case -NFS4ERR_FILE_OPEN:
 163                 return -EBUSY;
 164         default:
 165                 dprintk("%s could not handle NFSv4 error %d\n",
 166                                 __func__, -err);
 167                 break;
 168         }
 169         return -EIO;
 170 }
 171
 172 /*
 173  * This is our standard bitmap for GETATTR requests.
 174  */
 175 const u32 nfs4_fattr_bitmap[3] = {
 176         FATTR4_WORD0_TYPE
 177         | FATTR4_WORD0_CHANGE
 178         | FATTR4_WORD0_SIZE
 179         | FATTR4_WORD0_FSID
 180         | FATTR4_WORD0_FILEID,
 181         FATTR4_WORD1_MODE
 182         | FATTR4_WORD1_NUMLINKS
 183         | FATTR4_WORD1_OWNER
 184         | FATTR4_WORD1_OWNER_GROUP
 185         | FATTR4_WORD1_RAWDEV
 186         | FATTR4_WORD1_SPACE_USED
 187         | FATTR4_WORD1_TIME_ACCESS
 188         | FATTR4_WORD1_TIME_METADATA
 189         | FATTR4_WORD1_TIME_MODIFY
 190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 192         FATTR4_WORD2_SECURITY_LABEL
 193 #endif
 194 };
 195
 196 static const u32 nfs4_pnfs_open_bitmap[3] = {
 197         FATTR4_WORD0_TYPE
 198         | FATTR4_WORD0_CHANGE
 199         | FATTR4_WORD0_SIZE
 200         | FATTR4_WORD0_FSID
 201         | FATTR4_WORD0_FILEID,
 202         FATTR4_WORD1_MODE
 203         | FATTR4_WORD1_NUMLINKS
 204         | FATTR4_WORD1_OWNER
 205         | FATTR4_WORD1_OWNER_GROUP
 206         | FATTR4_WORD1_RAWDEV
 207         | FATTR4_WORD1_SPACE_USED
 208         | FATTR4_WORD1_TIME_ACCESS
 209         | FATTR4_WORD1_TIME_METADATA
 210         | FATTR4_WORD1_TIME_MODIFY,
 211         FATTR4_WORD2_MDSTHRESHOLD
 212 };
 213
 214 static const u32 nfs4_open_noattr_bitmap[3] = {
 215         FATTR4_WORD0_TYPE
 216         | FATTR4_WORD0_CHANGE
 217         | FATTR4_WORD0_FILEID,
 218 };
 219
 220 const u32 nfs4_statfs_bitmap[3] = {
 221         FATTR4_WORD0_FILES_AVAIL
 222         | FATTR4_WORD0_FILES_FREE
 223         | FATTR4_WORD0_FILES_TOTAL,
 224         FATTR4_WORD1_SPACE_AVAIL
 225         | FATTR4_WORD1_SPACE_FREE
 226         | FATTR4_WORD1_SPACE_TOTAL
 227 };
 228
 229 const u32 nfs4_pathconf_bitmap[3] = {
 230         FATTR4_WORD0_MAXLINK
 231         | FATTR4_WORD0_MAXNAME,
 232         0
 233 };
 234
 235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 236                         | FATTR4_WORD0_MAXREAD
 237                         | FATTR4_WORD0_MAXWRITE
 238                         | FATTR4_WORD0_LEASE_TIME,
 239                         FATTR4_WORD1_TIME_DELTA
 240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 241                         FATTR4_WORD2_LAYOUT_BLKSIZE
 242 };
 243
 244 const u32 nfs4_fs_locations_bitmap[3] = {
 245         FATTR4_WORD0_TYPE
 246         | FATTR4_WORD0_CHANGE
 247         | FATTR4_WORD0_SIZE
 248         | FATTR4_WORD0_FSID
 249         | FATTR4_WORD0_FILEID
 250         | FATTR4_WORD0_FS_LOCATIONS,
 251         FATTR4_WORD1_MODE
 252         | FATTR4_WORD1_NUMLINKS
 253         | FATTR4_WORD1_OWNER
 254         | FATTR4_WORD1_OWNER_GROUP
 255         | FATTR4_WORD1_RAWDEV
 256         | FATTR4_WORD1_SPACE_USED
 257         | FATTR4_WORD1_TIME_ACCESS
 258         | FATTR4_WORD1_TIME_METADATA
 259         | FATTR4_WORD1_TIME_MODIFY
 260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 261 };
 262
 263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 264                 struct nfs4_readdir_arg *readdir)
 265 {
 266         __be32 *start, *p;
 267
 268         if (cookie > 2) {
 269                 readdir->cookie = cookie;
 270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 271                 return;
 272         }
 273
 274         readdir->cookie = 0;
 275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 276         if (cookie == 2)
 277                 return;
 278
 279         /*
 280          * NFSv4 servers do not return entries for '.' and '..'
 281          * Therefore, we fake these entries here.  We let '.'
 282          * have cookie 0 and '..' have cookie 1.  Note that
 283          * when talking to the server, we always send cookie 0
 284          * instead of 1 or 2.
 285          */
 286         start = p = kmap_atomic(*readdir->pages);
 287
 288         if (cookie == 0) {
 289                 *p++ = xdr_one;                                  /* next */
 290                 *p++ = xdr_zero;                   /* cookie, first word */
 291                 *p++ = xdr_one;                   /* cookie, second word */
 292                 *p++ = xdr_one;                             /* entry len */
 293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 294                 p++;
 295                 *p++ = xdr_one;                         /* bitmap length */
 296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 297                 *p++ = htonl(8);              /* attribute buffer length */
 298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 299         }
 300
 301         *p++ = xdr_one;                                  /* next */
 302         *p++ = xdr_zero;                   /* cookie, first word */
 303         *p++ = xdr_two;                   /* cookie, second word */
 304         *p++ = xdr_two;                             /* entry len */
 305         memcpy(p, "..\0\0", 4);                         /* entry */
 306         p++;
 307         *p++ = xdr_one;                         /* bitmap length */
 308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309         *p++ = htonl(8);              /* attribute buffer length */
 310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 311
 312         readdir->pgbase = (char *)p - (char *)start;
 313         readdir->count -= readdir->pgbase;
 314         kunmap_atomic(start);
 315 }
 316
 317 static long nfs4_update_delay(long *timeout)
 318 {
 319         long ret;
 320         if (!timeout)
 321                 return NFS4_POLL_RETRY_MAX;
 322         if (*timeout <= 0)
 323                 *timeout = NFS4_POLL_RETRY_MIN;
 324         if (*timeout > NFS4_POLL_RETRY_MAX)
 325                 *timeout = NFS4_POLL_RETRY_MAX;
 326         ret = *timeout;
 327         *timeout <<= 1;
 328         return ret;
 329 }
 330
 331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 332 {
 333         int res = 0;
 334
 335         might_sleep();
 336
 337         freezable_schedule_timeout_killable_unsafe(
 338                 nfs4_update_delay(timeout));
 339         if (fatal_signal_pending(current))
 340                 res = -ERESTARTSYS;
 341         return res;
 342 }
 343
 344 /* This is the error handling routine for processes that are allowed
 345  * to sleep.
 346  */
 347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 348 {
 349         struct nfs_client *clp = server->nfs_client;
 350         struct nfs4_state *state = exception->state;
 351         struct inode *inode = exception->inode;
 352         int ret = errorcode;
 353
 354         exception->retry = 0;
 355         switch(errorcode) {
 356                 case 0:
 357                         return 0;
 358                 case -NFS4ERR_OPENMODE:
 359                 case -NFS4ERR_DELEG_REVOKED:
 360                 case -NFS4ERR_ADMIN_REVOKED:
 361                 case -NFS4ERR_BAD_STATEID:
 362                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 363                                 nfs4_inode_return_delegation(inode);
 364                                 exception->retry = 1;
 365                                 return 0;
 366                         }
 367                         if (state == NULL)
 368                                 break;
 369                         ret = nfs4_schedule_stateid_recovery(server, state);
 370                         if (ret < 0)
 371                                 break;
 372                         goto wait_on_recovery;
 373                 case -NFS4ERR_EXPIRED:
 374                         if (state != NULL) {
 375                                 ret = nfs4_schedule_stateid_recovery(server, state);
 376                                 if (ret < 0)
 377                                         break;
 378                         }
 379                 case -NFS4ERR_STALE_STATEID:
 380                 case -NFS4ERR_STALE_CLIENTID:
 381                         nfs4_schedule_lease_recovery(clp);
 382                         goto wait_on_recovery;
 383                 case -NFS4ERR_MOVED:
 384                         ret = nfs4_schedule_migration_recovery(server);
 385                         if (ret < 0)
 386                                 break;
 387                         goto wait_on_recovery;
 388                 case -NFS4ERR_LEASE_MOVED:
 389                         nfs4_schedule_lease_moved_recovery(clp);
 390                         goto wait_on_recovery;
 391 #if defined(CONFIG_NFS_V4_1)
 392                 case -NFS4ERR_BADSESSION:
 393                 case -NFS4ERR_BADSLOT:
 394                 case -NFS4ERR_BAD_HIGH_SLOT:
 395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 396                 case -NFS4ERR_DEADSESSION:
 397                 case -NFS4ERR_SEQ_FALSE_RETRY:
 398                 case -NFS4ERR_SEQ_MISORDERED:
 399                         dprintk("%s ERROR: %d Reset session\n", __func__,
 400                                 errorcode);
 401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 402                         goto wait_on_recovery;
 403 #endif /* defined(CONFIG_NFS_V4_1) */
 404                 case -NFS4ERR_FILE_OPEN:
 405                         if (exception->timeout > HZ) {
 406                                 /* We have retried a decent amount, time to
 407                                  * fail
 408                                  */
 409                                 ret = -EBUSY;
 410                                 break;
 411                         }
 412                 case -NFS4ERR_GRACE:
 413                 case -NFS4ERR_DELAY:
 414                         ret = nfs4_delay(server->client, &exception->timeout);
 415                         if (ret != 0)
 416                                 break;
 417                 case -NFS4ERR_RETRY_UNCACHED_REP:
 418                 case -NFS4ERR_OLD_STATEID:
 419                         exception->retry = 1;
 420                         break;
 421                 case -NFS4ERR_BADOWNER:
 422                         /* The following works around a Linux server bug! */
 423                 case -NFS4ERR_BADNAME:
 424                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 425                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 426                                 exception->retry = 1;
 427                                 printk(KERN_WARNING "NFS: v4 server %s "
 428                                                 "does not accept raw "
 429                                                 "uid/gids. "
 430                                                 "Reenabling the idmapper.\n",
 431                                                 server->nfs_client->cl_hostname);
 432                         }
 433         }
 434         /* We failed to handle the error */
 435         return nfs4_map_errors(ret);
 436 wait_on_recovery:
 437         ret = nfs4_wait_clnt_recover(clp);
 438         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 439                 return -EIO;
 440         if (ret == 0)
 441                 exception->retry = 1;
 442         return ret;
 443 }
 444
 445 /*
 446  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 447  * or 'false' otherwise.
 448  */
 449 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 450 {
 451         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 452
 453         if (flavor == RPC_AUTH_GSS_KRB5I ||
 454             flavor == RPC_AUTH_GSS_KRB5P)
 455                 return true;
 456
 457         return false;
 458 }
 459
 460 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 461 {
 462         spin_lock(&clp->cl_lock);
 463         if (time_before(clp->cl_last_renewal,timestamp))
 464                 clp->cl_last_renewal = timestamp;
 465         spin_unlock(&clp->cl_lock);
 466 }
 467
 468 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 469 {
 470         struct nfs_client *clp = server->nfs_client;
 471
 472         if (!nfs4_has_session(clp))
 473                 do_renew_lease(clp, timestamp);
 474 }
 475
 476 struct nfs4_call_sync_data {
 477         const struct nfs_server *seq_server;
 478         struct nfs4_sequence_args *seq_args;
 479         struct nfs4_sequence_res *seq_res;
 480 };
 481
 482 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 483                         struct nfs4_sequence_res *res, int cache_reply)
 484 {
 485         args->sa_slot = NULL;
 486         args->sa_cache_this = cache_reply;
 487         args->sa_privileged = 0;
 488
 489         res->sr_slot = NULL;
 490 }
 491
 492 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 493 {
 494         args->sa_privileged = 1;
 495 }
 496
 497 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 498                          struct nfs4_sequence_args *args,
 499                          struct nfs4_sequence_res *res,
 500                          struct rpc_task *task)
 501 {
 502         struct nfs4_slot *slot;
 503
 504         /* slot already allocated? */
 505         if (res->sr_slot != NULL)
 506                 goto out_start;
 507
 508         spin_lock(&tbl->slot_tbl_lock);
 509         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 510                 goto out_sleep;
 511
 512         slot = nfs4_alloc_slot(tbl);
 513         if (IS_ERR(slot)) {
 514                 if (slot == ERR_PTR(-ENOMEM))
 515                         task->tk_timeout = HZ >> 2;
 516                 goto out_sleep;
 517         }
 518         spin_unlock(&tbl->slot_tbl_lock);
 519
 520         args->sa_slot = slot;
 521         res->sr_slot = slot;
 522
 523 out_start:
 524         rpc_call_start(task);
 525         return 0;
 526
 527 out_sleep:
 528         if (args->sa_privileged)
 529                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 530                                 NULL, RPC_PRIORITY_PRIVILEGED);
 531         else
 532                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 533         spin_unlock(&tbl->slot_tbl_lock);
 534         return -EAGAIN;
 535 }
 536 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 537
 538 static int nfs40_sequence_done(struct rpc_task *task,
 539                                struct nfs4_sequence_res *res)
 540 {
 541         struct nfs4_slot *slot = res->sr_slot;
 542         struct nfs4_slot_table *tbl;
 543
 544         if (slot == NULL)
 545                 goto out;
 546
 547         tbl = slot->table;
 548         spin_lock(&tbl->slot_tbl_lock);
 549         if (!nfs41_wake_and_assign_slot(tbl, slot))
 550                 nfs4_free_slot(tbl, slot);
 551         spin_unlock(&tbl->slot_tbl_lock);
 552
 553         res->sr_slot = NULL;
 554 out:
 555         return 1;
 556 }
 557
 558 #if defined(CONFIG_NFS_V4_1)
 559
 560 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 561 {
 562         struct nfs4_session *session;
 563         struct nfs4_slot_table *tbl;
 564         struct nfs4_slot *slot = res->sr_slot;
 565         bool send_new_highest_used_slotid = false;
 566
 567         tbl = slot->table;
 568         session = tbl->session;
 569
 570         spin_lock(&tbl->slot_tbl_lock);
 571         /* Be nice to the server: try to ensure that the last transmitted
 572          * value for highest_user_slotid <= target_highest_slotid
 573          */
 574         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 575                 send_new_highest_used_slotid = true;
 576
 577         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 578                 send_new_highest_used_slotid = false;
 579                 goto out_unlock;
 580         }
 581         nfs4_free_slot(tbl, slot);
 582
 583         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 584                 send_new_highest_used_slotid = false;
 585 out_unlock:
 586         spin_unlock(&tbl->slot_tbl_lock);
 587         res->sr_slot = NULL;
 588         if (send_new_highest_used_slotid)
 589                 nfs41_notify_server(session->clp);
 590 }
 591
 592 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 593 {
 594         struct nfs4_session *session;
 595         struct nfs4_slot *slot = res->sr_slot;
 596         struct nfs_client *clp;
 597         bool interrupted = false;
 598         int ret = 1;
 599
 600         if (slot == NULL)
 601                 goto out_noaction;
 602         /* don't increment the sequence number if the task wasn't sent */
 603         if (!RPC_WAS_SENT(task))
 604                 goto out;
 605
 606         session = slot->table->session;
 607
 608         if (slot->interrupted) {
 609                 slot->interrupted = 0;
 610                 interrupted = true;
 611         }
 612
 613         trace_nfs4_sequence_done(session, res);
 614         /* Check the SEQUENCE operation status */
 615         switch (res->sr_status) {
 616         case 0:
 617                 /* Update the slot's sequence and clientid lease timer */
 618                 ++slot->seq_nr;
 619                 clp = session->clp;
 620                 do_renew_lease(clp, res->sr_timestamp);
 621                 /* Check sequence flags */
 622                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 623                 nfs41_update_target_slotid(slot->table, slot, res);
 624                 break;
 625         case 1:
 626                 /*
 627                  * sr_status remains 1 if an RPC level error occurred.
 628                  * The server may or may not have processed the sequence
 629                  * operation..
 630                  * Mark the slot as having hosted an interrupted RPC call.
 631                  */
 632                 slot->interrupted = 1;
 633                 goto out;
 634         case -NFS4ERR_DELAY:
 635                 /* The server detected a resend of the RPC call and
 636                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 637                  * of RFC5661.
 638                  */
 639                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 640                         __func__,
 641                         slot->slot_nr,
 642                         slot->seq_nr);
 643                 goto out_retry;
 644         case -NFS4ERR_BADSLOT:
 645                 /*
 646                  * The slot id we used was probably retired. Try again
 647                  * using a different slot id.
 648                  */
 649                 goto retry_nowait;
 650         case -NFS4ERR_SEQ_MISORDERED:
 651                 /*
 652                  * Was the last operation on this sequence interrupted?
 653                  * If so, retry after bumping the sequence number.
 654                  */
 655                 if (interrupted) {
 656                         ++slot->seq_nr;
 657                         goto retry_nowait;
 658                 }
 659                 /*
 660                  * Could this slot have been previously retired?
 661                  * If so, then the server may be expecting seq_nr = 1!
 662                  */
 663                 if (slot->seq_nr != 1) {
 664                         slot->seq_nr = 1;
 665                         goto retry_nowait;
 666                 }
 667                 break;
 668         case -NFS4ERR_SEQ_FALSE_RETRY:
 669                 ++slot->seq_nr;
 670                 goto retry_nowait;
 671         default:
 672                 /* Just update the slot sequence no. */
 673                 ++slot->seq_nr;
 674         }
 675 out:
 676         /* The session may be reset by one of the error handlers. */
 677         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 678         nfs41_sequence_free_slot(res);
 679 out_noaction:
 680         return ret;
 681 retry_nowait:
 682         if (rpc_restart_call_prepare(task)) {
 683                 task->tk_status = 0;
 684                 ret = 0;
 685         }
 686         goto out;
 687 out_retry:
 688         if (!rpc_restart_call(task))
 689                 goto out;
 690         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 691         return 0;
 692 }
 693 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 694
 695 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 696 {
 697         if (res->sr_slot == NULL)
 698                 return 1;
 699         if (!res->sr_slot->table->session)
 700                 return nfs40_sequence_done(task, res);
 701         return nfs41_sequence_done(task, res);
 702 }
 703 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 704
 705 int nfs41_setup_sequence(struct nfs4_session *session,
 706                                 struct nfs4_sequence_args *args,
 707                                 struct nfs4_sequence_res *res,
 708                                 struct rpc_task *task)
 709 {
 710         struct nfs4_slot *slot;
 711         struct nfs4_slot_table *tbl;
 712
 713         dprintk("--> %s\n", __func__);
 714         /* slot already allocated? */
 715         if (res->sr_slot != NULL)
 716                 goto out_success;
 717
 718         tbl = &session->fc_slot_table;
 719
 720         task->tk_timeout = 0;
 721
 722         spin_lock(&tbl->slot_tbl_lock);
 723         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 724             !args->sa_privileged) {
 725                 /* The state manager will wait until the slot table is empty */
 726                 dprintk("%s session is draining\n", __func__);
 727                 goto out_sleep;
 728         }
 729
 730         slot = nfs4_alloc_slot(tbl);
 731         if (IS_ERR(slot)) {
 732                 /* If out of memory, try again in 1/4 second */
 733                 if (slot == ERR_PTR(-ENOMEM))
 734                         task->tk_timeout = HZ >> 2;
 735                 dprintk("<-- %s: no free slots\n", __func__);
 736                 goto out_sleep;
 737         }
 738         spin_unlock(&tbl->slot_tbl_lock);
 739
 740         args->sa_slot = slot;
 741
 742         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 743                         slot->slot_nr, slot->seq_nr);
 744
 745         res->sr_slot = slot;
 746         res->sr_timestamp = jiffies;
 747         res->sr_status_flags = 0;
 748         /*
 749          * sr_status is only set in decode_sequence, and so will remain
 750          * set to 1 if an rpc level failure occurs.
 751          */
 752         res->sr_status = 1;
 753         trace_nfs4_setup_sequence(session, args);
 754 out_success:
 755         rpc_call_start(task);
 756         return 0;
 757 out_sleep:
 758         /* Privileged tasks are queued with top priority */
 759         if (args->sa_privileged)
 760                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 761                                 NULL, RPC_PRIORITY_PRIVILEGED);
 762         else
 763                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 764         spin_unlock(&tbl->slot_tbl_lock);
 765         return -EAGAIN;
 766 }
 767 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 768
 769 static int nfs4_setup_sequence(const struct nfs_server *server,
 770                                struct nfs4_sequence_args *args,
 771                                struct nfs4_sequence_res *res,
 772                                struct rpc_task *task)
 773 {
 774         struct nfs4_session *session = nfs4_get_session(server);
 775         int ret = 0;
 776
 777         if (!session)
 778                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 779                                             args, res, task);
 780
 781         dprintk("--> %s clp %p session %p sr_slot %u\n",
 782                 __func__, session->clp, session, res->sr_slot ?
 783                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 784
 785         ret = nfs41_setup_sequence(session, args, res, task);
 786
 787         dprintk("<-- %s status=%d\n", __func__, ret);
 788         return ret;
 789 }
 790
 791 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 792 {
 793         struct nfs4_call_sync_data *data = calldata;
 794         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 795
 796         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 797
 798         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 799 }
 800
 801 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 802 {
 803         struct nfs4_call_sync_data *data = calldata;
 804
 805         nfs41_sequence_done(task, data->seq_res);
 806 }
 807
 808 static const struct rpc_call_ops nfs41_call_sync_ops = {
 809         .rpc_call_prepare = nfs41_call_sync_prepare,
 810         .rpc_call_done = nfs41_call_sync_done,
 811 };
 812
 813 #else   /* !CONFIG_NFS_V4_1 */
 814
 815 static int nfs4_setup_sequence(const struct nfs_server *server,
 816                                struct nfs4_sequence_args *args,
 817                                struct nfs4_sequence_res *res,
 818                                struct rpc_task *task)
 819 {
 820         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 821                                     args, res, task);
 822 }
 823
 824 int nfs4_sequence_done(struct rpc_task *task,
 825                        struct nfs4_sequence_res *res)
 826 {
 827         return nfs40_sequence_done(task, res);
 828 }
 829 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 830
 831 #endif  /* !CONFIG_NFS_V4_1 */
 832
 833 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 834 {
 835         struct nfs4_call_sync_data *data = calldata;
 836         nfs4_setup_sequence(data->seq_server,
 837                                 data->seq_args, data->seq_res, task);
 838 }
 839
 840 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 841 {
 842         struct nfs4_call_sync_data *data = calldata;
 843         nfs4_sequence_done(task, data->seq_res);
 844 }
 845
 846 static const struct rpc_call_ops nfs40_call_sync_ops = {
 847         .rpc_call_prepare = nfs40_call_sync_prepare,
 848         .rpc_call_done = nfs40_call_sync_done,
 849 };
 850
 851 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 852                                    struct nfs_server *server,
 853                                    struct rpc_message *msg,
 854                                    struct nfs4_sequence_args *args,
 855                                    struct nfs4_sequence_res *res)
 856 {
 857         int ret;
 858         struct rpc_task *task;
 859         struct nfs_client *clp = server->nfs_client;
 860         struct nfs4_call_sync_data data = {
 861                 .seq_server = server,
 862                 .seq_args = args,
 863                 .seq_res = res,
 864         };
 865         struct rpc_task_setup task_setup = {
 866                 .rpc_client = clnt,
 867                 .rpc_message = msg,
 868                 .callback_ops = clp->cl_mvops->call_sync_ops,
 869                 .callback_data = &data
 870         };
 871
 872         task = rpc_run_task(&task_setup);
 873         if (IS_ERR(task))
 874                 ret = PTR_ERR(task);
 875         else {
 876                 ret = task->tk_status;
 877                 rpc_put_task(task);
 878         }
 879         return ret;
 880 }
 881
 882 int nfs4_call_sync(struct rpc_clnt *clnt,
 883                    struct nfs_server *server,
 884                    struct rpc_message *msg,
 885                    struct nfs4_sequence_args *args,
 886                    struct nfs4_sequence_res *res,
 887                    int cache_reply)
 888 {
 889         nfs4_init_sequence(args, res, cache_reply);
 890         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 891 }
 892
 893 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 894 {
 895         struct nfs_inode *nfsi = NFS_I(dir);
 896
 897         spin_lock(&dir->i_lock);
 898         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 899         if (!cinfo->atomic || cinfo->before != dir->i_version)
 900                 nfs_force_lookup_revalidate(dir);
 901         dir->i_version = cinfo->after;
 902         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 903         nfs_fscache_invalidate(dir);
 904         spin_unlock(&dir->i_lock);
 905 }
 906
 907 struct nfs4_opendata {
 908         struct kref kref;
 909         struct nfs_openargs o_arg;
 910         struct nfs_openres o_res;
 911         struct nfs_open_confirmargs c_arg;
 912         struct nfs_open_confirmres c_res;
 913         struct nfs4_string owner_name;
 914         struct nfs4_string group_name;
 915         struct nfs4_label *a_label;
 916         struct nfs_fattr f_attr;
 917         struct nfs4_label *f_label;
 918         struct dentry *dir;
 919         struct dentry *dentry;
 920         struct nfs4_state_owner *owner;
 921         struct nfs4_state *state;
 922         struct iattr attrs;
 923         unsigned long timestamp;
 924         unsigned int rpc_done : 1;
 925         unsigned int file_created : 1;
 926         unsigned int is_recover : 1;
 927         int rpc_status;
 928         int cancelled;
 929 };
 930
 931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 932                 int err, struct nfs4_exception *exception)
 933 {
 934         if (err != -EINVAL)
 935                 return false;
 936         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 937                 return false;
 938         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 939         exception->retry = 1;
 940         return true;
 941 }
 942
 943 static u32
 944 nfs4_map_atomic_open_share(struct nfs_server *server,
 945                 fmode_t fmode, int openflags)
 946 {
 947         u32 res = 0;
 948
 949         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
 950         case FMODE_READ:
 951                 res = NFS4_SHARE_ACCESS_READ;
 952                 break;
 953         case FMODE_WRITE:
 954                 res = NFS4_SHARE_ACCESS_WRITE;
 955                 break;
 956         case FMODE_READ|FMODE_WRITE:
 957                 res = NFS4_SHARE_ACCESS_BOTH;
 958         }
 959         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 960                 goto out;
 961         /* Want no delegation if we're using O_DIRECT */
 962         if (openflags & O_DIRECT)
 963                 res |= NFS4_SHARE_WANT_NO_DELEG;
 964 out:
 965         return res;
 966 }
 967
 968 static enum open_claim_type4
 969 nfs4_map_atomic_open_claim(struct nfs_server *server,
 970                 enum open_claim_type4 claim)
 971 {
 972         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 973                 return claim;
 974         switch (claim) {
 975         default:
 976                 return claim;
 977         case NFS4_OPEN_CLAIM_FH:
 978                 return NFS4_OPEN_CLAIM_NULL;
 979         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 980                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 981         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 982                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 983         }
 984 }
 985
 986 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 987 {
 988         p->o_res.f_attr = &p->f_attr;
 989         p->o_res.f_label = p->f_label;
 990         p->o_res.seqid = p->o_arg.seqid;
 991         p->c_res.seqid = p->c_arg.seqid;
 992         p->o_res.server = p->o_arg.server;
 993         p->o_res.access_request = p->o_arg.access;
 994         nfs_fattr_init(&p->f_attr);
 995         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 996 }
 997
 998 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 999                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1000                 const struct iattr *attrs,
1001                 struct nfs4_label *label,
1002                 enum open_claim_type4 claim,
1003                 gfp_t gfp_mask)
1004 {
1005         struct dentry *parent = dget_parent(dentry);
1006         struct inode *dir = d_inode(parent);
1007         struct nfs_server *server = NFS_SERVER(dir);
1008         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1009         struct nfs4_opendata *p;
1010
1011         p = kzalloc(sizeof(*p), gfp_mask);
1012         if (p == NULL)
1013                 goto err;
1014
1015         p->f_label = nfs4_label_alloc(server, gfp_mask);
1016         if (IS_ERR(p->f_label))
1017                 goto err_free_p;
1018
1019         p->a_label = nfs4_label_alloc(server, gfp_mask);
1020         if (IS_ERR(p->a_label))
1021                 goto err_free_f;
1022
1023         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1024         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1025         if (IS_ERR(p->o_arg.seqid))
1026                 goto err_free_label;
1027         nfs_sb_active(dentry->d_sb);
1028         p->dentry = dget(dentry);
1029         p->dir = parent;
1030         p->owner = sp;
1031         atomic_inc(&sp->so_count);
1032         p->o_arg.open_flags = flags;
1033         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1034         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1035                         fmode, flags);
1036         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1037          * will return permission denied for all bits until close */
1038         if (!(flags & O_EXCL)) {
1039                 /* ask server to check for all possible rights as results
1040                  * are cached */
1041                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1042                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1043         }
1044         p->o_arg.clientid = server->nfs_client->cl_clientid;
1045         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1046         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1047         p->o_arg.name = &dentry->d_name;
1048         p->o_arg.server = server;
1049         p->o_arg.bitmask = nfs4_bitmask(server, label);
1050         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1051         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1052         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1053         switch (p->o_arg.claim) {
1054         case NFS4_OPEN_CLAIM_NULL:
1055         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1056         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1057                 p->o_arg.fh = NFS_FH(dir);
1058                 break;
1059         case NFS4_OPEN_CLAIM_PREVIOUS:
1060         case NFS4_OPEN_CLAIM_FH:
1061         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1062         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1063                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1064         }
1065         if (attrs != NULL && attrs->ia_valid != 0) {
1066                 __u32 verf[2];
1067
1068                 p->o_arg.u.attrs = &p->attrs;
1069                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1070
1071                 verf[0] = jiffies;
1072                 verf[1] = current->pid;
1073                 memcpy(p->o_arg.u.verifier.data, verf,
1074                                 sizeof(p->o_arg.u.verifier.data));
1075         }
1076         p->c_arg.fh = &p->o_res.fh;
1077         p->c_arg.stateid = &p->o_res.stateid;
1078         p->c_arg.seqid = p->o_arg.seqid;
1079         nfs4_init_opendata_res(p);
1080         kref_init(&p->kref);
1081         return p;
1082
1083 err_free_label:
1084         nfs4_label_free(p->a_label);
1085 err_free_f:
1086         nfs4_label_free(p->f_label);
1087 err_free_p:
1088         kfree(p);
1089 err:
1090         dput(parent);
1091         return NULL;
1092 }
1093
1094 static void nfs4_opendata_free(struct kref *kref)
1095 {
1096         struct nfs4_opendata *p = container_of(kref,
1097                         struct nfs4_opendata, kref);
1098         struct super_block *sb = p->dentry->d_sb;
1099
1100         nfs_free_seqid(p->o_arg.seqid);
1101         if (p->state != NULL)
1102                 nfs4_put_open_state(p->state);
1103         nfs4_put_state_owner(p->owner);
1104
1105         nfs4_label_free(p->a_label);
1106         nfs4_label_free(p->f_label);
1107
1108         dput(p->dir);
1109         dput(p->dentry);
1110         nfs_sb_deactive(sb);
1111         nfs_fattr_free_names(&p->f_attr);
1112         kfree(p->f_attr.mdsthreshold);
1113         kfree(p);
1114 }
1115
1116 static void nfs4_opendata_put(struct nfs4_opendata *p)
1117 {
1118         if (p != NULL)
1119                 kref_put(&p->kref, nfs4_opendata_free);
1120 }
1121
1122 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1123 {
1124         int ret;
1125
1126         ret = rpc_wait_for_completion_task(task);
1127         return ret;
1128 }
1129
1130 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1131 {
1132         int ret = 0;
1133
1134         if (open_mode & (O_EXCL|O_TRUNC))
1135                 goto out;
1136         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1137                 case FMODE_READ:
1138                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1139                                 && state->n_rdonly != 0;
1140                         break;
1141                 case FMODE_WRITE:
1142                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1143                                 && state->n_wronly != 0;
1144                         break;
1145                 case FMODE_READ|FMODE_WRITE:
1146                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1147                                 && state->n_rdwr != 0;
1148         }
1149 out:
1150         return ret;
1151 }
1152
1153 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1154 {
1155         if (delegation == NULL)
1156                 return 0;
1157         if ((delegation->type & fmode) != fmode)
1158                 return 0;
1159         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1160                 return 0;
1161         nfs_mark_delegation_referenced(delegation);
1162         return 1;
1163 }
1164
1165 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1166 {
1167         switch (fmode) {
1168                 case FMODE_WRITE:
1169                         state->n_wronly++;
1170                         break;
1171                 case FMODE_READ:
1172                         state->n_rdonly++;
1173                         break;
1174                 case FMODE_READ|FMODE_WRITE:
1175                         state->n_rdwr++;
1176         }
1177         nfs4_state_set_mode_locked(state, state->state | fmode);
1178 }
1179
1180 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1181 {
1182         struct nfs_client *clp = state->owner->so_server->nfs_client;
1183         bool need_recover = false;
1184
1185         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1186                 need_recover = true;
1187         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1188                 need_recover = true;
1189         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1190                 need_recover = true;
1191         if (need_recover)
1192                 nfs4_state_mark_reclaim_nograce(clp, state);
1193 }
1194
1195 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1196                 nfs4_stateid *stateid)
1197 {
1198         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1199                 return true;
1200         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1201                 nfs_test_and_clear_all_open_stateid(state);
1202                 return true;
1203         }
1204         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1205                 return true;
1206         return false;
1207 }
1208
1209 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1210 {
1211         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1212                 return;
1213         if (state->n_wronly)
1214                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1215         if (state->n_rdonly)
1216                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1217         if (state->n_rdwr)
1218                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1219         set_bit(NFS_OPEN_STATE, &state->flags);
1220 }
1221
1222 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1223                 nfs4_stateid *stateid, fmode_t fmode)
1224 {
1225         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1226         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1227         case FMODE_WRITE:
1228                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1229                 break;
1230         case FMODE_READ:
1231                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1232                 break;
1233         case 0:
1234                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1235                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1236                 clear_bit(NFS_OPEN_STATE, &state->flags);
1237         }
1238         if (stateid == NULL)
1239                 return;
1240         /* Handle races with OPEN */
1241         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1242             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1243                 nfs_resync_open_stateid_locked(state);
1244                 return;
1245         }
1246         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1247                 nfs4_stateid_copy(&state->stateid, stateid);
1248         nfs4_stateid_copy(&state->open_stateid, stateid);
1249 }
1250
1251 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1252 {
1253         write_seqlock(&state->seqlock);
1254         nfs_clear_open_stateid_locked(state, stateid, fmode);
1255         write_sequnlock(&state->seqlock);
1256         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1257                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1258 }
1259
1260 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1261 {
1262         switch (fmode) {
1263                 case FMODE_READ:
1264                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1265                         break;
1266                 case FMODE_WRITE:
1267                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1268                         break;
1269                 case FMODE_READ|FMODE_WRITE:
1270                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1271         }
1272         if (!nfs_need_update_open_stateid(state, stateid))
1273                 return;
1274         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1275                 nfs4_stateid_copy(&state->stateid, stateid);
1276         nfs4_stateid_copy(&state->open_stateid, stateid);
1277 }
1278
1279 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1280 {
1281         /*
1282          * Protect the call to nfs4_state_set_mode_locked and
1283          * serialise the stateid update
1284          */
1285         write_seqlock(&state->seqlock);
1286         if (deleg_stateid != NULL) {
1287                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1288                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1289         }
1290         if (open_stateid != NULL)
1291                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1292         write_sequnlock(&state->seqlock);
1293         spin_lock(&state->owner->so_lock);
1294         update_open_stateflags(state, fmode);
1295         spin_unlock(&state->owner->so_lock);
1296 }
1297
1298 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1299 {
1300         struct nfs_inode *nfsi = NFS_I(state->inode);
1301         struct nfs_delegation *deleg_cur;
1302         int ret = 0;
1303
1304         fmode &= (FMODE_READ|FMODE_WRITE);
1305
1306         rcu_read_lock();
1307         deleg_cur = rcu_dereference(nfsi->delegation);
1308         if (deleg_cur == NULL)
1309                 goto no_delegation;
1310
1311         spin_lock(&deleg_cur->lock);
1312         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1313            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1314             (deleg_cur->type & fmode) != fmode)
1315                 goto no_delegation_unlock;
1316
1317         if (delegation == NULL)
1318                 delegation = &deleg_cur->stateid;
1319         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1320                 goto no_delegation_unlock;
1321
1322         nfs_mark_delegation_referenced(deleg_cur);
1323         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1324         ret = 1;
1325 no_delegation_unlock:
1326         spin_unlock(&deleg_cur->lock);
1327 no_delegation:
1328         rcu_read_unlock();
1329
1330         if (!ret && open_stateid != NULL) {
1331                 __update_open_stateid(state, open_stateid, NULL, fmode);
1332                 ret = 1;
1333         }
1334         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1335                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1336
1337         return ret;
1338 }
1339
1340 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1341                 const nfs4_stateid *stateid)
1342 {
1343         struct nfs4_state *state = lsp->ls_state;
1344         bool ret = false;
1345
1346         spin_lock(&state->state_lock);
1347         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1348                 goto out_noupdate;
1349         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1350                 goto out_noupdate;
1351         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1352         ret = true;
1353 out_noupdate:
1354         spin_unlock(&state->state_lock);
1355         return ret;
1356 }
1357
1358 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1359 {
1360         struct nfs_delegation *delegation;
1361
1362         rcu_read_lock();
1363         delegation = rcu_dereference(NFS_I(inode)->delegation);
1364         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1365                 rcu_read_unlock();
1366                 return;
1367         }
1368         rcu_read_unlock();
1369         nfs4_inode_return_delegation(inode);
1370 }
1371
1372 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1373 {
1374         struct nfs4_state *state = opendata->state;
1375         struct nfs_inode *nfsi = NFS_I(state->inode);
1376         struct nfs_delegation *delegation;
1377         int open_mode = opendata->o_arg.open_flags;
1378         fmode_t fmode = opendata->o_arg.fmode;
1379         nfs4_stateid stateid;
1380         int ret = -EAGAIN;
1381
1382         for (;;) {
1383                 spin_lock(&state->owner->so_lock);
1384                 if (can_open_cached(state, fmode, open_mode)) {
1385                         update_open_stateflags(state, fmode);
1386                         spin_unlock(&state->owner->so_lock);
1387                         goto out_return_state;
1388                 }
1389                 spin_unlock(&state->owner->so_lock);
1390                 rcu_read_lock();
1391                 delegation = rcu_dereference(nfsi->delegation);
1392                 if (!can_open_delegated(delegation, fmode)) {
1393                         rcu_read_unlock();
1394                         break;
1395                 }
1396                 /* Save the delegation */
1397                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1398                 rcu_read_unlock();
1399                 nfs_release_seqid(opendata->o_arg.seqid);
1400                 if (!opendata->is_recover) {
1401                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1402                         if (ret != 0)
1403                                 goto out;
1404                 }
1405                 ret = -EAGAIN;
1406
1407                 /* Try to update the stateid using the delegation */
1408                 if (update_open_stateid(state, NULL, &stateid, fmode))
1409                         goto out_return_state;
1410         }
1411 out:
1412         return ERR_PTR(ret);
1413 out_return_state:
1414         atomic_inc(&state->count);
1415         return state;
1416 }
1417
1418 static void
1419 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1420 {
1421         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1422         struct nfs_delegation *delegation;
1423         int delegation_flags = 0;
1424
1425         rcu_read_lock();
1426         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1427         if (delegation)
1428                 delegation_flags = delegation->flags;
1429         rcu_read_unlock();
1430         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1431                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1432                                    "returning a delegation for "
1433                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1434                                    clp->cl_hostname);
1435         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1436                 nfs_inode_set_delegation(state->inode,
1437                                          data->owner->so_cred,
1438                                          &data->o_res);
1439         else
1440                 nfs_inode_reclaim_delegation(state->inode,
1441                                              data->owner->so_cred,
1442                                              &data->o_res);
1443 }
1444
1445 /*
1446  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1447  * and update the nfs4_state.
