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 bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1131                 fmode_t fmode)
1132 {
1133         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1134         case FMODE_READ|FMODE_WRITE:
1135                 return state->n_rdwr != 0;
1136         case FMODE_WRITE:
1137                 return state->n_wronly != 0;
1138         case FMODE_READ:
1139                 return state->n_rdonly != 0;
1140         }
1141         WARN_ON_ONCE(1);
1142         return false;
1143 }
1144
1145 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1146 {
1147         int ret = 0;
1148
1149         if (open_mode & (O_EXCL|O_TRUNC))
1150                 goto out;
1151         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1152                 case FMODE_READ:
1153                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1154                                 && state->n_rdonly != 0;
1155                         break;
1156                 case FMODE_WRITE:
1157                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1158                                 && state->n_wronly != 0;
1159                         break;
1160                 case FMODE_READ|FMODE_WRITE:
1161                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1162                                 && state->n_rdwr != 0;
1163         }
1164 out:
1165         return ret;
1166 }
1167
1168 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1169                 enum open_claim_type4 claim)
1170 {
1171         if (delegation == NULL)
1172                 return 0;
1173         if ((delegation->type & fmode) != fmode)
1174                 return 0;
1175         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1176                 return 0;
1177         switch (claim) {
1178         case NFS4_OPEN_CLAIM_NULL:
1179         case NFS4_OPEN_CLAIM_FH:
1180                 break;
1181         case NFS4_OPEN_CLAIM_PREVIOUS:
1182                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1183                         break;
1184         default:
1185                 return 0;
1186         }
1187         nfs_mark_delegation_referenced(delegation);
1188         return 1;
1189 }
1190
1191 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1192 {
1193         switch (fmode) {
1194                 case FMODE_WRITE:
1195                         state->n_wronly++;
1196                         break;
1197                 case FMODE_READ:
1198                         state->n_rdonly++;
1199                         break;
1200                 case FMODE_READ|FMODE_WRITE:
1201                         state->n_rdwr++;
1202         }
1203         nfs4_state_set_mode_locked(state, state->state | fmode);
1204 }
1205
1206 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1207 {
1208         struct nfs_client *clp = state->owner->so_server->nfs_client;
1209         bool need_recover = false;
1210
1211         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1212                 need_recover = true;
1213         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1214                 need_recover = true;
1215         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1216                 need_recover = true;
1217         if (need_recover)
1218                 nfs4_state_mark_reclaim_nograce(clp, state);
1219 }
1220
1221 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1222                 nfs4_stateid *stateid)
1223 {
1224         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1225                 return true;
1226         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1227                 nfs_test_and_clear_all_open_stateid(state);
1228                 return true;
1229         }
1230         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1231                 return true;
1232         return false;
1233 }
1234
1235 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1236 {
1237         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1238                 return;
1239         if (state->n_wronly)
1240                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1241         if (state->n_rdonly)
1242                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1243         if (state->n_rdwr)
1244                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1245         set_bit(NFS_OPEN_STATE, &state->flags);
1246 }
1247
1248 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1249                 nfs4_stateid *arg_stateid,
1250                 nfs4_stateid *stateid, fmode_t fmode)
1251 {
1252         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1253         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1254         case FMODE_WRITE:
1255                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1256                 break;
1257         case FMODE_READ:
1258                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1259                 break;
1260         case 0:
1261                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1262                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1263                 clear_bit(NFS_OPEN_STATE, &state->flags);
1264         }
1265         if (stateid == NULL)
1266                 return;
1267         /* Handle races with OPEN */
1268         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1269             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1270             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1271                 nfs_resync_open_stateid_locked(state);
1272                 return;
1273         }
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 nfs_clear_open_stateid(struct nfs4_state *state,
1280         nfs4_stateid *arg_stateid,
1281         nfs4_stateid *stateid, fmode_t fmode)
1282 {
1283         write_seqlock(&state->seqlock);
1284         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1285         write_sequnlock(&state->seqlock);
1286         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1287                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1288 }
1289
1290 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1291 {
1292         switch (fmode) {
1293                 case FMODE_READ:
1294                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1295                         break;
1296                 case FMODE_WRITE:
1297                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1298                         break;
1299                 case FMODE_READ|FMODE_WRITE:
1300                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1301         }
1302         if (!nfs_need_update_open_stateid(state, stateid))
1303                 return;
1304         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1305                 nfs4_stateid_copy(&state->stateid, stateid);
1306         nfs4_stateid_copy(&state->open_stateid, stateid);
1307 }
1308
1309 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1310 {
1311         /*
1312          * Protect the call to nfs4_state_set_mode_locked and
1313          * serialise the stateid update
1314          */
1315         write_seqlock(&state->seqlock);
1316         if (deleg_stateid != NULL) {
1317                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1318                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1319         }
1320         if (open_stateid != NULL)
1321                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1322         write_sequnlock(&state->seqlock);
1323         spin_lock(&state->owner->so_lock);
1324         update_open_stateflags(state, fmode);
1325         spin_unlock(&state->owner->so_lock);
1326 }
1327
1328 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1329 {
1330         struct nfs_inode *nfsi = NFS_I(state->inode);
1331         struct nfs_delegation *deleg_cur;
1332         int ret = 0;
1333
1334         fmode &= (FMODE_READ|FMODE_WRITE);
1335
1336         rcu_read_lock();
1337         deleg_cur = rcu_dereference(nfsi->delegation);
1338         if (deleg_cur == NULL)
1339                 goto no_delegation;
1340
1341         spin_lock(&deleg_cur->lock);
1342         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1343            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1344             (deleg_cur->type & fmode) != fmode)
1345                 goto no_delegation_unlock;
1346
1347         if (delegation == NULL)
1348                 delegation = &deleg_cur->stateid;
1349         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1350                 goto no_delegation_unlock;
1351
1352         nfs_mark_delegation_referenced(deleg_cur);
1353         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1354         ret = 1;
1355 no_delegation_unlock:
1356         spin_unlock(&deleg_cur->lock);
1357 no_delegation:
1358         rcu_read_unlock();
1359
1360         if (!ret && open_stateid != NULL) {
1361                 __update_open_stateid(state, open_stateid, NULL, fmode);
1362                 ret = 1;
1363         }
1364         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1365                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1366
1367         return ret;
1368 }
1369
1370 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1371                 const nfs4_stateid *stateid)
1372 {
1373         struct nfs4_state *state = lsp->ls_state;
1374         bool ret = false;
1375
1376         spin_lock(&state->state_lock);
1377         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1378                 goto out_noupdate;
1379         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1380                 goto out_noupdate;
1381         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1382         ret = true;
1383 out_noupdate:
1384         spin_unlock(&state->state_lock);
1385         return ret;
1386 }
1387
1388 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1389 {
1390         struct nfs_delegation *delegation;
1391
1392         rcu_read_lock();
1393         delegation = rcu_dereference(NFS_I(inode)->delegation);
1394         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1395                 rcu_read_unlock();
1396                 return;
1397         }
1398         rcu_read_unlock();
1399         nfs4_inode_return_delegation(inode);
1400 }
1401
1402 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1403 {
1404         struct nfs4_state *state = opendata->state;
1405         struct nfs_inode *nfsi = NFS_I(state->inode);
1406         struct nfs_delegation *delegation;
1407         int open_mode = opendata->o_arg.open_flags;
1408         fmode_t fmode = opendata->o_arg.fmode;
1409         enum open_claim_type4 claim = opendata->o_arg.claim;
1410         nfs4_stateid stateid;
1411         int ret = -EAGAIN;
1412
1413         for (;;) {
1414                 spin_lock(&state->owner->so_lock);
1415                 if (can_open_cached(state, fmode, open_mode)) {
1416                         update_open_stateflags(state, fmode);
1417                         spin_unlock(&state->owner->so_lock);
1418                         goto out_return_state;
1419                 }
1420                 spin_unlock(&state->owner->so_lock);
1421                 rcu_read_lock();
1422                 delegation = rcu_dereference(nfsi->delegation);
1423                 if (!can_open_delegated(delegation, fmode, claim)) {
1424                         rcu_read_unlock();
1425                         break;
1426                 }
1427                 /* Save the delegation */
1428                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1429                 rcu_read_unlock();
1430                 nfs_release_seqid(opendata->o_arg.seqid);
1431                 if (!opendata->is_recover) {
1432                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1433                         if (ret != 0)
1434                                 goto out;
1435                 }
1436                 ret = -EAGAIN;
1437
1438                 /* Try to update the stateid using the delegation */
1439                 if (update_open_stateid(state, NULL, &stateid, fmode))
1440                         goto out_return_state;
1441         }
1442 out:
1443         return ERR_PTR(ret);
1444 out_return_state:
1445         atomic_inc(&state->count);
1446         return state;
1447 }
1448
1449 static void
1450 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1451 {
1452         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1453         struct nfs_delegation *delegation;
1454         int delegation_flags = 0;
1455
1456         rcu_read_lock();
1457         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1458         if (delegation)
1459                 delegation_flags = delegation->flags;
1460         rcu_read_unlock();
1461         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1462                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1463                                    "returning a delegation for "
1464                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1465                                    clp->cl_hostname);
1466         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1467                 nfs_inode_set_delegation(state->inode,
1468                                          data->owner->so_cred,
1469                                          &data->o_res);
1470         else
1471                 nfs_inode_reclaim_delegation(state->inode,
1472                                              data->owner->so_cred,
1473                                              &data->o_res);
1474 }
1475
1476 /*
1477  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1478  * and update the nfs4_state.
1479  */
1480 static struct nfs4_state *
1481 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1482 {
1483         struct inode *inode = data->state->inode;
1484         struct nfs4_state *state = data->state;
1485         int ret;
1486
1487         if (!data->rpc_done) {
1488                 if (data->rpc_status) {
1489                         ret = data->rpc_status;
1490                         goto err;
1491                 }
1492                 /* cached opens have already been processed */
1493                 goto update;
1494         }
1495
1496         ret = nfs_refresh_inode(inode, &data->f_attr);
1497         if (ret)
1498                 goto err;
1499
1500         if (data->o_res.delegation_type != 0)
1501                 nfs4_opendata_check_deleg(data, state);
1502 update:
1503         update_open_stateid(state, &data->o_res.stateid, NULL,
1504                             data->o_arg.fmode);
1505         atomic_inc(&state->count);
1506
1507         return state;
1508 err:
1509         return ERR_PTR(ret);
1510
1511 }
1512
1513 static struct nfs4_state *
1514 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1515 {
1516         struct inode *inode;
1517         struct nfs4_state *state = NULL;
1518         int ret;
1519
1520         if (!data->rpc_done) {
1521                 state = nfs4_try_open_cached(data);
1522                 goto out;
1523         }
1524
1525         ret = -EAGAIN;
1526         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1527                 goto err;
1528         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1529         ret = PTR_ERR(inode);
1530         if (IS_ERR(inode))
1531                 goto err;
1532         ret = -ENOMEM;
1533         state = nfs4_get_open_state(inode, data->owner);
1534         if (state == NULL)
1535                 goto err_put_inode;
1536         if (data->o_res.delegation_type != 0)
1537                 nfs4_opendata_check_deleg(data, state);
1538         update_open_stateid(state, &data->o_res.stateid, NULL,
1539                         data->o_arg.fmode);
1540         iput(inode);
1541 out:
1542         nfs_release_seqid(data->o_arg.seqid);
1543         return state;
1544 err_put_inode:
1545         iput(inode);
1546 err:
1547         return ERR_PTR(ret);
1548 }
1549
1550 static struct nfs4_state *
1551 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1552 {
1553         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1554                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1555         return _nfs4_opendata_to_nfs4_state(data);
1556 }
1557
1558 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1559 {
1560         struct nfs_inode *nfsi = NFS_I(state->inode);
1561         struct nfs_open_context *ctx;
1562
1563         spin_lock(&state->inode->i_lock);
1564         list_for_each_entry(ctx, &nfsi->open_files, list) {
1565                 if (ctx->state != state)
1566                         continue;
1567                 get_nfs_open_context(ctx);
1568                 spin_unlock(&state->inode->i_lock);
1569                 return ctx;
1570         }
1571         spin_unlock(&state->inode->i_lock);
1572         return ERR_PTR(-ENOENT);
1573 }
1574
1575 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1576                 struct nfs4_state *state, enum open_claim_type4 claim)
1577 {
1578         struct nfs4_opendata *opendata;
1579
1580         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1581                         NULL, NULL, claim, GFP_NOFS);
1582         if (opendata == NULL)
1583                 return ERR_PTR(-ENOMEM);
1584         opendata->state = state;
1585         atomic_inc(&state->count);
1586         return opendata;
1587 }
1588
1589 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1590                 fmode_t fmode)
1591 {
1592         struct nfs4_state *newstate;
1593         int ret;
1594
1595         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1596                 return 0;
1597         opendata->o_arg.open_flags = 0;
1598         opendata->o_arg.fmode = fmode;
1599         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1600                         NFS_SB(opendata->dentry->d_sb),
1601                         fmode, 0);
1602         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1603         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1604         nfs4_init_opendata_res(opendata);
1605         ret = _nfs4_recover_proc_open(opendata);
1606         if (ret != 0)
1607                 return ret;
1608         newstate = nfs4_opendata_to_nfs4_state(opendata);
1609         if (IS_ERR(newstate))
1610                 return PTR_ERR(newstate);
1611         if (newstate != opendata->state)
1612                 ret = -ESTALE;
1613         nfs4_close_state(newstate, fmode);
1614         return ret;
1615 }
1616
1617 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1618 {
1619         int ret;
1620
1621         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1622         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1623         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1624         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1625         /* memory barrier prior to reading state->n_* */
1626         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1627         clear_bit(NFS_OPEN_STATE, &state->flags);
1628         smp_rmb();
1629         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1630         if (ret != 0)
1631                 return ret;
1632         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1633         if (ret != 0)
1634                 return ret;
1635         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1636         if (ret != 0)
1637                 return ret;
1638         /*
1639          * We may have performed cached opens for all three recoveries.
1640          * Check if we need to update the current stateid.
1641          */
1642         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1643             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1644                 write_seqlock(&state->seqlock);
1645                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1646                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1647                 write_sequnlock(&state->seqlock);
1648         }
1649         return 0;
1650 }
1651
1652 /*
1653  * OPEN_RECLAIM:
1654  *      reclaim state on the server after a reboot.
1655  */
1656 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1657 {
1658         struct nfs_delegation *delegation;
1659         struct nfs4_opendata *opendata;
1660         fmode_t delegation_type = 0;
1661         int status;
1662
1663         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1664                         NFS4_OPEN_CLAIM_PREVIOUS);
1665         if (IS_ERR(opendata))
1666                 return PTR_ERR(opendata);
1667         rcu_read_lock();
1668         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1669         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1670                 delegation_type = delegation->type;
1671         rcu_read_unlock();
1672         opendata->o_arg.u.delegation_type = delegation_type;
1673         status = nfs4_open_recover(opendata, state);
1674         nfs4_opendata_put(opendata);
1675         return status;
1676 }
1677
1678 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1679 {
1680         struct nfs_server *server = NFS_SERVER(state->inode);
1681         struct nfs4_exception exception = { };
1682         int err;
1683         do {
1684                 err = _nfs4_do_open_reclaim(ctx, state);
1685                 trace_nfs4_open_reclaim(ctx, 0, err);
1686                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1687                         continue;
1688                 if (err != -NFS4ERR_DELAY)
1689                         break;
1690                 nfs4_handle_exception(server, err, &exception);
1691         } while (exception.retry);
1692         return err;
1693 }
1694
1695 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1696 {
1697         struct nfs_open_context *ctx;
1698         int ret;
1699
1700         ctx = nfs4_state_find_open_context(state);
1701         if (IS_ERR(ctx))
1702                 return -EAGAIN;
1703         ret = nfs4_do_open_reclaim(ctx, state);
1704         put_nfs_open_context(ctx);
1705         return ret;
1706 }
1707
1708 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1709 {
1710         switch (err) {
1711                 default:
1712                         printk(KERN_ERR "NFS: %s: unhandled error "
1713                                         "%d.\n", __func__, err);
1714                 case 0:
1715                 case -ENOENT:
1716                 case -EAGAIN:
1717                 case -ESTALE:
1718                         break;
1719                 case -NFS4ERR_BADSESSION:
1720                 case -NFS4ERR_BADSLOT:
1721                 case -NFS4ERR_BAD_HIGH_SLOT:
1722                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1723                 case -NFS4ERR_DEADSESSION:
1724                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1725                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1726                         return -EAGAIN;
1727                 case -NFS4ERR_STALE_CLIENTID:
1728                 case -NFS4ERR_STALE_STATEID:
1729                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1730                 case -NFS4ERR_EXPIRED:
1731                         /* Don't recall a delegation if it was lost */
1732                         nfs4_schedule_lease_recovery(server->nfs_client);
1733                         return -EAGAIN;
1734                 case -NFS4ERR_MOVED:
1735                         nfs4_schedule_migration_recovery(server);
1736                         return -EAGAIN;
1737                 case -NFS4ERR_LEASE_MOVED:
1738                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1739                         return -EAGAIN;
1740                 case -NFS4ERR_DELEG_REVOKED:
1741                 case -NFS4ERR_ADMIN_REVOKED:
1742                 case -NFS4ERR_BAD_STATEID:
1743                 case -NFS4ERR_OPENMODE:
1744                         nfs_inode_find_state_and_recover(state->inode,
1745                                         stateid);
1746                         nfs4_schedule_stateid_recovery(server, state);
1747                         return -EAGAIN;
1748                 case -NFS4ERR_DELAY:
1749                 case -NFS4ERR_GRACE:
1750                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1751                         ssleep(1);
1752                         return -EAGAIN;
1753                 case -ENOMEM:
1754                 case -NFS4ERR_DENIED:
1755                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1756                         return 0;
1757         }
1758         return err;
1759 }
1760
1761 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1762                 struct nfs4_state *state, const nfs4_stateid *stateid,
1763                 fmode_t type)
1764 {
1765         struct nfs_server *server = NFS_SERVER(state->inode);
1766         struct nfs4_opendata *opendata;
1767         int err = 0;
1768
1769         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1770                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1771         if (IS_ERR(opendata))
1772                 return PTR_ERR(opendata);
1773         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1774         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1775         switch (type & (FMODE_READ|FMODE_WRITE)) {
1776         case FMODE_READ|FMODE_WRITE:
1777         case FMODE_WRITE:
1778                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1779                 if (err)
1780                         break;
1781                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1782                 if (err)
1783                         break;
1784         case FMODE_READ:
1785                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1786         }
1787         nfs4_opendata_put(opendata);
1788         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1789 }
1790
1791 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1792 {
1793         struct nfs4_opendata *data = calldata;
1794
1795         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1796                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1797 }
1798
1799 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1800 {
1801         struct nfs4_opendata *data = calldata;
1802
1803         nfs40_sequence_done(task, &data->c_res.seq_res);
1804
1805         data->rpc_status = task->tk_status;
1806         if (data->rpc_status == 0) {
1807                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1808                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1809                 renew_lease(data->o_res.server, data->timestamp);
1810                 data->rpc_done = 1;
1811         }
1812 }
1813
1814 static void nfs4_open_confirm_release(void *calldata)
1815 {
1816         struct nfs4_opendata *data = calldata;
1817         struct nfs4_state *state = NULL;
1818
1819         /* If this request hasn't been cancelled, do nothing */
1820         if (data->cancelled == 0)
1821                 goto out_free;
1822         /* In case of error, no cleanup! */
1823         if (!data->rpc_done)
1824                 goto out_free;
1825         state = nfs4_opendata_to_nfs4_state(data);
1826         if (!IS_ERR(state))
1827                 nfs4_close_state(state, data->o_arg.fmode);
1828 out_free:
1829         nfs4_opendata_put(data);
1830 }
1831
1832 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1833         .rpc_call_prepare = nfs4_open_confirm_prepare,
1834         .rpc_call_done = nfs4_open_confirm_done,
1835         .rpc_release = nfs4_open_confirm_release,
1836 };
1837
1838 /*
1839  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1840  */
1841 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1842 {
1843         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1844         struct rpc_task *task;
1845         struct  rpc_message msg = {
1846                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1847                 .rpc_argp = &data->c_arg,
1848                 .rpc_resp = &data->c_res,
1849                 .rpc_cred = data->owner->so_cred,
1850         };
1851         struct rpc_task_setup task_setup_data = {
1852                 .rpc_client = server->client,
1853                 .rpc_message = &msg,
1854                 .callback_ops = &nfs4_open_confirm_ops,
1855                 .callback_data = data,
1856                 .workqueue = nfsiod_workqueue,
1857                 .flags = RPC_TASK_ASYNC,
1858         };
1859         int status;
1860
1861         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1862         kref_get(&data->kref);
1863         data->rpc_done = 0;
1864         data->rpc_status = 0;
1865         data->timestamp = jiffies;
1866         task = rpc_run_task(&task_setup_data);
1867         if (IS_ERR(task))
1868                 return PTR_ERR(task);
1869         status = nfs4_wait_for_completion_rpc_task(task);
1870         if (status != 0) {
1871                 data->cancelled = 1;
1872                 smp_wmb();
1873         } else
1874                 status = data->rpc_status;
1875         rpc_put_task(task);
1876         return status;
1877 }
1878
1879 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1880 {
1881         struct nfs4_opendata *data = calldata;
1882         struct nfs4_state_owner *sp = data->owner;
1883         struct nfs_client *clp = sp->so_server->nfs_client;
1884         enum open_claim_type4 claim = data->o_arg.claim;
1885
1886         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1887                 goto out_wait;
1888         /*
1889          * Check if we still need to send an OPEN call, or if we can use
1890          * a delegation instead.
1891          */
1892         if (data->state != NULL) {
1893                 struct nfs_delegation *delegation;
1894
1895                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1896                         goto out_no_action;
1897                 rcu_read_lock();
1898                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1899                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1900                         goto unlock_no_action;
1901                 rcu_read_unlock();
1902         }
1903         /* Update client id. */
1904         data->o_arg.clientid = clp->cl_clientid;
1905         switch (claim) {
1906         default:
1907                 break;
1908         case NFS4_OPEN_CLAIM_PREVIOUS:
1909         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1910         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1911                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1912         case NFS4_OPEN_CLAIM_FH:
1913                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1914                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1915         }
1916         data->timestamp = jiffies;
1917         if (nfs4_setup_sequence(data->o_arg.server,
1918                                 &data->o_arg.seq_args,
1919                                 &data->o_res.seq_res,
1920                                 task) != 0)
1921                 nfs_release_seqid(data->o_arg.seqid);
1922
1923         /* Set the create mode (note dependency on the session type) */
1924         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1925         if (data->o_arg.open_flags & O_EXCL) {
1926                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1927                 if (nfs4_has_persistent_session(clp))
1928                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1929                 else if (clp->cl_mvops->minor_version > 0)
1930                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1931         }
1932         return;
1933 unlock_no_action:
1934         rcu_read_unlock();
1935 out_no_action:
1936         task->tk_action = NULL;
1937 out_wait:
1938         nfs4_sequence_done(task, &data->o_res.seq_res);
1939 }
1940
1941 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1942 {
1943         struct nfs4_opendata *data = calldata;
1944
1945         data->rpc_status = task->tk_status;
1946
1947         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1948                 return;
1949
1950         if (task->tk_status == 0) {
1951                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1952                         switch (data->o_res.f_attr->mode & S_IFMT) {
1953                         case S_IFREG:
1954                                 break;
1955                         case S_IFLNK:
1956                                 data->rpc_status = -ELOOP;
1957                                 break;
1958                         case S_IFDIR:
1959                                 data->rpc_status = -EISDIR;
1960                                 break;
1961                         default:
1962                                 data->rpc_status = -ENOTDIR;
1963                         }
1964                 }
1965                 renew_lease(data->o_res.server, data->timestamp);
1966                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1967                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1968         }
1969         data->rpc_done = 1;
1970 }
1971
1972 static void nfs4_open_release(void *calldata)
1973 {
1974         struct nfs4_opendata *data = calldata;
1975         struct nfs4_state *state = NULL;
1976
1977         /* If this request hasn't been cancelled, do nothing */
1978         if (data->cancelled == 0)
1979                 goto out_free;
1980         /* In case of error, no cleanup! */
1981         if (data->rpc_status != 0 || !data->rpc_done)
1982                 goto out_free;
1983         /* In case we need an open_confirm, no cleanup! */
1984         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1985                 goto out_free;
1986         state = nfs4_opendata_to_nfs4_state(data);
1987         if (!IS_ERR(state))
1988                 nfs4_close_state(state, data->o_arg.fmode);
1989 out_free:
1990         nfs4_opendata_put(data);
1991 }
1992
1993 static const struct rpc_call_ops nfs4_open_ops = {
1994         .rpc_call_prepare = nfs4_open_prepare,
1995         .rpc_call_done = nfs4_open_done,
1996         .rpc_release = nfs4_open_release,
1997 };
1998
1999 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2000 {
2001         struct inode *dir = d_inode(data->dir);
2002         struct nfs_server *server = NFS_SERVER(dir);
2003         struct nfs_openargs *o_arg = &data->o_arg;
2004         struct nfs_openres *o_res = &data->o_res;
2005         struct rpc_task *task;
2006         struct rpc_message msg = {
2007                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2008                 .rpc_argp = o_arg,
2009                 .rpc_resp = o_res,
2010                 .rpc_cred = data->owner->so_cred,
2011         };
2012         struct rpc_task_setup task_setup_data = {
2013                 .rpc_client = server->client,
2014                 .rpc_message = &msg,
2015                 .callback_ops = &nfs4_open_ops,
2016                 .callback_data = data,
2017                 .workqueue = nfsiod_workqueue,
2018                 .flags = RPC_TASK_ASYNC,
2019         };
2020         int status;
2021
2022         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2023         kref_get(&data->kref);
2024         data->rpc_done = 0;
2025         data->rpc_status = 0;
2026         data->cancelled = 0;
2027         data->is_recover = 0;
2028         if (isrecover) {
2029                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2030                 data->is_recover = 1;
2031         }
2032         task = rpc_run_task(&task_setup_data);
2033         if (IS_ERR(task))
2034                 return PTR_ERR(task);
2035         status = nfs4_wait_for_completion_rpc_task(task);
2036         if (status != 0) {
2037                 data->cancelled = 1;
2038                 smp_wmb();
2039         } else
2040                 status = data->rpc_status;
2041         rpc_put_task(task);
2042
2043         return status;
2044 }
2045
2046 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2047 {
2048         struct inode *dir = d_inode(data->dir);
2049         struct nfs_openres *o_res = &data->o_res;
2050         int status;
2051
2052         status = nfs4_run_open_task(data, 1);
2053         if (status != 0 || !data->rpc_done)
2054                 return status;
2055
2056         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2057
2058         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2059                 status = _nfs4_proc_open_confirm(data);
2060                 if (status != 0)
2061                         return status;
2062         }
2063
2064         return status;
2065 }
2066
2067 /*
2068  * Additional permission checks in order to distinguish between an
2069  * open for read, and an open for execute. This works around the
2070  * fact that NFSv4 OPEN treats read and execute permissions as being
2071  * the same.