1448  */
1449 static struct nfs4_state *
1450 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1451 {
1452         struct inode *inode = data->state->inode;
1453         struct nfs4_state *state = data->state;
1454         int ret;
1455
1456         if (!data->rpc_done) {
1457                 if (data->rpc_status) {
1458                         ret = data->rpc_status;
1459                         goto err;
1460                 }
1461                 /* cached opens have already been processed */
1462                 goto update;
1463         }
1464
1465         ret = nfs_refresh_inode(inode, &data->f_attr);
1466         if (ret)
1467                 goto err;
1468
1469         if (data->o_res.delegation_type != 0)
1470                 nfs4_opendata_check_deleg(data, state);
1471 update:
1472         update_open_stateid(state, &data->o_res.stateid, NULL,
1473                             data->o_arg.fmode);
1474         atomic_inc(&state->count);
1475
1476         return state;
1477 err:
1478         return ERR_PTR(ret);
1479
1480 }
1481
1482 static struct nfs4_state *
1483 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1484 {
1485         struct inode *inode;
1486         struct nfs4_state *state = NULL;
1487         int ret;
1488
1489         if (!data->rpc_done) {
1490                 state = nfs4_try_open_cached(data);
1491                 goto out;
1492         }
1493
1494         ret = -EAGAIN;
1495         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1496                 goto err;
1497         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1498         ret = PTR_ERR(inode);
1499         if (IS_ERR(inode))
1500                 goto err;
1501         ret = -ENOMEM;
1502         state = nfs4_get_open_state(inode, data->owner);
1503         if (state == NULL)
1504                 goto err_put_inode;
1505         if (data->o_res.delegation_type != 0)
1506                 nfs4_opendata_check_deleg(data, state);
1507         update_open_stateid(state, &data->o_res.stateid, NULL,
1508                         data->o_arg.fmode);
1509         iput(inode);
1510 out:
1511         nfs_release_seqid(data->o_arg.seqid);
1512         return state;
1513 err_put_inode:
1514         iput(inode);
1515 err:
1516         return ERR_PTR(ret);
1517 }
1518
1519 static struct nfs4_state *
1520 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1521 {
1522         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1523                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1524         return _nfs4_opendata_to_nfs4_state(data);
1525 }
1526
1527 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1528 {
1529         struct nfs_inode *nfsi = NFS_I(state->inode);
1530         struct nfs_open_context *ctx;
1531
1532         spin_lock(&state->inode->i_lock);
1533         list_for_each_entry(ctx, &nfsi->open_files, list) {
1534                 if (ctx->state != state)
1535                         continue;
1536                 get_nfs_open_context(ctx);
1537                 spin_unlock(&state->inode->i_lock);
1538                 return ctx;
1539         }
1540         spin_unlock(&state->inode->i_lock);
1541         return ERR_PTR(-ENOENT);
1542 }
1543
1544 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1545                 struct nfs4_state *state, enum open_claim_type4 claim)
1546 {
1547         struct nfs4_opendata *opendata;
1548
1549         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1550                         NULL, NULL, claim, GFP_NOFS);
1551         if (opendata == NULL)
1552                 return ERR_PTR(-ENOMEM);
1553         opendata->state = state;
1554         atomic_inc(&state->count);
1555         return opendata;
1556 }
1557
1558 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1559 {
1560         struct nfs4_state *newstate;
1561         int ret;
1562
1563         if ((opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
1564              opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEG_CUR_FH) &&
1565             (opendata->o_arg.u.delegation_type & fmode) != fmode)
1566                 /* This mode can't have been delegated, so we must have
1567                  * a valid open_stateid to cover it - not need to reclaim.
1568                  */
1569                 return 0;
1570         opendata->o_arg.open_flags = 0;
1571         opendata->o_arg.fmode = fmode;
1572         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1573                         NFS_SB(opendata->dentry->d_sb),
1574                         fmode, 0);
1575         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1576         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1577         nfs4_init_opendata_res(opendata);
1578         ret = _nfs4_recover_proc_open(opendata);
1579         if (ret != 0)
1580                 return ret;
1581         newstate = nfs4_opendata_to_nfs4_state(opendata);
1582         if (IS_ERR(newstate))
1583                 return PTR_ERR(newstate);
1584         nfs4_close_state(newstate, fmode);
1585         *res = newstate;
1586         return 0;
1587 }
1588
1589 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1590 {
1591         struct nfs4_state *newstate;
1592         int ret;
1593
1594         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1595         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1596         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1597         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1598         /* memory barrier prior to reading state->n_* */
1599         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1600         clear_bit(NFS_OPEN_STATE, &state->flags);
1601         smp_rmb();
1602         if (state->n_rdwr != 0) {
1603                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1604                 if (ret != 0)
1605                         return ret;
1606                 if (newstate != state)
1607                         return -ESTALE;
1608         }
1609         if (state->n_wronly != 0) {
1610                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1611                 if (ret != 0)
1612                         return ret;
1613                 if (newstate != state)
1614                         return -ESTALE;
1615         }
1616         if (state->n_rdonly != 0) {
1617                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1618                 if (ret != 0)
1619                         return ret;
1620                 if (newstate != state)
1621                         return -ESTALE;
1622         }
1623         /*
1624          * We may have performed cached opens for all three recoveries.
1625          * Check if we need to update the current stateid.
1626          */
1627         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1628             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1629                 write_seqlock(&state->seqlock);
1630                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1631                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1632                 write_sequnlock(&state->seqlock);
1633         }
1634         return 0;
1635 }
1636
1637 /*
1638  * OPEN_RECLAIM:
1639  *      reclaim state on the server after a reboot.
1640  */
1641 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1642 {
1643         struct nfs_delegation *delegation;
1644         struct nfs4_opendata *opendata;
1645         fmode_t delegation_type = 0;
1646         int status;
1647
1648         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1649                         NFS4_OPEN_CLAIM_PREVIOUS);
1650         if (IS_ERR(opendata))
1651                 return PTR_ERR(opendata);
1652         rcu_read_lock();
1653         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1654         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1655                 delegation_type = delegation->type;
1656         rcu_read_unlock();
1657         opendata->o_arg.u.delegation_type = delegation_type;
1658         status = nfs4_open_recover(opendata, state);
1659         nfs4_opendata_put(opendata);
1660         return status;
1661 }
1662
1663 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1664 {
1665         struct nfs_server *server = NFS_SERVER(state->inode);
1666         struct nfs4_exception exception = { };
1667         int err;
1668         do {
1669                 err = _nfs4_do_open_reclaim(ctx, state);
1670                 trace_nfs4_open_reclaim(ctx, 0, err);
1671                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1672                         continue;
1673                 if (err != -NFS4ERR_DELAY)
1674                         break;
1675                 nfs4_handle_exception(server, err, &exception);
1676         } while (exception.retry);
1677         return err;
1678 }
1679
1680 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1681 {
1682         struct nfs_open_context *ctx;
1683         int ret;
1684
1685         ctx = nfs4_state_find_open_context(state);
1686         if (IS_ERR(ctx))
1687                 return -EAGAIN;
1688         ret = nfs4_do_open_reclaim(ctx, state);
1689         put_nfs_open_context(ctx);
1690         return ret;
1691 }
1692
1693 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1694 {
1695         switch (err) {
1696                 default:
1697                         printk(KERN_ERR "NFS: %s: unhandled error "
1698                                         "%d.\n", __func__, err);
1699                 case 0:
1700                 case -ENOENT:
1701                 case -EAGAIN:
1702                 case -ESTALE:
1703                         break;
1704                 case -NFS4ERR_BADSESSION:
1705                 case -NFS4ERR_BADSLOT:
1706                 case -NFS4ERR_BAD_HIGH_SLOT:
1707                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1708                 case -NFS4ERR_DEADSESSION:
1709                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1710                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1711                         return -EAGAIN;
1712                 case -NFS4ERR_STALE_CLIENTID:
1713                 case -NFS4ERR_STALE_STATEID:
1714                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1715                 case -NFS4ERR_EXPIRED:
1716                         /* Don't recall a delegation if it was lost */
1717                         nfs4_schedule_lease_recovery(server->nfs_client);
1718                         return -EAGAIN;
1719                 case -NFS4ERR_MOVED:
1720                         nfs4_schedule_migration_recovery(server);
1721                         return -EAGAIN;
1722                 case -NFS4ERR_LEASE_MOVED:
1723                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1724                         return -EAGAIN;
1725                 case -NFS4ERR_DELEG_REVOKED:
1726                 case -NFS4ERR_ADMIN_REVOKED:
1727                 case -NFS4ERR_BAD_STATEID:
1728                 case -NFS4ERR_OPENMODE:
1729                         nfs_inode_find_state_and_recover(state->inode,
1730                                         stateid);
1731                         nfs4_schedule_stateid_recovery(server, state);
1732                         return -EAGAIN;
1733                 case -NFS4ERR_DELAY:
1734                 case -NFS4ERR_GRACE:
1735                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1736                         ssleep(1);
1737                         return -EAGAIN;
1738                 case -ENOMEM:
1739                 case -NFS4ERR_DENIED:
1740                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1741                         return 0;
1742         }
1743         return err;
1744 }
1745
1746 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1747 {
1748         struct nfs_server *server = NFS_SERVER(state->inode);
1749         struct nfs4_opendata *opendata;
1750         int err;
1751
1752         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1753                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1754         if (IS_ERR(opendata))
1755                 return PTR_ERR(opendata);
1756         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1757         err = nfs4_open_recover(opendata, state);
1758         nfs4_opendata_put(opendata);
1759         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1760 }
1761
1762 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1763 {
1764         struct nfs4_opendata *data = calldata;
1765
1766         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1767                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1768 }
1769
1770 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1771 {
1772         struct nfs4_opendata *data = calldata;
1773
1774         nfs40_sequence_done(task, &data->c_res.seq_res);
1775
1776         data->rpc_status = task->tk_status;
1777         if (data->rpc_status == 0) {
1778                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1779                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1780                 renew_lease(data->o_res.server, data->timestamp);
1781                 data->rpc_done = 1;
1782         }
1783 }
1784
1785 static void nfs4_open_confirm_release(void *calldata)
1786 {
1787         struct nfs4_opendata *data = calldata;
1788         struct nfs4_state *state = NULL;
1789
1790         /* If this request hasn't been cancelled, do nothing */
1791         if (data->cancelled == 0)
1792                 goto out_free;
1793         /* In case of error, no cleanup! */
1794         if (!data->rpc_done)
1795                 goto out_free;
1796         state = nfs4_opendata_to_nfs4_state(data);
1797         if (!IS_ERR(state))
1798                 nfs4_close_state(state, data->o_arg.fmode);
1799 out_free:
1800         nfs4_opendata_put(data);
1801 }
1802
1803 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1804         .rpc_call_prepare = nfs4_open_confirm_prepare,
1805         .rpc_call_done = nfs4_open_confirm_done,
1806         .rpc_release = nfs4_open_confirm_release,
1807 };
1808
1809 /*
1810  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1811  */
1812 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1813 {
1814         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1815         struct rpc_task *task;
1816         struct  rpc_message msg = {
1817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1818                 .rpc_argp = &data->c_arg,
1819                 .rpc_resp = &data->c_res,
1820                 .rpc_cred = data->owner->so_cred,
1821         };
1822         struct rpc_task_setup task_setup_data = {
1823                 .rpc_client = server->client,
1824                 .rpc_message = &msg,
1825                 .callback_ops = &nfs4_open_confirm_ops,
1826                 .callback_data = data,
1827                 .workqueue = nfsiod_workqueue,
1828                 .flags = RPC_TASK_ASYNC,
1829         };
1830         int status;
1831
1832         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1833         kref_get(&data->kref);
1834         data->rpc_done = 0;
1835         data->rpc_status = 0;
1836         data->timestamp = jiffies;
1837         task = rpc_run_task(&task_setup_data);
1838         if (IS_ERR(task))
1839                 return PTR_ERR(task);
1840         status = nfs4_wait_for_completion_rpc_task(task);
1841         if (status != 0) {
1842                 data->cancelled = 1;
1843                 smp_wmb();
1844         } else
1845                 status = data->rpc_status;
1846         rpc_put_task(task);
1847         return status;
1848 }
1849
1850 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1851 {
1852         struct nfs4_opendata *data = calldata;
1853         struct nfs4_state_owner *sp = data->owner;
1854         struct nfs_client *clp = sp->so_server->nfs_client;
1855
1856         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1857                 goto out_wait;
1858         /*
1859          * Check if we still need to send an OPEN call, or if we can use
1860          * a delegation instead.
1861          */
1862         if (data->state != NULL) {
1863                 struct nfs_delegation *delegation;
1864
1865                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1866                         goto out_no_action;
1867                 rcu_read_lock();
1868                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1869                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1870                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1871                     can_open_delegated(delegation, data->o_arg.fmode))
1872                         goto unlock_no_action;
1873                 rcu_read_unlock();
1874         }
1875         /* Update client id. */
1876         data->o_arg.clientid = clp->cl_clientid;
1877         switch (data->o_arg.claim) {
1878         case NFS4_OPEN_CLAIM_PREVIOUS:
1879         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1880         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1881                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1882         case NFS4_OPEN_CLAIM_FH:
1883                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1884                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1885         }
1886         data->timestamp = jiffies;
1887         if (nfs4_setup_sequence(data->o_arg.server,
1888                                 &data->o_arg.seq_args,
1889                                 &data->o_res.seq_res,
1890                                 task) != 0)
1891                 nfs_release_seqid(data->o_arg.seqid);
1892
1893         /* Set the create mode (note dependency on the session type) */
1894         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1895         if (data->o_arg.open_flags & O_EXCL) {
1896                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1897                 if (nfs4_has_persistent_session(clp))
1898                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1899                 else if (clp->cl_mvops->minor_version > 0)
1900                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1901         }
1902         return;
1903 unlock_no_action:
1904         rcu_read_unlock();
1905 out_no_action:
1906         task->tk_action = NULL;
1907 out_wait:
1908         nfs4_sequence_done(task, &data->o_res.seq_res);
1909 }
1910
1911 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1912 {
1913         struct nfs4_opendata *data = calldata;
1914
1915         data->rpc_status = task->tk_status;
1916
1917         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1918                 return;
1919
1920         if (task->tk_status == 0) {
1921                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1922                         switch (data->o_res.f_attr->mode & S_IFMT) {
1923                         case S_IFREG:
1924                                 break;
1925                         case S_IFLNK:
1926                                 data->rpc_status = -ELOOP;
1927                                 break;
1928                         case S_IFDIR:
1929                                 data->rpc_status = -EISDIR;
1930                                 break;
1931                         default:
1932                                 data->rpc_status = -ENOTDIR;
1933                         }
1934                 }
1935                 renew_lease(data->o_res.server, data->timestamp);
1936                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1937                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1938         }
1939         data->rpc_done = 1;
1940 }
1941
1942 static void nfs4_open_release(void *calldata)
1943 {
1944         struct nfs4_opendata *data = calldata;
1945         struct nfs4_state *state = NULL;
1946
1947         /* If this request hasn't been cancelled, do nothing */
1948         if (data->cancelled == 0)
1949                 goto out_free;
1950         /* In case of error, no cleanup! */
1951         if (data->rpc_status != 0 || !data->rpc_done)
1952                 goto out_free;
1953         /* In case we need an open_confirm, no cleanup! */
1954         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1955                 goto out_free;
1956         state = nfs4_opendata_to_nfs4_state(data);
1957         if (!IS_ERR(state))
1958                 nfs4_close_state(state, data->o_arg.fmode);
1959 out_free:
1960         nfs4_opendata_put(data);
1961 }
1962
1963 static const struct rpc_call_ops nfs4_open_ops = {
1964         .rpc_call_prepare = nfs4_open_prepare,
1965         .rpc_call_done = nfs4_open_done,
1966         .rpc_release = nfs4_open_release,
1967 };
1968
1969 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1970 {
1971         struct inode *dir = d_inode(data->dir);
1972         struct nfs_server *server = NFS_SERVER(dir);
1973         struct nfs_openargs *o_arg = &data->o_arg;
1974         struct nfs_openres *o_res = &data->o_res;
1975         struct rpc_task *task;
1976         struct rpc_message msg = {
1977                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1978                 .rpc_argp = o_arg,
1979                 .rpc_resp = o_res,
1980                 .rpc_cred = data->owner->so_cred,
1981         };
1982         struct rpc_task_setup task_setup_data = {
1983                 .rpc_client = server->client,
1984                 .rpc_message = &msg,
1985                 .callback_ops = &nfs4_open_ops,
1986                 .callback_data = data,
1987                 .workqueue = nfsiod_workqueue,
1988                 .flags = RPC_TASK_ASYNC,
1989         };
1990         int status;
1991
1992         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1993         kref_get(&data->kref);
1994         data->rpc_done = 0;
1995         data->rpc_status = 0;
1996         data->cancelled = 0;
1997         data->is_recover = 0;
1998         if (isrecover) {
1999                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2000                 data->is_recover = 1;
2001         }
2002         task = rpc_run_task(&task_setup_data);
2003         if (IS_ERR(task))
2004                 return PTR_ERR(task);
2005         status = nfs4_wait_for_completion_rpc_task(task);
2006         if (status != 0) {
2007                 data->cancelled = 1;
2008                 smp_wmb();
2009         } else
2010                 status = data->rpc_status;
2011         rpc_put_task(task);
2012
2013         return status;
2014 }
2015
2016 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2017 {
2018         struct inode *dir = d_inode(data->dir);
2019         struct nfs_openres *o_res = &data->o_res;
2020         int status;
2021
2022         status = nfs4_run_open_task(data, 1);
2023         if (status != 0 || !data->rpc_done)
2024                 return status;
2025
2026         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2027
2028         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2029                 status = _nfs4_proc_open_confirm(data);
2030                 if (status != 0)
2031                         return status;
2032         }
2033
2034         return status;
2035 }
2036
2037 /*
2038  * Additional permission checks in order to distinguish between an
2039  * open for read, and an open for execute. This works around the
2040  * fact that NFSv4 OPEN treats read and execute permissions as being
2041  * the same.
2042  * Note that in the non-execute case, we want to turn off permission
2043  * checking if we just created a new file (POSIX open() semantics).
2044  */
2045 static int nfs4_opendata_access(struct rpc_cred *cred,
2046                                 struct nfs4_opendata *opendata,
2047                                 struct nfs4_state *state, fmode_t fmode,
2048                                 int openflags)
2049 {
2050         struct nfs_access_entry cache;
2051         u32 mask;
2052
2053         /* access call failed or for some reason the server doesn't
2054          * support any access modes -- defer access call until later */
2055         if (opendata->o_res.access_supported == 0)
2056                 return 0;
2057
2058         mask = 0;
2059         /*
2060          * Use openflags to check for exec, because fmode won't
2061          * always have FMODE_EXEC set when file open for exec.
2062          */
2063         if (openflags & __FMODE_EXEC) {
2064                 /* ONLY check for exec rights */
2065                 mask = MAY_EXEC;
2066         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2067                 mask = MAY_READ;
2068
2069         cache.cred = cred;
2070         cache.jiffies = jiffies;
2071         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2072         nfs_access_add_cache(state->inode, &cache);
2073
2074         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2075                 return 0;
2076
2077         /* even though OPEN succeeded, access is denied. Close the file */
2078         nfs4_close_state(state, fmode);
2079         return -EACCES;
2080 }
2081
2082 /*
2083  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2084  */
2085 static int _nfs4_proc_open(struct nfs4_opendata *data)
2086 {
2087         struct inode *dir = d_inode(data->dir);
2088         struct nfs_server *server = NFS_SERVER(dir);
2089         struct nfs_openargs *o_arg = &data->o_arg;
2090         struct nfs_openres *o_res = &data->o_res;
2091         int status;
2092
2093         status = nfs4_run_open_task(data, 0);
2094         if (!data->rpc_done)
2095                 return status;
2096         if (status != 0) {
2097                 if (status == -NFS4ERR_BADNAME &&
2098                                 !(o_arg->open_flags & O_CREAT))
2099                         return -ENOENT;
2100                 return status;
2101         }
2102
2103         nfs_fattr_map_and_free_names(server, &data->f_attr);
2104
2105         if (o_arg->open_flags & O_CREAT) {
2106                 update_changeattr(dir, &o_res->cinfo);
2107                 if (o_arg->open_flags & O_EXCL)
2108                         data->file_created = 1;
2109                 else if (o_res->cinfo.before != o_res->cinfo.after)
2110                         data->file_created = 1;
2111         }
2112         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2113                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2114         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2115                 status = _nfs4_proc_open_confirm(data);
2116                 if (status != 0)
2117                         return status;
2118         }
2119         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2120                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2121         return 0;
2122 }
2123
2124 static int nfs4_recover_expired_lease(struct nfs_server *server)
2125 {
2126         return nfs4_client_recover_expired_lease(server->nfs_client);
2127 }
2128
2129 /*
2130  * OPEN_EXPIRED:
2131  *      reclaim state on the server after a network partition.
2132  *      Assumes caller holds the appropriate lock
2133  */
2134 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2135 {
2136         struct nfs4_opendata *opendata;
2137         int ret;
2138
2139         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2140                         NFS4_OPEN_CLAIM_FH);
2141         if (IS_ERR(opendata))
2142                 return PTR_ERR(opendata);
2143         ret = nfs4_open_recover(opendata, state);
2144         if (ret == -ESTALE)
2145                 d_drop(ctx->dentry);
2146         nfs4_opendata_put(opendata);
2147         return ret;
2148 }
2149
2150 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2151 {
2152         struct nfs_server *server = NFS_SERVER(state->inode);
2153         struct nfs4_exception exception = { };
2154         int err;
2155
2156         do {
2157                 err = _nfs4_open_expired(ctx, state);
2158                 trace_nfs4_open_expired(ctx, 0, err);
2159                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2160                         continue;
2161                 switch (err) {
2162                 default:
2163                         goto out;
2164                 case -NFS4ERR_GRACE:
2165                 case -NFS4ERR_DELAY:
2166                         nfs4_handle_exception(server, err, &exception);
2167                         err = 0;
2168                 }
2169         } while (exception.retry);
2170 out:
2171         return err;
2172 }
2173
2174 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2175 {
2176         struct nfs_open_context *ctx;
2177         int ret;
2178
2179         ctx = nfs4_state_find_open_context(state);
2180         if (IS_ERR(ctx))
2181                 return -EAGAIN;
2182         ret = nfs4_do_open_expired(ctx, state);
2183         put_nfs_open_context(ctx);
2184         return ret;
2185 }
2186
2187 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2188 {
2189         nfs_remove_bad_delegation(state->inode);
2190         write_seqlock(&state->seqlock);
2191         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2192         write_sequnlock(&state->seqlock);
2193         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2194 }
2195
2196 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2197 {
2198         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2199                 nfs_finish_clear_delegation_stateid(state);
2200 }
2201
2202 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2203 {
2204         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2205         nfs40_clear_delegation_stateid(state);
2206         return nfs4_open_expired(sp, state);
2207 }
2208
2209 #if defined(CONFIG_NFS_V4_1)
2210 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2211 {
2212         struct nfs_server *server = NFS_SERVER(state->inode);
2213         nfs4_stateid stateid;
2214         struct nfs_delegation *delegation;
2215         struct rpc_cred *cred;
2216         int status;
2217
2218         /* Get the delegation credential for use by test/free_stateid */
2219         rcu_read_lock();
2220         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2221         if (delegation == NULL) {
2222                 rcu_read_unlock();
2223                 return;
2224         }
2225
2226         nfs4_stateid_copy(&stateid, &delegation->stateid);
2227         cred = get_rpccred(delegation->cred);
2228         rcu_read_unlock();
2229         status = nfs41_test_stateid(server, &stateid, cred);
2230         trace_nfs4_test_delegation_stateid(state, NULL, status);
2231
2232         if (status != NFS_OK) {
2233                 /* Free the stateid unless the server explicitly
2234                  * informs us the stateid is unrecognized. */
2235                 if (status != -NFS4ERR_BAD_STATEID)
2236                         nfs41_free_stateid(server, &stateid, cred);
2237                 nfs_finish_clear_delegation_stateid(state);
2238         }
2239
2240         put_rpccred(cred);
2241 }
2242
2243 /**
2244  * nfs41_check_open_stateid - possibly free an open stateid
2245  *
2246  * @state: NFSv4 state for an inode
2247  *
2248  * Returns NFS_OK if recovery for this stateid is now finished.