2072  * Note that in the non-execute case, we want to turn off permission
2073  * checking if we just created a new file (POSIX open() semantics).
2074  */
2075 static int nfs4_opendata_access(struct rpc_cred *cred,
2076                                 struct nfs4_opendata *opendata,
2077                                 struct nfs4_state *state, fmode_t fmode,
2078                                 int openflags)
2079 {
2080         struct nfs_access_entry cache;
2081         u32 mask;
2082
2083         /* access call failed or for some reason the server doesn't
2084          * support any access modes -- defer access call until later */
2085         if (opendata->o_res.access_supported == 0)
2086                 return 0;
2087
2088         mask = 0;
2089         /*
2090          * Use openflags to check for exec, because fmode won't
2091          * always have FMODE_EXEC set when file open for exec.
2092          */
2093         if (openflags & __FMODE_EXEC) {
2094                 /* ONLY check for exec rights */
2095                 mask = MAY_EXEC;
2096         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2097                 mask = MAY_READ;
2098
2099         cache.cred = cred;
2100         cache.jiffies = jiffies;
2101         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2102         nfs_access_add_cache(state->inode, &cache);
2103
2104         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2105                 return 0;
2106
2107         /* even though OPEN succeeded, access is denied. Close the file */
2108         nfs4_close_state(state, fmode);
2109         return -EACCES;
2110 }
2111
2112 /*
2113  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2114  */
2115 static int _nfs4_proc_open(struct nfs4_opendata *data)
2116 {
2117         struct inode *dir = d_inode(data->dir);
2118         struct nfs_server *server = NFS_SERVER(dir);
2119         struct nfs_openargs *o_arg = &data->o_arg;
2120         struct nfs_openres *o_res = &data->o_res;
2121         int status;
2122
2123         status = nfs4_run_open_task(data, 0);
2124         if (!data->rpc_done)
2125                 return status;
2126         if (status != 0) {
2127                 if (status == -NFS4ERR_BADNAME &&
2128                                 !(o_arg->open_flags & O_CREAT))
2129                         return -ENOENT;
2130                 return status;
2131         }
2132
2133         nfs_fattr_map_and_free_names(server, &data->f_attr);
2134
2135         if (o_arg->open_flags & O_CREAT) {
2136                 update_changeattr(dir, &o_res->cinfo);
2137                 if (o_arg->open_flags & O_EXCL)
2138                         data->file_created = 1;
2139                 else if (o_res->cinfo.before != o_res->cinfo.after)
2140                         data->file_created = 1;
2141         }
2142         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2143                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2144         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2145                 status = _nfs4_proc_open_confirm(data);
2146                 if (status != 0)
2147                         return status;
2148         }
2149         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2150                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2151         return 0;
2152 }
2153
2154 static int nfs4_recover_expired_lease(struct nfs_server *server)
2155 {
2156         return nfs4_client_recover_expired_lease(server->nfs_client);
2157 }
2158
2159 /*
2160  * OPEN_EXPIRED:
2161  *      reclaim state on the server after a network partition.
2162  *      Assumes caller holds the appropriate lock
2163  */
2164 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2165 {
2166         struct nfs4_opendata *opendata;
2167         int ret;
2168
2169         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2170                         NFS4_OPEN_CLAIM_FH);
2171         if (IS_ERR(opendata))
2172                 return PTR_ERR(opendata);
2173         ret = nfs4_open_recover(opendata, state);
2174         if (ret == -ESTALE)
2175                 d_drop(ctx->dentry);
2176         nfs4_opendata_put(opendata);
2177         return ret;
2178 }
2179
2180 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2181 {
2182         struct nfs_server *server = NFS_SERVER(state->inode);
2183         struct nfs4_exception exception = { };
2184         int err;
2185
2186         do {
2187                 err = _nfs4_open_expired(ctx, state);
2188                 trace_nfs4_open_expired(ctx, 0, err);
2189                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2190                         continue;
2191                 switch (err) {
2192                 default:
2193                         goto out;
2194                 case -NFS4ERR_GRACE:
2195                 case -NFS4ERR_DELAY:
2196                         nfs4_handle_exception(server, err, &exception);
2197                         err = 0;
2198                 }
2199         } while (exception.retry);
2200 out:
2201         return err;
2202 }
2203
2204 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2205 {
2206         struct nfs_open_context *ctx;
2207         int ret;
2208
2209         ctx = nfs4_state_find_open_context(state);
2210         if (IS_ERR(ctx))
2211                 return -EAGAIN;
2212         ret = nfs4_do_open_expired(ctx, state);
2213         put_nfs_open_context(ctx);
2214         return ret;
2215 }
2216
2217 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2218 {
2219         nfs_remove_bad_delegation(state->inode);
2220         write_seqlock(&state->seqlock);
2221         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2222         write_sequnlock(&state->seqlock);
2223         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2224 }
2225
2226 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2227 {
2228         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2229                 nfs_finish_clear_delegation_stateid(state);
2230 }
2231
2232 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2233 {
2234         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2235         nfs40_clear_delegation_stateid(state);
2236         return nfs4_open_expired(sp, state);
2237 }
2238
2239 #if defined(CONFIG_NFS_V4_1)
2240 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2241 {
2242         struct nfs_server *server = NFS_SERVER(state->inode);
2243         nfs4_stateid stateid;
2244         struct nfs_delegation *delegation;
2245         struct rpc_cred *cred;
2246         int status;
2247
2248         /* Get the delegation credential for use by test/free_stateid */
2249         rcu_read_lock();
2250         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2251         if (delegation == NULL) {
2252                 rcu_read_unlock();
2253                 return;
2254         }
2255
2256         nfs4_stateid_copy(&stateid, &delegation->stateid);
2257         cred = get_rpccred(delegation->cred);
2258         rcu_read_unlock();
2259         status = nfs41_test_stateid(server, &stateid, cred);
2260         trace_nfs4_test_delegation_stateid(state, NULL, status);
2261
2262         if (status != NFS_OK) {
2263                 /* Free the stateid unless the server explicitly
2264                  * informs us the stateid is unrecognized. */
2265                 if (status != -NFS4ERR_BAD_STATEID)
2266                         nfs41_free_stateid(server, &stateid, cred);
2267                 nfs_finish_clear_delegation_stateid(state);
2268         }
2269
2270         put_rpccred(cred);
2271 }
2272
2273 /**
2274  * nfs41_check_open_stateid - possibly free an open stateid
2275  *
2276  * @state: NFSv4 state for an inode
2277  *
2278  * Returns NFS_OK if recovery for this stateid is now finished.
2279  * Otherwise a negative NFS4ERR value is returned.
2280  */
2281 static int nfs41_check_open_stateid(struct nfs4_state *state)
2282 {
2283         struct nfs_server *server = NFS_SERVER(state->inode);
2284         nfs4_stateid *stateid = &state->open_stateid;
2285         struct rpc_cred *cred = state->owner->so_cred;
2286         int status;
2287
2288         /* If a state reset has been done, test_stateid is unneeded */
2289         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2290             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2291             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2292                 return -NFS4ERR_BAD_STATEID;
2293
2294         status = nfs41_test_stateid(server, stateid, cred);
2295         trace_nfs4_test_open_stateid(state, NULL, status);
2296         if (status != NFS_OK) {
2297                 /* Free the stateid unless the server explicitly
2298                  * informs us the stateid is unrecognized. */
2299                 if (status != -NFS4ERR_BAD_STATEID)
2300                         nfs41_free_stateid(server, stateid, cred);
2301
2302                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2303                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2304                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2305                 clear_bit(NFS_OPEN_STATE, &state->flags);
2306         }
2307         return status;
2308 }
2309
2310 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2311 {
2312         int status;
2313
2314         nfs41_check_delegation_stateid(state);
2315         status = nfs41_check_open_stateid(state);
2316         if (status != NFS_OK)
2317                 status = nfs4_open_expired(sp, state);
2318         return status;
2319 }
2320 #endif
2321
2322 /*
2323  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2324  * fields corresponding to attributes that were used to store the verifier.
2325  * Make sure we clobber those fields in the later setattr call
2326  */
2327 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2328                                 struct iattr *sattr, struct nfs4_label **label)
2329 {
2330         const u32 *attrset = opendata->o_res.attrset;
2331
2332         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2333             !(sattr->ia_valid & ATTR_ATIME_SET))
2334                 sattr->ia_valid |= ATTR_ATIME;
2335
2336         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2337             !(sattr->ia_valid & ATTR_MTIME_SET))
2338                 sattr->ia_valid |= ATTR_MTIME;
2339
2340         /* Except MODE, it seems harmless of setting twice. */
2341         if ((attrset[1] & FATTR4_WORD1_MODE))
2342                 sattr->ia_valid &= ~ATTR_MODE;
2343
2344         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2345                 *label = NULL;
2346 }
2347
2348 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2349                 fmode_t fmode,
2350                 int flags,
2351                 struct nfs_open_context *ctx)
2352 {
2353         struct nfs4_state_owner *sp = opendata->owner;
2354         struct nfs_server *server = sp->so_server;
2355         struct dentry *dentry;
2356         struct nfs4_state *state;
2357         unsigned int seq;
2358         int ret;
2359
2360         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2361
2362         ret = _nfs4_proc_open(opendata);
2363         if (ret != 0)
2364                 goto out;
2365
2366         state = nfs4_opendata_to_nfs4_state(opendata);
2367         ret = PTR_ERR(state);
2368         if (IS_ERR(state))
2369                 goto out;
2370         if (server->caps & NFS_CAP_POSIX_LOCK)
2371                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2372
2373         dentry = opendata->dentry;
2374         if (d_really_is_negative(dentry)) {
2375                 /* FIXME: Is this d_drop() ever needed? */
2376                 d_drop(dentry);
2377                 dentry = d_add_unique(dentry, igrab(state->inode));
2378                 if (dentry == NULL) {
2379                         dentry = opendata->dentry;
2380                 } else if (dentry != ctx->dentry) {
2381                         dput(ctx->dentry);
2382                         ctx->dentry = dget(dentry);
2383                 }
2384                 nfs_set_verifier(dentry,
2385                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2386         }
2387
2388         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2389         if (ret != 0)
2390                 goto out;
2391
2392         ctx->state = state;
2393         if (d_inode(dentry) == state->inode) {
2394                 nfs_inode_attach_open_context(ctx);
2395                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2396                         nfs4_schedule_stateid_recovery(server, state);
2397         }
2398 out:
2399         return ret;
2400 }
2401
2402 /*
2403  * Returns a referenced nfs4_state
2404  */
2405 static int _nfs4_do_open(struct inode *dir,
2406                         struct nfs_open_context *ctx,
2407                         int flags,
2408                         struct iattr *sattr,
2409                         struct nfs4_label *label,
2410                         int *opened)
2411 {
2412         struct nfs4_state_owner  *sp;
2413         struct nfs4_state     *state = NULL;
2414         struct nfs_server       *server = NFS_SERVER(dir);
2415         struct nfs4_opendata *opendata;
2416         struct dentry *dentry = ctx->dentry;
2417         struct rpc_cred *cred = ctx->cred;
2418         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2419         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2420         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2421         struct nfs4_label *olabel = NULL;
2422         int status;
2423
2424         /* Protect against reboot recovery conflicts */
2425         status = -ENOMEM;
2426         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2427         if (sp == NULL) {
2428                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2429                 goto out_err;
2430         }
2431         status = nfs4_recover_expired_lease(server);
2432         if (status != 0)
2433                 goto err_put_state_owner;
2434         if (d_really_is_positive(dentry))
2435                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2436         status = -ENOMEM;
2437         if (d_really_is_positive(dentry))
2438                 claim = NFS4_OPEN_CLAIM_FH;
2439         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2440                         label, claim, GFP_KERNEL);
2441         if (opendata == NULL)
2442                 goto err_put_state_owner;
2443
2444         if (label) {
2445                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2446                 if (IS_ERR(olabel)) {
2447                         status = PTR_ERR(olabel);
2448                         goto err_opendata_put;
2449                 }
2450         }
2451
2452         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2453                 if (!opendata->f_attr.mdsthreshold) {
2454                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2455                         if (!opendata->f_attr.mdsthreshold)
2456                                 goto err_free_label;
2457                 }
2458                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2459         }
2460         if (d_really_is_positive(dentry))
2461                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2462
2463         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2464         if (status != 0)
2465                 goto err_free_label;
2466         state = ctx->state;
2467
2468         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2469             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2470                 nfs4_exclusive_attrset(opendata, sattr, &label);
2471
2472                 nfs_fattr_init(opendata->o_res.f_attr);
2473                 status = nfs4_do_setattr(state->inode, cred,
2474                                 opendata->o_res.f_attr, sattr,
2475                                 state, label, olabel);
2476                 if (status == 0) {
2477                         nfs_setattr_update_inode(state->inode, sattr,
2478                                         opendata->o_res.f_attr);
2479                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2480                 }
2481         }
2482         if (opened && opendata->file_created)
2483                 *opened |= FILE_CREATED;
2484
2485         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2486                 *ctx_th = opendata->f_attr.mdsthreshold;
2487                 opendata->f_attr.mdsthreshold = NULL;
2488         }
2489
2490         nfs4_label_free(olabel);
2491
2492         nfs4_opendata_put(opendata);
2493         nfs4_put_state_owner(sp);
2494         return 0;
2495 err_free_label:
2496         nfs4_label_free(olabel);
2497 err_opendata_put:
2498         nfs4_opendata_put(opendata);
2499 err_put_state_owner:
2500         nfs4_put_state_owner(sp);
2501 out_err:
2502         return status;
2503 }
2504
2505
2506 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2507                                         struct nfs_open_context *ctx,
2508                                         int flags,
2509                                         struct iattr *sattr,
2510                                         struct nfs4_label *label,
2511                                         int *opened)
2512 {
2513         struct nfs_server *server = NFS_SERVER(dir);
2514         struct nfs4_exception exception = { };
2515         struct nfs4_state *res;
2516         int status;
2517
2518         do {
2519                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2520                 res = ctx->state;
2521                 trace_nfs4_open_file(ctx, flags, status);
2522                 if (status == 0)
2523                         break;
2524                 /* NOTE: BAD_SEQID means the server and client disagree about the
2525                  * book-keeping w.r.t. state-changing operations
2526                  * (OPEN/CLOSE/LOCK/LOCKU...)