2249  * Otherwise a negative NFS4ERR value is returned.
2250  */
2251 static int nfs41_check_open_stateid(struct nfs4_state *state)
2252 {
2253         struct nfs_server *server = NFS_SERVER(state->inode);
2254         nfs4_stateid *stateid = &state->open_stateid;
2255         struct rpc_cred *cred = state->owner->so_cred;
2256         int status;
2257
2258         /* If a state reset has been done, test_stateid is unneeded */
2259         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2260             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2261             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2262                 return -NFS4ERR_BAD_STATEID;
2263
2264         status = nfs41_test_stateid(server, stateid, cred);
2265         trace_nfs4_test_open_stateid(state, NULL, status);
2266         if (status != NFS_OK) {
2267                 /* Free the stateid unless the server explicitly
2268                  * informs us the stateid is unrecognized. */
2269                 if (status != -NFS4ERR_BAD_STATEID)
2270                         nfs41_free_stateid(server, stateid, cred);
2271
2272                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2273                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2274                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2275                 clear_bit(NFS_OPEN_STATE, &state->flags);
2276         }
2277         return status;
2278 }
2279
2280 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2281 {
2282         int status;
2283
2284         nfs41_check_delegation_stateid(state);
2285         status = nfs41_check_open_stateid(state);
2286         if (status != NFS_OK)
2287                 status = nfs4_open_expired(sp, state);
2288         return status;
2289 }
2290 #endif
2291
2292 /*
2293  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2294  * fields corresponding to attributes that were used to store the verifier.
2295  * Make sure we clobber those fields in the later setattr call
2296  */
2297 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2298 {
2299         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2300             !(sattr->ia_valid & ATTR_ATIME_SET))
2301                 sattr->ia_valid |= ATTR_ATIME;
2302
2303         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2304             !(sattr->ia_valid & ATTR_MTIME_SET))
2305                 sattr->ia_valid |= ATTR_MTIME;
2306 }
2307
2308 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2309                 fmode_t fmode,
2310                 int flags,
2311                 struct nfs_open_context *ctx)
2312 {
2313         struct nfs4_state_owner *sp = opendata->owner;
2314         struct nfs_server *server = sp->so_server;
2315         struct dentry *dentry;
2316         struct nfs4_state *state;
2317         unsigned int seq;
2318         int ret;
2319
2320         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2321
2322         ret = _nfs4_proc_open(opendata);
2323         if (ret != 0)
2324                 goto out;
2325
2326         state = nfs4_opendata_to_nfs4_state(opendata);
2327         ret = PTR_ERR(state);
2328         if (IS_ERR(state))
2329                 goto out;
2330         if (server->caps & NFS_CAP_POSIX_LOCK)
2331                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2332
2333         dentry = opendata->dentry;
2334         if (d_really_is_negative(dentry)) {
2335                 /* FIXME: Is this d_drop() ever needed? */
2336                 d_drop(dentry);
2337                 dentry = d_add_unique(dentry, igrab(state->inode));
2338                 if (dentry == NULL) {
2339                         dentry = opendata->dentry;
2340                 } else if (dentry != ctx->dentry) {
2341                         dput(ctx->dentry);
2342                         ctx->dentry = dget(dentry);
2343                 }
2344                 nfs_set_verifier(dentry,
2345                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2346         }
2347
2348         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2349         if (ret != 0)
2350                 goto out;
2351
2352         ctx->state = state;
2353         if (d_inode(dentry) == state->inode) {
2354                 nfs_inode_attach_open_context(ctx);
2355                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2356                         nfs4_schedule_stateid_recovery(server, state);
2357         }
2358 out:
2359         return ret;
2360 }
2361
2362 /*
2363  * Returns a referenced nfs4_state
2364  */
2365 static int _nfs4_do_open(struct inode *dir,
2366                         struct nfs_open_context *ctx,
2367                         int flags,
2368                         struct iattr *sattr,
2369                         struct nfs4_label *label,
2370                         int *opened)
2371 {
2372         struct nfs4_state_owner  *sp;
2373         struct nfs4_state     *state = NULL;
2374         struct nfs_server       *server = NFS_SERVER(dir);
2375         struct nfs4_opendata *opendata;
2376         struct dentry *dentry = ctx->dentry;
2377         struct rpc_cred *cred = ctx->cred;
2378         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2379         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2380         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2381         struct nfs4_label *olabel = NULL;
2382         int status;
2383
2384         /* Protect against reboot recovery conflicts */
2385         status = -ENOMEM;
2386         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2387         if (sp == NULL) {
2388                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2389                 goto out_err;
2390         }
2391         status = nfs4_recover_expired_lease(server);
2392         if (status != 0)
2393                 goto err_put_state_owner;
2394         if (d_really_is_positive(dentry))
2395                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2396         status = -ENOMEM;
2397         if (d_really_is_positive(dentry))
2398                 claim = NFS4_OPEN_CLAIM_FH;
2399         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2400                         label, claim, GFP_KERNEL);
2401         if (opendata == NULL)
2402                 goto err_put_state_owner;
2403
2404         if (label) {
2405                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2406                 if (IS_ERR(olabel)) {
2407                         status = PTR_ERR(olabel);
2408                         goto err_opendata_put;
2409                 }
2410         }
2411
2412         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2413                 if (!opendata->f_attr.mdsthreshold) {
2414                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2415                         if (!opendata->f_attr.mdsthreshold)
2416                                 goto err_free_label;
2417                 }
2418                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2419         }
2420         if (d_really_is_positive(dentry))
2421                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2422
2423         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2424         if (status != 0)
2425                 goto err_free_label;
2426         state = ctx->state;
2427
2428         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2429             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2430                 nfs4_exclusive_attrset(opendata, sattr);
2431
2432                 nfs_fattr_init(opendata->o_res.f_attr);
2433                 status = nfs4_do_setattr(state->inode, cred,
2434                                 opendata->o_res.f_attr, sattr,
2435                                 state, label, olabel);
2436                 if (status == 0) {
2437                         nfs_setattr_update_inode(state->inode, sattr,
2438                                         opendata->o_res.f_attr);
2439                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2440                 }
2441         }
2442         if (opendata->file_created)
2443                 *opened |= FILE_CREATED;
2444
2445         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2446                 *ctx_th = opendata->f_attr.mdsthreshold;
2447                 opendata->f_attr.mdsthreshold = NULL;
2448         }
2449
2450         nfs4_label_free(olabel);
2451
2452         nfs4_opendata_put(opendata);
2453         nfs4_put_state_owner(sp);
2454         return 0;
2455 err_free_label:
2456         nfs4_label_free(olabel);
2457 err_opendata_put:
2458         nfs4_opendata_put(opendata);
2459 err_put_state_owner:
2460         nfs4_put_state_owner(sp);
2461 out_err:
2462         return status;
2463 }
2464
2465
2466 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2467                                         struct nfs_open_context *ctx,
2468                                         int flags,
2469                                         struct iattr *sattr,
2470                                         struct nfs4_label *label,
2471                                         int *opened)
2472 {
2473         struct nfs_server *server = NFS_SERVER(dir);
2474         struct nfs4_exception exception = { };
2475         struct nfs4_state *res;
2476         int status;
2477
2478         do {
2479                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2480                 res = ctx->state;
2481                 trace_nfs4_open_file(ctx, flags, status);
2482                 if (status == 0)
2483                         break;
2484                 /* NOTE: BAD_SEQID means the server and client disagree about the
2485                  * book-keeping w.r.t. state-changing operations
2486                  * (OPEN/CLOSE/LOCK/LOCKU...)
2487                  * It is actually a sign of a bug on the client or on the server.
2488                  *
2489                  * If we receive a BAD_SEQID error in the particular case of
2490                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2491                  * have unhashed the old state_owner for us, and that we can
2492                  * therefore safely retry using a new one. We should still warn
2493                  * the user though...
2494                  */
2495                 if (status == -NFS4ERR_BAD_SEQID) {
2496                         pr_warn_ratelimited("NFS: v4 server %s "
2497                                         " returned a bad sequence-id error!\n",
2498                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2499                         exception.retry = 1;
2500                         continue;
2501                 }
2502                 /*
2503                  * BAD_STATEID on OPEN means that the server cancelled our
2504                  * state before it received the OPEN_CONFIRM.
2505                  * Recover by retrying the request as per the discussion
2506                  * on Page 181 of RFC3530.
2507                  */
2508                 if (status == -NFS4ERR_BAD_STATEID) {
2509                         exception.retry = 1;
2510                         continue;
2511                 }
2512                 if (status == -EAGAIN) {
2513                         /* We must have found a delegation */
2514                         exception.retry = 1;
2515                         continue;
2516                 }
2517                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2518                         continue;
2519                 res = ERR_PTR(nfs4_handle_exception(server,
2520                                         status, &exception));
2521         } while (exception.retry);
2522         return res;
2523 }
2524
2525 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2526                             struct nfs_fattr *fattr, struct iattr *sattr,
2527                             struct nfs4_state *state, struct nfs4_label *ilabel,
2528                             struct nfs4_label *olabel)
2529 {
2530         struct nfs_server *server = NFS_SERVER(inode);
2531         struct nfs_setattrargs  arg = {
2532                 .fh             = NFS_FH(inode),
2533                 .iap            = sattr,
2534                 .server         = server,
2535                 .bitmask = server->attr_bitmask,
2536                 .label          = ilabel,
2537         };
2538         struct nfs_setattrres  res = {
2539                 .fattr          = fattr,
2540                 .label          = olabel,
2541                 .server         = server,
2542         };
2543         struct rpc_message msg = {
2544                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2545                 .rpc_argp       = &arg,
2546                 .rpc_resp       = &res,
2547                 .rpc_cred       = cred,
2548         };
2549         unsigned long timestamp = jiffies;
2550         fmode_t fmode;
2551         bool truncate;
2552         int status;
2553
2554         arg.bitmask = nfs4_bitmask(server, ilabel);
2555         if (ilabel)
2556                 arg.bitmask = nfs4_bitmask(server, olabel);
2557
2558         nfs_fattr_init(fattr);
2559
2560         /* Servers should only apply open mode checks for file size changes */
2561         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2562         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2563
2564         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2565                 /* Use that stateid */
2566         } else if (truncate && state != NULL) {
2567                 struct nfs_lockowner lockowner = {
2568                         .l_owner = current->files,
2569                         .l_pid = current->tgid,
2570                 };
2571                 if (!nfs4_valid_open_stateid(state))
2572                         return -EBADF;
2573                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2574                                 &lockowner) == -EIO)
2575                         return -EBADF;
2576         } else
2577                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2578
2579         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2580         if (status == 0 && state != NULL)
2581                 renew_lease(server, timestamp);
2582         return status;
2583 }
2584
2585 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2586                            struct nfs_fattr *fattr, struct iattr *sattr,
2587                            struct nfs4_state *state, struct nfs4_label *ilabel,
2588                            struct nfs4_label *olabel)
2589 {
2590         struct nfs_server *server = NFS_SERVER(inode);
2591         struct nfs4_exception exception = {
2592                 .state = state,
2593                 .inode = inode,
2594         };
2595         int err;
2596         do {
2597                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2598                 trace_nfs4_setattr(inode, err);
2599                 switch (err) {
2600                 case -NFS4ERR_OPENMODE:
2601                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2602                                 pr_warn_once("NFSv4: server %s is incorrectly "
2603                                                 "applying open mode checks to "
2604                                                 "a SETATTR that is not "
2605                                                 "changing file size.\n",
2606                                                 server->nfs_client->cl_hostname);
2607                         }
2608                         if (state && !(state->state & FMODE_WRITE)) {
2609                                 err = -EBADF;
2610                                 if (sattr->ia_valid & ATTR_OPEN)
2611                                         err = -EACCES;
2612                                 goto out;
2613                         }
2614                 }
2615                 err = nfs4_handle_exception(server, err, &exception);
2616         } while (exception.retry);
2617 out:
2618         return err;
2619 }
2620
2621 struct nfs4_closedata {
2622         struct inode *inode;
2623         struct nfs4_state *state;
2624         struct nfs_closeargs arg;
2625         struct nfs_closeres res;
2626         struct nfs_fattr fattr;
2627         unsigned long timestamp;
2628         bool roc;
2629         u32 roc_barrier;
2630 };
2631
2632 static void nfs4_free_closedata(void *data)
2633 {
2634         struct nfs4_closedata *calldata = data;
2635         struct nfs4_state_owner *sp = calldata->state->owner;
2636         struct super_block *sb = calldata->state->inode->i_sb;
2637
2638         if (calldata->roc)
2639                 pnfs_roc_release(calldata->state->inode);
2640         nfs4_put_open_state(calldata->state);
2641         nfs_free_seqid(calldata->arg.seqid);
2642         nfs4_put_state_owner(sp);
2643         nfs_sb_deactive(sb);
2644         kfree(calldata);
2645 }
2646
2647 static void nfs4_close_done(struct rpc_task *task, void *data)
2648 {
2649         struct nfs4_closedata *calldata = data;
2650         struct nfs4_state *state = calldata->state;
2651         struct nfs_server *server = NFS_SERVER(calldata->inode);
2652         nfs4_stateid *res_stateid = NULL;
2653
2654         dprintk("%s: begin!\n", __func__);
2655         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2656                 return;
2657         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2658         /* hmm. we are done with the inode, and in the process of freeing
2659          * the state_owner. we keep this around to process errors
2660          */
2661         switch (task->tk_status) {
2662                 case 0:
2663                         res_stateid = &calldata->res.stateid;
2664                         if (calldata->arg.fmode == 0 && calldata->roc)
2665                                 pnfs_roc_set_barrier(state->inode,
2666                                                      calldata->roc_barrier);
2667                         renew_lease(server, calldata->timestamp);
2668                         break;
2669                 case -NFS4ERR_ADMIN_REVOKED:
2670                 case -NFS4ERR_STALE_STATEID:
2671                 case -NFS4ERR_OLD_STATEID:
2672                 case -NFS4ERR_BAD_STATEID:
2673                 case -NFS4ERR_EXPIRED:
2674                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2675                                                 &state->open_stateid)) {
2676                                 rpc_restart_call_prepare(task);
2677                                 goto out_release;
2678                         }
2679                         if (calldata->arg.fmode == 0)
2680                                 break;
2681                 default:
2682                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2683                                 rpc_restart_call_prepare(task);
2684                                 goto out_release;
2685                         }
2686         }
2687         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2688 out_release:
2689         nfs_release_seqid(calldata->arg.seqid);
2690         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2691         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2692 }
2693
2694 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2695 {
2696         struct nfs4_closedata *calldata = data;
2697         struct nfs4_state *state = calldata->state;
2698         struct inode *inode = calldata->inode;
2699         bool is_rdonly, is_wronly, is_rdwr;
2700         int call_close = 0;
2701
2702         dprintk("%s: begin!\n", __func__);
2703         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2704                 goto out_wait;
2705
2706         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2707         spin_lock(&state->owner->so_lock);
2708         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2709         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2710         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2711         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2712         /* Calculate the change in open mode */
2713         calldata->arg.fmode = 0;
2714         if (state->n_rdwr == 0) {
2715                 if (state->n_rdonly == 0)
2716                         call_close |= is_rdonly;
2717                 else if (is_rdonly)
2718                         calldata->arg.fmode |= FMODE_READ;
2719                 if (state->n_wronly == 0)
2720                         call_close |= is_wronly;
2721                 else if (is_wronly)
2722                         calldata->arg.fmode |= FMODE_WRITE;
2723         } else if (is_rdwr)
2724                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2725
2726         if (calldata->arg.fmode == 0)
2727                 call_close |= is_rdwr;
2728
2729         if (!nfs4_valid_open_stateid(state))
2730                 call_close = 0;
2731         spin_unlock(&state->owner->so_lock);
2732
2733         if (!call_close) {
2734                 /* Note: exit _without_ calling nfs4_close_done */
2735                 goto out_no_action;
2736         }
2737
2738         if (calldata->arg.fmode == 0) {
2739                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2740                 if (calldata->roc)
2741                         pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2742         }
2743         calldata->arg.share_access =
2744                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2745                                 calldata->arg.fmode, 0);
2746
2747         nfs_fattr_init(calldata->res.fattr);
2748         calldata->timestamp = jiffies;
2749         if (nfs4_setup_sequence(NFS_SERVER(inode),
2750                                 &calldata->arg.seq_args,
2751                                 &calldata->res.seq_res,
2752                                 task) != 0)
2753                 nfs_release_seqid(calldata->arg.seqid);
2754         dprintk("%s: done!\n", __func__);
2755         return;
2756 out_no_action:
2757         task->tk_action = NULL;
2758 out_wait:
2759         nfs4_sequence_done(task, &calldata->res.seq_res);
2760 }
2761
2762 static const struct rpc_call_ops nfs4_close_ops = {
2763         .rpc_call_prepare = nfs4_close_prepare,
2764         .rpc_call_done = nfs4_close_done,
2765         .rpc_release = nfs4_free_closedata,
2766 };
2767
2768 static bool nfs4_roc(struct inode *inode)
2769 {
2770         if (!nfs_have_layout(inode))
2771                 return false;
2772         return pnfs_roc(inode);
2773 }
2774
2775 /*
2776  * It is possible for data to be read/written from a mem-mapped file
2777  * after the sys_close call (which hits the vfs layer as a flush).
2778  * This means that we can't safely call nfsv4 close on a file until
2779  * the inode is cleared. This in turn means that we are not good
2780  * NFSv4 citizens - we do not indicate to the server to update the file's
2781  * share state even when we are done with one of the three share
2782  * stateid's in the inode.
2783  *
2784  * NOTE: Caller must be holding the sp->so_owner semaphore!
2785  */
2786 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2787 {
2788         struct nfs_server *server = NFS_SERVER(state->inode);
2789         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2790         struct nfs4_closedata *calldata;
2791         struct nfs4_state_owner *sp = state->owner;
2792         struct rpc_task *task;
2793         struct rpc_message msg = {
2794                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2795                 .rpc_cred = state->owner->so_cred,
2796         };
2797         struct rpc_task_setup task_setup_data = {
2798                 .rpc_client = server->client,
2799                 .rpc_message = &msg,
2800                 .callback_ops = &nfs4_close_ops,
2801                 .workqueue = nfsiod_workqueue,
2802                 .flags = RPC_TASK_ASYNC,
2803         };
2804         int status = -ENOMEM;
2805
2806         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2807                 &task_setup_data.rpc_client, &msg);
2808
2809         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2810         if (calldata == NULL)
2811                 goto out;
2812         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2813         calldata->inode = state->inode;
2814         calldata->state = state;
2815         calldata->arg.fh = NFS_FH(state->inode);
2816         /* Serialization for the sequence id */
2817         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2818         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2819         if (IS_ERR(calldata->arg.seqid))
2820                 goto out_free_calldata;
2821         calldata->arg.fmode = 0;
2822         calldata->arg.bitmask = server->cache_consistency_bitmask;
2823         calldata->res.fattr = &calldata->fattr;
2824         calldata->res.seqid = calldata->arg.seqid;
2825         calldata->res.server = server;
2826         calldata->roc = nfs4_roc(state->inode);
2827         nfs_sb_active(calldata->inode->i_sb);
2828
2829         msg.rpc_argp = &calldata->arg;
2830         msg.rpc_resp = &calldata->res;
2831         task_setup_data.callback_data = calldata;
2832         task = rpc_run_task(&task_setup_data);
2833         if (IS_ERR(task))
2834                 return PTR_ERR(task);
2835         status = 0;
2836         if (wait)
2837                 status = rpc_wait_for_completion_task(task);
2838         rpc_put_task(task);
2839         return status;
2840 out_free_calldata:
2841         kfree(calldata);
2842 out:
2843         nfs4_put_open_state(state);
2844         nfs4_put_state_owner(sp);
2845         return status;
2846 }
2847
2848 static struct inode *
2849 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2850                 int open_flags, struct iattr *attr, int *opened)
2851 {
2852         struct nfs4_state *state;
2853         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2854
2855         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2856
2857         /* Protect against concurrent sillydeletes */
2858         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2859
2860         nfs4_label_release_security(label);
2861
2862         if (IS_ERR(state))
2863                 return ERR_CAST(state);
2864         return state->inode;
2865 }
2866
2867 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2868 {
2869         if (ctx->state == NULL)
2870                 return;
2871         if (is_sync)
2872                 nfs4_close_sync(ctx->state, ctx->mode);
2873         else
2874                 nfs4_close_state(ctx->state, ctx->mode);
2875 }
2876
2877 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2878 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2879 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2880
2881 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2882 {
2883         struct nfs4_server_caps_arg args = {
2884                 .fhandle = fhandle,
2885         };
2886         struct nfs4_server_caps_res res = {};
2887         struct rpc_message msg = {
2888                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2889                 .rpc_argp = &args,
2890                 .rpc_resp = &res,
2891         };
2892         int status;
2893
2894         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2895         if (status == 0) {
2896                 /* Sanity check the server answers */
2897                 switch (server->nfs_client->cl_minorversion) {
2898                 case 0:
2899                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2900                         res.attr_bitmask[2] = 0;
2901                         break;
2902                 case 1:
2903                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2904                         break;
2905                 case 2:
2906                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2907                 }
2908                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2909                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2910                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2911                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2912                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2913                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2914                                 NFS_CAP_SECURITY_LABEL);
2915                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2916                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2917                         server->caps |= NFS_CAP_ACLS;
2918                 if (res.has_links != 0)
2919                         server->caps |= NFS_CAP_HARDLINKS;
2920                 if (res.has_symlinks != 0)
2921                         server->caps |= NFS_CAP_SYMLINKS;
2922                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2923                         server->caps |= NFS_CAP_FILEID;
2924                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2925                         server->caps |= NFS_CAP_MODE;
2926                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2927                         server->caps |= NFS_CAP_NLINK;
2928                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2929                         server->caps |= NFS_CAP_OWNER;
2930                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2931                         server->caps |= NFS_CAP_OWNER_GROUP;
2932                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2933                         server->caps |= NFS_CAP_ATIME;
2934                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2935                         server->caps |= NFS_CAP_CTIME;
2936                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2937                         server->caps |= NFS_CAP_MTIME;
2938 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2939                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2940                         server->caps |= NFS_CAP_SECURITY_LABEL;
2941 #endif
2942                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2943                                 sizeof(server->attr_bitmask));
2944                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2945
2946                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2947                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2948                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2949                 server->cache_consistency_bitmask[2] = 0;
2950                 server->acl_bitmask = res.acl_bitmask;
2951                 server->fh_expire_type = res.fh_expire_type;
2952         }
2953
2954         return status;
2955 }
2956
2957 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2958 {
2959         struct nfs4_exception exception = { };
2960         int err;
2961         do {
2962                 err = nfs4_handle_exception(server,
2963                                 _nfs4_server_capabilities(server, fhandle),
2964                                 &exception);
2965         } while (exception.retry);
2966         return err;
2967 }
2968
2969 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2970                 struct nfs_fsinfo *info)
2971 {
2972         u32 bitmask[3];
2973         struct nfs4_lookup_root_arg args = {
2974                 .bitmask = bitmask,
2975         };
2976         struct nfs4_lookup_res res = {
2977                 .server = server,
2978                 .fattr = info->fattr,
2979                 .fh = fhandle,
2980         };
2981         struct rpc_message msg = {
2982                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2983                 .rpc_argp = &args,
2984                 .rpc_resp = &res,
2985         };
2986
2987         bitmask[0] = nfs4_fattr_bitmap[0];
2988         bitmask[1] = nfs4_fattr_bitmap[1];
2989         /*
2990          * Process the label in the upcoming getfattr
2991          */
2992         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2993
2994         nfs_fattr_init(info->fattr);
2995         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2996 }
2997
2998 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2999                 struct nfs_fsinfo *info)
3000 {
3001         struct nfs4_exception exception = { };
3002         int err;
3003         do {
3004                 err = _nfs4_lookup_root(server, fhandle, info);
3005                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3006                 switch (err) {
3007                 case 0:
3008                 case -NFS4ERR_WRONGSEC:
3009                         goto out;
3010                 default:
3011                         err = nfs4_handle_exception(server, err, &exception);
3012                 }
3013         } while (exception.retry);
3014 out:
3015         return err;
3016 }
3017
3018 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3019                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3020 {
3021         struct rpc_auth_create_args auth_args = {
3022                 .pseudoflavor = flavor,
3023         };
3024         struct rpc_auth *auth;
3025         int ret;
3026
3027         auth = rpcauth_create(&auth_args, server->client);
3028         if (IS_ERR(auth)) {
3029                 ret = -EACCES;
3030                 goto out;
3031         }
3032         ret = nfs4_lookup_root(server, fhandle, info);
3033 out:
3034         return ret;
3035 }
3036
3037 /*
3038  * Retry pseudoroot lookup with various security flavors.  We do this when:
3039  *
3040  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3041  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3042  *
3043  * Returns zero on success, or a negative NFS4ERR value, or a
3044  * negative errno value.
3045  */
3046 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3047                               struct nfs_fsinfo *info)
3048 {
3049         /* Per 3530bis 15.33.5 */
3050         static const rpc_authflavor_t flav_array[] = {
3051                 RPC_AUTH_GSS_KRB5P,
3052                 RPC_AUTH_GSS_KRB5I,
3053                 RPC_AUTH_GSS_KRB5,
3054                 RPC_AUTH_UNIX,                  /* courtesy */
3055                 RPC_AUTH_NULL,
3056         };
3057         int status = -EPERM;
3058         size_t i;
3059
3060         if (server->auth_info.flavor_len > 0) {
3061                 /* try each flavor specified by user */
3062                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3063                         status = nfs4_lookup_root_sec(server, fhandle, info,
3064                                                 server->auth_info.flavors[i]);
3065                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3066                                 continue;
3067                         break;
3068                 }
3069         } else {
3070                 /* no flavors specified by user, try default list */
3071                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3072                         status = nfs4_lookup_root_sec(server, fhandle, info,
3073                                                       flav_array[i]);
3074                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3075                                 continue;
3076                         break;
3077                 }
3078         }
3079
3080         /*
3081          * -EACCESS could mean that the user doesn't have correct permissions
3082          * to access the mount.  It could also mean that we tried to mount
3083          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3084          * existing mount programs don't handle -EACCES very well so it should
3085          * be mapped to -EPERM instead.
3086          */
3087         if (status == -EACCES)
3088                 status = -EPERM;
3089         return status;
3090 }
3091
3092 static int nfs4_do_find_root_sec(struct nfs_server *server,
3093                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3094 {
3095         int mv = server->nfs_client->cl_minorversion;
3096         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3097 }
3098
3099 /**
3100  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3101  * @server: initialized nfs_server handle
3102  * @fhandle: we fill in the pseudo-fs root file handle
3103  * @info: we fill in an FSINFO struct
3104  * @auth_probe: probe the auth flavours
3105  *
3106  * Returns zero on success, or a negative errno.
3107  */
3108 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3109                          struct nfs_fsinfo *info,
3110                          bool auth_probe)
3111 {
3112         int status = 0;
3113
3114         if (!auth_probe)
3115                 status = nfs4_lookup_root(server, fhandle, info);
3116
3117         if (auth_probe || status == NFS4ERR_WRONGSEC)
3118                 status = nfs4_do_find_root_sec(server, fhandle, info);
3119
3120         if (status == 0)
3121                 status = nfs4_server_capabilities(server, fhandle);
3122         if (status == 0)
3123                 status = nfs4_do_fsinfo(server, fhandle, info);
3124
3125         return nfs4_map_errors(status);
3126 }
3127
3128 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3129                               struct nfs_fsinfo *info)
3130 {
3131         int error;
3132         struct nfs_fattr *fattr = info->fattr;
3133         struct nfs4_label *label = NULL;
3134
3135         error = nfs4_server_capabilities(server, mntfh);
3136         if (error < 0) {
3137                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3138                 return error;
3139         }
3140
3141         label = nfs4_label_alloc(server, GFP_KERNEL);
3142         if (IS_ERR(label))
3143                 return PTR_ERR(label);
3144
3145         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3146         if (error < 0) {
3147                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3148                 goto err_free_label;
3149         }
3150
3151         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3152             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3153                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3154
3155 err_free_label:
3156         nfs4_label_free(label);
3157
3158         return error;
3159 }
3160
3161 /*
3162  * Get locations and (maybe) other attributes of a referral.
3163  * Note that we'll actually follow the referral later when
3164  * we detect fsid mismatch in inode revalidation
3165  */
3166 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3167                              const struct qstr *name, struct nfs_fattr *fattr,
3168                              struct nfs_fh *fhandle)
3169 {
3170         int status = -ENOMEM;
3171         struct page *page = NULL;
3172         struct nfs4_fs_locations *locations = NULL;
3173
3174         page = alloc_page(GFP_KERNEL);
3175         if (page == NULL)
3176                 goto out;
3177         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3178         if (locations == NULL)
3179                 goto out;
3180
3181         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3182         if (status != 0)
3183                 goto out;
3184
3185         /*
3186          * If the fsid didn't change, this is a migration event, not a
3187          * referral.  Cause us to drop into the exception handler, which
3188          * will kick off migration recovery.
3189          */
3190         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3191                 dprintk("%s: server did not return a different fsid for"
3192                         " a referral at %s\n", __func__, name->name);
3193                 status = -NFS4ERR_MOVED;
3194                 goto out;
3195         }
3196         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3197         nfs_fixup_referral_attributes(&locations->fattr);
3198
3199         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3200         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3201         memset(fhandle, 0, sizeof(struct nfs_fh));
3202 out:
3203         if (page)
3204                 __free_page(page);
3205         kfree(locations);
3206         return status;
3207 }
3208
3209 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3210                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3211 {
3212         struct nfs4_getattr_arg args = {
3213                 .fh = fhandle,
3214                 .bitmask = server->attr_bitmask,
3215         };
3216         struct nfs4_getattr_res res = {
3217                 .fattr = fattr,
3218                 .label = label,
3219                 .server = server,
3220         };
3221         struct rpc_message msg = {
3222                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3223                 .rpc_argp = &args,
3224                 .rpc_resp = &res,
3225         };
3226
3227         args.bitmask = nfs4_bitmask(server, label);
3228
3229         nfs_fattr_init(fattr);
3230         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3231 }
3232
3233 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3234                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3235 {
3236         struct nfs4_exception exception = { };
3237         int err;
3238         do {
3239                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3240                 trace_nfs4_getattr(server, fhandle, fattr, err);
3241                 err = nfs4_handle_exception(server, err,
3242                                 &exception);
3243         } while (exception.retry);
3244         return err;
3245 }
3246
3247 /*
3248  * The file is not closed if it is opened due to the a request to change
3249  * the size of the file. The open call will not be needed once the
3250  * VFS layer lookup-intents are implemented.
3251  *
3252  * Close is called when the inode is destroyed.
3253  * If we haven't opened the file for O_WRONLY, we
3254  * need to in the size_change case to obtain a stateid.
3255  *
3256  * Got race?
3257  * Because OPEN is always done by name in nfsv4, it is
3258  * possible that we opened a different file by the same
3259  * name.  We can recognize this race condition, but we
3260  * can't do anything about it besides returning an error.
3261  *
3262  * This will be fixed with VFS changes (lookup-intent).
3263  */
3264 static int
3265 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3266                   struct iattr *sattr)
3267 {
3268         struct inode *inode = d_inode(dentry);
3269         struct rpc_cred *cred = NULL;
3270         struct nfs4_state *state = NULL;
3271         struct nfs4_label *label = NULL;
3272         int status;
3273
3274         if (pnfs_ld_layoutret_on_setattr(inode) &&
3275             sattr->ia_valid & ATTR_SIZE &&
3276             sattr->ia_size < i_size_read(inode))
3277                 pnfs_commit_and_return_layout(inode);
3278
3279         nfs_fattr_init(fattr);
3280
3281         /* Deal with open(O_TRUNC) */
3282         if (sattr->ia_valid & ATTR_OPEN)
3283                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3284
3285         /* Optimization: if the end result is no change, don't RPC */
3286         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3287                 return 0;
3288
3289         /* Search for an existing open(O_WRITE) file */
3290         if (sattr->ia_valid & ATTR_FILE) {
3291                 struct nfs_open_context *ctx;
3292
3293                 ctx = nfs_file_open_context(sattr->ia_file);
3294                 if (ctx) {
3295                         cred = ctx->cred;
3296                         state = ctx->state;
3297                 }
3298         }
3299
3300         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3301         if (IS_ERR(label))
3302                 return PTR_ERR(label);
3303
3304         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3305         if (status == 0) {
3306                 nfs_setattr_update_inode(inode, sattr, fattr);
3307                 nfs_setsecurity(inode, fattr, label);
3308         }
3309         nfs4_label_free(label);
3310         return status;
3311 }
3312
3313 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3314                 const struct qstr *name, struct nfs_fh *fhandle,
3315                 struct nfs_fattr *fattr, struct nfs4_label *label)
3316 {
3317         struct nfs_server *server = NFS_SERVER(dir);
3318         int                    status;
3319         struct nfs4_lookup_arg args = {
3320                 .bitmask = server->attr_bitmask,
3321                 .dir_fh = NFS_FH(dir),
3322                 .name = name,
3323         };
3324         struct nfs4_lookup_res res = {
3325                 .server = server,
3326                 .fattr = fattr,
3327                 .label = label,
3328                 .fh = fhandle,
3329         };
3330         struct rpc_message msg = {
3331                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3332                 .rpc_argp = &args,
3333                 .rpc_resp = &res,
3334         };
3335
3336         args.bitmask = nfs4_bitmask(server, label);
3337
3338         nfs_fattr_init(fattr);
3339
3340         dprintk("NFS call  lookup %s\n", name->name);
3341         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3342         dprintk("NFS reply lookup: %d\n", status);
3343         return status;
3344 }
3345
3346 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3347 {
3348         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3349                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3350         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3351         fattr->nlink = 2;
3352 }
3353
3354 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3355                                    struct qstr *name, struct nfs_fh *fhandle,
3356                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3357 {
3358         struct nfs4_exception exception = { };
3359         struct rpc_clnt *client = *clnt;
3360         int err;
3361         do {
3362                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3363                 trace_nfs4_lookup(dir, name, err);
3364                 switch (err) {
3365                 case -NFS4ERR_BADNAME:
3366                         err = -ENOENT;
3367                         goto out;
3368                 case -NFS4ERR_MOVED:
3369                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3370                         if (err == -NFS4ERR_MOVED)
3371                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3372                         goto out;
3373                 case -NFS4ERR_WRONGSEC:
3374                         err = -EPERM;
3375                         if (client != *clnt)
3376                                 goto out;
3377                         client = nfs4_negotiate_security(client, dir, name);
3378                         if (IS_ERR(client))
3379                                 return PTR_ERR(client);
3380
3381                         exception.retry = 1;
3382                         break;
3383                 default:
3384                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3385                 }
3386         } while (exception.retry);
3387
3388 out:
3389         if (err == 0)
3390                 *clnt = client;
3391         else if (client != *clnt)
3392                 rpc_shutdown_client(client);
3393
3394         return err;
3395 }
3396
3397 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3398                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3399                             struct nfs4_label *label)
3400 {
3401         int status;
3402         struct rpc_clnt *client = NFS_CLIENT(dir);
3403
3404         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3405         if (client != NFS_CLIENT(dir)) {
3406                 rpc_shutdown_client(client);
3407                 nfs_fixup_secinfo_attributes(fattr);
3408         }
3409         return status;
3410 }
3411
3412 struct rpc_clnt *
3413 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3414                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3415 {
3416         struct rpc_clnt *client = NFS_CLIENT(dir);
3417         int status;
3418
3419         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3420         if (status < 0)
3421                 return ERR_PTR(status);
3422         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3423 }
3424
3425 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3426 {
3427         struct nfs_server *server = NFS_SERVER(inode);
3428         struct nfs4_accessargs args = {
3429                 .fh = NFS_FH(inode),
3430                 .bitmask = server->cache_consistency_bitmask,
3431         };
3432         struct nfs4_accessres res = {
3433                 .server = server,
3434         };
3435         struct rpc_message msg = {
3436                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3437                 .rpc_argp = &args,
3438                 .rpc_resp = &res,
3439                 .rpc_cred = entry->cred,
3440         };
3441         int mode = entry->mask;
3442         int status = 0;
3443
3444         /*
3445          * Determine which access bits we want to ask for...
3446          */
3447         if (mode & MAY_READ)
3448                 args.access |= NFS4_ACCESS_READ;
3449         if (S_ISDIR(inode->i_mode)) {
3450                 if (mode & MAY_WRITE)
3451                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3452                 if (mode & MAY_EXEC)
3453                         args.access |= NFS4_ACCESS_LOOKUP;
3454         } else {
3455                 if (mode & MAY_WRITE)
3456                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3457                 if (mode & MAY_EXEC)
3458                         args.access |= NFS4_ACCESS_EXECUTE;
3459         }
3460
3461         res.fattr = nfs_alloc_fattr();
3462         if (res.fattr == NULL)
3463                 return -ENOMEM;
3464
3465         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3466         if (!status) {
3467                 nfs_access_set_mask(entry, res.access);
3468                 nfs_refresh_inode(inode, res.fattr);
3469         }
3470         nfs_free_fattr(res.fattr);
3471         return status;
3472 }
3473
3474 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3475 {
3476         struct nfs4_exception exception = { };
3477         int err;
3478         do {
3479                 err = _nfs4_proc_access(inode, entry);
3480                 trace_nfs4_access(inode, err);
3481                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3482                                 &exception);
3483         } while (exception.retry);
3484         return err;
3485 }
3486
3487 /*
3488  * TODO: For the time being, we don't try to get any attributes
3489  * along with any of the zero-copy operations READ, READDIR,
3490  * READLINK, WRITE.
3491  *
3492  * In the case of the first three, we want to put the GETATTR
3493  * after the read-type operation -- this is because it is hard
3494  * to predict the length of a GETATTR response in v4, and thus
3495  * align the READ data correctly.  This means that the GETATTR
3496  * may end up partially falling into the page cache, and we should
3497  * shift it into the 'tail' of the xdr_buf before processing.
3498  * To do this efficiently, we need to know the total length
3499  * of data received, which doesn't seem to be available outside
3500  * of the RPC layer.
3501  *
3502  * In the case of WRITE, we also want to put the GETATTR after
3503  * the operation -- in this case because we want to make sure
3504  * we get the post-operation mtime and size.
3505  *
3506  * Both of these changes to the XDR layer would in fact be quite
3507  * minor, but I decided to leave them for a subsequent patch.
3508  */
3509 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3510                 unsigned int pgbase, unsigned int pglen)
3511 {
3512         struct nfs4_readlink args = {
3513                 .fh       = NFS_FH(inode),
3514                 .pgbase   = pgbase,
3515                 .pglen    = pglen,
3516                 .pages    = &page,
3517         };
3518         struct nfs4_readlink_res res;
3519         struct rpc_message msg = {
3520                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3521                 .rpc_argp = &args,
3522                 .rpc_resp = &res,
3523         };
3524
3525         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3526 }
3527
3528 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3529                 unsigned int pgbase, unsigned int pglen)
3530 {
3531         struct nfs4_exception exception = { };
3532         int err;
3533         do {
3534                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3535                 trace_nfs4_readlink(inode, err);
3536                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3537                                 &exception);
3538         } while (exception.retry);
3539         return err;
3540 }
3541
3542 /*
3543  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3544  */
3545 static int
3546 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3547                  int flags)
3548 {
3549         struct nfs4_label l, *ilabel = NULL;
3550         struct nfs_open_context *ctx;
3551         struct nfs4_state *state;
3552         int opened = 0;
3553         int status = 0;
3554
3555         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3556         if (IS_ERR(ctx))
3557                 return PTR_ERR(ctx);
3558
3559         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3560
3561         sattr->ia_mode &= ~current_umask();
3562         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3563         if (IS_ERR(state)) {
3564                 status = PTR_ERR(state);
3565                 goto out;
3566         }
3567 out:
3568         nfs4_label_release_security(ilabel);
3569         put_nfs_open_context(ctx);
3570         return status;
3571 }
3572
3573 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3574 {
3575         struct nfs_server *server = NFS_SERVER(dir);
3576         struct nfs_removeargs args = {
3577                 .fh = NFS_FH(dir),
3578                 .name = *name,
3579         };
3580         struct nfs_removeres res = {
3581                 .server = server,
3582         };
3583         struct rpc_message msg = {
3584                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3585                 .rpc_argp = &args,
3586                 .rpc_resp = &res,
3587         };
3588         int status;
3589
3590         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3591         if (status == 0)
3592                 update_changeattr(dir, &res.cinfo);
3593         return status;
3594 }
3595
3596 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3597 {
3598         struct nfs4_exception exception = { };
3599         int err;
3600         do {
3601                 err = _nfs4_proc_remove(dir, name);
3602                 trace_nfs4_remove(dir, name, err);
3603                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3604                                 &exception);
3605         } while (exception.retry);
3606         return err;
3607 }
3608
3609 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3610 {
3611         struct nfs_server *server = NFS_SERVER(dir);
3612         struct nfs_removeargs *args = msg->rpc_argp;
3613         struct nfs_removeres *res = msg->rpc_resp;
3614
3615         res->server = server;
3616         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3617         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3618
3619         nfs_fattr_init(res->dir_attr);
3620 }
3621
3622 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3623 {
3624         nfs4_setup_sequence(NFS_SERVER(data->dir),
3625                         &data->args.seq_args,
3626                         &data->res.seq_res,
3627                         task);
3628 }
3629
3630 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3631 {
3632         struct nfs_unlinkdata *data = task->tk_calldata;
3633         struct nfs_removeres *res = &data->res;
3634
3635         if (!nfs4_sequence_done(task, &res->seq_res))
3636                 return 0;
3637         if (nfs4_async_handle_error(task, res->server, NULL,
3638                                     &data->timeout) == -EAGAIN)
3639                 return 0;
3640         update_changeattr(dir, &res->cinfo);
3641         return 1;
3642 }
3643
3644 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3645 {
3646         struct nfs_server *server = NFS_SERVER(dir);
3647         struct nfs_renameargs *arg = msg->rpc_argp;
3648         struct nfs_renameres *res = msg->rpc_resp;
3649
3650         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3651         res->server = server;
3652         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3653 }
3654
3655 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3656 {
3657         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3658                         &data->args.seq_args,
3659                         &data->res.seq_res,
3660                         task);
3661 }
3662
3663 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3664                                  struct inode *new_dir)
3665 {
3666         struct nfs_renamedata *data = task->tk_calldata;
3667         struct nfs_renameres *res = &data->res;
3668
3669         if (!nfs4_sequence_done(task, &res->seq_res))
3670                 return 0;
3671         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3672                 return 0;
3673
3674         update_changeattr(old_dir, &res->old_cinfo);
3675         update_changeattr(new_dir, &res->new_cinfo);
3676         return 1;
3677 }
3678
3679 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3680 {
3681         struct nfs_server *server = NFS_SERVER(inode);
3682         struct nfs4_link_arg arg = {
3683                 .fh     = NFS_FH(inode),
3684                 .dir_fh = NFS_FH(dir),
3685                 .name   = name,
3686                 .bitmask = server->attr_bitmask,
3687         };
3688         struct nfs4_link_res res = {
3689                 .server = server,
3690                 .label = NULL,
3691         };
3692         struct rpc_message msg = {
3693                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3694                 .rpc_argp = &arg,
3695                 .rpc_resp = &res,
3696         };
3697         int status = -ENOMEM;
3698
3699         res.fattr = nfs_alloc_fattr();
3700         if (res.fattr == NULL)
3701                 goto out;
3702
3703         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3704         if (IS_ERR(res.label)) {
3705                 status = PTR_ERR(res.label);
3706                 goto out;
3707         }
3708         arg.bitmask = nfs4_bitmask(server, res.label);
3709
3710         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3711         if (!status) {
3712                 update_changeattr(dir, &res.cinfo);
3713                 status = nfs_post_op_update_inode(inode, res.fattr);
3714                 if (!status)
3715                         nfs_setsecurity(inode, res.fattr, res.label);
3716         }
3717
3718
3719         nfs4_label_free(res.label);
3720
3721 out:
3722         nfs_free_fattr(res.fattr);
3723         return status;
3724 }
3725
3726 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3727 {
3728         struct nfs4_exception exception = { };
3729         int err;
3730         do {
3731                 err = nfs4_handle_exception(NFS_SERVER(inode),
3732                                 _nfs4_proc_link(inode, dir, name),
3733                                 &exception);
3734         } while (exception.retry);
3735         return err;
3736 }
3737
3738 struct nfs4_createdata {
3739         struct rpc_message msg;
3740         struct nfs4_create_arg arg;
3741         struct nfs4_create_res res;
3742         struct nfs_fh fh;
3743         struct nfs_fattr fattr;
3744         struct nfs4_label *label;
3745 };
3746
3747 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3748                 struct qstr *name, struct iattr *sattr, u32 ftype)
3749 {
3750         struct nfs4_createdata *data;
3751
3752         data = kzalloc(sizeof(*data), GFP_KERNEL);
3753         if (data != NULL) {
3754                 struct nfs_server *server = NFS_SERVER(dir);
3755
3756                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3757                 if (IS_ERR(data->label))
3758                         goto out_free;
3759
3760                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3761                 data->msg.rpc_argp = &data->arg;
3762                 data->msg.rpc_resp = &data->res;
3763                 data->arg.dir_fh = NFS_FH(dir);
3764                 data->arg.server = server;
3765                 data->arg.name = name;
3766                 data->arg.attrs = sattr;
3767                 data->arg.ftype = ftype;
3768                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3769                 data->res.server = server;
3770                 data->res.fh = &data->fh;
3771                 data->res.fattr = &data->fattr;
3772                 data->res.label = data->label;
3773                 nfs_fattr_init(data->res.fattr);
3774         }
3775         return data;
3776 out_free:
3777         kfree(data);
3778         return NULL;
3779 }
3780
3781 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3782 {
3783         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3784                                     &data->arg.seq_args, &data->res.seq_res, 1);
3785         if (status == 0) {
3786                 update_changeattr(dir, &data->res.dir_cinfo);
3787                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3788         }
3789         return status;
3790 }
3791
3792 static void nfs4_free_createdata(struct nfs4_createdata *data)
3793 {
3794         nfs4_label_free(data->label);
3795         kfree(data);
3796 }
3797
3798 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3799                 struct page *page, unsigned int len, struct iattr *sattr,
3800                 struct nfs4_label *label)
3801 {
3802         struct nfs4_createdata *data;
3803         int status = -ENAMETOOLONG;
3804
3805         if (len > NFS4_MAXPATHLEN)
3806                 goto out;
3807
3808         status = -ENOMEM;
3809         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3810         if (data == NULL)
3811                 goto out;
3812
3813         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3814         data->arg.u.symlink.pages = &page;
3815         data->arg.u.symlink.len = len;
3816         data->arg.label = label;
3817
3818         status = nfs4_do_create(dir, dentry, data);
3819
3820         nfs4_free_createdata(data);
3821 out:
3822         return status;
3823 }
3824
3825 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3826                 struct page *page, unsigned int len, struct iattr *sattr)
3827 {
3828         struct nfs4_exception exception = { };
3829         struct nfs4_label l, *label = NULL;
3830         int err;
3831
3832         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3833
3834         do {
3835                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3836                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3837                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3838                                 &exception);
3839         } while (exception.retry);
3840
3841         nfs4_label_release_security(label);
3842         return err;
3843 }
3844
3845 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3846                 struct iattr *sattr, struct nfs4_label *label)
3847 {
3848         struct nfs4_createdata *data;
3849         int status = -ENOMEM;
3850
3851         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3852         if (data == NULL)
3853                 goto out;
3854
3855         data->arg.label = label;
3856         status = nfs4_do_create(dir, dentry, data);
3857
3858         nfs4_free_createdata(data);
3859 out:
3860         return status;
3861 }
3862
3863 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3864                 struct iattr *sattr)
3865 {
3866         struct nfs4_exception exception = { };
3867         struct nfs4_label l, *label = NULL;
3868         int err;
3869
3870         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3871
3872         sattr->ia_mode &= ~current_umask();
3873         do {
3874                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3875                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3876                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3877                                 &exception);
3878         } while (exception.retry);
3879         nfs4_label_release_security(label);
3880
3881         return err;
3882 }
3883
3884 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3885                 u64 cookie, struct page **pages, unsigned int count, int plus)
3886 {
3887         struct inode            *dir = d_inode(dentry);
3888         struct nfs4_readdir_arg args = {
3889                 .fh = NFS_FH(dir),
3890                 .pages = pages,
3891                 .pgbase = 0,
3892                 .count = count,
3893                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3894                 .plus = plus,
3895         };
3896         struct nfs4_readdir_res res;
3897         struct rpc_message msg = {
3898                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3899                 .rpc_argp = &args,
3900                 .rpc_resp = &res,
3901                 .rpc_cred = cred,
3902         };
3903         int                     status;
3904
3905         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3906                         dentry,
3907                         (unsigned long long)cookie);
3908         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3909         res.pgbase = args.pgbase;
3910         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3911         if (status >= 0) {
3912                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3913                 status += args.pgbase;
3914         }
3915
3916         nfs_invalidate_atime(dir);
3917
3918         dprintk("%s: returns %d\n", __func__, status);
3919         return status;
3920 }
3921
3922 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3923                 u64 cookie, struct page **pages, unsigned int count, int plus)
3924 {
3925         struct nfs4_exception exception = { };
3926         int err;
3927         do {
3928                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3929                                 pages, count, plus);
3930                 trace_nfs4_readdir(d_inode(dentry), err);
3931                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3932                                 &exception);
3933         } while (exception.retry);
3934         return err;
3935 }
3936
3937 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3938                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3939 {
3940         struct nfs4_createdata *data;
3941         int mode = sattr->ia_mode;
3942         int status = -ENOMEM;
3943
3944         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3945         if (data == NULL)
3946                 goto out;
3947
3948         if (S_ISFIFO(mode))
3949                 data->arg.ftype = NF4FIFO;
3950         else if (S_ISBLK(mode)) {
3951                 data->arg.ftype = NF4BLK;
3952                 data->arg.u.device.specdata1 = MAJOR(rdev);
3953                 data->arg.u.device.specdata2 = MINOR(rdev);
3954         }
3955         else if (S_ISCHR(mode)) {
3956                 data->arg.ftype = NF4CHR;
3957                 data->arg.u.device.specdata1 = MAJOR(rdev);
3958                 data->arg.u.device.specdata2 = MINOR(rdev);
3959         } else if (!S_ISSOCK(mode)) {
3960                 status = -EINVAL;
3961                 goto out_free;
3962         }
3963
3964         data->arg.label = label;
3965         status = nfs4_do_create(dir, dentry, data);
3966 out_free:
3967         nfs4_free_createdata(data);
3968 out:
3969         return status;
3970 }
3971
3972 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3973                 struct iattr *sattr, dev_t rdev)
3974 {
3975         struct nfs4_exception exception = { };
3976         struct nfs4_label l, *label = NULL;
3977         int err;
3978
3979         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3980
3981         sattr->ia_mode &= ~current_umask();
3982         do {
3983                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3984                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3985                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3986                                 &exception);
3987         } while (exception.retry);
3988
3989         nfs4_label_release_security(label);
3990
3991         return err;
3992 }
3993
3994 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3995                  struct nfs_fsstat *fsstat)
3996 {
3997         struct nfs4_statfs_arg args = {
3998                 .fh = fhandle,
3999                 .bitmask = server->attr_bitmask,
4000         };
4001         struct nfs4_statfs_res res = {
4002                 .fsstat = fsstat,
4003         };
4004         struct rpc_message msg = {
4005                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4006                 .rpc_argp = &args,
4007                 .rpc_resp = &res,
4008         };
4009
4010         nfs_fattr_init(fsstat->fattr);
4011         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4012 }
4013
4014 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4015 {
4016         struct nfs4_exception exception = { };
4017         int err;
4018         do {
4019                 err = nfs4_handle_exception(server,
4020                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4021                                 &exception);
4022         } while (exception.retry);
4023         return err;
4024 }
4025
4026 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4027                 struct nfs_fsinfo *fsinfo)
4028 {
4029         struct nfs4_fsinfo_arg args = {
4030                 .fh = fhandle,
4031                 .bitmask = server->attr_bitmask,
4032         };
4033         struct nfs4_fsinfo_res res = {
4034                 .fsinfo = fsinfo,
4035         };
4036         struct rpc_message msg = {
4037                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4038                 .rpc_argp = &args,
4039                 .rpc_resp = &res,
4040         };
4041
4042         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4043 }
4044
4045 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4046 {
4047         struct nfs4_exception exception = { };
4048         unsigned long now = jiffies;
4049         int err;
4050
4051         do {
4052                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4053                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4054                 if (err == 0) {
4055                         struct nfs_client *clp = server->nfs_client;
4056
4057                         spin_lock(&clp->cl_lock);
4058                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4059                         clp->cl_last_renewal = now;
4060                         spin_unlock(&clp->cl_lock);
4061                         break;
4062                 }
4063                 err = nfs4_handle_exception(server, err, &exception);
4064         } while (exception.retry);
4065         return err;
4066 }
4067
4068 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4069 {
4070         int error;
4071
4072         nfs_fattr_init(fsinfo->fattr);
4073         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4074         if (error == 0) {
4075                 /* block layout checks this! */
4076                 server->pnfs_blksize = fsinfo->blksize;
4077                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4078         }
4079
4080         return error;
4081 }
4082
4083 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4084                 struct nfs_pathconf *pathconf)
4085 {
4086         struct nfs4_pathconf_arg args = {
4087                 .fh = fhandle,
4088                 .bitmask = server->attr_bitmask,
4089         };
4090         struct nfs4_pathconf_res res = {
4091                 .pathconf = pathconf,
4092         };
4093         struct rpc_message msg = {
4094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4095                 .rpc_argp = &args,
4096                 .rpc_resp = &res,
4097         };
4098
4099         /* None of the pathconf attributes are mandatory to implement */
4100         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4101                 memset(pathconf, 0, sizeof(*pathconf));
4102                 return 0;
4103         }
4104
4105         nfs_fattr_init(pathconf->fattr);
4106         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4107 }
4108
4109 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4110                 struct nfs_pathconf *pathconf)
4111 {
4112         struct nfs4_exception exception = { };
4113         int err;
4114
4115         do {
4116                 err = nfs4_handle_exception(server,
4117                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4118                                 &exception);
4119         } while (exception.retry);
4120         return err;
4121 }
4122
4123 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4124                 const struct nfs_open_context *ctx,
4125                 const struct nfs_lock_context *l_ctx,
4126                 fmode_t fmode)
4127 {
4128         const struct nfs_lockowner *lockowner = NULL;
4129
4130         if (l_ctx != NULL)
4131                 lockowner = &l_ctx->lockowner;
4132         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4133 }
4134 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4135
4136 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4137                 const struct nfs_open_context *ctx,
4138                 const struct nfs_lock_context *l_ctx,
4139                 fmode_t fmode)
4140 {
4141         nfs4_stateid current_stateid;
4142
4143         /* If the current stateid represents a lost lock, then exit */
4144         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4145                 return true;
4146         return nfs4_stateid_match(stateid, &current_stateid);
4147 }
4148
4149 static bool nfs4_error_stateid_expired(int err)
4150 {
4151         switch (err) {
4152         case -NFS4ERR_DELEG_REVOKED:
4153         case -NFS4ERR_ADMIN_REVOKED:
4154         case -NFS4ERR_BAD_STATEID:
4155         case -NFS4ERR_STALE_STATEID:
4156         case -NFS4ERR_OLD_STATEID:
4157         case -NFS4ERR_OPENMODE:
4158         case -NFS4ERR_EXPIRED:
4159                 return true;
4160         }
4161         return false;
4162 }
4163
4164 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4165 {
4166         nfs_invalidate_atime(hdr->inode);
4167 }
4168
4169 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4170 {
4171         struct nfs_server *server = NFS_SERVER(hdr->inode);
4172
4173         trace_nfs4_read(hdr, task->tk_status);
4174         if (nfs4_async_handle_error(task, server,
4175                                     hdr->args.context->state,
4176                                     NULL) == -EAGAIN) {
4177                 rpc_restart_call_prepare(task);
4178                 return -EAGAIN;
4179         }
4180
4181         __nfs4_read_done_cb(hdr);
4182         if (task->tk_status > 0)
4183                 renew_lease(server, hdr->timestamp);
4184         return 0;
4185 }
4186
4187 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4188                 struct nfs_pgio_args *args)
4189 {
4190
4191         if (!nfs4_error_stateid_expired(task->tk_status) ||
4192                 nfs4_stateid_is_current(&args->stateid,
4193                                 args->context,
4194                                 args->lock_context,
4195                                 FMODE_READ))
4196                 return false;
4197         rpc_restart_call_prepare(task);
4198         return true;
4199 }
4200
4201 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4202 {
4203
4204         dprintk("--> %s\n", __func__);
4205
4206         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4207                 return -EAGAIN;
4208         if (nfs4_read_stateid_changed(task, &hdr->args))
4209                 return -EAGAIN;
4210         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4211                                     nfs4_read_done_cb(task, hdr);
4212 }
4213
4214 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4215                                  struct rpc_message *msg)
4216 {
4217         hdr->timestamp   = jiffies;
4218         hdr->pgio_done_cb = nfs4_read_done_cb;
4219         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4220         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4221 }
4222
4223 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4224                                       struct nfs_pgio_header *hdr)
4225 {
4226         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4227                         &hdr->args.seq_args,
4228                         &hdr->res.seq_res,
4229                         task))
4230                 return 0;
4231         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4232                                 hdr->args.lock_context,
4233                                 hdr->rw_ops->rw_mode) == -EIO)
4234                 return -EIO;
4235         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4236                 return -EIO;
4237         return 0;
4238 }
4239
4240 static int nfs4_write_done_cb(struct rpc_task *task,
4241                               struct nfs_pgio_header *hdr)
4242 {
4243         struct inode *inode = hdr->inode;
4244
4245         trace_nfs4_write(hdr, task->tk_status);
4246         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4247                                     hdr->args.context->state,
4248                                     NULL) == -EAGAIN) {
4249                 rpc_restart_call_prepare(task);
4250                 return -EAGAIN;
4251         }
4252         if (task->tk_status >= 0) {
4253                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4254                 nfs_writeback_update_inode(hdr);
4255         }
4256         return 0;
4257 }
4258
4259 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4260                 struct nfs_pgio_args *args)
4261 {
4262
4263         if (!nfs4_error_stateid_expired(task->tk_status) ||
4264                 nfs4_stateid_is_current(&args->stateid,
4265                                 args->context,
4266                                 args->lock_context,
4267                                 FMODE_WRITE))
4268                 return false;
4269         rpc_restart_call_prepare(task);
4270         return true;
4271 }
4272
4273 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4274 {
4275         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4276                 return -EAGAIN;
4277         if (nfs4_write_stateid_changed(task, &hdr->args))
4278                 return -EAGAIN;
4279         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4280                 nfs4_write_done_cb(task, hdr);
4281 }
4282
4283 static
4284 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4285 {
4286         /* Don't request attributes for pNFS or O_DIRECT writes */
4287         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4288                 return false;
4289         /* Otherwise, request attributes if and only if we don't hold
4290          * a delegation
4291          */
4292         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4293 }
4294
4295 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4296                                   struct rpc_message *msg)
4297 {
4298         struct nfs_server *server = NFS_SERVER(hdr->inode);
4299
4300         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4301                 hdr->args.bitmask = NULL;
4302                 hdr->res.fattr = NULL;
4303         } else
4304                 hdr->args.bitmask = server->cache_consistency_bitmask;
4305
4306         if (!hdr->pgio_done_cb)
4307                 hdr->pgio_done_cb = nfs4_write_done_cb;
4308         hdr->res.server = server;
4309         hdr->timestamp   = jiffies;
4310
4311         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4312         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4313 }
4314
4315 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4316 {
4317         nfs4_setup_sequence(NFS_SERVER(data->inode),
4318                         &data->args.seq_args,
4319                         &data->res.seq_res,
4320                         task);
4321 }
4322
4323 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4324 {
4325         struct inode *inode = data->inode;
4326
4327         trace_nfs4_commit(data, task->tk_status);
4328         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4329                                     NULL, NULL) == -EAGAIN) {
4330                 rpc_restart_call_prepare(task);
4331                 return -EAGAIN;
4332         }
4333         return 0;
4334 }
4335
4336 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4337 {
4338         if (!nfs4_sequence_done(task, &data->res.seq_res))
4339                 return -EAGAIN;
4340         return data->commit_done_cb(task, data);
4341 }
4342
4343 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4344 {
4345         struct nfs_server *server = NFS_SERVER(data->inode);
4346
4347         if (data->commit_done_cb == NULL)
4348                 data->commit_done_cb = nfs4_commit_done_cb;
4349         data->res.server = server;
4350         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4351         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4352 }
4353
4354 struct nfs4_renewdata {
4355         struct nfs_client       *client;
4356         unsigned long           timestamp;
4357 };
4358
4359 /*
4360  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4361  * standalone procedure for queueing an asynchronous RENEW.
4362  */
4363 static void nfs4_renew_release(void *calldata)
4364 {
4365         struct nfs4_renewdata *data = calldata;
4366         struct nfs_client *clp = data->client;
4367
4368         if (atomic_read(&clp->cl_count) > 1)
4369                 nfs4_schedule_state_renewal(clp);
4370         nfs_put_client(clp);
4371         kfree(data);
4372 }
4373
4374 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4375 {
4376         struct nfs4_renewdata *data = calldata;
4377         struct nfs_client *clp = data->client;
4378         unsigned long timestamp = data->timestamp;
4379
4380         trace_nfs4_renew_async(clp, task->tk_status);
4381         switch (task->tk_status) {
4382         case 0:
4383                 break;
4384         case -NFS4ERR_LEASE_MOVED:
4385                 nfs4_schedule_lease_moved_recovery(clp);
4386                 break;
4387         default:
4388                 /* Unless we're shutting down, schedule state recovery! */
4389                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4390                         return;
4391                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4392                         nfs4_schedule_lease_recovery(clp);
4393                         return;
4394                 }
4395                 nfs4_schedule_path_down_recovery(clp);
4396         }
4397         do_renew_lease(clp, timestamp);
4398 }
4399
4400 static const struct rpc_call_ops nfs4_renew_ops = {
4401         .rpc_call_done = nfs4_renew_done,
4402         .rpc_release = nfs4_renew_release,
4403 };
4404
4405 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4406 {
4407         struct rpc_message msg = {
4408                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4409                 .rpc_argp       = clp,
4410                 .rpc_cred       = cred,
4411         };
4412         struct nfs4_renewdata *data;
4413
4414         if (renew_flags == 0)
4415                 return 0;
4416         if (!atomic_inc_not_zero(&clp->cl_count))
4417                 return -EIO;
4418         data = kmalloc(sizeof(*data), GFP_NOFS);
4419         if (data == NULL)
4420                 return -ENOMEM;
4421         data->client = clp;
4422         data->timestamp = jiffies;
4423         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4424                         &nfs4_renew_ops, data);
4425 }
4426
4427 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4428 {
4429         struct rpc_message msg = {
4430                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4431                 .rpc_argp       = clp,
4432                 .rpc_cred       = cred,
4433         };
4434         unsigned long now = jiffies;
4435         int status;
4436
4437         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4438         if (status < 0)
4439                 return status;
4440         do_renew_lease(clp, now);
4441         return 0;
4442 }
4443
4444 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4445 {
4446         return server->caps & NFS_CAP_ACLS;
4447 }
4448
4449 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4450  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4451  * the stack.