2527                  * It is actually a sign of a bug on the client or on the server.
2528                  *
2529                  * If we receive a BAD_SEQID error in the particular case of
2530                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2531                  * have unhashed the old state_owner for us, and that we can
2532                  * therefore safely retry using a new one. We should still warn
2533                  * the user though...
2534                  */
2535                 if (status == -NFS4ERR_BAD_SEQID) {
2536                         pr_warn_ratelimited("NFS: v4 server %s "
2537                                         " returned a bad sequence-id error!\n",
2538                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2539                         exception.retry = 1;
2540                         continue;
2541                 }
2542                 /*
2543                  * BAD_STATEID on OPEN means that the server cancelled our
2544                  * state before it received the OPEN_CONFIRM.
2545                  * Recover by retrying the request as per the discussion
2546                  * on Page 181 of RFC3530.
2547                  */
2548                 if (status == -NFS4ERR_BAD_STATEID) {
2549                         exception.retry = 1;
2550                         continue;
2551                 }
2552                 if (status == -EAGAIN) {
2553                         /* We must have found a delegation */
2554                         exception.retry = 1;
2555                         continue;
2556                 }
2557                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2558                         continue;
2559                 res = ERR_PTR(nfs4_handle_exception(server,
2560                                         status, &exception));
2561         } while (exception.retry);
2562         return res;
2563 }
2564
2565 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2566                             struct nfs_fattr *fattr, struct iattr *sattr,
2567                             struct nfs4_state *state, struct nfs4_label *ilabel,
2568                             struct nfs4_label *olabel)
2569 {
2570         struct nfs_server *server = NFS_SERVER(inode);
2571         struct nfs_setattrargs  arg = {
2572                 .fh             = NFS_FH(inode),
2573                 .iap            = sattr,
2574                 .server         = server,
2575                 .bitmask = server->attr_bitmask,
2576                 .label          = ilabel,
2577         };
2578         struct nfs_setattrres  res = {
2579                 .fattr          = fattr,
2580                 .label          = olabel,
2581                 .server         = server,
2582         };
2583         struct rpc_message msg = {
2584                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2585                 .rpc_argp       = &arg,
2586                 .rpc_resp       = &res,
2587                 .rpc_cred       = cred,
2588         };
2589         unsigned long timestamp = jiffies;
2590         fmode_t fmode;
2591         bool truncate;
2592         int status;
2593
2594         arg.bitmask = nfs4_bitmask(server, ilabel);
2595         if (ilabel)
2596                 arg.bitmask = nfs4_bitmask(server, olabel);
2597
2598         nfs_fattr_init(fattr);
2599
2600         /* Servers should only apply open mode checks for file size changes */
2601         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2602         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2603
2604         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2605                 /* Use that stateid */
2606         } else if (truncate && state != NULL) {
2607                 struct nfs_lockowner lockowner = {
2608                         .l_owner = current->files,
2609                         .l_pid = current->tgid,
2610                 };
2611                 if (!nfs4_valid_open_stateid(state))
2612                         return -EBADF;
2613                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2614                                 &lockowner) == -EIO)
2615                         return -EBADF;
2616         } else
2617                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2618
2619         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2620         if (status == 0 && state != NULL)
2621                 renew_lease(server, timestamp);
2622         return status;
2623 }
2624
2625 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2626                            struct nfs_fattr *fattr, struct iattr *sattr,
2627                            struct nfs4_state *state, struct nfs4_label *ilabel,
2628                            struct nfs4_label *olabel)
2629 {
2630         struct nfs_server *server = NFS_SERVER(inode);
2631         struct nfs4_exception exception = {
2632                 .state = state,
2633                 .inode = inode,
2634         };
2635         int err;
2636         do {
2637                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2638                 trace_nfs4_setattr(inode, err);
2639                 switch (err) {
2640                 case -NFS4ERR_OPENMODE:
2641                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2642                                 pr_warn_once("NFSv4: server %s is incorrectly "
2643                                                 "applying open mode checks to "
2644                                                 "a SETATTR that is not "
2645                                                 "changing file size.\n",
2646                                                 server->nfs_client->cl_hostname);
2647                         }
2648                         if (state && !(state->state & FMODE_WRITE)) {
2649                                 err = -EBADF;
2650                                 if (sattr->ia_valid & ATTR_OPEN)
2651                                         err = -EACCES;
2652                                 goto out;
2653                         }
2654                 }
2655                 err = nfs4_handle_exception(server, err, &exception);
2656         } while (exception.retry);
2657 out:
2658         return err;
2659 }
2660
2661 struct nfs4_closedata {
2662         struct inode *inode;
2663         struct nfs4_state *state;
2664         struct nfs_closeargs arg;
2665         struct nfs_closeres res;
2666         struct nfs_fattr fattr;
2667         unsigned long timestamp;
2668         bool roc;
2669         u32 roc_barrier;
2670 };
2671
2672 static void nfs4_free_closedata(void *data)
2673 {
2674         struct nfs4_closedata *calldata = data;
2675         struct nfs4_state_owner *sp = calldata->state->owner;
2676         struct super_block *sb = calldata->state->inode->i_sb;
2677
2678         if (calldata->roc)
2679                 pnfs_roc_release(calldata->state->inode);
2680         nfs4_put_open_state(calldata->state);
2681         nfs_free_seqid(calldata->arg.seqid);
2682         nfs4_put_state_owner(sp);
2683         nfs_sb_deactive(sb);
2684         kfree(calldata);
2685 }
2686
2687 static void nfs4_close_done(struct rpc_task *task, void *data)
2688 {
2689         struct nfs4_closedata *calldata = data;
2690         struct nfs4_state *state = calldata->state;
2691         struct nfs_server *server = NFS_SERVER(calldata->inode);
2692         nfs4_stateid *res_stateid = NULL;
2693
2694         dprintk("%s: begin!\n", __func__);
2695         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2696                 return;
2697         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2698         /* hmm. we are done with the inode, and in the process of freeing
2699          * the state_owner. we keep this around to process errors
2700          */
2701         switch (task->tk_status) {
2702                 case 0:
2703                         res_stateid = &calldata->res.stateid;
2704                         if (calldata->roc)
2705                                 pnfs_roc_set_barrier(state->inode,
2706                                                      calldata->roc_barrier);
2707                         renew_lease(server, calldata->timestamp);
2708                         break;
2709                 case -NFS4ERR_ADMIN_REVOKED:
2710                 case -NFS4ERR_STALE_STATEID:
2711                 case -NFS4ERR_OLD_STATEID:
2712                 case -NFS4ERR_BAD_STATEID:
2713                 case -NFS4ERR_EXPIRED:
2714                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2715                                                 &state->open_stateid)) {
2716                                 rpc_restart_call_prepare(task);
2717                                 goto out_release;
2718                         }
2719                         if (calldata->arg.fmode == 0)
2720                                 break;
2721                 default:
2722                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2723                                 rpc_restart_call_prepare(task);
2724                                 goto out_release;
2725                         }
2726         }
2727         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2728                         res_stateid, calldata->arg.fmode);
2729 out_release:
2730         nfs_release_seqid(calldata->arg.seqid);
2731         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2732         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2733 }
2734
2735 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2736 {
2737         struct nfs4_closedata *calldata = data;
2738         struct nfs4_state *state = calldata->state;
2739         struct inode *inode = calldata->inode;
2740         bool is_rdonly, is_wronly, is_rdwr;
2741         int call_close = 0;
2742
2743         dprintk("%s: begin!\n", __func__);
2744         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2745                 goto out_wait;
2746
2747         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2748         spin_lock(&state->owner->so_lock);
2749         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2750         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2751         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2752         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2753         /* Calculate the change in open mode */
2754         calldata->arg.fmode = 0;
2755         if (state->n_rdwr == 0) {
2756                 if (state->n_rdonly == 0)
2757                         call_close |= is_rdonly;
2758                 else if (is_rdonly)
2759                         calldata->arg.fmode |= FMODE_READ;
2760                 if (state->n_wronly == 0)
2761                         call_close |= is_wronly;
2762                 else if (is_wronly)
2763                         calldata->arg.fmode |= FMODE_WRITE;
2764         } else if (is_rdwr)
2765                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2766
2767         if (calldata->arg.fmode == 0)
2768                 call_close |= is_rdwr;
2769
2770         if (!nfs4_valid_open_stateid(state))
2771                 call_close = 0;
2772         spin_unlock(&state->owner->so_lock);
2773
2774         if (!call_close) {
2775                 /* Note: exit _without_ calling nfs4_close_done */
2776                 goto out_no_action;
2777         }
2778
2779         if (calldata->arg.fmode == 0)
2780                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2781         if (calldata->roc)
2782                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2783
2784         calldata->arg.share_access =
2785                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2786                                 calldata->arg.fmode, 0);
2787
2788         nfs_fattr_init(calldata->res.fattr);
2789         calldata->timestamp = jiffies;
2790         if (nfs4_setup_sequence(NFS_SERVER(inode),
2791                                 &calldata->arg.seq_args,
2792                                 &calldata->res.seq_res,
2793                                 task) != 0)
2794                 nfs_release_seqid(calldata->arg.seqid);
2795         dprintk("%s: done!\n", __func__);
2796         return;
2797 out_no_action:
2798         task->tk_action = NULL;
2799 out_wait:
2800         nfs4_sequence_done(task, &calldata->res.seq_res);
2801 }
2802
2803 static const struct rpc_call_ops nfs4_close_ops = {
2804         .rpc_call_prepare = nfs4_close_prepare,
2805         .rpc_call_done = nfs4_close_done,
2806         .rpc_release = nfs4_free_closedata,
2807 };
2808
2809 static bool nfs4_roc(struct inode *inode)
2810 {
2811         if (!nfs_have_layout(inode))
2812                 return false;
2813         return pnfs_roc(inode);
2814 }
2815
2816 /*
2817  * It is possible for data to be read/written from a mem-mapped file
2818  * after the sys_close call (which hits the vfs layer as a flush).
2819  * This means that we can't safely call nfsv4 close on a file until
2820  * the inode is cleared. This in turn means that we are not good
2821  * NFSv4 citizens - we do not indicate to the server to update the file's
2822  * share state even when we are done with one of the three share
2823  * stateid's in the inode.
2824  *
2825  * NOTE: Caller must be holding the sp->so_owner semaphore!