4452  */
4453 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4454
4455 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4456                 struct page **pages, unsigned int *pgbase)
4457 {
4458         struct page *newpage, **spages;
4459         int rc = 0;
4460         size_t len;
4461         spages = pages;
4462
4463         do {
4464                 len = min_t(size_t, PAGE_SIZE, buflen);
4465                 newpage = alloc_page(GFP_KERNEL);
4466
4467                 if (newpage == NULL)
4468                         goto unwind;
4469                 memcpy(page_address(newpage), buf, len);
4470                 buf += len;
4471                 buflen -= len;
4472                 *pages++ = newpage;
4473                 rc++;
4474         } while (buflen != 0);
4475
4476         return rc;
4477
4478 unwind:
4479         for(; rc > 0; rc--)
4480                 __free_page(spages[rc-1]);
4481         return -ENOMEM;
4482 }
4483
4484 struct nfs4_cached_acl {
4485         int cached;
4486         size_t len;
4487         char data[0];
4488 };
4489
4490 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4491 {
4492         struct nfs_inode *nfsi = NFS_I(inode);
4493
4494         spin_lock(&inode->i_lock);
4495         kfree(nfsi->nfs4_acl);
4496         nfsi->nfs4_acl = acl;
4497         spin_unlock(&inode->i_lock);
4498 }
4499
4500 static void nfs4_zap_acl_attr(struct inode *inode)
4501 {
4502         nfs4_set_cached_acl(inode, NULL);
4503 }
4504
4505 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4506 {
4507         struct nfs_inode *nfsi = NFS_I(inode);
4508         struct nfs4_cached_acl *acl;
4509         int ret = -ENOENT;
4510
4511         spin_lock(&inode->i_lock);
4512         acl = nfsi->nfs4_acl;
4513         if (acl == NULL)
4514                 goto out;
4515         if (buf == NULL) /* user is just asking for length */
4516                 goto out_len;
4517         if (acl->cached == 0)
4518                 goto out;
4519         ret = -ERANGE; /* see getxattr(2) man page */
4520         if (acl->len > buflen)
4521                 goto out;
4522         memcpy(buf, acl->data, acl->len);
4523 out_len:
4524         ret = acl->len;
4525 out:
4526         spin_unlock(&inode->i_lock);
4527         return ret;
4528 }
4529
4530 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4531 {
4532         struct nfs4_cached_acl *acl;
4533         size_t buflen = sizeof(*acl) + acl_len;
4534
4535         if (buflen <= PAGE_SIZE) {
4536                 acl = kmalloc(buflen, GFP_KERNEL);
4537                 if (acl == NULL)
4538                         goto out;
4539                 acl->cached = 1;
4540                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4541         } else {
4542                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4543                 if (acl == NULL)
4544                         goto out;
4545                 acl->cached = 0;
4546         }
4547         acl->len = acl_len;
4548 out:
4549         nfs4_set_cached_acl(inode, acl);
4550 }
4551
4552 /*
4553  * The getxattr API returns the required buffer length when called with a
4554  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4555  * the required buf.  On a NULL buf, we send a page of data to the server
4556  * guessing that the ACL request can be serviced by a page. If so, we cache
4557  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4558  * the cache. If not so, we throw away the page, and cache the required
4559  * length. The next getxattr call will then produce another round trip to
4560  * the server, this time with the input buf of the required size.
4561  */
4562 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4563 {
4564         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4565         struct nfs_getaclargs args = {
4566                 .fh = NFS_FH(inode),
4567                 .acl_pages = pages,
4568                 .acl_len = buflen,
4569         };
4570         struct nfs_getaclres res = {
4571                 .acl_len = buflen,
4572         };
4573         struct rpc_message msg = {
4574                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4575                 .rpc_argp = &args,
4576                 .rpc_resp = &res,
4577         };
4578         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4579         int ret = -ENOMEM, i;
4580
4581         /* As long as we're doing a round trip to the server anyway,
4582          * let's be prepared for a page of acl data. */
4583         if (npages == 0)
4584                 npages = 1;
4585         if (npages > ARRAY_SIZE(pages))
4586                 return -ERANGE;
4587
4588         for (i = 0; i < npages; i++) {
4589                 pages[i] = alloc_page(GFP_KERNEL);
4590                 if (!pages[i])
4591                         goto out_free;
4592         }
4593
4594         /* for decoding across pages */
4595         res.acl_scratch = alloc_page(GFP_KERNEL);
4596         if (!res.acl_scratch)
4597                 goto out_free;
4598
4599         args.acl_len = npages * PAGE_SIZE;
4600         args.acl_pgbase = 0;
4601
4602         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4603                 __func__, buf, buflen, npages, args.acl_len);
4604         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4605                              &msg, &args.seq_args, &res.seq_res, 0);
4606         if (ret)
4607                 goto out_free;
4608
4609         /* Handle the case where the passed-in buffer is too short */
4610         if (res.acl_flags & NFS4_ACL_TRUNC) {
4611                 /* Did the user only issue a request for the acl length? */
4612                 if (buf == NULL)
4613                         goto out_ok;
4614                 ret = -ERANGE;
4615                 goto out_free;
4616         }
4617         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4618         if (buf) {
4619                 if (res.acl_len > buflen) {
4620                         ret = -ERANGE;
4621                         goto out_free;
4622                 }
4623                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4624         }
4625 out_ok:
4626         ret = res.acl_len;
4627 out_free:
4628         for (i = 0; i < npages; i++)
4629                 if (pages[i])
4630                         __free_page(pages[i]);
4631         if (res.acl_scratch)
4632                 __free_page(res.acl_scratch);
4633         return ret;
4634 }
4635
4636 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4637 {
4638         struct nfs4_exception exception = { };
4639         ssize_t ret;
4640         do {
4641                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4642                 trace_nfs4_get_acl(inode, ret);
4643                 if (ret >= 0)
4644                         break;
4645                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4646         } while (exception.retry);
4647         return ret;
4648 }
4649
4650 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4651 {
4652         struct nfs_server *server = NFS_SERVER(inode);
4653         int ret;
4654
4655         if (!nfs4_server_supports_acls(server))
4656                 return -EOPNOTSUPP;
4657         ret = nfs_revalidate_inode(server, inode);
4658         if (ret < 0)
4659                 return ret;
4660         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4661                 nfs_zap_acl_cache(inode);
4662         ret = nfs4_read_cached_acl(inode, buf, buflen);
4663         if (ret != -ENOENT)
4664                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4665                  * but no cached acl data, just the acl length */
4666                 return ret;
4667         return nfs4_get_acl_uncached(inode, buf, buflen);
4668 }
4669
4670 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4671 {
4672         struct nfs_server *server = NFS_SERVER(inode);
4673         struct page *pages[NFS4ACL_MAXPAGES];
4674         struct nfs_setaclargs arg = {
4675                 .fh             = NFS_FH(inode),
4676                 .acl_pages      = pages,
4677                 .acl_len        = buflen,
4678         };
4679         struct nfs_setaclres res;
4680         struct rpc_message msg = {
4681                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4682                 .rpc_argp       = &arg,
4683                 .rpc_resp       = &res,
4684         };
4685         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4686         int ret, i;
4687
4688         if (!nfs4_server_supports_acls(server))
4689                 return -EOPNOTSUPP;
4690         if (npages > ARRAY_SIZE(pages))
4691                 return -ERANGE;
4692         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4693         if (i < 0)
4694                 return i;
4695         nfs4_inode_return_delegation(inode);
4696         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4697
4698         /*
4699          * Free each page after tx, so the only ref left is
4700          * held by the network stack
4701          */
4702         for (; i > 0; i--)
4703                 put_page(pages[i-1]);
4704
4705         /*
4706          * Acl update can result in inode attribute update.
4707          * so mark the attribute cache invalid.
4708          */
4709         spin_lock(&inode->i_lock);
4710         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4711         spin_unlock(&inode->i_lock);
4712         nfs_access_zap_cache(inode);
4713         nfs_zap_acl_cache(inode);
4714         return ret;
4715 }
4716
4717 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4718 {
4719         struct nfs4_exception exception = { };
4720         int err;
4721         do {
4722                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4723                 trace_nfs4_set_acl(inode, err);
4724                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4725                                 &exception);
4726         } while (exception.retry);
4727         return err;
4728 }
4729
4730 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4731 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4732                                         size_t buflen)
4733 {
4734         struct nfs_server *server = NFS_SERVER(inode);
4735         struct nfs_fattr fattr;
4736         struct nfs4_label label = {0, 0, buflen, buf};
4737
4738         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4739         struct nfs4_getattr_arg arg = {
4740                 .fh             = NFS_FH(inode),
4741                 .bitmask        = bitmask,
4742         };
4743         struct nfs4_getattr_res res = {
4744                 .fattr          = &fattr,
4745                 .label          = &label,
4746                 .server         = server,
4747         };
4748         struct rpc_message msg = {
4749                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4750                 .rpc_argp       = &arg,
4751                 .rpc_resp       = &res,
4752         };
4753         int ret;
4754
4755         nfs_fattr_init(&fattr);
4756
4757         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4758         if (ret)
4759                 return ret;
4760         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4761                 return -ENOENT;
4762         if (buflen < label.len)
4763                 return -ERANGE;
4764         return 0;
4765 }
4766
4767 static int nfs4_get_security_label(struct inode *inode, void *buf,
4768                                         size_t buflen)
4769 {
4770         struct nfs4_exception exception = { };
4771         int err;
4772
4773         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4774                 return -EOPNOTSUPP;
4775
4776         do {
4777                 err = _nfs4_get_security_label(inode, buf, buflen);
4778                 trace_nfs4_get_security_label(inode, err);
4779                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4780                                 &exception);
4781         } while (exception.retry);
4782         return err;
4783 }
4784
4785 static int _nfs4_do_set_security_label(struct inode *inode,
4786                 struct nfs4_label *ilabel,
4787                 struct nfs_fattr *fattr,
4788                 struct nfs4_label *olabel)
4789 {
4790
4791         struct iattr sattr = {0};
4792         struct nfs_server *server = NFS_SERVER(inode);
4793         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4794         struct nfs_setattrargs arg = {
4795                 .fh             = NFS_FH(inode),
4796                 .iap            = &sattr,
4797                 .server         = server,
4798                 .bitmask        = bitmask,
4799                 .label          = ilabel,
4800         };
4801         struct nfs_setattrres res = {
4802                 .fattr          = fattr,
4803                 .label          = olabel,
4804                 .server         = server,
4805         };
4806         struct rpc_message msg = {
4807                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4808                 .rpc_argp       = &arg,
4809                 .rpc_resp       = &res,
4810         };
4811         int status;
4812
4813         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4814
4815         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4816         if (status)
4817                 dprintk("%s failed: %d\n", __func__, status);
4818
4819         return status;
4820 }
4821
4822 static int nfs4_do_set_security_label(struct inode *inode,
4823                 struct nfs4_label *ilabel,
4824                 struct nfs_fattr *fattr,
4825                 struct nfs4_label *olabel)
4826 {
4827         struct nfs4_exception exception = { };
4828         int err;
4829
4830         do {
4831                 err = _nfs4_do_set_security_label(inode, ilabel,
4832                                 fattr, olabel);
4833                 trace_nfs4_set_security_label(inode, err);
4834                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4835                                 &exception);
4836         } while (exception.retry);
4837         return err;
4838 }
4839
4840 static int
4841 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4842 {
4843         struct nfs4_label ilabel, *olabel = NULL;
4844         struct nfs_fattr fattr;
4845         struct rpc_cred *cred;
4846         struct inode *inode = d_inode(dentry);
4847         int status;
4848
4849         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4850                 return -EOPNOTSUPP;
4851
4852         nfs_fattr_init(&fattr);
4853
4854         ilabel.pi = 0;
4855         ilabel.lfs = 0;
4856         ilabel.label = (char *)buf;
4857         ilabel.len = buflen;
4858
4859         cred = rpc_lookup_cred();
4860         if (IS_ERR(cred))
4861                 return PTR_ERR(cred);
4862
4863         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4864         if (IS_ERR(olabel)) {
4865                 status = -PTR_ERR(olabel);
4866                 goto out;
4867         }
4868
4869         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4870         if (status == 0)
4871                 nfs_setsecurity(inode, &fattr, olabel);
4872
4873         nfs4_label_free(olabel);
4874 out:
4875         put_rpccred(cred);
4876         return status;
4877 }
4878 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4879
4880
4881 static int
4882 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4883                         struct nfs4_state *state, long *timeout)
4884 {
4885         struct nfs_client *clp = server->nfs_client;
4886
4887         if (task->tk_status >= 0)
4888                 return 0;
4889         switch(task->tk_status) {
4890                 case -NFS4ERR_DELEG_REVOKED:
4891                 case -NFS4ERR_ADMIN_REVOKED:
4892                 case -NFS4ERR_BAD_STATEID:
4893                 case -NFS4ERR_OPENMODE:
4894                         if (state == NULL)
4895                                 break;
4896                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4897                                 goto recovery_failed;
4898                         goto wait_on_recovery;
4899                 case -NFS4ERR_EXPIRED:
4900                         if (state != NULL) {
4901                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4902                                         goto recovery_failed;
4903                         }
4904                 case -NFS4ERR_STALE_STATEID:
4905                 case -NFS4ERR_STALE_CLIENTID:
4906                         nfs4_schedule_lease_recovery(clp);
4907                         goto wait_on_recovery;
4908                 case -NFS4ERR_MOVED:
4909                         if (nfs4_schedule_migration_recovery(server) < 0)
4910                                 goto recovery_failed;
4911                         goto wait_on_recovery;
4912                 case -NFS4ERR_LEASE_MOVED:
4913                         nfs4_schedule_lease_moved_recovery(clp);
4914                         goto wait_on_recovery;
4915 #if defined(CONFIG_NFS_V4_1)
4916                 case -NFS4ERR_BADSESSION:
4917                 case -NFS4ERR_BADSLOT:
4918                 case -NFS4ERR_BAD_HIGH_SLOT:
4919                 case -NFS4ERR_DEADSESSION:
4920                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4921                 case -NFS4ERR_SEQ_FALSE_RETRY:
4922                 case -NFS4ERR_SEQ_MISORDERED:
4923                         dprintk("%s ERROR %d, Reset session\n", __func__,
4924                                 task->tk_status);
4925                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4926                         goto wait_on_recovery;
4927 #endif /* CONFIG_NFS_V4_1 */
4928                 case -NFS4ERR_DELAY:
4929                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4930                         rpc_delay(task, nfs4_update_delay(timeout));
4931                         goto restart_call;
4932                 case -NFS4ERR_GRACE:
4933                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4934                 case -NFS4ERR_RETRY_UNCACHED_REP:
4935                 case -NFS4ERR_OLD_STATEID:
4936                         goto restart_call;
4937         }
4938         task->tk_status = nfs4_map_errors(task->tk_status);
4939         return 0;
4940 recovery_failed:
4941         task->tk_status = -EIO;
4942         return 0;
4943 wait_on_recovery:
4944         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4945         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4946                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4947         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4948                 goto recovery_failed;
4949 restart_call:
4950         task->tk_status = 0;
4951         return -EAGAIN;
4952 }
4953
4954 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4955                                     nfs4_verifier *bootverf)
4956 {
4957         __be32 verf[2];
4958
4959         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4960                 /* An impossible timestamp guarantees this value
4961                  * will never match a generated boot time. */
4962                 verf[0] = 0;
4963                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4964         } else {
4965                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4966                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4967                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4968         }
4969         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4970 }
4971
4972 static int
4973 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
4974 {
4975         int result;
4976         size_t len;
4977         char *str;
4978         bool retried = false;
4979
4980         if (clp->cl_owner_id != NULL)
4981                 return 0;
4982 retry:
4983         rcu_read_lock();
4984         len = 10 + strlen(clp->cl_ipaddr) + 1 +
4985                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
4986                 1 +
4987                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
4988                 1;
4989         rcu_read_unlock();
4990
4991         if (len > NFS4_OPAQUE_LIMIT + 1)
4992                 return -EINVAL;
4993
4994         /*
4995          * Since this string is allocated at mount time, and held until the
4996          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
4997          * about a memory-reclaim deadlock.
4998          */
4999         str = kmalloc(len, GFP_KERNEL);
5000         if (!str)
5001                 return -ENOMEM;
5002
5003         rcu_read_lock();
5004         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5005                         clp->cl_ipaddr,
5006                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5007                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5008         rcu_read_unlock();
5009
5010         /* Did something change? */
5011         if (result >= len) {
5012                 kfree(str);
5013                 if (retried)
5014                         return -EINVAL;
5015                 retried = true;
5016                 goto retry;
5017         }
5018         clp->cl_owner_id = str;
5019         return 0;
5020 }
5021
5022 static int
5023 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5024 {
5025         int result;
5026         size_t len;
5027         char *str;
5028
5029         len = 10 + 10 + 1 + 10 + 1 +
5030                 strlen(nfs4_client_id_uniquifier) + 1 +
5031                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5032
5033         if (len > NFS4_OPAQUE_LIMIT + 1)
5034                 return -EINVAL;
5035
5036         /*
5037          * Since this string is allocated at mount time, and held until the
5038          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5039          * about a memory-reclaim deadlock.
5040          */
5041         str = kmalloc(len, GFP_KERNEL);
5042         if (!str)
5043                 return -ENOMEM;
5044
5045         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5046                         clp->rpc_ops->version, clp->cl_minorversion,
5047                         nfs4_client_id_uniquifier,
5048                         clp->cl_rpcclient->cl_nodename);
5049         if (result >= len) {
5050                 kfree(str);
5051                 return -EINVAL;
5052         }
5053         clp->cl_owner_id = str;
5054         return 0;
5055 }
5056
5057 static int
5058 nfs4_init_uniform_client_string(struct nfs_client *clp)
5059 {
5060         int result;
5061         size_t len;
5062         char *str;
5063
5064         if (clp->cl_owner_id != NULL)
5065                 return 0;
5066
5067         if (nfs4_client_id_uniquifier[0] != '\0')
5068                 return nfs4_init_uniquifier_client_string(clp);
5069
5070         len = 10 + 10 + 1 + 10 + 1 +
5071                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5072
5073         if (len > NFS4_OPAQUE_LIMIT + 1)
5074                 return -EINVAL;
5075
5076         /*
5077          * Since this string is allocated at mount time, and held until the
5078          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5079          * about a memory-reclaim deadlock.
5080          */
5081         str = kmalloc(len, GFP_KERNEL);
5082         if (!str)
5083                 return -ENOMEM;
5084
5085         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5086                         clp->rpc_ops->version, clp->cl_minorversion,
5087                         clp->cl_rpcclient->cl_nodename);
5088         if (result >= len) {
5089                 kfree(str);
5090                 return -EINVAL;
5091         }
5092         clp->cl_owner_id = str;
5093         return 0;
5094 }
5095
5096 /*
5097  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5098  * services.  Advertise one based on the address family of the
5099  * clientaddr.
5100  */
5101 static unsigned int
5102 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5103 {
5104         if (strchr(clp->cl_ipaddr, ':') != NULL)
5105                 return scnprintf(buf, len, "tcp6");
5106         else
5107                 return scnprintf(buf, len, "tcp");
5108 }
5109
5110 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5111 {
5112         struct nfs4_setclientid *sc = calldata;
5113
5114         if (task->tk_status == 0)
5115                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5116 }
5117
5118 static const struct rpc_call_ops nfs4_setclientid_ops = {
5119         .rpc_call_done = nfs4_setclientid_done,
5120 };
5121
5122 /**
5123  * nfs4_proc_setclientid - Negotiate client ID
5124  * @clp: state data structure
5125  * @program: RPC program for NFSv4 callback service
5126  * @port: IP port number for NFS4 callback service
5127  * @cred: RPC credential to use for this call
5128  * @res: where to place the result
5129  *
5130  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5131  */
5132 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5133                 unsigned short port, struct rpc_cred *cred,
5134                 struct nfs4_setclientid_res *res)
5135 {
5136         nfs4_verifier sc_verifier;
5137         struct nfs4_setclientid setclientid = {
5138                 .sc_verifier = &sc_verifier,
5139                 .sc_prog = program,
5140                 .sc_clnt = clp,
5141         };
5142         struct rpc_message msg = {
5143                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5144                 .rpc_argp = &setclientid,
5145                 .rpc_resp = res,
5146                 .rpc_cred = cred,
5147         };
5148         struct rpc_task *task;
5149         struct rpc_task_setup task_setup_data = {
5150                 .rpc_client = clp->cl_rpcclient,
5151                 .rpc_message = &msg,
5152                 .callback_ops = &nfs4_setclientid_ops,
5153                 .callback_data = &setclientid,
5154                 .flags = RPC_TASK_TIMEOUT,
5155         };
5156         int status;
5157
5158         /* nfs_client_id4 */
5159         nfs4_init_boot_verifier(clp, &sc_verifier);
5160
5161         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5162                 status = nfs4_init_uniform_client_string(clp);
5163         else
5164                 status = nfs4_init_nonuniform_client_string(clp);
5165
5166         if (status)
5167                 goto out;
5168
5169         /* cb_client4 */
5170         setclientid.sc_netid_len =
5171                                 nfs4_init_callback_netid(clp,
5172                                                 setclientid.sc_netid,
5173                                                 sizeof(setclientid.sc_netid));
5174         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5175                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5176                                 clp->cl_ipaddr, port >> 8, port & 255);
5177
5178         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5179                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5180                 clp->cl_owner_id);
5181         task = rpc_run_task(&task_setup_data);
5182         if (IS_ERR(task)) {
5183                 status = PTR_ERR(task);
5184                 goto out;
5185         }
5186         status = task->tk_status;
5187         if (setclientid.sc_cred) {
5188                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5189                 put_rpccred(setclientid.sc_cred);
5190         }
5191         rpc_put_task(task);
5192 out:
5193         trace_nfs4_setclientid(clp, status);
5194         dprintk("NFS reply setclientid: %d\n", status);
5195         return status;
5196 }
5197
5198 /**
5199  * nfs4_proc_setclientid_confirm - Confirm client ID
5200  * @clp: state data structure
5201  * @res: result of a previous SETCLIENTID
5202  * @cred: RPC credential to use for this call
5203  *
5204  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5205  */
5206 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5207                 struct nfs4_setclientid_res *arg,
5208                 struct rpc_cred *cred)
5209 {
5210         struct rpc_message msg = {
5211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5212                 .rpc_argp = arg,
5213                 .rpc_cred = cred,
5214         };
5215         int status;
5216
5217         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5218                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5219                 clp->cl_clientid);
5220         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5221         trace_nfs4_setclientid_confirm(clp, status);
5222         dprintk("NFS reply setclientid_confirm: %d\n", status);
5223         return status;
5224 }
5225
5226 struct nfs4_delegreturndata {
5227         struct nfs4_delegreturnargs args;
5228         struct nfs4_delegreturnres res;
5229         struct nfs_fh fh;
5230         nfs4_stateid stateid;
5231         unsigned long timestamp;
5232         struct nfs_fattr fattr;
5233         int rpc_status;
5234         struct inode *inode;
5235         bool roc;
5236         u32 roc_barrier;
5237 };
5238
5239 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5240 {
5241         struct nfs4_delegreturndata *data = calldata;
5242
5243         if (!nfs4_sequence_done(task, &data->res.seq_res))
5244                 return;
5245
5246         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5247         switch (task->tk_status) {
5248         case 0:
5249                 renew_lease(data->res.server, data->timestamp);
5250         case -NFS4ERR_ADMIN_REVOKED:
5251         case -NFS4ERR_DELEG_REVOKED:
5252         case -NFS4ERR_BAD_STATEID:
5253         case -NFS4ERR_OLD_STATEID:
5254         case -NFS4ERR_STALE_STATEID:
5255         case -NFS4ERR_EXPIRED:
5256                 task->tk_status = 0;
5257                 if (data->roc)
5258                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5259                 break;
5260         default:
5261                 if (nfs4_async_handle_error(task, data->res.server,
5262                                             NULL, NULL) == -EAGAIN) {
5263                         rpc_restart_call_prepare(task);
5264                         return;
5265                 }
5266         }
5267         data->rpc_status = task->tk_status;
5268 }
5269
5270 static void nfs4_delegreturn_release(void *calldata)
5271 {
5272         struct nfs4_delegreturndata *data = calldata;
5273         struct inode *inode = data->inode;
5274
5275         if (inode) {
5276                 if (data->roc)
5277                         pnfs_roc_release(inode);
5278                 nfs_iput_and_deactive(inode);
5279         }
5280         kfree(calldata);
5281 }
5282
5283 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5284 {
5285         struct nfs4_delegreturndata *d_data;
5286
5287         d_data = (struct nfs4_delegreturndata *)data;
5288
5289         if (d_data->roc)
5290                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5291
5292         nfs4_setup_sequence(d_data->res.server,
5293                         &d_data->args.seq_args,
5294                         &d_data->res.seq_res,
5295                         task);
5296 }
5297
5298 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5299         .rpc_call_prepare = nfs4_delegreturn_prepare,
5300         .rpc_call_done = nfs4_delegreturn_done,
5301         .rpc_release = nfs4_delegreturn_release,
5302 };
5303
5304 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5305 {
5306         struct nfs4_delegreturndata *data;
5307         struct nfs_server *server = NFS_SERVER(inode);
5308         struct rpc_task *task;
5309         struct rpc_message msg = {
5310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5311                 .rpc_cred = cred,
5312         };
5313         struct rpc_task_setup task_setup_data = {
5314                 .rpc_client = server->client,
5315                 .rpc_message = &msg,
5316                 .callback_ops = &nfs4_delegreturn_ops,
5317                 .flags = RPC_TASK_ASYNC,
5318         };
5319         int status = 0;
5320
5321         data = kzalloc(sizeof(*data), GFP_NOFS);
5322         if (data == NULL)
5323                 return -ENOMEM;
5324         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5325         data->args.fhandle = &data->fh;
5326         data->args.stateid = &data->stateid;
5327         data->args.bitmask = server->cache_consistency_bitmask;
5328         nfs_copy_fh(&data->fh, NFS_FH(inode));
5329         nfs4_stateid_copy(&data->stateid, stateid);
5330         data->res.fattr = &data->fattr;
5331         data->res.server = server;
5332         nfs_fattr_init(data->res.fattr);
5333         data->timestamp = jiffies;
5334         data->rpc_status = 0;
5335         data->inode = nfs_igrab_and_active(inode);
5336         if (data->inode)
5337                 data->roc = nfs4_roc(inode);
5338
5339         task_setup_data.callback_data = data;
5340         msg.rpc_argp = &data->args;
5341         msg.rpc_resp = &data->res;
5342         task = rpc_run_task(&task_setup_data);
5343         if (IS_ERR(task))
5344                 return PTR_ERR(task);
5345         if (!issync)
5346                 goto out;
5347         status = nfs4_wait_for_completion_rpc_task(task);
5348         if (status != 0)
5349                 goto out;
5350         status = data->rpc_status;
5351         if (status == 0)
5352                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5353         else
5354                 nfs_refresh_inode(inode, &data->fattr);
5355 out:
5356         rpc_put_task(task);
5357         return status;
5358 }
5359
5360 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5361 {
5362         struct nfs_server *server = NFS_SERVER(inode);
5363         struct nfs4_exception exception = { };
5364         int err;
5365         do {
5366                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5367                 trace_nfs4_delegreturn(inode, err);
5368                 switch (err) {
5369                         case -NFS4ERR_STALE_STATEID:
5370                         case -NFS4ERR_EXPIRED:
5371                         case 0:
5372                                 return 0;
5373                 }
5374                 err = nfs4_handle_exception(server, err, &exception);
5375         } while (exception.retry);
5376         return err;
5377 }
5378
5379 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5380 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5381
5382 /*
5383  * sleep, with exponential backoff, and retry the LOCK operation.
5384  */
5385 static unsigned long
5386 nfs4_set_lock_task_retry(unsigned long timeout)
5387 {
5388         freezable_schedule_timeout_killable_unsafe(timeout);
5389         timeout <<= 1;
5390         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5391                 return NFS4_LOCK_MAXTIMEOUT;
5392         return timeout;
5393 }
5394
5395 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5396 {
5397         struct inode *inode = state->inode;
5398         struct nfs_server *server = NFS_SERVER(inode);
5399         struct nfs_client *clp = server->nfs_client;
5400         struct nfs_lockt_args arg = {
5401                 .fh = NFS_FH(inode),
5402                 .fl = request,
5403         };
5404         struct nfs_lockt_res res = {
5405                 .denied = request,
5406         };
5407         struct rpc_message msg = {
5408                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5409                 .rpc_argp       = &arg,
5410                 .rpc_resp       = &res,
5411                 .rpc_cred       = state->owner->so_cred,
5412         };
5413         struct nfs4_lock_state *lsp;
5414         int status;
5415
5416         arg.lock_owner.clientid = clp->cl_clientid;
5417         status = nfs4_set_lock_state(state, request);
5418         if (status != 0)
5419                 goto out;
5420         lsp = request->fl_u.nfs4_fl.owner;
5421         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5422         arg.lock_owner.s_dev = server->s_dev;
5423         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5424         switch (status) {
5425                 case 0:
5426                         request->fl_type = F_UNLCK;
5427                         break;
5428                 case -NFS4ERR_DENIED:
5429                         status = 0;
5430         }
5431         request->fl_ops->fl_release_private(request);
5432         request->fl_ops = NULL;
5433 out:
5434         return status;
5435 }
5436
5437 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5438 {
5439         struct nfs4_exception exception = { };
5440         int err;
5441
5442         do {
5443                 err = _nfs4_proc_getlk(state, cmd, request);
5444                 trace_nfs4_get_lock(request, state, cmd, err);
5445                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5446                                 &exception);
5447         } while (exception.retry);
5448         return err;
5449 }
5450
5451 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5452 {
5453         int res = 0;
5454         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5455                 case FL_POSIX:
5456                         res = posix_lock_inode_wait(inode, fl);
5457                         break;
5458                 case FL_FLOCK:
5459                         res = flock_lock_inode_wait(inode, fl);
5460                         break;
5461                 default:
5462                         BUG();
5463         }
5464         return res;
5465 }
5466
5467 struct nfs4_unlockdata {
5468         struct nfs_locku_args arg;
5469         struct nfs_locku_res res;
5470         struct nfs4_lock_state *lsp;
5471         struct nfs_open_context *ctx;
5472         struct file_lock fl;
5473         const struct nfs_server *server;
5474         unsigned long timestamp;
5475 };
5476
5477 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5478                 struct nfs_open_context *ctx,
5479                 struct nfs4_lock_state *lsp,
5480                 struct nfs_seqid *seqid)
5481 {
5482         struct nfs4_unlockdata *p;
5483         struct inode *inode = lsp->ls_state->inode;
5484
5485         p = kzalloc(sizeof(*p), GFP_NOFS);
5486         if (p == NULL)
5487                 return NULL;
5488         p->arg.fh = NFS_FH(inode);
5489         p->arg.fl = &p->fl;
5490         p->arg.seqid = seqid;
5491         p->res.seqid = seqid;
5492         p->lsp = lsp;
5493         atomic_inc(&lsp->ls_count);
5494         /* Ensure we don't close file until we're done freeing locks! */
5495         p->ctx = get_nfs_open_context(ctx);
5496         memcpy(&p->fl, fl, sizeof(p->fl));
5497         p->server = NFS_SERVER(inode);
5498         return p;
5499 }
5500
5501 static void nfs4_locku_release_calldata(void *data)
5502 {
5503         struct nfs4_unlockdata *calldata = data;
5504         nfs_free_seqid(calldata->arg.seqid);
5505         nfs4_put_lock_state(calldata->lsp);
5506         put_nfs_open_context(calldata->ctx);
5507         kfree(calldata);
5508 }
5509
5510 static void nfs4_locku_done(struct rpc_task *task, void *data)
5511 {
5512         struct nfs4_unlockdata *calldata = data;
5513
5514         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5515                 return;
5516         switch (task->tk_status) {
5517                 case 0:
5518                         renew_lease(calldata->server, calldata->timestamp);
5519                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5520                         if (nfs4_update_lock_stateid(calldata->lsp,
5521                                         &calldata->res.stateid))
5522                                 break;
5523                 case -NFS4ERR_BAD_STATEID:
5524                 case -NFS4ERR_OLD_STATEID:
5525                 case -NFS4ERR_STALE_STATEID:
5526                 case -NFS4ERR_EXPIRED:
5527                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5528                                                 &calldata->lsp->ls_stateid))
5529                                 rpc_restart_call_prepare(task);
5530                         break;
5531                 default:
5532                         if (nfs4_async_handle_error(task, calldata->server,
5533                                                     NULL, NULL) == -EAGAIN)
5534                                 rpc_restart_call_prepare(task);
5535         }
5536         nfs_release_seqid(calldata->arg.seqid);
5537 }
5538
5539 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5540 {
5541         struct nfs4_unlockdata *calldata = data;
5542
5543         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5544                 goto out_wait;
5545         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5546         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5547                 /* Note: exit _without_ running nfs4_locku_done */
5548                 goto out_no_action;
5549         }
5550         calldata->timestamp = jiffies;
5551         if (nfs4_setup_sequence(calldata->server,
5552                                 &calldata->arg.seq_args,
5553                                 &calldata->res.seq_res,
5554                                 task) != 0)
5555                 nfs_release_seqid(calldata->arg.seqid);
5556         return;
5557 out_no_action:
5558         task->tk_action = NULL;
5559 out_wait:
5560         nfs4_sequence_done(task, &calldata->res.seq_res);
5561 }
5562
5563 static const struct rpc_call_ops nfs4_locku_ops = {
5564         .rpc_call_prepare = nfs4_locku_prepare,
5565         .rpc_call_done = nfs4_locku_done,
5566         .rpc_release = nfs4_locku_release_calldata,
5567 };
5568
5569 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5570                 struct nfs_open_context *ctx,
5571                 struct nfs4_lock_state *lsp,
5572                 struct nfs_seqid *seqid)
5573 {
5574         struct nfs4_unlockdata *data;
5575         struct rpc_message msg = {
5576                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5577                 .rpc_cred = ctx->cred,
5578         };
5579         struct rpc_task_setup task_setup_data = {
5580                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5581                 .rpc_message = &msg,
5582                 .callback_ops = &nfs4_locku_ops,
5583                 .workqueue = nfsiod_workqueue,
5584                 .flags = RPC_TASK_ASYNC,
5585         };
5586
5587         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5588                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5589
5590         /* Ensure this is an unlock - when canceling a lock, the
5591          * canceled lock is passed in, and it won't be an unlock.
5592          */
5593         fl->fl_type = F_UNLCK;
5594
5595         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5596         if (data == NULL) {
5597                 nfs_free_seqid(seqid);
5598                 return ERR_PTR(-ENOMEM);
5599         }
5600
5601         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5602         msg.rpc_argp = &data->arg;
5603         msg.rpc_resp = &data->res;
5604         task_setup_data.callback_data = data;
5605         return rpc_run_task(&task_setup_data);
5606 }
5607
5608 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5609 {
5610         struct inode *inode = state->inode;
5611         struct nfs4_state_owner *sp = state->owner;
5612         struct nfs_inode *nfsi = NFS_I(inode);
5613         struct nfs_seqid *seqid;
5614         struct nfs4_lock_state *lsp;
5615         struct rpc_task *task;
5616         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5617         int status = 0;
5618         unsigned char fl_flags = request->fl_flags;
5619
5620         status = nfs4_set_lock_state(state, request);
5621         /* Unlock _before_ we do the RPC call */
5622         request->fl_flags |= FL_EXISTS;
5623         /* Exclude nfs_delegation_claim_locks() */
5624         mutex_lock(&sp->so_delegreturn_mutex);
5625         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5626         down_read(&nfsi->rwsem);
5627         if (do_vfs_lock(inode, request) == -ENOENT) {
5628                 up_read(&nfsi->rwsem);
5629                 mutex_unlock(&sp->so_delegreturn_mutex);
5630                 goto out;
5631         }
5632         up_read(&nfsi->rwsem);
5633         mutex_unlock(&sp->so_delegreturn_mutex);
5634         if (status != 0)
5635                 goto out;
5636         /* Is this a delegated lock? */
5637         lsp = request->fl_u.nfs4_fl.owner;
5638         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5639                 goto out;
5640         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5641         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5642         status = -ENOMEM;
5643         if (IS_ERR(seqid))
5644                 goto out;
5645         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5646         status = PTR_ERR(task);
5647         if (IS_ERR(task))
5648                 goto out;
5649         status = nfs4_wait_for_completion_rpc_task(task);
5650         rpc_put_task(task);
5651 out:
5652         request->fl_flags = fl_flags;
5653         trace_nfs4_unlock(request, state, F_SETLK, status);
5654         return status;
5655 }
5656
5657 struct nfs4_lockdata {
5658         struct nfs_lock_args arg;
5659         struct nfs_lock_res res;
5660         struct nfs4_lock_state *lsp;
5661         struct nfs_open_context *ctx;
5662         struct file_lock fl;
5663         unsigned long timestamp;
5664         int rpc_status;
5665         int cancelled;
5666         struct nfs_server *server;
5667 };
5668
5669 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5670                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5671                 gfp_t gfp_mask)
5672 {
5673         struct nfs4_lockdata *p;
5674         struct inode *inode = lsp->ls_state->inode;
5675         struct nfs_server *server = NFS_SERVER(inode);
5676         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5677
5678         p = kzalloc(sizeof(*p), gfp_mask);
5679         if (p == NULL)
5680                 return NULL;
5681
5682         p->arg.fh = NFS_FH(inode);
5683         p->arg.fl = &p->fl;
5684         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5685         if (IS_ERR(p->arg.open_seqid))
5686                 goto out_free;
5687         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5688         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5689         if (IS_ERR(p->arg.lock_seqid))
5690                 goto out_free_seqid;
5691         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5692         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5693         p->arg.lock_owner.s_dev = server->s_dev;
5694         p->res.lock_seqid = p->arg.lock_seqid;
5695         p->lsp = lsp;
5696         p->server = server;
5697         atomic_inc(&lsp->ls_count);
5698         p->ctx = get_nfs_open_context(ctx);
5699         get_file(fl->fl_file);
5700         memcpy(&p->fl, fl, sizeof(p->fl));
5701         return p;
5702 out_free_seqid:
5703         nfs_free_seqid(p->arg.open_seqid);
5704 out_free:
5705         kfree(p);
5706         return NULL;
5707 }
5708
5709 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5710 {
5711         struct nfs4_lockdata *data = calldata;
5712         struct nfs4_state *state = data->lsp->ls_state;
5713
5714         dprintk("%s: begin!\n", __func__);
5715         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5716                 goto out_wait;
5717         /* Do we need to do an open_to_lock_owner? */
5718         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5719                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5720                         goto out_release_lock_seqid;
5721                 }
5722                 nfs4_stateid_copy(&data->arg.open_stateid,
5723                                 &state->open_stateid);
5724                 data->arg.new_lock_owner = 1;
5725                 data->res.open_seqid = data->arg.open_seqid;
5726         } else {
5727                 data->arg.new_lock_owner = 0;
5728                 nfs4_stateid_copy(&data->arg.lock_stateid,
5729                                 &data->lsp->ls_stateid);
5730         }
5731         if (!nfs4_valid_open_stateid(state)) {
5732                 data->rpc_status = -EBADF;
5733                 task->tk_action = NULL;
5734                 goto out_release_open_seqid;
5735         }
5736         data->timestamp = jiffies;
5737         if (nfs4_setup_sequence(data->server,
5738                                 &data->arg.seq_args,
5739                                 &data->res.seq_res,
5740                                 task) == 0)
5741                 return;
5742 out_release_open_seqid:
5743         nfs_release_seqid(data->arg.open_seqid);
5744 out_release_lock_seqid:
5745         nfs_release_seqid(data->arg.lock_seqid);
5746 out_wait:
5747         nfs4_sequence_done(task, &data->res.seq_res);
5748         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5749 }
5750
5751 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5752 {
5753         struct nfs4_lockdata *data = calldata;
5754         struct nfs4_lock_state *lsp = data->lsp;
5755
5756         dprintk("%s: begin!\n", __func__);
5757
5758         if (!nfs4_sequence_done(task, &data->res.seq_res))
5759                 return;
5760
5761         data->rpc_status = task->tk_status;
5762         switch (task->tk_status) {
5763         case 0:
5764                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5765                                 data->timestamp);
5766                 if (data->arg.new_lock) {
5767                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5768                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5769                                 rpc_restart_call_prepare(task);
5770                                 break;
5771                         }
5772                 }
5773                 if (data->arg.new_lock_owner != 0) {
5774                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5775                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5776                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5777                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5778                         rpc_restart_call_prepare(task);
5779                 break;
5780         case -NFS4ERR_BAD_STATEID:
5781         case -NFS4ERR_OLD_STATEID:
5782         case -NFS4ERR_STALE_STATEID:
5783         case -NFS4ERR_EXPIRED:
5784                 if (data->arg.new_lock_owner != 0) {
5785                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5786                                                 &lsp->ls_state->open_stateid))
5787                                 rpc_restart_call_prepare(task);
5788                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5789                                                 &lsp->ls_stateid))
5790                                 rpc_restart_call_prepare(task);
5791         }
5792         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5793 }
5794
5795 static void nfs4_lock_release(void *calldata)
5796 {
5797         struct nfs4_lockdata *data = calldata;
5798
5799         dprintk("%s: begin!\n", __func__);
5800         nfs_free_seqid(data->arg.open_seqid);
5801         if (data->cancelled != 0) {
5802                 struct rpc_task *task;
5803                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5804                                 data->arg.lock_seqid);
5805                 if (!IS_ERR(task))
5806                         rpc_put_task_async(task);
5807                 dprintk("%s: cancelling lock!\n", __func__);
5808         } else
5809                 nfs_free_seqid(data->arg.lock_seqid);
5810         nfs4_put_lock_state(data->lsp);
5811         put_nfs_open_context(data->ctx);
5812         fput(data->fl.fl_file);
5813         kfree(data);
5814         dprintk("%s: done!\n", __func__);
5815 }
5816
5817 static const struct rpc_call_ops nfs4_lock_ops = {
5818         .rpc_call_prepare = nfs4_lock_prepare,
5819         .rpc_call_done = nfs4_lock_done,
5820         .rpc_release = nfs4_lock_release,
5821 };
5822
5823 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5824 {
5825         switch (error) {
5826         case -NFS4ERR_ADMIN_REVOKED:
5827         case -NFS4ERR_BAD_STATEID:
5828                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5829                 if (new_lock_owner != 0 ||
5830                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5831                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5832                 break;
5833         case -NFS4ERR_STALE_STATEID:
5834                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5835         case -NFS4ERR_EXPIRED:
5836                 nfs4_schedule_lease_recovery(server->nfs_client);
5837         };
5838 }
5839
5840 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5841 {
5842         struct nfs4_lockdata *data;
5843         struct rpc_task *task;
5844         struct rpc_message msg = {
5845                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5846                 .rpc_cred = state->owner->so_cred,
5847         };
5848         struct rpc_task_setup task_setup_data = {
5849                 .rpc_client = NFS_CLIENT(state->inode),
5850                 .rpc_message = &msg,
5851                 .callback_ops = &nfs4_lock_ops,
5852                 .workqueue = nfsiod_workqueue,
5853                 .flags = RPC_TASK_ASYNC,
5854         };
5855         int ret;
5856
5857         dprintk("%s: begin!\n", __func__);
5858         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5859                         fl->fl_u.nfs4_fl.owner,
5860                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5861         if (data == NULL)
5862                 return -ENOMEM;
5863         if (IS_SETLKW(cmd))
5864                 data->arg.block = 1;
5865         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5866         msg.rpc_argp = &data->arg;
5867         msg.rpc_resp = &data->res;
5868         task_setup_data.callback_data = data;
5869         if (recovery_type > NFS_LOCK_NEW) {
5870                 if (recovery_type == NFS_LOCK_RECLAIM)
5871                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5872                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5873         } else
5874                 data->arg.new_lock = 1;
5875         task = rpc_run_task(&task_setup_data);
5876         if (IS_ERR(task))
5877                 return PTR_ERR(task);
5878         ret = nfs4_wait_for_completion_rpc_task(task);
5879         if (ret == 0) {
5880                 ret = data->rpc_status;
5881                 if (ret)
5882                         nfs4_handle_setlk_error(data->server, data->lsp,
5883                                         data->arg.new_lock_owner, ret);
5884         } else
5885                 data->cancelled = 1;
5886         rpc_put_task(task);
5887         dprintk("%s: done, ret = %d!\n", __func__, ret);
5888         return ret;
5889 }
5890
5891 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5892 {
5893         struct nfs_server *server = NFS_SERVER(state->inode);
5894         struct nfs4_exception exception = {
5895                 .inode = state->inode,
5896         };
5897         int err;
5898
5899         do {
5900                 /* Cache the lock if possible... */
5901                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5902                         return 0;
5903                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5904                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5905                 if (err != -NFS4ERR_DELAY)
5906                         break;
5907                 nfs4_handle_exception(server, err, &exception);
5908         } while (exception.retry);
5909         return err;
5910 }
5911
5912 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5913 {
5914         struct nfs_server *server = NFS_SERVER(state->inode);
5915         struct nfs4_exception exception = {
5916                 .inode = state->inode,
5917         };
5918         int err;
5919
5920         err = nfs4_set_lock_state(state, request);
5921         if (err != 0)
5922                 return err;
5923         if (!recover_lost_locks) {
5924                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5925                 return 0;
5926         }
5927         do {
5928                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5929                         return 0;
5930                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5931                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5932                 switch (err) {
5933                 default:
5934                         goto out;
5935                 case -NFS4ERR_GRACE:
5936                 case -NFS4ERR_DELAY:
5937                         nfs4_handle_exception(server, err, &exception);
5938                         err = 0;
5939                 }
5940         } while (exception.retry);
5941 out:
5942         return err;
5943 }
5944
5945 #if defined(CONFIG_NFS_V4_1)
5946 /**
5947  * nfs41_check_expired_locks - possibly free a lock stateid
5948  *
5949  * @state: NFSv4 state for an inode
5950  *
5951  * Returns NFS_OK if recovery for this stateid is now finished.
5952  * Otherwise a negative NFS4ERR value is returned.
5953  */
5954 static int nfs41_check_expired_locks(struct nfs4_state *state)
5955 {
5956         int status, ret = -NFS4ERR_BAD_STATEID;
5957         struct nfs4_lock_state *lsp;
5958         struct nfs_server *server = NFS_SERVER(state->inode);
5959
5960         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5961                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5962                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5963
5964                         status = nfs41_test_stateid(server,
5965                                         &lsp->ls_stateid,
5966                                         cred);
5967                         trace_nfs4_test_lock_stateid(state, lsp, status);
5968                         if (status != NFS_OK) {
5969                                 /* Free the stateid unless the server
5970                                  * informs us the stateid is unrecognized. */
5971                                 if (status != -NFS4ERR_BAD_STATEID)
5972                                         nfs41_free_stateid(server,
5973                                                         &lsp->ls_stateid,
5974                                                         cred);
5975                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5976                                 ret = status;
5977                         }
5978                 }
5979         };
5980
5981         return ret;
5982 }
5983
5984 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5985 {
5986         int status = NFS_OK;
5987
5988         if (test_bit(LK_STATE_IN_USE, &state->flags))
5989                 status = nfs41_check_expired_locks(state);
5990         if (status != NFS_OK)
5991                 status = nfs4_lock_expired(state, request);
5992         return status;
5993 }
5994 #endif
5995
5996 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5997 {
5998         struct nfs_inode *nfsi = NFS_I(state->inode);
5999         unsigned char fl_flags = request->fl_flags;
6000         int status = -ENOLCK;
6001
6002         if ((fl_flags & FL_POSIX) &&
6003                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6004                 goto out;
6005         /* Is this a delegated open? */
6006         status = nfs4_set_lock_state(state, request);
6007         if (status != 0)
6008                 goto out;
6009         request->fl_flags |= FL_ACCESS;
6010         status = do_vfs_lock(state->inode, request);
6011         if (status < 0)
6012                 goto out;
6013         down_read(&nfsi->rwsem);
6014         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6015                 /* Yes: cache locks! */
6016                 /* ...but avoid races with delegation recall... */
6017                 request->fl_flags = fl_flags & ~FL_SLEEP;
6018                 status = do_vfs_lock(state->inode, request);
6019                 up_read(&nfsi->rwsem);
6020                 goto out;
6021         }
6022         up_read(&nfsi->rwsem);
6023         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6024 out:
6025         request->fl_flags = fl_flags;
6026         return status;
6027 }
6028
6029 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6030 {
6031         struct nfs4_exception exception = {
6032                 .state = state,
6033                 .inode = state->inode,
6034         };
6035         int err;
6036
6037         do {
6038                 err = _nfs4_proc_setlk(state, cmd, request);
6039                 trace_nfs4_set_lock(request, state, cmd, err);
6040                 if (err == -NFS4ERR_DENIED)
6041                         err = -EAGAIN;
6042                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6043                                 err, &exception);
6044         } while (exception.retry);
6045         return err;
6046 }
6047
6048 static int
6049 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6050 {
6051         struct nfs_open_context *ctx;
6052         struct nfs4_state *state;
6053         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6054         int status;
6055
6056         /* verify open state */
6057         ctx = nfs_file_open_context(filp);
6058         state = ctx->state;
6059
6060         if (request->fl_start < 0 || request->fl_end < 0)
6061                 return -EINVAL;
6062
6063         if (IS_GETLK(cmd)) {
6064                 if (state != NULL)
6065                         return nfs4_proc_getlk(state, F_GETLK, request);
6066                 return 0;
6067         }
6068
6069         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6070                 return -EINVAL;
6071
6072         if (request->fl_type == F_UNLCK) {
6073                 if (state != NULL)
6074                         return nfs4_proc_unlck(state, cmd, request);
6075                 return 0;
6076         }
6077
6078         if (state == NULL)
6079                 return -ENOLCK;
6080         /*
6081          * Don't rely on the VFS having checked the file open mode,
6082          * since it won't do this for flock() locks.
6083          */
6084         switch (request->fl_type) {
6085         case F_RDLCK:
6086                 if (!(filp->f_mode & FMODE_READ))
6087                         return -EBADF;
6088                 break;
6089         case F_WRLCK:
6090                 if (!(filp->f_mode & FMODE_WRITE))
6091                         return -EBADF;
6092         }
6093
6094         do {
6095                 status = nfs4_proc_setlk(state, cmd, request);
6096                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6097                         break;
6098                 timeout = nfs4_set_lock_task_retry(timeout);
6099                 status = -ERESTARTSYS;
6100                 if (signalled())
6101                         break;
6102         } while(status < 0);
6103         return status;
6104 }
6105
6106 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6107 {
6108         struct nfs_server *server = NFS_SERVER(state->inode);
6109         int err;
6110
6111         err = nfs4_set_lock_state(state, fl);
6112         if (err != 0)
6113                 return err;
6114         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6115         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6116 }
6117
6118 struct nfs_release_lockowner_data {
6119         struct nfs4_lock_state *lsp;
6120         struct nfs_server *server;
6121         struct nfs_release_lockowner_args args;
6122         struct nfs_release_lockowner_res res;
6123         unsigned long timestamp;
6124 };
6125
6126 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6127 {
6128         struct nfs_release_lockowner_data *data = calldata;
6129         struct nfs_server *server = data->server;
6130         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6131                              &data->args.seq_args, &data->res.seq_res, task);
6132         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6133         data->timestamp = jiffies;
6134 }
6135
6136 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6137 {
6138         struct nfs_release_lockowner_data *data = calldata;
6139         struct nfs_server *server = data->server;
6140
6141         nfs40_sequence_done(task, &data->res.seq_res);
6142
6143         switch (task->tk_status) {
6144         case 0:
6145                 renew_lease(server, data->timestamp);
6146                 break;
6147         case -NFS4ERR_STALE_CLIENTID:
6148         case -NFS4ERR_EXPIRED:
6149                 nfs4_schedule_lease_recovery(server->nfs_client);
6150                 break;
6151         case -NFS4ERR_LEASE_MOVED:
6152         case -NFS4ERR_DELAY:
6153                 if (nfs4_async_handle_error(task, server,
6154                                             NULL, NULL) == -EAGAIN)
6155                         rpc_restart_call_prepare(task);
6156         }
6157 }
6158
6159 static void nfs4_release_lockowner_release(void *calldata)
6160 {
6161         struct nfs_release_lockowner_data *data = calldata;
6162         nfs4_free_lock_state(data->server, data->lsp);
6163         kfree(calldata);
6164 }
6165
6166 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6167         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6168         .rpc_call_done = nfs4_release_lockowner_done,
6169         .rpc_release = nfs4_release_lockowner_release,
6170 };
6171
6172 static void
6173 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6174 {
6175         struct nfs_release_lockowner_data *data;
6176         struct rpc_message msg = {
6177                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6178         };
6179
6180         if (server->nfs_client->cl_mvops->minor_version != 0)
6181                 return;
6182
6183         data = kmalloc(sizeof(*data), GFP_NOFS);
6184         if (!data)
6185                 return;
6186         data->lsp = lsp;
6187         data->server = server;
6188         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6189         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6190         data->args.lock_owner.s_dev = server->s_dev;
6191
6192         msg.rpc_argp = &data->args;
6193         msg.rpc_resp = &data->res;
6194         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6195         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6196 }
6197
6198 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6199
6200 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6201                                    const void *buf, size_t buflen,
6202                                    int flags, int type)
6203 {
6204         if (strcmp(key, "") != 0)
6205                 return -EINVAL;
6206
6207         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6208 }
6209
6210 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6211                                    void *buf, size_t buflen, int type)
6212 {
6213         if (strcmp(key, "") != 0)
6214                 return -EINVAL;
6215
6216         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6217 }
6218
6219 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6220                                        size_t list_len, const char *name,
6221                                        size_t name_len, int type)
6222 {
6223         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6224
6225         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6226                 return 0;
6227
6228         if (list && len <= list_len)
6229                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6230         return len;
6231 }
6232
6233 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6234 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6235 {
6236         return server->caps & NFS_CAP_SECURITY_LABEL;
6237 }
6238
6239 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6240                                    const void *buf, size_t buflen,
6241                                    int flags, int type)
6242 {
6243         if (security_ismaclabel(key))
6244                 return nfs4_set_security_label(dentry, buf, buflen);
6245
6246         return -EOPNOTSUPP;
6247 }
6248
6249 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6250                                    void *buf, size_t buflen, int type)
6251 {
6252         if (security_ismaclabel(key))
6253                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6254         return -EOPNOTSUPP;
6255 }
6256
6257 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6258                                        size_t list_len, const char *name,
6259                                        size_t name_len, int type)
6260 {
6261         size_t len = 0;
6262
6263         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6264                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6265                 if (list && len <= list_len)
6266                         security_inode_listsecurity(d_inode(dentry), list, len);
6267         }
6268         return len;
6269 }
6270
6271 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6272         .prefix = XATTR_SECURITY_PREFIX,
6273         .list   = nfs4_xattr_list_nfs4_label,
6274         .get    = nfs4_xattr_get_nfs4_label,
6275         .set    = nfs4_xattr_set_nfs4_label,
6276 };
6277 #endif
6278
6279
6280 /*
6281  * nfs_fhget will use either the mounted_on_fileid or the fileid
6282  */
6283 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6284 {
6285         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6286                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6287               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6288               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6289                 return;
6290
6291         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6292                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6293         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6294         fattr->nlink = 2;
6295 }
6296
6297 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6298                                    const struct qstr *name,
6299                                    struct nfs4_fs_locations *fs_locations,
6300                                    struct page *page)
6301 {
6302         struct nfs_server *server = NFS_SERVER(dir);
6303         u32 bitmask[3] = {
6304                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6305         };
6306         struct nfs4_fs_locations_arg args = {
6307                 .dir_fh = NFS_FH(dir),
6308                 .name = name,
6309                 .page = page,
6310                 .bitmask = bitmask,
6311         };
6312         struct nfs4_fs_locations_res res = {
6313                 .fs_locations = fs_locations,
6314         };
6315         struct rpc_message msg = {
6316                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6317                 .rpc_argp = &args,
6318                 .rpc_resp = &res,
6319         };
6320         int status;
6321
6322         dprintk("%s: start\n", __func__);
6323
6324         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6325          * is not supported */
6326         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6327                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6328         else
6329                 bitmask[0] |= FATTR4_WORD0_FILEID;
6330
6331         nfs_fattr_init(&fs_locations->fattr);
6332         fs_locations->server = server;
6333         fs_locations->nlocations = 0;
6334         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6335         dprintk("%s: returned status = %d\n", __func__, status);
6336         return status;
6337 }
6338
6339 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6340                            const struct qstr *name,
6341                            struct nfs4_fs_locations *fs_locations,
6342                            struct page *page)
6343 {
6344         struct nfs4_exception exception = { };
6345         int err;
6346         do {
6347                 err = _nfs4_proc_fs_locations(client, dir, name,
6348                                 fs_locations, page);
6349                 trace_nfs4_get_fs_locations(dir, name, err);
6350                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6351                                 &exception);
6352         } while (exception.retry);
6353         return err;
6354 }
6355
6356 /*
6357  * This operation also signals the server that this client is
6358  * performing migration recovery.  The server can stop returning
6359  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6360  * appended to this compound to identify the client ID which is
6361  * performing recovery.
6362  */
6363 static int _nfs40_proc_get_locations(struct inode *inode,
6364                                      struct nfs4_fs_locations *locations,
6365                                      struct page *page, struct rpc_cred *cred)
6366 {
6367         struct nfs_server *server = NFS_SERVER(inode);
6368         struct rpc_clnt *clnt = server->client;
6369         u32 bitmask[2] = {
6370                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6371         };
6372         struct nfs4_fs_locations_arg args = {
6373                 .clientid       = server->nfs_client->cl_clientid,
6374                 .fh             = NFS_FH(inode),
6375                 .page           = page,
6376                 .bitmask        = bitmask,
6377                 .migration      = 1,            /* skip LOOKUP */
6378                 .renew          = 1,            /* append RENEW */
6379         };
6380         struct nfs4_fs_locations_res res = {
6381                 .fs_locations   = locations,
6382                 .migration      = 1,
6383                 .renew          = 1,
6384         };
6385         struct rpc_message msg = {
6386                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6387                 .rpc_argp       = &args,
6388                 .rpc_resp       = &res,
6389                 .rpc_cred       = cred,
6390         };
6391         unsigned long now = jiffies;
6392         int status;
6393
6394         nfs_fattr_init(&locations->fattr);
6395         locations->server = server;
6396         locations->nlocations = 0;
6397
6398         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6399         nfs4_set_sequence_privileged(&args.seq_args);
6400         status = nfs4_call_sync_sequence(clnt, server, &msg,
6401                                         &args.seq_args, &res.seq_res);
6402         if (status)
6403                 return status;
6404
6405         renew_lease(server, now);
6406         return 0;
6407 }
6408
6409 #ifdef CONFIG_NFS_V4_1
6410
6411 /*
6412  * This operation also signals the server that this client is
6413  * performing migration recovery.  The server can stop asserting
6414  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6415  * performing this operation is identified in the SEQUENCE
6416  * operation in this compound.
6417  *
6418  * When the client supports GETATTR(fs_locations_info), it can
6419  * be plumbed in here.
6420  */
6421 static int _nfs41_proc_get_locations(struct inode *inode,
6422                                      struct nfs4_fs_locations *locations,
6423                                      struct page *page, struct rpc_cred *cred)
6424 {
6425         struct nfs_server *server = NFS_SERVER(inode);
6426         struct rpc_clnt *clnt = server->client;
6427         u32 bitmask[2] = {
6428                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6429         };
6430         struct nfs4_fs_locations_arg args = {
6431                 .fh             = NFS_FH(inode),
6432                 .page           = page,
6433                 .bitmask        = bitmask,
6434                 .migration      = 1,            /* skip LOOKUP */
6435         };
6436         struct nfs4_fs_locations_res res = {
6437                 .fs_locations   = locations,
6438                 .migration      = 1,
6439         };
6440         struct rpc_message msg = {
6441                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6442                 .rpc_argp       = &args,
6443                 .rpc_resp       = &res,
6444                 .rpc_cred       = cred,
6445         };
6446         int status;
6447
6448         nfs_fattr_init(&locations->fattr);
6449         locations->server = server;
6450         locations->nlocations = 0;
6451
6452         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6453         nfs4_set_sequence_privileged(&args.seq_args);
6454         status = nfs4_call_sync_sequence(clnt, server, &msg,
6455                                         &args.seq_args, &res.seq_res);
6456         if (status == NFS4_OK &&
6457             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6458                 status = -NFS4ERR_LEASE_MOVED;
6459         return status;
6460 }
6461
6462 #endif  /* CONFIG_NFS_V4_1 */
6463
6464 /**
6465  * nfs4_proc_get_locations - discover locations for a migrated FSID
6466  * @inode: inode on FSID that is migrating
6467  * @locations: result of query
6468  * @page: buffer
6469  * @cred: credential to use for this operation
6470  *
6471  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6472  * operation failed, or a negative errno if a local error occurred.
6473  *
6474  * On success, "locations" is filled in, but if the server has
6475  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6476  * asserted.
6477  *
6478  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6479  * from this client that require migration recovery.
6480  */
6481 int nfs4_proc_get_locations(struct inode *inode,
6482                             struct nfs4_fs_locations *locations,
6483                             struct page *page, struct rpc_cred *cred)
6484 {
6485         struct nfs_server *server = NFS_SERVER(inode);
6486         struct nfs_client *clp = server->nfs_client;
6487         const struct nfs4_mig_recovery_ops *ops =
6488                                         clp->cl_mvops->mig_recovery_ops;
6489         struct nfs4_exception exception = { };
6490         int status;
6491
6492         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6493                 (unsigned long long)server->fsid.major,
6494                 (unsigned long long)server->fsid.minor,
6495                 clp->cl_hostname);
6496         nfs_display_fhandle(NFS_FH(inode), __func__);
6497
6498         do {
6499                 status = ops->get_locations(inode, locations, page, cred);
6500                 if (status != -NFS4ERR_DELAY)
6501                         break;
6502                 nfs4_handle_exception(server, status, &exception);
6503         } while (exception.retry);
6504         return status;
6505 }
6506
6507 /*
6508  * This operation also signals the server that this client is
6509  * performing "lease moved" recovery.  The server can stop
6510  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6511  * is appended to this compound to identify the client ID which is
6512  * performing recovery.
6513  */
6514 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6515 {
6516         struct nfs_server *server = NFS_SERVER(inode);
6517         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6518         struct rpc_clnt *clnt = server->client;
6519         struct nfs4_fsid_present_arg args = {
6520                 .fh             = NFS_FH(inode),
6521                 .clientid       = clp->cl_clientid,
6522                 .renew          = 1,            /* append RENEW */
6523         };
6524         struct nfs4_fsid_present_res res = {
6525                 .renew          = 1,
6526         };
6527         struct rpc_message msg = {
6528                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6529                 .rpc_argp       = &args,
6530                 .rpc_resp       = &res,
6531                 .rpc_cred       = cred,
6532         };
6533         unsigned long now = jiffies;
6534         int status;
6535
6536         res.fh = nfs_alloc_fhandle();
6537         if (res.fh == NULL)
6538                 return -ENOMEM;
6539
6540         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6541         nfs4_set_sequence_privileged(&args.seq_args);
6542         status = nfs4_call_sync_sequence(clnt, server, &msg,
6543                                                 &args.seq_args, &res.seq_res);
6544         nfs_free_fhandle(res.fh);
6545         if (status)
6546                 return status;
6547
6548         do_renew_lease(clp, now);
6549         return 0;
6550 }
6551
6552 #ifdef CONFIG_NFS_V4_1
6553
6554 /*
6555  * This operation also signals the server that this client is
6556  * performing "lease moved" recovery.  The server can stop asserting
6557  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6558  * this operation is identified in the SEQUENCE operation in this
6559  * compound.
6560  */
6561 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6562 {
6563         struct nfs_server *server = NFS_SERVER(inode);
6564         struct rpc_clnt *clnt = server->client;
6565         struct nfs4_fsid_present_arg args = {
6566                 .fh             = NFS_FH(inode),
6567         };
6568         struct nfs4_fsid_present_res res = {
6569         };
6570         struct rpc_message msg = {
6571                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6572                 .rpc_argp       = &args,
6573                 .rpc_resp       = &res,
6574                 .rpc_cred       = cred,
6575         };
6576         int status;
6577
6578         res.fh = nfs_alloc_fhandle();
6579         if (res.fh == NULL)
6580                 return -ENOMEM;
6581
6582         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6583         nfs4_set_sequence_privileged(&args.seq_args);
6584         status = nfs4_call_sync_sequence(clnt, server, &msg,
6585                                                 &args.seq_args, &res.seq_res);
6586         nfs_free_fhandle(res.fh);
6587         if (status == NFS4_OK &&
6588             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6589                 status = -NFS4ERR_LEASE_MOVED;
6590         return status;
6591 }
6592
6593 #endif  /* CONFIG_NFS_V4_1 */
6594
6595 /**
6596  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6597  * @inode: inode on FSID to check
6598  * @cred: credential to use for this operation
6599  *
6600  * Server indicates whether the FSID is present, moved, or not
6601  * recognized.  This operation is necessary to clear a LEASE_MOVED
6602  * condition for this client ID.
6603  *
6604  * Returns NFS4_OK if the FSID is present on this server,
6605  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6606  *  NFS4ERR code if some error occurred on the server, or a
6607  *  negative errno if a local failure occurred.