2826  */
2827 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2828 {
2829         struct nfs_server *server = NFS_SERVER(state->inode);
2830         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2831         struct nfs4_closedata *calldata;
2832         struct nfs4_state_owner *sp = state->owner;
2833         struct rpc_task *task;
2834         struct rpc_message msg = {
2835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2836                 .rpc_cred = state->owner->so_cred,
2837         };
2838         struct rpc_task_setup task_setup_data = {
2839                 .rpc_client = server->client,
2840                 .rpc_message = &msg,
2841                 .callback_ops = &nfs4_close_ops,
2842                 .workqueue = nfsiod_workqueue,
2843                 .flags = RPC_TASK_ASYNC,
2844         };
2845         int status = -ENOMEM;
2846
2847         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2848                 &task_setup_data.rpc_client, &msg);
2849
2850         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2851         if (calldata == NULL)
2852                 goto out;
2853         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2854         calldata->inode = state->inode;
2855         calldata->state = state;
2856         calldata->arg.fh = NFS_FH(state->inode);
2857         /* Serialization for the sequence id */
2858         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2859         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2860         if (IS_ERR(calldata->arg.seqid))
2861                 goto out_free_calldata;
2862         calldata->arg.fmode = 0;
2863         calldata->arg.bitmask = server->cache_consistency_bitmask;
2864         calldata->res.fattr = &calldata->fattr;
2865         calldata->res.seqid = calldata->arg.seqid;
2866         calldata->res.server = server;
2867         calldata->roc = nfs4_roc(state->inode);
2868         nfs_sb_active(calldata->inode->i_sb);
2869
2870         msg.rpc_argp = &calldata->arg;
2871         msg.rpc_resp = &calldata->res;
2872         task_setup_data.callback_data = calldata;
2873         task = rpc_run_task(&task_setup_data);
2874         if (IS_ERR(task))
2875                 return PTR_ERR(task);
2876         status = 0;
2877         if (wait)
2878                 status = rpc_wait_for_completion_task(task);
2879         rpc_put_task(task);
2880         return status;
2881 out_free_calldata:
2882         kfree(calldata);
2883 out:
2884         nfs4_put_open_state(state);
2885         nfs4_put_state_owner(sp);
2886         return status;
2887 }
2888
2889 static struct inode *
2890 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2891                 int open_flags, struct iattr *attr, int *opened)
2892 {
2893         struct nfs4_state *state;
2894         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2895
2896         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2897
2898         /* Protect against concurrent sillydeletes */
2899         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2900
2901         nfs4_label_release_security(label);
2902
2903         if (IS_ERR(state))
2904                 return ERR_CAST(state);
2905         return state->inode;
2906 }
2907
2908 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2909 {
2910         if (ctx->state == NULL)
2911                 return;
2912         if (is_sync)
2913                 nfs4_close_sync(ctx->state, ctx->mode);
2914         else
2915                 nfs4_close_state(ctx->state, ctx->mode);
2916 }
2917
2918 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2919 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2920 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2921
2922 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2923 {
2924         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
2925         struct nfs4_server_caps_arg args = {
2926                 .fhandle = fhandle,
2927                 .bitmask = bitmask,
2928         };
2929         struct nfs4_server_caps_res res = {};
2930         struct rpc_message msg = {
2931                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2932                 .rpc_argp = &args,
2933                 .rpc_resp = &res,
2934         };
2935         int status;
2936
2937         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
2938                      FATTR4_WORD0_FH_EXPIRE_TYPE |
2939                      FATTR4_WORD0_LINK_SUPPORT |
2940                      FATTR4_WORD0_SYMLINK_SUPPORT |
2941                      FATTR4_WORD0_ACLSUPPORT;
2942         if (minorversion)
2943                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
2944
2945         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2946         if (status == 0) {
2947                 /* Sanity check the server answers */
2948                 switch (minorversion) {
2949                 case 0:
2950                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2951                         res.attr_bitmask[2] = 0;
2952                         break;
2953                 case 1:
2954                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2955                         break;
2956                 case 2:
2957                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2958                 }
2959                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2960                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2961                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2962                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2963                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2964                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2965                                 NFS_CAP_SECURITY_LABEL);
2966                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2967                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2968                         server->caps |= NFS_CAP_ACLS;
2969                 if (res.has_links != 0)
2970                         server->caps |= NFS_CAP_HARDLINKS;
2971                 if (res.has_symlinks != 0)
2972                         server->caps |= NFS_CAP_SYMLINKS;
2973                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2974                         server->caps |= NFS_CAP_FILEID;
2975                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2976                         server->caps |= NFS_CAP_MODE;
2977                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2978                         server->caps |= NFS_CAP_NLINK;
2979                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2980                         server->caps |= NFS_CAP_OWNER;
2981                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2982                         server->caps |= NFS_CAP_OWNER_GROUP;
2983                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2984                         server->caps |= NFS_CAP_ATIME;
2985                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2986                         server->caps |= NFS_CAP_CTIME;
2987                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2988                         server->caps |= NFS_CAP_MTIME;
2989 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2990                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2991                         server->caps |= NFS_CAP_SECURITY_LABEL;
2992 #endif
2993                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2994                                 sizeof(server->attr_bitmask));
2995                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2996
2997                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2998                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2999                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3000                 server->cache_consistency_bitmask[2] = 0;
3001                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3002                         sizeof(server->exclcreat_bitmask));
3003                 server->acl_bitmask = res.acl_bitmask;
3004                 server->fh_expire_type = res.fh_expire_type;
3005         }
3006
3007         return status;
3008 }
3009
3010 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3011 {
3012         struct nfs4_exception exception = { };
3013         int err;
3014         do {
3015                 err = nfs4_handle_exception(server,
3016                                 _nfs4_server_capabilities(server, fhandle),
3017                                 &exception);
3018         } while (exception.retry);
3019         return err;
3020 }
3021
3022 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3023                 struct nfs_fsinfo *info)
3024 {
3025         u32 bitmask[3];
3026         struct nfs4_lookup_root_arg args = {
3027                 .bitmask = bitmask,
3028         };
3029         struct nfs4_lookup_res res = {
3030                 .server = server,
3031                 .fattr = info->fattr,
3032                 .fh = fhandle,
3033         };
3034         struct rpc_message msg = {
3035                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3036                 .rpc_argp = &args,
3037                 .rpc_resp = &res,
3038         };
3039
3040         bitmask[0] = nfs4_fattr_bitmap[0];
3041         bitmask[1] = nfs4_fattr_bitmap[1];
3042         /*
3043          * Process the label in the upcoming getfattr
3044          */
3045         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3046
3047         nfs_fattr_init(info->fattr);
3048         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3049 }
3050
3051 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3052                 struct nfs_fsinfo *info)
3053 {
3054         struct nfs4_exception exception = { };
3055         int err;
3056         do {
3057                 err = _nfs4_lookup_root(server, fhandle, info);
3058                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3059                 switch (err) {
3060                 case 0:
3061                 case -NFS4ERR_WRONGSEC:
3062                         goto out;
3063                 default:
3064                         err = nfs4_handle_exception(server, err, &exception);
3065                 }
3066         } while (exception.retry);
3067 out:
3068         return err;
3069 }
3070
3071 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3072                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3073 {
3074         struct rpc_auth_create_args auth_args = {
3075                 .pseudoflavor = flavor,
3076         };
3077         struct rpc_auth *auth;
3078         int ret;
3079
3080         auth = rpcauth_create(&auth_args, server->client);
3081         if (IS_ERR(auth)) {
3082                 ret = -EACCES;
3083                 goto out;
3084         }
3085         ret = nfs4_lookup_root(server, fhandle, info);
3086 out:
3087         return ret;
3088 }
3089
3090 /*
3091  * Retry pseudoroot lookup with various security flavors.  We do this when:
3092  *
3093  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3094  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3095  *
3096  * Returns zero on success, or a negative NFS4ERR value, or a
3097  * negative errno value.
3098  */
3099 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3100                               struct nfs_fsinfo *info)
3101 {
3102         /* Per 3530bis 15.33.5 */
3103         static const rpc_authflavor_t flav_array[] = {
3104                 RPC_AUTH_GSS_KRB5P,
3105                 RPC_AUTH_GSS_KRB5I,
3106                 RPC_AUTH_GSS_KRB5,
3107                 RPC_AUTH_UNIX,                  /* courtesy */
3108                 RPC_AUTH_NULL,
3109         };
3110         int status = -EPERM;
3111         size_t i;
3112
3113         if (server->auth_info.flavor_len > 0) {
3114                 /* try each flavor specified by user */
3115                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3116                         status = nfs4_lookup_root_sec(server, fhandle, info,
3117                                                 server->auth_info.flavors[i]);
3118                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3119                                 continue;
3120                         break;
3121                 }
3122         } else {
3123                 /* no flavors specified by user, try default list */
3124                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3125                         status = nfs4_lookup_root_sec(server, fhandle, info,
3126                                                       flav_array[i]);
3127                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3128                                 continue;
3129                         break;
3130                 }
3131         }
3132
3133         /*
3134          * -EACCESS could mean that the user doesn't have correct permissions
3135          * to access the mount.  It could also mean that we tried to mount
3136          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3137          * existing mount programs don't handle -EACCES very well so it should
3138          * be mapped to -EPERM instead.
3139          */
3140         if (status == -EACCES)
3141                 status = -EPERM;
3142         return status;
3143 }
3144
3145 static int nfs4_do_find_root_sec(struct nfs_server *server,
3146                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3147 {
3148         int mv = server->nfs_client->cl_minorversion;
3149         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3150 }
3151
3152 /**
3153  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3154  * @server: initialized nfs_server handle
3155  * @fhandle: we fill in the pseudo-fs root file handle
3156  * @info: we fill in an FSINFO struct
3157  * @auth_probe: probe the auth flavours
3158  *
3159  * Returns zero on success, or a negative errno.
3160  */
3161 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3162                          struct nfs_fsinfo *info,
3163                          bool auth_probe)
3164 {
3165         int status = 0;
3166
3167         if (!auth_probe)
3168                 status = nfs4_lookup_root(server, fhandle, info);
3169
3170         if (auth_probe || status == NFS4ERR_WRONGSEC)
3171                 status = nfs4_do_find_root_sec(server, fhandle, info);
3172
3173         if (status == 0)
3174                 status = nfs4_server_capabilities(server, fhandle);
3175         if (status == 0)
3176                 status = nfs4_do_fsinfo(server, fhandle, info);
3177
3178         return nfs4_map_errors(status);
3179 }
3180
3181 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3182                               struct nfs_fsinfo *info)
3183 {
3184         int error;
3185         struct nfs_fattr *fattr = info->fattr;
3186         struct nfs4_label *label = NULL;
3187
3188         error = nfs4_server_capabilities(server, mntfh);
3189         if (error < 0) {
3190                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3191                 return error;
3192         }
3193
3194         label = nfs4_label_alloc(server, GFP_KERNEL);
3195         if (IS_ERR(label))
3196                 return PTR_ERR(label);
3197
3198         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3199         if (error < 0) {
3200                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3201                 goto err_free_label;
3202         }
3203
3204         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3205             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3206                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3207
3208 err_free_label:
3209         nfs4_label_free(label);
3210
3211         return error;
3212 }
3213
3214 /*
3215  * Get locations and (maybe) other attributes of a referral.
3216  * Note that we'll actually follow the referral later when
3217  * we detect fsid mismatch in inode revalidation
3218  */
3219 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3220                              const struct qstr *name, struct nfs_fattr *fattr,
3221                              struct nfs_fh *fhandle)
3222 {
3223         int status = -ENOMEM;
3224         struct page *page = NULL;
3225         struct nfs4_fs_locations *locations = NULL;
3226
3227         page = alloc_page(GFP_KERNEL);
3228         if (page == NULL)
3229                 goto out;
3230         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3231         if (locations == NULL)
3232                 goto out;
3233
3234         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3235         if (status != 0)
3236                 goto out;
3237
3238         /*
3239          * If the fsid didn't change, this is a migration event, not a
3240          * referral.  Cause us to drop into the exception handler, which
3241          * will kick off migration recovery.
3242          */
3243         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3244                 dprintk("%s: server did not return a different fsid for"
3245                         " a referral at %s\n", __func__, name->name);
3246                 status = -NFS4ERR_MOVED;
3247                 goto out;
3248         }
3249         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3250         nfs_fixup_referral_attributes(&locations->fattr);
3251
3252         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3253         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3254         memset(fhandle, 0, sizeof(struct nfs_fh));
3255 out:
3256         if (page)
3257                 __free_page(page);
3258         kfree(locations);
3259         return status;
3260 }
3261
3262 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3263                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3264 {
3265         struct nfs4_getattr_arg args = {
3266                 .fh = fhandle,
3267                 .bitmask = server->attr_bitmask,
3268         };
3269         struct nfs4_getattr_res res = {
3270                 .fattr = fattr,
3271                 .label = label,
3272                 .server = server,
3273         };
3274         struct rpc_message msg = {
3275                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3276                 .rpc_argp = &args,
3277                 .rpc_resp = &res,
3278         };
3279
3280         args.bitmask = nfs4_bitmask(server, label);
3281
3282         nfs_fattr_init(fattr);
3283         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3284 }
3285
3286 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3287                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3288 {
3289         struct nfs4_exception exception = { };
3290         int err;
3291         do {
3292                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3293                 trace_nfs4_getattr(server, fhandle, fattr, err);
3294                 err = nfs4_handle_exception(server, err,
3295                                 &exception);
3296         } while (exception.retry);
3297         return err;
3298 }
3299
3300 /*
3301  * The file is not closed if it is opened due to the a request to change
3302  * the size of the file. The open call will not be needed once the
3303  * VFS layer lookup-intents are implemented.
3304  *
3305  * Close is called when the inode is destroyed.
3306  * If we haven't opened the file for O_WRONLY, we
3307  * need to in the size_change case to obtain a stateid.
3308  *
3309  * Got race?
3310  * Because OPEN is always done by name in nfsv4, it is
3311  * possible that we opened a different file by the same
3312  * name.  We can recognize this race condition, but we
3313  * can't do anything about it besides returning an error.
3314  *
3315  * This will be fixed with VFS changes (lookup-intent).