6608  */
6609 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6610 {
6611         struct nfs_server *server = NFS_SERVER(inode);
6612         struct nfs_client *clp = server->nfs_client;
6613         const struct nfs4_mig_recovery_ops *ops =
6614                                         clp->cl_mvops->mig_recovery_ops;
6615         struct nfs4_exception exception = { };
6616         int status;
6617
6618         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6619                 (unsigned long long)server->fsid.major,
6620                 (unsigned long long)server->fsid.minor,
6621                 clp->cl_hostname);
6622         nfs_display_fhandle(NFS_FH(inode), __func__);
6623
6624         do {
6625                 status = ops->fsid_present(inode, cred);
6626                 if (status != -NFS4ERR_DELAY)
6627                         break;
6628                 nfs4_handle_exception(server, status, &exception);
6629         } while (exception.retry);
6630         return status;
6631 }
6632
6633 /**
6634  * If 'use_integrity' is true and the state managment nfs_client
6635  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6636  * and the machine credential as per RFC3530bis and RFC5661 Security
6637  * Considerations sections. Otherwise, just use the user cred with the
6638  * filesystem's rpc_client.
6639  */
6640 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6641 {
6642         int status;
6643         struct nfs4_secinfo_arg args = {
6644                 .dir_fh = NFS_FH(dir),
6645                 .name   = name,
6646         };
6647         struct nfs4_secinfo_res res = {
6648                 .flavors     = flavors,
6649         };
6650         struct rpc_message msg = {
6651                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6652                 .rpc_argp = &args,
6653                 .rpc_resp = &res,
6654         };
6655         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6656         struct rpc_cred *cred = NULL;
6657
6658         if (use_integrity) {
6659                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6660                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6661                 msg.rpc_cred = cred;
6662         }
6663
6664         dprintk("NFS call  secinfo %s\n", name->name);
6665
6666         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6667                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6668
6669         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6670                                 &res.seq_res, 0);
6671         dprintk("NFS reply  secinfo: %d\n", status);
6672
6673         if (cred)
6674                 put_rpccred(cred);
6675
6676         return status;
6677 }
6678
6679 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6680                       struct nfs4_secinfo_flavors *flavors)
6681 {
6682         struct nfs4_exception exception = { };
6683         int err;
6684         do {
6685                 err = -NFS4ERR_WRONGSEC;
6686
6687                 /* try to use integrity protection with machine cred */
6688                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6689                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6690
6691                 /*
6692                  * if unable to use integrity protection, or SECINFO with
6693                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6694                  * disallowed by spec, but exists in deployed servers) use
6695                  * the current filesystem's rpc_client and the user cred.
6696                  */
6697                 if (err == -NFS4ERR_WRONGSEC)
6698                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6699
6700                 trace_nfs4_secinfo(dir, name, err);
6701                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6702                                 &exception);
6703         } while (exception.retry);
6704         return err;
6705 }
6706
6707 #ifdef CONFIG_NFS_V4_1
6708 /*
6709  * Check the exchange flags returned by the server for invalid flags, having
6710  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6711  * DS flags set.
6712  */
6713 static int nfs4_check_cl_exchange_flags(u32 flags)
6714 {
6715         if (flags & ~EXCHGID4_FLAG_MASK_R)
6716                 goto out_inval;
6717         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6718             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6719                 goto out_inval;
6720         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6721                 goto out_inval;
6722         return NFS_OK;
6723 out_inval:
6724         return -NFS4ERR_INVAL;
6725 }
6726
6727 static bool
6728 nfs41_same_server_scope(struct nfs41_server_scope *a,
6729                         struct nfs41_server_scope *b)
6730 {
6731         if (a->server_scope_sz == b->server_scope_sz &&
6732             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6733                 return true;
6734
6735         return false;
6736 }
6737
6738 /*
6739  * nfs4_proc_bind_conn_to_session()
6740  *
6741  * The 4.1 client currently uses the same TCP connection for the
6742  * fore and backchannel.
6743  */
6744 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6745 {
6746         int status;
6747         struct nfs41_bind_conn_to_session_args args = {
6748                 .client = clp,
6749                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6750         };
6751         struct nfs41_bind_conn_to_session_res res;
6752         struct rpc_message msg = {
6753                 .rpc_proc =
6754                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6755                 .rpc_argp = &args,
6756                 .rpc_resp = &res,
6757                 .rpc_cred = cred,
6758         };
6759
6760         dprintk("--> %s\n", __func__);
6761
6762         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6763         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6764                 args.dir = NFS4_CDFC4_FORE;
6765
6766         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6767         trace_nfs4_bind_conn_to_session(clp, status);
6768         if (status == 0) {
6769                 if (memcmp(res.sessionid.data,
6770                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6771                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6772                         status = -EIO;
6773                         goto out;
6774                 }
6775                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6776                         dprintk("NFS: %s: Unexpected direction from server\n",
6777                                 __func__);
6778                         status = -EIO;
6779                         goto out;
6780                 }
6781                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6782                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6783                                 __func__);
6784                         status = -EIO;
6785                         goto out;
6786                 }
6787         }
6788 out:
6789         dprintk("<-- %s status= %d\n", __func__, status);
6790         return status;
6791 }
6792
6793 /*
6794  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6795  * and operations we'd like to see to enable certain features in the allow map
6796  */
6797 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6798         .how = SP4_MACH_CRED,
6799         .enforce.u.words = {
6800                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6801                       1 << (OP_EXCHANGE_ID - 32) |
6802                       1 << (OP_CREATE_SESSION - 32) |
6803                       1 << (OP_DESTROY_SESSION - 32) |
6804                       1 << (OP_DESTROY_CLIENTID - 32)
6805         },
6806         .allow.u.words = {
6807                 [0] = 1 << (OP_CLOSE) |
6808                       1 << (OP_LOCKU) |
6809                       1 << (OP_COMMIT),
6810                 [1] = 1 << (OP_SECINFO - 32) |
6811                       1 << (OP_SECINFO_NO_NAME - 32) |
6812                       1 << (OP_TEST_STATEID - 32) |
6813                       1 << (OP_FREE_STATEID - 32) |
6814                       1 << (OP_WRITE - 32)
6815         }
6816 };
6817
6818 /*
6819  * Select the state protection mode for client `clp' given the server results
6820  * from exchange_id in `sp'.
6821  *
6822  * Returns 0 on success, negative errno otherwise.
6823  */
6824 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6825                                  struct nfs41_state_protection *sp)
6826 {
6827         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6828                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6829                       1 << (OP_EXCHANGE_ID - 32) |
6830                       1 << (OP_CREATE_SESSION - 32) |
6831                       1 << (OP_DESTROY_SESSION - 32) |
6832                       1 << (OP_DESTROY_CLIENTID - 32)
6833         };
6834         unsigned int i;
6835
6836         if (sp->how == SP4_MACH_CRED) {
6837                 /* Print state protect result */
6838                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6839                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6840                         if (test_bit(i, sp->enforce.u.longs))
6841                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6842                         if (test_bit(i, sp->allow.u.longs))
6843                                 dfprintk(MOUNT, "  allow op %d\n", i);
6844                 }
6845
6846                 /* make sure nothing is on enforce list that isn't supported */
6847                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6848                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6849                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6850                                 return -EINVAL;
6851                         }
6852                 }
6853
6854                 /*
6855                  * Minimal mode - state operations are allowed to use machine
6856                  * credential.  Note this already happens by default, so the
6857                  * client doesn't have to do anything more than the negotiation.
6858                  *
6859                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6860                  *       we're already using the machine cred for exchange_id
6861                  *       and will never use a different cred.
6862                  */
6863                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6864                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6865                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6866                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6867                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6868                         dfprintk(MOUNT, "  minimal mode enabled\n");
6869                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6870                 } else {
6871                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6872                         return -EINVAL;
6873                 }
6874
6875                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6876                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6877                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6878                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6879                 }
6880
6881                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6882                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6883                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6884                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6885                 }
6886
6887                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6888                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6889                         dfprintk(MOUNT, "  stateid mode enabled\n");
6890                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6891                 }
6892
6893                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6894                         dfprintk(MOUNT, "  write mode enabled\n");
6895                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6896                 }
6897
6898                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6899                         dfprintk(MOUNT, "  commit mode enabled\n");
6900                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6901                 }
6902         }
6903
6904         return 0;
6905 }
6906
6907 /*
6908  * _nfs4_proc_exchange_id()
6909  *
6910  * Wrapper for EXCHANGE_ID operation.
6911  */
6912 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6913         u32 sp4_how)
6914 {
6915         nfs4_verifier verifier;
6916         struct nfs41_exchange_id_args args = {
6917                 .verifier = &verifier,
6918                 .client = clp,
6919 #ifdef CONFIG_NFS_V4_1_MIGRATION
6920                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6921                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6922                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6923 #else
6924                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6925                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6926 #endif
6927         };
6928         struct nfs41_exchange_id_res res = {
6929                 0
6930         };
6931         int status;
6932         struct rpc_message msg = {
6933                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6934                 .rpc_argp = &args,
6935                 .rpc_resp = &res,
6936                 .rpc_cred = cred,
6937         };
6938
6939         nfs4_init_boot_verifier(clp, &verifier);
6940
6941         status = nfs4_init_uniform_client_string(clp);
6942         if (status)
6943                 goto out;
6944
6945         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6946                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6947                 clp->cl_owner_id);
6948
6949         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6950                                         GFP_NOFS);
6951         if (unlikely(res.server_owner == NULL)) {
6952                 status = -ENOMEM;
6953                 goto out;
6954         }
6955
6956         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6957                                         GFP_NOFS);
6958         if (unlikely(res.server_scope == NULL)) {
6959                 status = -ENOMEM;
6960                 goto out_server_owner;
6961         }
6962
6963         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6964         if (unlikely(res.impl_id == NULL)) {
6965                 status = -ENOMEM;
6966                 goto out_server_scope;
6967         }
6968
6969         switch (sp4_how) {
6970         case SP4_NONE:
6971                 args.state_protect.how = SP4_NONE;
6972                 break;
6973
6974         case SP4_MACH_CRED:
6975                 args.state_protect = nfs4_sp4_mach_cred_request;
6976                 break;
6977
6978         default:
6979                 /* unsupported! */
6980                 WARN_ON_ONCE(1);
6981                 status = -EINVAL;
6982                 goto out_impl_id;
6983         }
6984
6985         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6986         trace_nfs4_exchange_id(clp, status);
6987         if (status == 0)
6988                 status = nfs4_check_cl_exchange_flags(res.flags);
6989
6990         if (status == 0)
6991                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6992
6993         if (status == 0) {
6994                 clp->cl_clientid = res.clientid;
6995                 clp->cl_exchange_flags = res.flags;
6996                 /* Client ID is not confirmed */
6997                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6998                         clear_bit(NFS4_SESSION_ESTABLISHED,
6999                                         &clp->cl_session->session_state);
7000                         clp->cl_seqid = res.seqid;
7001                 }
7002
7003                 kfree(clp->cl_serverowner);
7004                 clp->cl_serverowner = res.server_owner;
7005                 res.server_owner = NULL;
7006
7007                 /* use the most recent implementation id */
7008                 kfree(clp->cl_implid);
7009                 clp->cl_implid = res.impl_id;
7010                 res.impl_id = NULL;
7011
7012                 if (clp->cl_serverscope != NULL &&
7013                     !nfs41_same_server_scope(clp->cl_serverscope,
7014                                              res.server_scope)) {
7015                         dprintk("%s: server_scope mismatch detected\n",
7016                                 __func__);
7017                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7018                         kfree(clp->cl_serverscope);
7019                         clp->cl_serverscope = NULL;
7020                 }
7021
7022                 if (clp->cl_serverscope == NULL) {
7023                         clp->cl_serverscope = res.server_scope;
7024                         res.server_scope = NULL;
7025                 }
7026         }
7027
7028 out_impl_id:
7029         kfree(res.impl_id);
7030 out_server_scope:
7031         kfree(res.server_scope);
7032 out_server_owner:
7033         kfree(res.server_owner);
7034 out:
7035         if (clp->cl_implid != NULL)
7036                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7037                         "domain: %s, name: %s, date: %llu,%u\n",
7038                         clp->cl_implid->domain, clp->cl_implid->name,
7039                         clp->cl_implid->date.seconds,
7040                         clp->cl_implid->date.nseconds);
7041         dprintk("NFS reply exchange_id: %d\n", status);
7042         return status;
7043 }
7044
7045 /*
7046  * nfs4_proc_exchange_id()
7047  *
7048  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7049  *
7050  * Since the clientid has expired, all compounds using sessions
7051  * associated with the stale clientid will be returning
7052  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7053  * be in some phase of session reset.
7054  *
7055  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7056  */
7057 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7058 {
7059         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7060         int status;
7061
7062         /* try SP4_MACH_CRED if krb5i/p */
7063         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7064             authflavor == RPC_AUTH_GSS_KRB5P) {
7065                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7066                 if (!status)
7067                         return 0;
7068         }
7069
7070         /* try SP4_NONE */
7071         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7072 }
7073
7074 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7075                 struct rpc_cred *cred)
7076 {
7077         struct rpc_message msg = {
7078                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7079                 .rpc_argp = clp,
7080                 .rpc_cred = cred,
7081         };
7082         int status;
7083
7084         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7085         trace_nfs4_destroy_clientid(clp, status);
7086         if (status)
7087                 dprintk("NFS: Got error %d from the server %s on "
7088                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7089         return status;
7090 }
7091
7092 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7093                 struct rpc_cred *cred)
7094 {
7095         unsigned int loop;
7096         int ret;
7097
7098         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7099                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7100                 switch (ret) {
7101                 case -NFS4ERR_DELAY:
7102                 case -NFS4ERR_CLIENTID_BUSY:
7103                         ssleep(1);
7104                         break;
7105                 default:
7106                         return ret;
7107                 }
7108         }
7109         return 0;
7110 }
7111
7112 int nfs4_destroy_clientid(struct nfs_client *clp)
7113 {
7114         struct rpc_cred *cred;
7115         int ret = 0;
7116
7117         if (clp->cl_mvops->minor_version < 1)
7118                 goto out;
7119         if (clp->cl_exchange_flags == 0)
7120                 goto out;
7121         if (clp->cl_preserve_clid)
7122                 goto out;
7123         cred = nfs4_get_clid_cred(clp);
7124         ret = nfs4_proc_destroy_clientid(clp, cred);
7125         if (cred)
7126                 put_rpccred(cred);
7127         switch (ret) {
7128         case 0:
7129         case -NFS4ERR_STALE_CLIENTID:
7130                 clp->cl_exchange_flags = 0;
7131         }
7132 out:
7133         return ret;
7134 }
7135
7136 struct nfs4_get_lease_time_data {
7137         struct nfs4_get_lease_time_args *args;
7138         struct nfs4_get_lease_time_res *res;
7139         struct nfs_client *clp;
7140 };
7141
7142 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7143                                         void *calldata)
7144 {
7145         struct nfs4_get_lease_time_data *data =
7146                         (struct nfs4_get_lease_time_data *)calldata;
7147
7148         dprintk("--> %s\n", __func__);
7149         /* just setup sequence, do not trigger session recovery
7150            since we're invoked within one */
7151         nfs41_setup_sequence(data->clp->cl_session,
7152                         &data->args->la_seq_args,
7153                         &data->res->lr_seq_res,
7154                         task);
7155         dprintk("<-- %s\n", __func__);
7156 }
7157
7158 /*
7159  * Called from nfs4_state_manager thread for session setup, so don't recover
7160  * from sequence operation or clientid errors.
7161  */
7162 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7163 {
7164         struct nfs4_get_lease_time_data *data =
7165                         (struct nfs4_get_lease_time_data *)calldata;
7166
7167         dprintk("--> %s\n", __func__);
7168         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7169                 return;
7170         switch (task->tk_status) {
7171         case -NFS4ERR_DELAY:
7172         case -NFS4ERR_GRACE:
7173                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7174                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7175                 task->tk_status = 0;
7176                 /* fall through */
7177         case -NFS4ERR_RETRY_UNCACHED_REP:
7178                 rpc_restart_call_prepare(task);
7179                 return;
7180         }
7181         dprintk("<-- %s\n", __func__);
7182 }
7183
7184 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7185         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7186         .rpc_call_done = nfs4_get_lease_time_done,
7187 };
7188
7189 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7190 {
7191         struct rpc_task *task;
7192         struct nfs4_get_lease_time_args args;
7193         struct nfs4_get_lease_time_res res = {
7194                 .lr_fsinfo = fsinfo,
7195         };
7196         struct nfs4_get_lease_time_data data = {
7197                 .args = &args,
7198                 .res = &res,
7199                 .clp = clp,
7200         };
7201         struct rpc_message msg = {
7202                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7203                 .rpc_argp = &args,
7204                 .rpc_resp = &res,
7205         };
7206         struct rpc_task_setup task_setup = {
7207                 .rpc_client = clp->cl_rpcclient,
7208                 .rpc_message = &msg,
7209                 .callback_ops = &nfs4_get_lease_time_ops,
7210                 .callback_data = &data,
7211                 .flags = RPC_TASK_TIMEOUT,
7212         };
7213         int status;
7214
7215         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7216         nfs4_set_sequence_privileged(&args.la_seq_args);
7217         dprintk("--> %s\n", __func__);
7218         task = rpc_run_task(&task_setup);
7219
7220         if (IS_ERR(task))
7221                 status = PTR_ERR(task);
7222         else {
7223                 status = task->tk_status;
7224                 rpc_put_task(task);
7225         }
7226         dprintk("<-- %s return %d\n", __func__, status);
7227
7228         return status;
7229 }
7230
7231 /*
7232  * Initialize the values to be used by the client in CREATE_SESSION
7233  * If nfs4_init_session set the fore channel request and response sizes,
7234  * use them.
7235  *
7236  * Set the back channel max_resp_sz_cached to zero to force the client to
7237  * always set csa_cachethis to FALSE because the current implementation
7238  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7239  */
7240 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7241 {
7242         unsigned int max_rqst_sz, max_resp_sz;
7243
7244         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7245         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7246
7247         /* Fore channel attributes */
7248         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7249         args->fc_attrs.max_resp_sz = max_resp_sz;
7250         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7251         args->fc_attrs.max_reqs = max_session_slots;
7252
7253         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7254                 "max_ops=%u max_reqs=%u\n",
7255                 __func__,
7256                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7257                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7258
7259         /* Back channel attributes */
7260         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7261         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7262         args->bc_attrs.max_resp_sz_cached = 0;
7263         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7264         args->bc_attrs.max_reqs = 1;
7265
7266         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7267                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7268                 __func__,
7269                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7270                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7271                 args->bc_attrs.max_reqs);
7272 }
7273
7274 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7275                 struct nfs41_create_session_res *res)
7276 {
7277         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7278         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7279
7280         if (rcvd->max_resp_sz > sent->max_resp_sz)
7281                 return -EINVAL;
7282         /*
7283          * Our requested max_ops is the minimum we need; we're not
7284          * prepared to break up compounds into smaller pieces than that.
7285          * So, no point even trying to continue if the server won't
7286          * cooperate:
7287          */
7288         if (rcvd->max_ops < sent->max_ops)
7289                 return -EINVAL;
7290         if (rcvd->max_reqs == 0)
7291                 return -EINVAL;
7292         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7293                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7294         return 0;
7295 }
7296
7297 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7298                 struct nfs41_create_session_res *res)
7299 {
7300         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7301         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7302
7303         if (!(res->flags & SESSION4_BACK_CHAN))
7304                 goto out;
7305         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7306                 return -EINVAL;
7307         if (rcvd->max_resp_sz < sent->max_resp_sz)
7308                 return -EINVAL;
7309         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7310                 return -EINVAL;
7311         /* These would render the backchannel useless: */
7312         if (rcvd->max_ops != sent->max_ops)
7313                 return -EINVAL;
7314         if (rcvd->max_reqs != sent->max_reqs)
7315                 return -EINVAL;
7316 out:
7317         return 0;
7318 }
7319
7320 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7321                                      struct nfs41_create_session_res *res)
7322 {
7323         int ret;
7324
7325         ret = nfs4_verify_fore_channel_attrs(args, res);
7326         if (ret)
7327                 return ret;
7328         return nfs4_verify_back_channel_attrs(args, res);
7329 }
7330
7331 static void nfs4_update_session(struct nfs4_session *session,
7332                 struct nfs41_create_session_res *res)
7333 {
7334         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7335         /* Mark client id and session as being confirmed */
7336         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7337         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7338         session->flags = res->flags;
7339         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7340         if (res->flags & SESSION4_BACK_CHAN)
7341                 memcpy(&session->bc_attrs, &res->bc_attrs,
7342                                 sizeof(session->bc_attrs));
7343 }
7344
7345 static int _nfs4_proc_create_session(struct nfs_client *clp,
7346                 struct rpc_cred *cred)
7347 {
7348         struct nfs4_session *session = clp->cl_session;
7349         struct nfs41_create_session_args args = {
7350                 .client = clp,
7351                 .clientid = clp->cl_clientid,
7352                 .seqid = clp->cl_seqid,
7353                 .cb_program = NFS4_CALLBACK,
7354         };
7355         struct nfs41_create_session_res res;
7356
7357         struct rpc_message msg = {
7358                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7359                 .rpc_argp = &args,
7360                 .rpc_resp = &res,
7361                 .rpc_cred = cred,
7362         };
7363         int status;
7364
7365         nfs4_init_channel_attrs(&args);
7366         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7367
7368         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7369         trace_nfs4_create_session(clp, status);
7370
7371         if (!status) {
7372                 /* Verify the session's negotiated channel_attrs values */
7373                 status = nfs4_verify_channel_attrs(&args, &res);
7374                 /* Increment the clientid slot sequence id */
7375                 if (clp->cl_seqid == res.seqid)
7376                         clp->cl_seqid++;
7377                 if (status)
7378                         goto out;
7379                 nfs4_update_session(session, &res);
7380         }
7381 out:
7382         return status;
7383 }
7384
7385 /*
7386  * Issues a CREATE_SESSION operation to the server.
7387  * It is the responsibility of the caller to verify the session is
7388  * expired before calling this routine.
7389  */
7390 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7391 {
7392         int status;
7393         unsigned *ptr;
7394         struct nfs4_session *session = clp->cl_session;
7395
7396         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7397
7398         status = _nfs4_proc_create_session(clp, cred);
7399         if (status)
7400                 goto out;
7401
7402         /* Init or reset the session slot tables */
7403         status = nfs4_setup_session_slot_tables(session);
7404         dprintk("slot table setup returned %d\n", status);
7405         if (status)
7406                 goto out;
7407
7408         ptr = (unsigned *)&session->sess_id.data[0];
7409         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7410                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7411 out:
7412         dprintk("<-- %s\n", __func__);
7413         return status;
7414 }
7415
7416 /*
7417  * Issue the over-the-wire RPC DESTROY_SESSION.
7418  * The caller must serialize access to this routine.
7419  */
7420 int nfs4_proc_destroy_session(struct nfs4_session *session,
7421                 struct rpc_cred *cred)
7422 {
7423         struct rpc_message msg = {
7424                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7425                 .rpc_argp = session,
7426                 .rpc_cred = cred,
7427         };
7428         int status = 0;
7429
7430         dprintk("--> nfs4_proc_destroy_session\n");
7431
7432         /* session is still being setup */
7433         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7434                 return 0;
7435
7436         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7437         trace_nfs4_destroy_session(session->clp, status);
7438
7439         if (status)
7440                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7441                         "Session has been destroyed regardless...\n", status);
7442
7443         dprintk("<-- nfs4_proc_destroy_session\n");
7444         return status;
7445 }
7446
7447 /*
7448  * Renew the cl_session lease.
7449  */
7450 struct nfs4_sequence_data {
7451         struct nfs_client *clp;
7452         struct nfs4_sequence_args args;
7453         struct nfs4_sequence_res res;
7454 };
7455
7456 static void nfs41_sequence_release(void *data)
7457 {
7458         struct nfs4_sequence_data *calldata = data;
7459         struct nfs_client *clp = calldata->clp;
7460
7461         if (atomic_read(&clp->cl_count) > 1)
7462                 nfs4_schedule_state_renewal(clp);
7463         nfs_put_client(clp);
7464         kfree(calldata);
7465 }
7466
7467 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7468 {
7469         switch(task->tk_status) {
7470         case -NFS4ERR_DELAY:
7471                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7472                 return -EAGAIN;
7473         default:
7474                 nfs4_schedule_lease_recovery(clp);
7475         }
7476         return 0;
7477 }
7478
7479 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7480 {
7481         struct nfs4_sequence_data *calldata = data;
7482         struct nfs_client *clp = calldata->clp;
7483
7484         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7485                 return;
7486
7487         trace_nfs4_sequence(clp, task->tk_status);
7488         if (task->tk_status < 0) {
7489                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7490                 if (atomic_read(&clp->cl_count) == 1)
7491                         goto out;
7492
7493                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7494                         rpc_restart_call_prepare(task);
7495                         return;
7496                 }
7497         }
7498         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7499 out:
7500         dprintk("<-- %s\n", __func__);
7501 }
7502
7503 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7504 {
7505         struct nfs4_sequence_data *calldata = data;
7506         struct nfs_client *clp = calldata->clp;
7507         struct nfs4_sequence_args *args;
7508         struct nfs4_sequence_res *res;
7509
7510         args = task->tk_msg.rpc_argp;
7511         res = task->tk_msg.rpc_resp;
7512
7513         nfs41_setup_sequence(clp->cl_session, args, res, task);
7514 }
7515
7516 static const struct rpc_call_ops nfs41_sequence_ops = {
7517         .rpc_call_done = nfs41_sequence_call_done,
7518         .rpc_call_prepare = nfs41_sequence_prepare,
7519         .rpc_release = nfs41_sequence_release,
7520 };
7521
7522 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7523                 struct rpc_cred *cred,
7524                 bool is_privileged)
7525 {
7526         struct nfs4_sequence_data *calldata;
7527         struct rpc_message msg = {
7528                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7529                 .rpc_cred = cred,
7530         };
7531         struct rpc_task_setup task_setup_data = {
7532                 .rpc_client = clp->cl_rpcclient,
7533                 .rpc_message = &msg,
7534                 .callback_ops = &nfs41_sequence_ops,
7535                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7536         };
7537
7538         if (!atomic_inc_not_zero(&clp->cl_count))
7539                 return ERR_PTR(-EIO);
7540         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7541         if (calldata == NULL) {
7542                 nfs_put_client(clp);
7543                 return ERR_PTR(-ENOMEM);
7544         }
7545         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7546         if (is_privileged)
7547                 nfs4_set_sequence_privileged(&calldata->args);
7548         msg.rpc_argp = &calldata->args;
7549         msg.rpc_resp = &calldata->res;
7550         calldata->clp = clp;
7551         task_setup_data.callback_data = calldata;
7552
7553         return rpc_run_task(&task_setup_data);
7554 }
7555
7556 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7557 {
7558         struct rpc_task *task;
7559         int ret = 0;
7560
7561         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7562                 return -EAGAIN;
7563         task = _nfs41_proc_sequence(clp, cred, false);
7564         if (IS_ERR(task))
7565                 ret = PTR_ERR(task);
7566         else
7567                 rpc_put_task_async(task);
7568         dprintk("<-- %s status=%d\n", __func__, ret);
7569         return ret;
7570 }
7571
7572 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7573 {
7574         struct rpc_task *task;
7575         int ret;
7576
7577         task = _nfs41_proc_sequence(clp, cred, true);
7578         if (IS_ERR(task)) {
7579                 ret = PTR_ERR(task);
7580                 goto out;
7581         }
7582         ret = rpc_wait_for_completion_task(task);
7583         if (!ret)
7584                 ret = task->tk_status;
7585         rpc_put_task(task);
7586 out:
7587         dprintk("<-- %s status=%d\n", __func__, ret);
7588         return ret;
7589 }
7590
7591 struct nfs4_reclaim_complete_data {
7592         struct nfs_client *clp;
7593         struct nfs41_reclaim_complete_args arg;
7594         struct nfs41_reclaim_complete_res res;
7595 };
7596
7597 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7598 {
7599         struct nfs4_reclaim_complete_data *calldata = data;
7600
7601         nfs41_setup_sequence(calldata->clp->cl_session,
7602                         &calldata->arg.seq_args,
7603                         &calldata->res.seq_res,
7604                         task);
7605 }
7606
7607 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7608 {
7609         switch(task->tk_status) {
7610         case 0:
7611         case -NFS4ERR_COMPLETE_ALREADY:
7612         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7613                 break;
7614         case -NFS4ERR_DELAY:
7615                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7616                 /* fall through */
7617         case -NFS4ERR_RETRY_UNCACHED_REP:
7618                 return -EAGAIN;
7619         default:
7620                 nfs4_schedule_lease_recovery(clp);
7621         }
7622         return 0;
7623 }
7624
7625 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7626 {
7627         struct nfs4_reclaim_complete_data *calldata = data;
7628         struct nfs_client *clp = calldata->clp;
7629         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7630
7631         dprintk("--> %s\n", __func__);
7632         if (!nfs41_sequence_done(task, res))
7633                 return;
7634
7635         trace_nfs4_reclaim_complete(clp, task->tk_status);
7636         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7637                 rpc_restart_call_prepare(task);
7638                 return;
7639         }
7640         dprintk("<-- %s\n", __func__);
7641 }
7642
7643 static void nfs4_free_reclaim_complete_data(void *data)
7644 {
7645         struct nfs4_reclaim_complete_data *calldata = data;
7646
7647         kfree(calldata);
7648 }
7649
7650 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7651         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7652         .rpc_call_done = nfs4_reclaim_complete_done,
7653         .rpc_release = nfs4_free_reclaim_complete_data,
7654 };
7655
7656 /*
7657  * Issue a global reclaim complete.
7658  */
7659 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7660                 struct rpc_cred *cred)
7661 {
7662         struct nfs4_reclaim_complete_data *calldata;
7663         struct rpc_task *task;
7664         struct rpc_message msg = {
7665                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7666                 .rpc_cred = cred,
7667         };
7668         struct rpc_task_setup task_setup_data = {
7669                 .rpc_client = clp->cl_rpcclient,
7670                 .rpc_message = &msg,
7671                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7672                 .flags = RPC_TASK_ASYNC,
7673         };
7674         int status = -ENOMEM;
7675
7676         dprintk("--> %s\n", __func__);
7677         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7678         if (calldata == NULL)
7679                 goto out;
7680         calldata->clp = clp;
7681         calldata->arg.one_fs = 0;
7682
7683         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7684         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7685         msg.rpc_argp = &calldata->arg;
7686         msg.rpc_resp = &calldata->res;
7687         task_setup_data.callback_data = calldata;
7688         task = rpc_run_task(&task_setup_data);
7689         if (IS_ERR(task)) {
7690                 status = PTR_ERR(task);
7691                 goto out;
7692         }
7693         status = nfs4_wait_for_completion_rpc_task(task);
7694         if (status == 0)
7695                 status = task->tk_status;
7696         rpc_put_task(task);
7697         return 0;
7698 out:
7699         dprintk("<-- %s status=%d\n", __func__, status);
7700         return status;
7701 }
7702
7703 static void
7704 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7705 {
7706         struct nfs4_layoutget *lgp = calldata;
7707         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7708         struct nfs4_session *session = nfs4_get_session(server);
7709
7710         dprintk("--> %s\n", __func__);
7711         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7712          * right now covering the LAYOUTGET we are about to send.
7713          * However, that is not so catastrophic, and there seems
7714          * to be no way to prevent it completely.
7715          */
7716         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7717                                 &lgp->res.seq_res, task))
7718                 return;
7719         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7720                                           NFS_I(lgp->args.inode)->layout,
7721                                           &lgp->args.range,
7722                                           lgp->args.ctx->state)) {
7723                 rpc_exit(task, NFS4_OK);
7724         }
7725 }
7726
7727 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7728 {
7729         struct nfs4_layoutget *lgp = calldata;
7730         struct inode *inode = lgp->args.inode;
7731         struct nfs_server *server = NFS_SERVER(inode);
7732         struct pnfs_layout_hdr *lo;
7733         struct nfs4_state *state = NULL;
7734         unsigned long timeo, now, giveup;
7735
7736         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7737
7738         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7739                 goto out;
7740
7741         switch (task->tk_status) {
7742         case 0:
7743                 goto out;
7744         /*
7745          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7746          * (or clients) writing to the same RAID stripe
7747          */
7748         case -NFS4ERR_LAYOUTTRYLATER:
7749         /*
7750          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7751          * existing layout before getting a new one).
7752          */
7753         case -NFS4ERR_RECALLCONFLICT:
7754                 timeo = rpc_get_timeout(task->tk_client);
7755                 giveup = lgp->args.timestamp + timeo;
7756                 now = jiffies;
7757                 if (time_after(giveup, now)) {
7758                         unsigned long delay;
7759
7760                         /* Delay for:
7761                          * - Not less then NFS4_POLL_RETRY_MIN.
7762                          * - One last time a jiffie before we give up
7763                          * - exponential backoff (time_now minus start_attempt)
7764                          */
7765                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7766                                     min((giveup - now - 1),
7767                                         now - lgp->args.timestamp));
7768
7769                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7770                                 __func__, delay);
7771                         rpc_delay(task, delay);
7772                         task->tk_status = 0;
7773                         rpc_restart_call_prepare(task);
7774                         goto out; /* Do not call nfs4_async_handle_error() */
7775                 }
7776                 break;
7777         case -NFS4ERR_EXPIRED:
7778         case -NFS4ERR_BAD_STATEID:
7779                 spin_lock(&inode->i_lock);
7780                 lo = NFS_I(inode)->layout;
7781                 if (!lo || list_empty(&lo->plh_segs)) {
7782                         spin_unlock(&inode->i_lock);
7783                         /* If the open stateid was bad, then recover it. */
7784                         state = lgp->args.ctx->state;
7785                 } else {
7786                         LIST_HEAD(head);
7787
7788                         /*
7789                          * Mark the bad layout state as invalid, then retry
7790                          * with the current stateid.
7791                          */
7792                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7793                         spin_unlock(&inode->i_lock);
7794                         pnfs_free_lseg_list(&head);
7795
7796                         task->tk_status = 0;
7797                         rpc_restart_call_prepare(task);
7798                 }
7799         }
7800         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7801                 rpc_restart_call_prepare(task);
7802 out:
7803         dprintk("<-- %s\n", __func__);
7804 }
7805
7806 static size_t max_response_pages(struct nfs_server *server)
7807 {
7808         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7809         return nfs_page_array_len(0, max_resp_sz);
7810 }
7811
7812 static void nfs4_free_pages(struct page **pages, size_t size)
7813 {
7814         int i;
7815
7816         if (!pages)
7817                 return;
7818
7819         for (i = 0; i < size; i++) {
7820                 if (!pages[i])
7821                         break;
7822                 __free_page(pages[i]);
7823         }
7824         kfree(pages);
7825 }
7826
7827 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7828 {
7829         struct page **pages;
7830         int i;
7831
7832         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7833         if (!pages) {
7834                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7835                 return NULL;
7836         }
7837
7838         for (i = 0; i < size; i++) {
7839                 pages[i] = alloc_page(gfp_flags);
7840                 if (!pages[i]) {
7841                         dprintk("%s: failed to allocate page\n", __func__);
7842                         nfs4_free_pages(pages, size);
7843                         return NULL;
7844                 }
7845         }
7846
7847         return pages;
7848 }
7849
7850 static void nfs4_layoutget_release(void *calldata)
7851 {
7852         struct nfs4_layoutget *lgp = calldata;
7853         struct inode *inode = lgp->args.inode;
7854         struct nfs_server *server = NFS_SERVER(inode);
7855         size_t max_pages = max_response_pages(server);
7856
7857         dprintk("--> %s\n", __func__);
7858         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7859         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7860         put_nfs_open_context(lgp->args.ctx);
7861         kfree(calldata);
7862         dprintk("<-- %s\n", __func__);
7863 }
7864
7865 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7866         .rpc_call_prepare = nfs4_layoutget_prepare,
7867         .rpc_call_done = nfs4_layoutget_done,
7868         .rpc_release = nfs4_layoutget_release,
7869 };
7870
7871 struct pnfs_layout_segment *
7872 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7873 {
7874         struct inode *inode = lgp->args.inode;
7875         struct nfs_server *server = NFS_SERVER(inode);
7876         size_t max_pages = max_response_pages(server);
7877         struct rpc_task *task;
7878         struct rpc_message msg = {
7879                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7880                 .rpc_argp = &lgp->args,
7881                 .rpc_resp = &lgp->res,
7882                 .rpc_cred = lgp->cred,
7883         };
7884         struct rpc_task_setup task_setup_data = {
7885                 .rpc_client = server->client,
7886                 .rpc_message = &msg,
7887                 .callback_ops = &nfs4_layoutget_call_ops,
7888                 .callback_data = lgp,
7889                 .flags = RPC_TASK_ASYNC,
7890         };
7891         struct pnfs_layout_segment *lseg = NULL;
7892         int status = 0;
7893
7894         dprintk("--> %s\n", __func__);
7895
7896         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7897         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7898
7899         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7900         if (!lgp->args.layout.pages) {
7901                 nfs4_layoutget_release(lgp);
7902                 return ERR_PTR(-ENOMEM);
7903         }
7904         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7905         lgp->args.timestamp = jiffies;
7906
7907         lgp->res.layoutp = &lgp->args.layout;
7908         lgp->res.seq_res.sr_slot = NULL;
7909         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7910
7911         task = rpc_run_task(&task_setup_data);
7912         if (IS_ERR(task))
7913                 return ERR_CAST(task);
7914         status = nfs4_wait_for_completion_rpc_task(task);
7915         if (status == 0)
7916                 status = task->tk_status;
7917         trace_nfs4_layoutget(lgp->args.ctx,
7918                         &lgp->args.range,
7919                         &lgp->res.range,
7920                         status);
7921         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7922         if (status == 0 && lgp->res.layoutp->len)
7923                 lseg = pnfs_layout_process(lgp);
7924         rpc_put_task(task);
7925         dprintk("<-- %s status=%d\n", __func__, status);
7926         if (status)
7927                 return ERR_PTR(status);
7928         return lseg;
7929 }
7930
7931 static void
7932 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7933 {
7934         struct nfs4_layoutreturn *lrp = calldata;
7935
7936         dprintk("--> %s\n", __func__);
7937         nfs41_setup_sequence(lrp->clp->cl_session,
7938                         &lrp->args.seq_args,
7939                         &lrp->res.seq_res,
7940                         task);
7941 }
7942
7943 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7944 {
7945         struct nfs4_layoutreturn *lrp = calldata;
7946         struct nfs_server *server;
7947
7948         dprintk("--> %s\n", __func__);
7949
7950         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7951                 return;
7952
7953         server = NFS_SERVER(lrp->args.inode);
7954         switch (task->tk_status) {
7955         default:
7956                 task->tk_status = 0;
7957         case 0:
7958                 break;
7959         case -NFS4ERR_DELAY:
7960                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7961                         break;
7962                 rpc_restart_call_prepare(task);
7963                 return;
7964         }
7965         dprintk("<-- %s\n", __func__);
7966 }
7967
7968 static void nfs4_layoutreturn_release(void *calldata)
7969 {
7970         struct nfs4_layoutreturn *lrp = calldata;
7971         struct pnfs_layout_hdr *lo = lrp->args.layout;
7972         LIST_HEAD(freeme);
7973
7974         dprintk("--> %s\n", __func__);
7975         spin_lock(&lo->plh_inode->i_lock);
7976         if (lrp->res.lrs_present)
7977                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7978         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
7979         pnfs_clear_layoutreturn_waitbit(lo);
7980         lo->plh_block_lgets--;
7981         spin_unlock(&lo->plh_inode->i_lock);
7982         pnfs_free_lseg_list(&freeme);
7983         pnfs_put_layout_hdr(lrp->args.layout);
7984         nfs_iput_and_deactive(lrp->inode);
7985         kfree(calldata);
7986         dprintk("<-- %s\n", __func__);
7987 }
7988
7989 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7990         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7991         .rpc_call_done = nfs4_layoutreturn_done,
7992         .rpc_release = nfs4_layoutreturn_release,
7993 };
7994
7995 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7996 {
7997         struct rpc_task *task;
7998         struct rpc_message msg = {
7999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8000                 .rpc_argp = &lrp->args,
8001                 .rpc_resp = &lrp->res,
8002                 .rpc_cred = lrp->cred,
8003         };
8004         struct rpc_task_setup task_setup_data = {
8005                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8006                 .rpc_message = &msg,
8007                 .callback_ops = &nfs4_layoutreturn_call_ops,
8008                 .callback_data = lrp,
8009         };
8010         int status = 0;
8011
8012         dprintk("--> %s\n", __func__);
8013         if (!sync) {
8014                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8015                 if (!lrp->inode) {
8016                         nfs4_layoutreturn_release(lrp);
8017                         return -EAGAIN;
8018                 }
8019                 task_setup_data.flags |= RPC_TASK_ASYNC;
8020         }
8021         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8022         task = rpc_run_task(&task_setup_data);
8023         if (IS_ERR(task))
8024                 return PTR_ERR(task);
8025         if (sync)
8026                 status = task->tk_status;
8027         trace_nfs4_layoutreturn(lrp->args.inode, status);
8028         dprintk("<-- %s status=%d\n", __func__, status);
8029         rpc_put_task(task);
8030         return status;
8031 }
8032
8033 static int
8034 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8035                 struct pnfs_device *pdev,
8036                 struct rpc_cred *cred)
8037 {
8038         struct nfs4_getdeviceinfo_args args = {
8039                 .pdev = pdev,
8040                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8041                         NOTIFY_DEVICEID4_DELETE,
8042         };
8043         struct nfs4_getdeviceinfo_res res = {
8044                 .pdev = pdev,
8045         };
8046         struct rpc_message msg = {
8047                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8048                 .rpc_argp = &args,
8049                 .rpc_resp = &res,
8050                 .rpc_cred = cred,
8051         };
8052         int status;
8053
8054         dprintk("--> %s\n", __func__);
8055         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8056         if (res.notification & ~args.notify_types)
8057                 dprintk("%s: unsupported notification\n", __func__);
8058         if (res.notification != args.notify_types)
8059                 pdev->nocache = 1;
8060
8061         dprintk("<-- %s status=%d\n", __func__, status);
8062
8063         return status;
8064 }
8065
8066 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8067                 struct pnfs_device *pdev,
8068                 struct rpc_cred *cred)
8069 {
8070         struct nfs4_exception exception = { };
8071         int err;
8072
8073         do {
8074                 err = nfs4_handle_exception(server,
8075                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8076                                         &exception);
8077         } while (exception.retry);
8078         return err;
8079 }
8080 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8081
8082 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8083 {
8084         struct nfs4_layoutcommit_data *data = calldata;
8085         struct nfs_server *server = NFS_SERVER(data->args.inode);
8086         struct nfs4_session *session = nfs4_get_session(server);
8087
8088         nfs41_setup_sequence(session,
8089                         &data->args.seq_args,
8090                         &data->res.seq_res,
8091                         task);
8092 }
8093
8094 static void
8095 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8096 {
8097         struct nfs4_layoutcommit_data *data = calldata;
8098         struct nfs_server *server = NFS_SERVER(data->args.inode);
8099
8100         if (!nfs41_sequence_done(task, &data->res.seq_res))
8101                 return;
8102
8103         switch (task->tk_status) { /* Just ignore these failures */
8104         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8105         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8106         case -NFS4ERR_BADLAYOUT:     /* no layout */
8107         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8108                 task->tk_status = 0;
8109         case 0:
8110                 break;
8111         default:
8112                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8113                         rpc_restart_call_prepare(task);
8114                         return;
8115                 }
8116         }
8117 }
8118
8119 static void nfs4_layoutcommit_release(void *calldata)
8120 {
8121         struct nfs4_layoutcommit_data *data = calldata;
8122
8123         pnfs_cleanup_layoutcommit(data);
8124         nfs_post_op_update_inode_force_wcc(data->args.inode,
8125                                            data->res.fattr);
8126         put_rpccred(data->cred);
8127         nfs_iput_and_deactive(data->inode);
8128         kfree(data);
8129 }
8130
8131 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8132         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8133         .rpc_call_done = nfs4_layoutcommit_done,
8134         .rpc_release = nfs4_layoutcommit_release,
8135 };
8136
8137 int
8138 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8139 {
8140         struct rpc_message msg = {
8141                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8142                 .rpc_argp = &data->args,
8143                 .rpc_resp = &data->res,
8144                 .rpc_cred = data->cred,
8145         };
8146         struct rpc_task_setup task_setup_data = {
8147                 .task = &data->task,
8148                 .rpc_client = NFS_CLIENT(data->args.inode),
8149                 .rpc_message = &msg,
8150                 .callback_ops = &nfs4_layoutcommit_ops,
8151                 .callback_data = data,
8152         };
8153         struct rpc_task *task;
8154         int status = 0;
8155
8156         dprintk("NFS: initiating layoutcommit call. sync %d "
8157                 "lbw: %llu inode %lu\n", sync,
8158                 data->args.lastbytewritten,
8159                 data->args.inode->i_ino);
8160
8161         if (!sync) {
8162                 data->inode = nfs_igrab_and_active(data->args.inode);
8163                 if (data->inode == NULL) {
8164                         nfs4_layoutcommit_release(data);
8165                         return -EAGAIN;
8166                 }
8167                 task_setup_data.flags = RPC_TASK_ASYNC;
8168         }
8169         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8170         task = rpc_run_task(&task_setup_data);
8171         if (IS_ERR(task))
8172                 return PTR_ERR(task);
8173         if (sync)
8174                 status = task->tk_status;
8175         trace_nfs4_layoutcommit(data->args.inode, status);
8176         dprintk("%s: status %d\n", __func__, status);
8177         rpc_put_task(task);
8178         return status;
8179 }
8180
8181 /**
8182  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8183  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8184  */
8185 static int
8186 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8187                     struct nfs_fsinfo *info,
8188                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8189 {
8190         struct nfs41_secinfo_no_name_args args = {
8191                 .style = SECINFO_STYLE_CURRENT_FH,
8192         };
8193         struct nfs4_secinfo_res res = {
8194                 .flavors = flavors,
8195         };
8196         struct rpc_message msg = {
8197                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8198                 .rpc_argp = &args,
8199                 .rpc_resp = &res,
8200         };
8201         struct rpc_clnt *clnt = server->client;
8202         struct rpc_cred *cred = NULL;
8203         int status;
8204
8205         if (use_integrity) {
8206                 clnt = server->nfs_client->cl_rpcclient;
8207                 cred = nfs4_get_clid_cred(server->nfs_client);
8208                 msg.rpc_cred = cred;
8209         }
8210
8211         dprintk("--> %s\n", __func__);
8212         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8213                                 &res.seq_res, 0);
8214         dprintk("<-- %s status=%d\n", __func__, status);
8215
8216         if (cred)
8217                 put_rpccred(cred);
8218
8219         return status;
8220 }
8221
8222 static int
8223 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8224                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8225 {
8226         struct nfs4_exception exception = { };
8227         int err;
8228         do {
8229                 /* first try using integrity protection */
8230                 err = -NFS4ERR_WRONGSEC;
8231
8232                 /* try to use integrity protection with machine cred */
8233                 if (_nfs4_is_integrity_protected(server->nfs_client))
8234                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8235                                                           flavors, true);
8236
8237                 /*
8238                  * if unable to use integrity protection, or SECINFO with
8239                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8240                  * disallowed by spec, but exists in deployed servers) use
8241                  * the current filesystem's rpc_client and the user cred.
8242                  */
8243                 if (err == -NFS4ERR_WRONGSEC)
8244                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8245                                                           flavors, false);
8246
8247                 switch (err) {
8248                 case 0:
8249                 case -NFS4ERR_WRONGSEC:
8250                 case -ENOTSUPP:
8251                         goto out;
8252                 default:
8253                         err = nfs4_handle_exception(server, err, &exception);
8254                 }
8255         } while (exception.retry);
8256 out:
8257         return err;
8258 }
8259
8260 static int
8261 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8262                     struct nfs_fsinfo *info)
8263 {
8264         int err;
8265         struct page *page;
8266         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8267         struct nfs4_secinfo_flavors *flavors;
8268         struct nfs4_secinfo4 *secinfo;
8269         int i;
8270
8271         page = alloc_page(GFP_KERNEL);
8272         if (!page) {
8273                 err = -ENOMEM;
8274                 goto out;
8275         }
8276
8277         flavors = page_address(page);
8278         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8279
8280         /*
8281          * Fall back on "guess and check" method if
8282          * the server doesn't support SECINFO_NO_NAME
8283          */
8284         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8285                 err = nfs4_find_root_sec(server, fhandle, info);
8286                 goto out_freepage;
8287         }
8288         if (err)
8289                 goto out_freepage;
8290
8291         for (i = 0; i < flavors->num_flavors; i++) {
8292                 secinfo = &flavors->flavors[i];
8293
8294                 switch (secinfo->flavor) {
8295                 case RPC_AUTH_NULL:
8296                 case RPC_AUTH_UNIX:
8297                 case RPC_AUTH_GSS:
8298                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8299                                         &secinfo->flavor_info);
8300                         break;
8301                 default:
8302                         flavor = RPC_AUTH_MAXFLAVOR;
8303                         break;
8304                 }
8305
8306                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8307                         flavor = RPC_AUTH_MAXFLAVOR;
8308
8309                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8310                         err = nfs4_lookup_root_sec(server, fhandle,
8311                                                    info, flavor);
8312                         if (!err)
8313                                 break;
8314                 }
8315         }
8316
8317         if (flavor == RPC_AUTH_MAXFLAVOR)
8318                 err = -EPERM;
8319
8320 out_freepage:
8321         put_page(page);
8322         if (err == -EACCES)
8323                 return -EPERM;
8324 out:
8325         return err;
8326 }
8327
8328 static int _nfs41_test_stateid(struct nfs_server *server,
8329                 nfs4_stateid *stateid,
8330                 struct rpc_cred *cred)
8331 {
8332         int status;
8333         struct nfs41_test_stateid_args args = {
8334                 .stateid = stateid,
8335         };
8336         struct nfs41_test_stateid_res res;
8337         struct rpc_message msg = {
8338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8339                 .rpc_argp = &args,
8340                 .rpc_resp = &res,
8341                 .rpc_cred = cred,
8342         };
8343         struct rpc_clnt *rpc_client = server->client;
8344
8345         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8346                 &rpc_client, &msg);
8347
8348         dprintk("NFS call  test_stateid %p\n", stateid);
8349         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8350         nfs4_set_sequence_privileged(&args.seq_args);
8351         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8352                         &args.seq_args, &res.seq_res);
8353         if (status != NFS_OK) {
8354                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8355                 return status;
8356         }
8357         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8358         return -res.status;
8359 }
8360
8361 /**
8362  * nfs41_test_stateid - perform a TEST_STATEID operation
8363  *
8364  * @server: server / transport on which to perform the operation
8365  * @stateid: state ID to test
8366  * @cred: credential
8367  *
8368  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8369  * Otherwise a negative NFS4ERR value is returned if the operation
8370  * failed or the state ID is not currently valid.
8371  */
8372 static int nfs41_test_stateid(struct nfs_server *server,
8373                 nfs4_stateid *stateid,
8374                 struct rpc_cred *cred)
8375 {
8376         struct nfs4_exception exception = { };
8377         int err;
8378         do {
8379                 err = _nfs41_test_stateid(server, stateid, cred);
8380                 if (err != -NFS4ERR_DELAY)
8381                         break;
8382                 nfs4_handle_exception(server, err, &exception);
8383         } while (exception.retry);
8384         return err;
8385 }
8386
8387 struct nfs_free_stateid_data {
8388         struct nfs_server *server;
8389         struct nfs41_free_stateid_args args;
8390         struct nfs41_free_stateid_res res;
8391 };
8392
8393 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8394 {
8395         struct nfs_free_stateid_data *data = calldata;
8396         nfs41_setup_sequence(nfs4_get_session(data->server),
8397                         &data->args.seq_args,
8398                         &data->res.seq_res,
8399                         task);
8400 }
8401
8402 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8403 {
8404         struct nfs_free_stateid_data *data = calldata;
8405
8406         nfs41_sequence_done(task, &data->res.seq_res);
8407
8408         switch (task->tk_status) {
8409         case -NFS4ERR_DELAY:
8410                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8411                         rpc_restart_call_prepare(task);
8412         }
8413 }
8414
8415 static void nfs41_free_stateid_release(void *calldata)
8416 {
8417         kfree(calldata);
8418 }
8419
8420 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8421         .rpc_call_prepare = nfs41_free_stateid_prepare,
8422         .rpc_call_done = nfs41_free_stateid_done,
8423         .rpc_release = nfs41_free_stateid_release,
8424 };
8425
8426 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8427                 nfs4_stateid *stateid,
8428                 struct rpc_cred *cred,
8429                 bool privileged)
8430 {
8431         struct rpc_message msg = {
8432                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8433                 .rpc_cred = cred,
8434         };
8435         struct rpc_task_setup task_setup = {
8436                 .rpc_client = server->client,
8437                 .rpc_message = &msg,
8438                 .callback_ops = &nfs41_free_stateid_ops,
8439                 .flags = RPC_TASK_ASYNC,
8440         };
8441         struct nfs_free_stateid_data *data;
8442
8443         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8444                 &task_setup.rpc_client, &msg);
8445
8446         dprintk("NFS call  free_stateid %p\n", stateid);
8447         data = kmalloc(sizeof(*data), GFP_NOFS);
8448         if (!data)
8449                 return ERR_PTR(-ENOMEM);
8450         data->server = server;
8451         nfs4_stateid_copy(&data->args.stateid, stateid);
8452
8453         task_setup.callback_data = data;
8454
8455         msg.rpc_argp = &data->args;
8456         msg.rpc_resp = &data->res;
8457         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8458         if (privileged)
8459                 nfs4_set_sequence_privileged(&data->args.seq_args);
8460
8461         return rpc_run_task(&task_setup);
8462 }
8463
8464 /**
8465  * nfs41_free_stateid - perform a FREE_STATEID operation
8466  *
8467  * @server: server / transport on which to perform the operation
8468  * @stateid: state ID to release
8469  * @cred: credential
8470  *
8471  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8472  * negative NFS4ERR value is returned.
8473  */
8474 static int nfs41_free_stateid(struct nfs_server *server,
8475                 nfs4_stateid *stateid,
8476                 struct rpc_cred *cred)
8477 {
8478         struct rpc_task *task;
8479         int ret;
8480
8481         task = _nfs41_free_stateid(server, stateid, cred, true);
8482         if (IS_ERR(task))
8483                 return PTR_ERR(task);
8484         ret = rpc_wait_for_completion_task(task);
8485         if (!ret)
8486                 ret = task->tk_status;
8487         rpc_put_task(task);
8488         return ret;
8489 }
8490
8491 static void
8492 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8493 {
8494         struct rpc_task *task;
8495         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8496
8497         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8498         nfs4_free_lock_state(server, lsp);
8499         if (IS_ERR(task))
8500                 return;
8501         rpc_put_task(task);
8502 }
8503
8504 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8505                 const nfs4_stateid *s2)
8506 {
8507         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8508                 return false;
8509
8510         if (s1->seqid == s2->seqid)
8511                 return true;
8512         if (s1->seqid == 0 || s2->seqid == 0)
8513                 return true;
8514
8515         return false;
8516 }
8517
8518 #endif /* CONFIG_NFS_V4_1 */
8519
8520 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8521                 const nfs4_stateid *s2)
8522 {
8523         return nfs4_stateid_match(s1, s2);
8524 }
8525
8526
8527 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8528         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8529         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8530         .recover_open   = nfs4_open_reclaim,
8531         .recover_lock   = nfs4_lock_reclaim,
8532         .establish_clid = nfs4_init_clientid,
8533         .detect_trunking = nfs40_discover_server_trunking,
8534 };
8535
8536 #if defined(CONFIG_NFS_V4_1)
8537 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8538         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8539         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8540         .recover_open   = nfs4_open_reclaim,
8541         .recover_lock   = nfs4_lock_reclaim,
8542         .establish_clid = nfs41_init_clientid,
8543         .reclaim_complete = nfs41_proc_reclaim_complete,
8544         .detect_trunking = nfs41_discover_server_trunking,
8545 };
8546 #endif /* CONFIG_NFS_V4_1 */
8547
8548 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8549         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8550         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8551         .recover_open   = nfs40_open_expired,
8552         .recover_lock   = nfs4_lock_expired,
8553         .establish_clid = nfs4_init_clientid,
8554 };
8555
8556 #if defined(CONFIG_NFS_V4_1)
8557 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8558         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8559         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8560         .recover_open   = nfs41_open_expired,
8561         .recover_lock   = nfs41_lock_expired,
8562         .establish_clid = nfs41_init_clientid,
8563 };
8564 #endif /* CONFIG_NFS_V4_1 */
8565
8566 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8567         .sched_state_renewal = nfs4_proc_async_renew,
8568         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8569         .renew_lease = nfs4_proc_renew,
8570 };
8571
8572 #if defined(CONFIG_NFS_V4_1)
8573 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8574         .sched_state_renewal = nfs41_proc_async_sequence,
8575         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8576         .renew_lease = nfs4_proc_sequence,
8577 };
8578 #endif
8579
8580 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8581         .get_locations = _nfs40_proc_get_locations,
8582         .fsid_present = _nfs40_proc_fsid_present,
8583 };
8584
8585 #if defined(CONFIG_NFS_V4_1)
8586 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8587         .get_locations = _nfs41_proc_get_locations,
8588         .fsid_present = _nfs41_proc_fsid_present,
8589 };
8590 #endif  /* CONFIG_NFS_V4_1 */
8591
8592 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8593         .minor_version = 0,
8594         .init_caps = NFS_CAP_READDIRPLUS
8595                 | NFS_CAP_ATOMIC_OPEN
8596                 | NFS_CAP_POSIX_LOCK,
8597         .init_client = nfs40_init_client,
8598         .shutdown_client = nfs40_shutdown_client,
8599         .match_stateid = nfs4_match_stateid,
8600         .find_root_sec = nfs4_find_root_sec,
8601         .free_lock_state = nfs4_release_lockowner,
8602         .alloc_seqid = nfs_alloc_seqid,
8603         .call_sync_ops = &nfs40_call_sync_ops,
8604         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8605         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8606         .state_renewal_ops = &nfs40_state_renewal_ops,
8607         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8608 };
8609
8610 #if defined(CONFIG_NFS_V4_1)
8611 static struct nfs_seqid *
8612 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8613 {
8614         return NULL;
8615 }
8616
8617 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8618         .minor_version = 1,
8619         .init_caps = NFS_CAP_READDIRPLUS
8620                 | NFS_CAP_ATOMIC_OPEN
8621                 | NFS_CAP_POSIX_LOCK
8622                 | NFS_CAP_STATEID_NFSV41
8623                 | NFS_CAP_ATOMIC_OPEN_V1,
8624         .init_client = nfs41_init_client,
8625         .shutdown_client = nfs41_shutdown_client,
8626         .match_stateid = nfs41_match_stateid,
8627         .find_root_sec = nfs41_find_root_sec,
8628         .free_lock_state = nfs41_free_lock_state,
8629         .alloc_seqid = nfs_alloc_no_seqid,
8630         .call_sync_ops = &nfs41_call_sync_ops,
8631         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8632         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8633         .state_renewal_ops = &nfs41_state_renewal_ops,
8634         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8635 };
8636 #endif
8637
8638 #if defined(CONFIG_NFS_V4_2)
8639 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8640         .minor_version = 2,
8641         .init_caps = NFS_CAP_READDIRPLUS
8642                 | NFS_CAP_ATOMIC_OPEN
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         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8661 };
8662 #endif
8663
8664 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8665         [0] = &nfs_v4_0_minor_ops,
8666 #if defined(CONFIG_NFS_V4_1)
8667         [1] = &nfs_v4_1_minor_ops,
8668 #endif
8669 #if defined(CONFIG_NFS_V4_2)
8670         [2] = &nfs_v4_2_minor_ops,
8671 #endif
8672 };
8673
8674 static const struct inode_operations nfs4_dir_inode_operations = {
8675         .create         = nfs_create,
8676         .lookup         = nfs_lookup,
8677         .atomic_open    = nfs_atomic_open,
8678         .link           = nfs_link,
8679         .unlink         = nfs_unlink,
8680         .symlink        = nfs_symlink,
8681         .mkdir          = nfs_mkdir,
8682         .rmdir          = nfs_rmdir,
8683         .mknod          = nfs_mknod,
8684         .rename         = nfs_rename,
8685         .permission     = nfs_permission,
8686         .getattr        = nfs_getattr,
8687         .setattr        = nfs_setattr,
8688         .getxattr       = generic_getxattr,
8689         .setxattr       = generic_setxattr,
8690         .listxattr      = generic_listxattr,
8691         .removexattr    = generic_removexattr,
8692 };
8693
8694 static const struct inode_operations nfs4_file_inode_operations = {
8695         .permission     = nfs_permission,
8696         .getattr        = nfs_getattr,
8697         .setattr        = nfs_setattr,
8698         .getxattr       = generic_getxattr,
8699         .setxattr       = generic_setxattr,
8700         .listxattr      = generic_listxattr,
8701         .removexattr    = generic_removexattr,
8702 };
8703
8704 const struct nfs_rpc_ops nfs_v4_clientops = {
8705         .version        = 4,                    /* protocol version */
8706         .dentry_ops     = &nfs4_dentry_operations,
8707         .dir_inode_ops  = &nfs4_dir_inode_operations,
8708         .file_inode_ops = &nfs4_file_inode_operations,
8709         .file_ops       = &nfs4_file_operations,
8710         .getroot        = nfs4_proc_get_root,
8711         .submount       = nfs4_submount,
8712         .try_mount      = nfs4_try_mount,
8713         .getattr        = nfs4_proc_getattr,
8714         .setattr        = nfs4_proc_setattr,
8715         .lookup         = nfs4_proc_lookup,
8716         .access         = nfs4_proc_access,
8717         .readlink       = nfs4_proc_readlink,
8718         .create         = nfs4_proc_create,
8719         .remove         = nfs4_proc_remove,
8720         .unlink_setup   = nfs4_proc_unlink_setup,
8721         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8722         .unlink_done    = nfs4_proc_unlink_done,
8723         .rename_setup   = nfs4_proc_rename_setup,
8724         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8725         .rename_done    = nfs4_proc_rename_done,
8726         .link           = nfs4_proc_link,
8727         .symlink        = nfs4_proc_symlink,
8728         .mkdir          = nfs4_proc_mkdir,
8729         .rmdir          = nfs4_proc_remove,
8730         .readdir        = nfs4_proc_readdir,
8731         .mknod          = nfs4_proc_mknod,
8732         .statfs         = nfs4_proc_statfs,
8733         .fsinfo         = nfs4_proc_fsinfo,
8734         .pathconf       = nfs4_proc_pathconf,
8735         .set_capabilities = nfs4_server_capabilities,
8736         .decode_dirent  = nfs4_decode_dirent,
8737         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8738         .read_setup     = nfs4_proc_read_setup,
8739         .read_done      = nfs4_read_done,
8740         .write_setup    = nfs4_proc_write_setup,
8741         .write_done     = nfs4_write_done,
8742         .commit_setup   = nfs4_proc_commit_setup,
8743         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8744         .commit_done    = nfs4_commit_done,
8745         .lock           = nfs4_proc_lock,
8746         .clear_acl_cache = nfs4_zap_acl_attr,
8747         .close_context  = nfs4_close_context,
8748         .open_context   = nfs4_atomic_open,
8749         .have_delegation = nfs4_have_delegation,
8750         .return_delegation = nfs4_inode_return_delegation,
8751         .alloc_client   = nfs4_alloc_client,
8752         .init_client    = nfs4_init_client,
8753         .free_client    = nfs4_free_client,
8754         .create_server  = nfs4_create_server,
8755         .clone_server   = nfs_clone_server,
8756 };
8757
8758 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8759         .prefix = XATTR_NAME_NFSV4_ACL,
8760         .list   = nfs4_xattr_list_nfs4_acl,
8761         .get    = nfs4_xattr_get_nfs4_acl,
8762         .set    = nfs4_xattr_set_nfs4_acl,
8763 };
8764
8765 const struct xattr_handler *nfs4_xattr_handlers[] = {
8766         &nfs4_xattr_nfs4_acl_handler,
8767 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8768         &nfs4_xattr_nfs4_label_handler,
8769 #endif
8770         NULL
8771 };
8772
8773 /*
8774  * Local variables:
8775  *  c-basic-offset: 8
8776  * End:
8777  */