3316  */
3317 static int
3318 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3319                   struct iattr *sattr)
3320 {
3321         struct inode *inode = d_inode(dentry);
3322         struct rpc_cred *cred = NULL;
3323         struct nfs4_state *state = NULL;
3324         struct nfs4_label *label = NULL;
3325         int status;
3326
3327         if (pnfs_ld_layoutret_on_setattr(inode) &&
3328             sattr->ia_valid & ATTR_SIZE &&
3329             sattr->ia_size < i_size_read(inode))
3330                 pnfs_commit_and_return_layout(inode);
3331
3332         nfs_fattr_init(fattr);
3333
3334         /* Deal with open(O_TRUNC) */
3335         if (sattr->ia_valid & ATTR_OPEN)
3336                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3337
3338         /* Optimization: if the end result is no change, don't RPC */
3339         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3340                 return 0;
3341
3342         /* Search for an existing open(O_WRITE) file */
3343         if (sattr->ia_valid & ATTR_FILE) {
3344                 struct nfs_open_context *ctx;
3345
3346                 ctx = nfs_file_open_context(sattr->ia_file);
3347                 if (ctx) {
3348                         cred = ctx->cred;
3349                         state = ctx->state;
3350                 }
3351         }
3352
3353         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3354         if (IS_ERR(label))
3355                 return PTR_ERR(label);
3356
3357         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3358         if (status == 0) {
3359                 nfs_setattr_update_inode(inode, sattr, fattr);
3360                 nfs_setsecurity(inode, fattr, label);
3361         }
3362         nfs4_label_free(label);
3363         return status;
3364 }
3365
3366 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3367                 const struct qstr *name, struct nfs_fh *fhandle,
3368                 struct nfs_fattr *fattr, struct nfs4_label *label)
3369 {
3370         struct nfs_server *server = NFS_SERVER(dir);
3371         int                    status;
3372         struct nfs4_lookup_arg args = {
3373                 .bitmask = server->attr_bitmask,
3374                 .dir_fh = NFS_FH(dir),
3375                 .name = name,
3376         };
3377         struct nfs4_lookup_res res = {
3378                 .server = server,
3379                 .fattr = fattr,
3380                 .label = label,
3381                 .fh = fhandle,
3382         };
3383         struct rpc_message msg = {
3384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3385                 .rpc_argp = &args,
3386                 .rpc_resp = &res,
3387         };
3388
3389         args.bitmask = nfs4_bitmask(server, label);
3390
3391         nfs_fattr_init(fattr);
3392
3393         dprintk("NFS call  lookup %s\n", name->name);
3394         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3395         dprintk("NFS reply lookup: %d\n", status);
3396         return status;
3397 }
3398
3399 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3400 {
3401         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3402                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3403         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3404         fattr->nlink = 2;
3405 }
3406
3407 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3408                                    struct qstr *name, struct nfs_fh *fhandle,
3409                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3410 {
3411         struct nfs4_exception exception = { };
3412         struct rpc_clnt *client = *clnt;
3413         int err;
3414         do {
3415                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3416                 trace_nfs4_lookup(dir, name, err);
3417                 switch (err) {
3418                 case -NFS4ERR_BADNAME:
3419                         err = -ENOENT;
3420                         goto out;
3421                 case -NFS4ERR_MOVED:
3422                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3423                         if (err == -NFS4ERR_MOVED)
3424                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3425                         goto out;
3426                 case -NFS4ERR_WRONGSEC:
3427                         err = -EPERM;
3428                         if (client != *clnt)
3429                                 goto out;
3430                         client = nfs4_negotiate_security(client, dir, name);
3431                         if (IS_ERR(client))
3432                                 return PTR_ERR(client);
3433
3434                         exception.retry = 1;
3435                         break;
3436                 default:
3437                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3438                 }
3439         } while (exception.retry);
3440
3441 out:
3442         if (err == 0)
3443                 *clnt = client;
3444         else if (client != *clnt)
3445                 rpc_shutdown_client(client);
3446
3447         return err;
3448 }
3449
3450 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3451                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3452                             struct nfs4_label *label)
3453 {
3454         int status;
3455         struct rpc_clnt *client = NFS_CLIENT(dir);
3456
3457         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3458         if (client != NFS_CLIENT(dir)) {
3459                 rpc_shutdown_client(client);
3460                 nfs_fixup_secinfo_attributes(fattr);
3461         }
3462         return status;
3463 }
3464
3465 struct rpc_clnt *
3466 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3467                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3468 {
3469         struct rpc_clnt *client = NFS_CLIENT(dir);
3470         int status;
3471
3472         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3473         if (status < 0)
3474                 return ERR_PTR(status);
3475         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3476 }
3477
3478 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3479 {
3480         struct nfs_server *server = NFS_SERVER(inode);
3481         struct nfs4_accessargs args = {
3482                 .fh = NFS_FH(inode),
3483                 .bitmask = server->cache_consistency_bitmask,
3484         };
3485         struct nfs4_accessres res = {
3486                 .server = server,
3487         };
3488         struct rpc_message msg = {
3489                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3490                 .rpc_argp = &args,
3491                 .rpc_resp = &res,
3492                 .rpc_cred = entry->cred,
3493         };
3494         int mode = entry->mask;
3495         int status = 0;
3496
3497         /*
3498          * Determine which access bits we want to ask for...
3499          */
3500         if (mode & MAY_READ)
3501                 args.access |= NFS4_ACCESS_READ;
3502         if (S_ISDIR(inode->i_mode)) {
3503                 if (mode & MAY_WRITE)
3504                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3505                 if (mode & MAY_EXEC)
3506                         args.access |= NFS4_ACCESS_LOOKUP;
3507         } else {
3508                 if (mode & MAY_WRITE)
3509                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3510                 if (mode & MAY_EXEC)
3511                         args.access |= NFS4_ACCESS_EXECUTE;
3512         }
3513
3514         res.fattr = nfs_alloc_fattr();
3515         if (res.fattr == NULL)
3516                 return -ENOMEM;
3517
3518         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3519         if (!status) {
3520                 nfs_access_set_mask(entry, res.access);
3521                 nfs_refresh_inode(inode, res.fattr);
3522         }
3523         nfs_free_fattr(res.fattr);
3524         return status;
3525 }
3526
3527 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3528 {
3529         struct nfs4_exception exception = { };
3530         int err;
3531         do {
3532                 err = _nfs4_proc_access(inode, entry);
3533                 trace_nfs4_access(inode, err);
3534                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3535                                 &exception);
3536         } while (exception.retry);
3537         return err;
3538 }
3539
3540 /*
3541  * TODO: For the time being, we don't try to get any attributes
3542  * along with any of the zero-copy operations READ, READDIR,
3543  * READLINK, WRITE.
3544  *
3545  * In the case of the first three, we want to put the GETATTR
3546  * after the read-type operation -- this is because it is hard
3547  * to predict the length of a GETATTR response in v4, and thus
3548  * align the READ data correctly.  This means that the GETATTR
3549  * may end up partially falling into the page cache, and we should
3550  * shift it into the 'tail' of the xdr_buf before processing.
3551  * To do this efficiently, we need to know the total length
3552  * of data received, which doesn't seem to be available outside
3553  * of the RPC layer.
3554  *
3555  * In the case of WRITE, we also want to put the GETATTR after
3556  * the operation -- in this case because we want to make sure
3557  * we get the post-operation mtime and size.
3558  *
3559  * Both of these changes to the XDR layer would in fact be quite
3560  * minor, but I decided to leave them for a subsequent patch.
3561  */
3562 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3563                 unsigned int pgbase, unsigned int pglen)
3564 {
3565         struct nfs4_readlink args = {
3566                 .fh       = NFS_FH(inode),
3567                 .pgbase   = pgbase,
3568                 .pglen    = pglen,
3569                 .pages    = &page,
3570         };
3571         struct nfs4_readlink_res res;
3572         struct rpc_message msg = {
3573                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3574                 .rpc_argp = &args,
3575                 .rpc_resp = &res,
3576         };
3577
3578         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3579 }
3580
3581 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3582                 unsigned int pgbase, unsigned int pglen)
3583 {
3584         struct nfs4_exception exception = { };
3585         int err;
3586         do {
3587                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3588                 trace_nfs4_readlink(inode, err);
3589                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3590                                 &exception);
3591         } while (exception.retry);
3592         return err;
3593 }
3594
3595 /*
3596  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3597  */
3598 static int
3599 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3600                  int flags)
3601 {
3602         struct nfs4_label l, *ilabel = NULL;
3603         struct nfs_open_context *ctx;
3604         struct nfs4_state *state;
3605         int status = 0;
3606
3607         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3608         if (IS_ERR(ctx))
3609                 return PTR_ERR(ctx);
3610
3611         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3612
3613         sattr->ia_mode &= ~current_umask();
3614         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3615         if (IS_ERR(state)) {
3616                 status = PTR_ERR(state);
3617                 goto out;
3618         }
3619 out:
3620         nfs4_label_release_security(ilabel);
3621         put_nfs_open_context(ctx);
3622         return status;
3623 }
3624
3625 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3626 {
3627         struct nfs_server *server = NFS_SERVER(dir);
3628         struct nfs_removeargs args = {
3629                 .fh = NFS_FH(dir),
3630                 .name = *name,
3631         };
3632         struct nfs_removeres res = {
3633                 .server = server,
3634         };
3635         struct rpc_message msg = {
3636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3637                 .rpc_argp = &args,
3638                 .rpc_resp = &res,
3639         };
3640         int status;
3641
3642         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3643         if (status == 0)
3644                 update_changeattr(dir, &res.cinfo);
3645         return status;
3646 }
3647
3648 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3649 {
3650         struct nfs4_exception exception = { };
3651         int err;
3652         do {
3653                 err = _nfs4_proc_remove(dir, name);
3654                 trace_nfs4_remove(dir, name, err);
3655                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3656                                 &exception);
3657         } while (exception.retry);
3658         return err;
3659 }
3660
3661 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3662 {
3663         struct nfs_server *server = NFS_SERVER(dir);
3664         struct nfs_removeargs *args = msg->rpc_argp;
3665         struct nfs_removeres *res = msg->rpc_resp;
3666
3667         res->server = server;
3668         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3669         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3670
3671         nfs_fattr_init(res->dir_attr);
3672 }
3673
3674 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3675 {
3676         nfs4_setup_sequence(NFS_SERVER(data->dir),
3677                         &data->args.seq_args,
3678                         &data->res.seq_res,
3679                         task);
3680 }
3681
3682 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3683 {
3684         struct nfs_unlinkdata *data = task->tk_calldata;
3685         struct nfs_removeres *res = &data->res;
3686
3687         if (!nfs4_sequence_done(task, &res->seq_res))
3688                 return 0;
3689         if (nfs4_async_handle_error(task, res->server, NULL,
3690                                     &data->timeout) == -EAGAIN)
3691                 return 0;
3692         update_changeattr(dir, &res->cinfo);
3693         return 1;
3694 }
3695
3696 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3697 {
3698         struct nfs_server *server = NFS_SERVER(dir);
3699         struct nfs_renameargs *arg = msg->rpc_argp;
3700         struct nfs_renameres *res = msg->rpc_resp;
3701
3702         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3703         res->server = server;
3704         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3705 }
3706
3707 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3708 {
3709         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3710                         &data->args.seq_args,
3711                         &data->res.seq_res,
3712                         task);
3713 }
3714
3715 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3716                                  struct inode *new_dir)
3717 {
3718         struct nfs_renamedata *data = task->tk_calldata;
3719         struct nfs_renameres *res = &data->res;
3720
3721         if (!nfs4_sequence_done(task, &res->seq_res))
3722                 return 0;
3723         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3724                 return 0;
3725
3726         update_changeattr(old_dir, &res->old_cinfo);
3727         update_changeattr(new_dir, &res->new_cinfo);
3728         return 1;
3729 }
3730
3731 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3732 {
3733         struct nfs_server *server = NFS_SERVER(inode);
3734         struct nfs4_link_arg arg = {
3735                 .fh     = NFS_FH(inode),
3736                 .dir_fh = NFS_FH(dir),
3737                 .name   = name,
3738                 .bitmask = server->attr_bitmask,
3739         };
3740         struct nfs4_link_res res = {
3741                 .server = server,
3742                 .label = NULL,
3743         };
3744         struct rpc_message msg = {
3745                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3746                 .rpc_argp = &arg,
3747                 .rpc_resp = &res,
3748         };
3749         int status = -ENOMEM;
3750
3751         res.fattr = nfs_alloc_fattr();
3752         if (res.fattr == NULL)
3753                 goto out;
3754
3755         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3756         if (IS_ERR(res.label)) {
3757                 status = PTR_ERR(res.label);
3758                 goto out;
3759         }
3760         arg.bitmask = nfs4_bitmask(server, res.label);
3761
3762         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3763         if (!status) {
3764                 update_changeattr(dir, &res.cinfo);
3765                 status = nfs_post_op_update_inode(inode, res.fattr);
3766                 if (!status)
3767                         nfs_setsecurity(inode, res.fattr, res.label);
3768         }
3769
3770
3771         nfs4_label_free(res.label);
3772
3773 out:
3774         nfs_free_fattr(res.fattr);
3775         return status;
3776 }
3777
3778 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3779 {
3780         struct nfs4_exception exception = { };
3781         int err;
3782         do {
3783                 err = nfs4_handle_exception(NFS_SERVER(inode),
3784                                 _nfs4_proc_link(inode, dir, name),
3785                                 &exception);
3786         } while (exception.retry);
3787         return err;
3788 }
3789
3790 struct nfs4_createdata {
3791         struct rpc_message msg;
3792         struct nfs4_create_arg arg;
3793         struct nfs4_create_res res;
3794         struct nfs_fh fh;
3795         struct nfs_fattr fattr;
3796         struct nfs4_label *label;
3797 };
3798
3799 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3800                 struct qstr *name, struct iattr *sattr, u32 ftype)
3801 {
3802         struct nfs4_createdata *data;
3803
3804         data = kzalloc(sizeof(*data), GFP_KERNEL);
3805         if (data != NULL) {
3806                 struct nfs_server *server = NFS_SERVER(dir);
3807
3808                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3809                 if (IS_ERR(data->label))
3810                         goto out_free;
3811
3812                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3813                 data->msg.rpc_argp = &data->arg;
3814                 data->msg.rpc_resp = &data->res;
3815                 data->arg.dir_fh = NFS_FH(dir);
3816                 data->arg.server = server;
3817                 data->arg.name = name;
3818                 data->arg.attrs = sattr;
3819                 data->arg.ftype = ftype;
3820                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3821                 data->res.server = server;
3822                 data->res.fh = &data->fh;
3823                 data->res.fattr = &data->fattr;
3824                 data->res.label = data->label;
3825                 nfs_fattr_init(data->res.fattr);
3826         }
3827         return data;
3828 out_free:
3829         kfree(data);
3830         return NULL;
3831 }
3832
3833 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3834 {
3835         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3836                                     &data->arg.seq_args, &data->res.seq_res, 1);
3837         if (status == 0) {
3838                 update_changeattr(dir, &data->res.dir_cinfo);
3839                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3840         }
3841         return status;
3842 }
3843
3844 static void nfs4_free_createdata(struct nfs4_createdata *data)
3845 {
3846         nfs4_label_free(data->label);
3847         kfree(data);
3848 }
3849
3850 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3851                 struct page *page, unsigned int len, struct iattr *sattr,
3852                 struct nfs4_label *label)
3853 {
3854         struct nfs4_createdata *data;
3855         int status = -ENAMETOOLONG;
3856
3857         if (len > NFS4_MAXPATHLEN)
3858                 goto out;
3859
3860         status = -ENOMEM;
3861         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3862         if (data == NULL)
3863                 goto out;
3864
3865         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3866         data->arg.u.symlink.pages = &page;
3867         data->arg.u.symlink.len = len;
3868         data->arg.label = label;
3869
3870         status = nfs4_do_create(dir, dentry, data);
3871
3872         nfs4_free_createdata(data);
3873 out:
3874         return status;
3875 }
3876
3877 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3878                 struct page *page, unsigned int len, struct iattr *sattr)
3879 {
3880         struct nfs4_exception exception = { };
3881         struct nfs4_label l, *label = NULL;
3882         int err;
3883
3884         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3885
3886         do {
3887                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3888                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3889                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3890                                 &exception);
3891         } while (exception.retry);
3892
3893         nfs4_label_release_security(label);
3894         return err;
3895 }
3896
3897 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3898                 struct iattr *sattr, struct nfs4_label *label)
3899 {
3900         struct nfs4_createdata *data;
3901         int status = -ENOMEM;
3902
3903         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3904         if (data == NULL)
3905                 goto out;
3906
3907         data->arg.label = label;
3908         status = nfs4_do_create(dir, dentry, data);
3909
3910         nfs4_free_createdata(data);
3911 out:
3912         return status;
3913 }
3914
3915 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3916                 struct iattr *sattr)
3917 {
3918         struct nfs4_exception exception = { };
3919         struct nfs4_label l, *label = NULL;
3920         int err;
3921
3922         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3923
3924         sattr->ia_mode &= ~current_umask();
3925         do {
3926                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3927                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3928                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3929                                 &exception);
3930         } while (exception.retry);
3931         nfs4_label_release_security(label);
3932
3933         return err;
3934 }
3935
3936 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3937                 u64 cookie, struct page **pages, unsigned int count, int plus)
3938 {
3939         struct inode            *dir = d_inode(dentry);
3940         struct nfs4_readdir_arg args = {
3941                 .fh = NFS_FH(dir),
3942                 .pages = pages,
3943                 .pgbase = 0,
3944                 .count = count,
3945                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3946                 .plus = plus,
3947         };
3948         struct nfs4_readdir_res res;
3949         struct rpc_message msg = {
3950                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3951                 .rpc_argp = &args,
3952                 .rpc_resp = &res,
3953                 .rpc_cred = cred,
3954         };
3955         int                     status;
3956
3957         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3958                         dentry,
3959                         (unsigned long long)cookie);
3960         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3961         res.pgbase = args.pgbase;
3962         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3963         if (status >= 0) {
3964                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3965                 status += args.pgbase;
3966         }
3967
3968         nfs_invalidate_atime(dir);
3969
3970         dprintk("%s: returns %d\n", __func__, status);
3971         return status;
3972 }
3973
3974 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3975                 u64 cookie, struct page **pages, unsigned int count, int plus)
3976 {
3977         struct nfs4_exception exception = { };
3978         int err;
3979         do {
3980                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3981                                 pages, count, plus);
3982                 trace_nfs4_readdir(d_inode(dentry), err);
3983                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3984                                 &exception);
3985         } while (exception.retry);
3986         return err;
3987 }
3988
3989 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3990                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3991 {
3992         struct nfs4_createdata *data;
3993         int mode = sattr->ia_mode;
3994         int status = -ENOMEM;
3995
3996         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3997         if (data == NULL)
3998                 goto out;
3999
4000         if (S_ISFIFO(mode))
4001                 data->arg.ftype = NF4FIFO;
4002         else if (S_ISBLK(mode)) {
4003                 data->arg.ftype = NF4BLK;
4004                 data->arg.u.device.specdata1 = MAJOR(rdev);
4005                 data->arg.u.device.specdata2 = MINOR(rdev);
4006         }
4007         else if (S_ISCHR(mode)) {
4008                 data->arg.ftype = NF4CHR;
4009                 data->arg.u.device.specdata1 = MAJOR(rdev);
4010                 data->arg.u.device.specdata2 = MINOR(rdev);
4011         } else if (!S_ISSOCK(mode)) {
4012                 status = -EINVAL;
4013                 goto out_free;
4014         }
4015
4016         data->arg.label = label;
4017         status = nfs4_do_create(dir, dentry, data);
4018 out_free:
4019         nfs4_free_createdata(data);
4020 out:
4021         return status;
4022 }
4023
4024 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4025                 struct iattr *sattr, dev_t rdev)
4026 {
4027         struct nfs4_exception exception = { };
4028         struct nfs4_label l, *label = NULL;
4029         int err;
4030
4031         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4032
4033         sattr->ia_mode &= ~current_umask();
4034         do {
4035                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4036                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4037                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4038                                 &exception);
4039         } while (exception.retry);
4040
4041         nfs4_label_release_security(label);
4042
4043         return err;
4044 }
4045
4046 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4047                  struct nfs_fsstat *fsstat)
4048 {
4049         struct nfs4_statfs_arg args = {
4050                 .fh = fhandle,
4051                 .bitmask = server->attr_bitmask,
4052         };
4053         struct nfs4_statfs_res res = {
4054                 .fsstat = fsstat,
4055         };
4056         struct rpc_message msg = {
4057                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4058                 .rpc_argp = &args,
4059                 .rpc_resp = &res,
4060         };
4061
4062         nfs_fattr_init(fsstat->fattr);
4063         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4064 }
4065
4066 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4067 {
4068         struct nfs4_exception exception = { };
4069         int err;
4070         do {
4071                 err = nfs4_handle_exception(server,
4072                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4073                                 &exception);
4074         } while (exception.retry);
4075         return err;
4076 }
4077
4078 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4079                 struct nfs_fsinfo *fsinfo)
4080 {
4081         struct nfs4_fsinfo_arg args = {
4082                 .fh = fhandle,
4083                 .bitmask = server->attr_bitmask,
4084         };
4085         struct nfs4_fsinfo_res res = {
4086                 .fsinfo = fsinfo,
4087         };
4088         struct rpc_message msg = {
4089                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4090                 .rpc_argp = &args,
4091                 .rpc_resp = &res,
4092         };
4093
4094         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4095 }
4096
4097 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4098 {
4099         struct nfs4_exception exception = { };
4100         unsigned long now = jiffies;
4101         int err;
4102
4103         do {
4104                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4105                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4106                 if (err == 0) {
4107                         struct nfs_client *clp = server->nfs_client;
4108
4109                         spin_lock(&clp->cl_lock);
4110                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4111                         clp->cl_last_renewal = now;
4112                         spin_unlock(&clp->cl_lock);
4113                         break;
4114                 }
4115                 err = nfs4_handle_exception(server, err, &exception);
4116         } while (exception.retry);
4117         return err;
4118 }
4119
4120 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4121 {
4122         int error;
4123
4124         nfs_fattr_init(fsinfo->fattr);
4125         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4126         if (error == 0) {
4127                 /* block layout checks this! */
4128                 server->pnfs_blksize = fsinfo->blksize;
4129                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4130         }
4131
4132         return error;
4133 }
4134
4135 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4136                 struct nfs_pathconf *pathconf)
4137 {
4138         struct nfs4_pathconf_arg args = {
4139                 .fh = fhandle,
4140                 .bitmask = server->attr_bitmask,
4141         };
4142         struct nfs4_pathconf_res res = {
4143                 .pathconf = pathconf,
4144         };
4145         struct rpc_message msg = {
4146                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4147                 .rpc_argp = &args,
4148                 .rpc_resp = &res,
4149         };
4150
4151         /* None of the pathconf attributes are mandatory to implement */
4152         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4153                 memset(pathconf, 0, sizeof(*pathconf));
4154                 return 0;
4155         }
4156
4157         nfs_fattr_init(pathconf->fattr);
4158         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4159 }
4160
4161 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4162                 struct nfs_pathconf *pathconf)
4163 {
4164         struct nfs4_exception exception = { };
4165         int err;
4166
4167         do {
4168                 err = nfs4_handle_exception(server,
4169                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4170                                 &exception);
4171         } while (exception.retry);
4172         return err;
4173 }
4174
4175 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4176                 const struct nfs_open_context *ctx,
4177                 const struct nfs_lock_context *l_ctx,
4178                 fmode_t fmode)
4179 {
4180         const struct nfs_lockowner *lockowner = NULL;
4181
4182         if (l_ctx != NULL)
4183                 lockowner = &l_ctx->lockowner;
4184         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4185 }
4186 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4187
4188 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4189                 const struct nfs_open_context *ctx,
4190                 const struct nfs_lock_context *l_ctx,
4191                 fmode_t fmode)
4192 {
4193         nfs4_stateid current_stateid;
4194
4195         /* If the current stateid represents a lost lock, then exit */
4196         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4197                 return true;
4198         return nfs4_stateid_match(stateid, &current_stateid);
4199 }
4200
4201 static bool nfs4_error_stateid_expired(int err)
4202 {
4203         switch (err) {
4204         case -NFS4ERR_DELEG_REVOKED:
4205         case -NFS4ERR_ADMIN_REVOKED:
4206         case -NFS4ERR_BAD_STATEID:
4207         case -NFS4ERR_STALE_STATEID:
4208         case -NFS4ERR_OLD_STATEID:
4209         case -NFS4ERR_OPENMODE:
4210         case -NFS4ERR_EXPIRED:
4211                 return true;
4212         }
4213         return false;
4214 }
4215
4216 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4217 {
4218         nfs_invalidate_atime(hdr->inode);
4219 }
4220
4221 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4222 {
4223         struct nfs_server *server = NFS_SERVER(hdr->inode);
4224
4225         trace_nfs4_read(hdr, task->tk_status);
4226         if (nfs4_async_handle_error(task, server,
4227                                     hdr->args.context->state,
4228                                     NULL) == -EAGAIN) {
4229                 rpc_restart_call_prepare(task);
4230                 return -EAGAIN;
4231         }
4232
4233         __nfs4_read_done_cb(hdr);
4234         if (task->tk_status > 0)
4235                 renew_lease(server, hdr->timestamp);
4236         return 0;
4237 }
4238
4239 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4240                 struct nfs_pgio_args *args)
4241 {
4242
4243         if (!nfs4_error_stateid_expired(task->tk_status) ||
4244                 nfs4_stateid_is_current(&args->stateid,
4245                                 args->context,
4246                                 args->lock_context,
4247                                 FMODE_READ))
4248                 return false;
4249         rpc_restart_call_prepare(task);
4250         return true;
4251 }
4252
4253 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4254 {
4255
4256         dprintk("--> %s\n", __func__);
4257
4258         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4259                 return -EAGAIN;
4260         if (nfs4_read_stateid_changed(task, &hdr->args))
4261                 return -EAGAIN;
4262         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4263                                     nfs4_read_done_cb(task, hdr);
4264 }
4265
4266 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4267                                  struct rpc_message *msg)
4268 {
4269         hdr->timestamp   = jiffies;
4270         hdr->pgio_done_cb = nfs4_read_done_cb;
4271         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4272         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4273 }
4274
4275 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4276                                       struct nfs_pgio_header *hdr)
4277 {
4278         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4279                         &hdr->args.seq_args,
4280                         &hdr->res.seq_res,
4281                         task))
4282                 return 0;
4283         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4284                                 hdr->args.lock_context,
4285                                 hdr->rw_ops->rw_mode) == -EIO)
4286                 return -EIO;
4287         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4288                 return -EIO;
4289         return 0;
4290 }
4291
4292 static int nfs4_write_done_cb(struct rpc_task *task,
4293                               struct nfs_pgio_header *hdr)
4294 {
4295         struct inode *inode = hdr->inode;
4296
4297         trace_nfs4_write(hdr, task->tk_status);
4298         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4299                                     hdr->args.context->state,
4300                                     NULL) == -EAGAIN) {
4301                 rpc_restart_call_prepare(task);
4302                 return -EAGAIN;
4303         }
4304         if (task->tk_status >= 0) {
4305                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4306                 nfs_writeback_update_inode(hdr);
4307         }
4308         return 0;
4309 }
4310
4311 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4312                 struct nfs_pgio_args *args)
4313 {
4314
4315         if (!nfs4_error_stateid_expired(task->tk_status) ||
4316                 nfs4_stateid_is_current(&args->stateid,
4317                                 args->context,
4318                                 args->lock_context,
4319                                 FMODE_WRITE))
4320                 return false;
4321         rpc_restart_call_prepare(task);
4322         return true;
4323 }
4324
4325 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4326 {
4327         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4328                 return -EAGAIN;
4329         if (nfs4_write_stateid_changed(task, &hdr->args))
4330                 return -EAGAIN;
4331         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4332                 nfs4_write_done_cb(task, hdr);
4333 }
4334
4335 static
4336 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4337 {
4338         /* Don't request attributes for pNFS or O_DIRECT writes */
4339         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4340                 return false;
4341         /* Otherwise, request attributes if and only if we don't hold
4342          * a delegation
4343          */
4344         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4345 }
4346
4347 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4348                                   struct rpc_message *msg)
4349 {
4350         struct nfs_server *server = NFS_SERVER(hdr->inode);
4351
4352         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4353                 hdr->args.bitmask = NULL;
4354                 hdr->res.fattr = NULL;
4355         } else
4356                 hdr->args.bitmask = server->cache_consistency_bitmask;
4357
4358         if (!hdr->pgio_done_cb)
4359                 hdr->pgio_done_cb = nfs4_write_done_cb;
4360         hdr->res.server = server;
4361         hdr->timestamp   = jiffies;
4362
4363         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4364         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4365 }
4366
4367 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4368 {
4369         nfs4_setup_sequence(NFS_SERVER(data->inode),
4370                         &data->args.seq_args,
4371                         &data->res.seq_res,
4372                         task);
4373 }
4374
4375 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4376 {
4377         struct inode *inode = data->inode;
4378
4379         trace_nfs4_commit(data, task->tk_status);
4380         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4381                                     NULL, NULL) == -EAGAIN) {
4382                 rpc_restart_call_prepare(task);
4383                 return -EAGAIN;
4384         }
4385         return 0;
4386 }
4387
4388 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4389 {
4390         if (!nfs4_sequence_done(task, &data->res.seq_res))
4391                 return -EAGAIN;
4392         return data->commit_done_cb(task, data);
4393 }
4394
4395 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4396 {
4397         struct nfs_server *server = NFS_SERVER(data->inode);
4398
4399         if (data->commit_done_cb == NULL)
4400                 data->commit_done_cb = nfs4_commit_done_cb;
4401         data->res.server = server;
4402         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4403         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4404 }
4405
4406 struct nfs4_renewdata {
4407         struct nfs_client       *client;
4408         unsigned long           timestamp;
4409 };
4410
4411 /*
4412  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4413  * standalone procedure for queueing an asynchronous RENEW.
4414  */
4415 static void nfs4_renew_release(void *calldata)
4416 {
4417         struct nfs4_renewdata *data = calldata;
4418         struct nfs_client *clp = data->client;
4419
4420         if (atomic_read(&clp->cl_count) > 1)
4421                 nfs4_schedule_state_renewal(clp);
4422         nfs_put_client(clp);
4423         kfree(data);
4424 }
4425
4426 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4427 {
4428         struct nfs4_renewdata *data = calldata;
4429         struct nfs_client *clp = data->client;
4430         unsigned long timestamp = data->timestamp;
4431
4432         trace_nfs4_renew_async(clp, task->tk_status);
4433         switch (task->tk_status) {
4434         case 0:
4435                 break;
4436         case -NFS4ERR_LEASE_MOVED:
4437                 nfs4_schedule_lease_moved_recovery(clp);
4438                 break;
4439         default:
4440                 /* Unless we're shutting down, schedule state recovery! */
4441                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4442                         return;
4443                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4444                         nfs4_schedule_lease_recovery(clp);
4445                         return;
4446                 }
4447                 nfs4_schedule_path_down_recovery(clp);
4448         }
4449         do_renew_lease(clp, timestamp);
4450 }
4451
4452 static const struct rpc_call_ops nfs4_renew_ops = {
4453         .rpc_call_done = nfs4_renew_done,
4454         .rpc_release = nfs4_renew_release,
4455 };
4456
4457 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4458 {
4459         struct rpc_message msg = {
4460                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4461                 .rpc_argp       = clp,
4462                 .rpc_cred       = cred,
4463         };
4464         struct nfs4_renewdata *data;
4465
4466         if (renew_flags == 0)
4467                 return 0;
4468         if (!atomic_inc_not_zero(&clp->cl_count))
4469                 return -EIO;
4470         data = kmalloc(sizeof(*data), GFP_NOFS);
4471         if (data == NULL)
4472                 return -ENOMEM;
4473         data->client = clp;
4474         data->timestamp = jiffies;
4475         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4476                         &nfs4_renew_ops, data);
4477 }
4478
4479 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4480 {
4481         struct rpc_message msg = {
4482                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4483                 .rpc_argp       = clp,
4484                 .rpc_cred       = cred,
4485         };
4486         unsigned long now = jiffies;
4487         int status;
4488
4489         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4490         if (status < 0)
4491                 return status;
4492         do_renew_lease(clp, now);
4493         return 0;
4494 }
4495
4496 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4497 {
4498         return server->caps & NFS_CAP_ACLS;
4499 }
4500
4501 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4502  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4503  * the stack.
4504  */
4505 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4506
4507 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4508                 struct page **pages, unsigned int *pgbase)
4509 {
4510         struct page *newpage, **spages;
4511         int rc = 0;
4512         size_t len;
4513         spages = pages;
4514
4515         do {
4516                 len = min_t(size_t, PAGE_SIZE, buflen);
4517                 newpage = alloc_page(GFP_KERNEL);
4518
4519                 if (newpage == NULL)
4520                         goto unwind;
4521                 memcpy(page_address(newpage), buf, len);
4522                 buf += len;
4523                 buflen -= len;
4524                 *pages++ = newpage;
4525                 rc++;
4526         } while (buflen != 0);
4527
4528         return rc;
4529
4530 unwind:
4531         for(; rc > 0; rc--)
4532                 __free_page(spages[rc-1]);
4533         return -ENOMEM;
4534 }
4535
4536 struct nfs4_cached_acl {
4537         int cached;
4538         size_t len;
4539         char data[0];
4540 };
4541
4542 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4543 {
4544         struct nfs_inode *nfsi = NFS_I(inode);
4545
4546         spin_lock(&inode->i_lock);
4547         kfree(nfsi->nfs4_acl);
4548         nfsi->nfs4_acl = acl;
4549         spin_unlock(&inode->i_lock);
4550 }
4551
4552 static void nfs4_zap_acl_attr(struct inode *inode)
4553 {
4554         nfs4_set_cached_acl(inode, NULL);
4555 }
4556
4557 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4558 {
4559         struct nfs_inode *nfsi = NFS_I(inode);
4560         struct nfs4_cached_acl *acl;
4561         int ret = -ENOENT;
4562
4563         spin_lock(&inode->i_lock);
4564         acl = nfsi->nfs4_acl;
4565         if (acl == NULL)
4566                 goto out;
4567         if (buf == NULL) /* user is just asking for length */
4568                 goto out_len;
4569         if (acl->cached == 0)
4570                 goto out;
4571         ret = -ERANGE; /* see getxattr(2) man page */
4572         if (acl->len > buflen)
4573                 goto out;
4574         memcpy(buf, acl->data, acl->len);
4575 out_len:
4576         ret = acl->len;
4577 out:
4578         spin_unlock(&inode->i_lock);
4579         return ret;
4580 }
4581
4582 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4583 {
4584         struct nfs4_cached_acl *acl;
4585         size_t buflen = sizeof(*acl) + acl_len;
4586
4587         if (buflen <= PAGE_SIZE) {
4588                 acl = kmalloc(buflen, GFP_KERNEL);
4589                 if (acl == NULL)
4590                         goto out;
4591                 acl->cached = 1;
4592                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4593         } else {
4594                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4595                 if (acl == NULL)
4596                         goto out;
4597                 acl->cached = 0;
4598         }
4599         acl->len = acl_len;
4600 out:
4601         nfs4_set_cached_acl(inode, acl);
4602 }
4603
4604 /*
4605  * The getxattr API returns the required buffer length when called with a
4606  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4607  * the required buf.  On a NULL buf, we send a page of data to the server
4608  * guessing that the ACL request can be serviced by a page. If so, we cache
4609  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4610  * the cache. If not so, we throw away the page, and cache the required
4611  * length. The next getxattr call will then produce another round trip to
4612  * the server, this time with the input buf of the required size.
4613  */
4614 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4615 {
4616         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4617         struct nfs_getaclargs args = {
4618                 .fh = NFS_FH(inode),
4619                 .acl_pages = pages,
4620                 .acl_len = buflen,
4621         };
4622         struct nfs_getaclres res = {
4623                 .acl_len = buflen,
4624         };
4625         struct rpc_message msg = {
4626                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4627                 .rpc_argp = &args,
4628                 .rpc_resp = &res,
4629         };
4630         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4631         int ret = -ENOMEM, i;
4632
4633         /* As long as we're doing a round trip to the server anyway,
4634          * let's be prepared for a page of acl data. */
4635         if (npages == 0)
4636                 npages = 1;
4637         if (npages > ARRAY_SIZE(pages))
4638                 return -ERANGE;
4639
4640         for (i = 0; i < npages; i++) {
4641                 pages[i] = alloc_page(GFP_KERNEL);
4642                 if (!pages[i])
4643                         goto out_free;
4644         }
4645
4646         /* for decoding across pages */
4647         res.acl_scratch = alloc_page(GFP_KERNEL);
4648         if (!res.acl_scratch)
4649                 goto out_free;
4650
4651         args.acl_len = npages * PAGE_SIZE;
4652         args.acl_pgbase = 0;
4653
4654         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4655                 __func__, buf, buflen, npages, args.acl_len);
4656         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4657                              &msg, &args.seq_args, &res.seq_res, 0);
4658         if (ret)
4659                 goto out_free;
4660
4661         /* Handle the case where the passed-in buffer is too short */
4662         if (res.acl_flags & NFS4_ACL_TRUNC) {
4663                 /* Did the user only issue a request for the acl length? */
4664                 if (buf == NULL)
4665                         goto out_ok;
4666                 ret = -ERANGE;
4667                 goto out_free;
4668         }
4669         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4670         if (buf) {
4671                 if (res.acl_len > buflen) {
4672                         ret = -ERANGE;
4673                         goto out_free;
4674                 }
4675                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4676         }
4677 out_ok:
4678         ret = res.acl_len;
4679 out_free:
4680         for (i = 0; i < npages; i++)
4681                 if (pages[i])
4682                         __free_page(pages[i]);
4683         if (res.acl_scratch)
4684                 __free_page(res.acl_scratch);
4685         return ret;
4686 }
4687
4688 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4689 {
4690         struct nfs4_exception exception = { };
4691         ssize_t ret;
4692         do {
4693                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4694                 trace_nfs4_get_acl(inode, ret);
4695                 if (ret >= 0)
4696                         break;
4697                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4698         } while (exception.retry);
4699         return ret;
4700 }
4701
4702 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4703 {
4704         struct nfs_server *server = NFS_SERVER(inode);
4705         int ret;
4706
4707         if (!nfs4_server_supports_acls(server))
4708                 return -EOPNOTSUPP;
4709         ret = nfs_revalidate_inode(server, inode);
4710         if (ret < 0)
4711                 return ret;
4712         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4713                 nfs_zap_acl_cache(inode);
4714         ret = nfs4_read_cached_acl(inode, buf, buflen);
4715         if (ret != -ENOENT)
4716                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4717                  * but no cached acl data, just the acl length */
4718                 return ret;
4719         return nfs4_get_acl_uncached(inode, buf, buflen);
4720 }
4721
4722 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4723 {
4724         struct nfs_server *server = NFS_SERVER(inode);
4725         struct page *pages[NFS4ACL_MAXPAGES];
4726         struct nfs_setaclargs arg = {
4727                 .fh             = NFS_FH(inode),
4728                 .acl_pages      = pages,
4729                 .acl_len        = buflen,
4730         };
4731         struct nfs_setaclres res;
4732         struct rpc_message msg = {
4733                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4734                 .rpc_argp       = &arg,
4735                 .rpc_resp       = &res,
4736         };
4737         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4738         int ret, i;
4739
4740         if (!nfs4_server_supports_acls(server))
4741                 return -EOPNOTSUPP;
4742         if (npages > ARRAY_SIZE(pages))
4743                 return -ERANGE;
4744         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4745         if (i < 0)
4746                 return i;
4747         nfs4_inode_return_delegation(inode);
4748         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4749
4750         /*
4751          * Free each page after tx, so the only ref left is
4752          * held by the network stack
4753          */
4754         for (; i > 0; i--)
4755                 put_page(pages[i-1]);
4756
4757         /*
4758          * Acl update can result in inode attribute update.
4759          * so mark the attribute cache invalid.
4760          */
4761         spin_lock(&inode->i_lock);
4762         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4763         spin_unlock(&inode->i_lock);
4764         nfs_access_zap_cache(inode);
4765         nfs_zap_acl_cache(inode);
4766         return ret;
4767 }
4768
4769 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4770 {
4771         struct nfs4_exception exception = { };
4772         int err;
4773         do {
4774                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4775                 trace_nfs4_set_acl(inode, err);
4776                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4777                                 &exception);
4778         } while (exception.retry);
4779         return err;
4780 }
4781
4782 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4783 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4784                                         size_t buflen)
4785 {
4786         struct nfs_server *server = NFS_SERVER(inode);
4787         struct nfs_fattr fattr;
4788         struct nfs4_label label = {0, 0, buflen, buf};
4789
4790         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4791         struct nfs4_getattr_arg arg = {
4792                 .fh             = NFS_FH(inode),
4793                 .bitmask        = bitmask,
4794         };
4795         struct nfs4_getattr_res res = {
4796                 .fattr          = &fattr,
4797                 .label          = &label,
4798                 .server         = server,
4799         };
4800         struct rpc_message msg = {
4801                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4802                 .rpc_argp       = &arg,
4803                 .rpc_resp       = &res,
4804         };
4805         int ret;
4806
4807         nfs_fattr_init(&fattr);
4808
4809         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4810         if (ret)
4811                 return ret;
4812         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4813                 return -ENOENT;
4814         if (buflen < label.len)
4815                 return -ERANGE;
4816         return 0;
4817 }
4818
4819 static int nfs4_get_security_label(struct inode *inode, void *buf,
4820                                         size_t buflen)
4821 {
4822         struct nfs4_exception exception = { };
4823         int err;
4824
4825         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4826                 return -EOPNOTSUPP;
4827
4828         do {
4829                 err = _nfs4_get_security_label(inode, buf, buflen);
4830                 trace_nfs4_get_security_label(inode, err);
4831                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4832                                 &exception);
4833         } while (exception.retry);
4834         return err;
4835 }
4836
4837 static int _nfs4_do_set_security_label(struct inode *inode,
4838                 struct nfs4_label *ilabel,
4839                 struct nfs_fattr *fattr,
4840                 struct nfs4_label *olabel)
4841 {
4842
4843         struct iattr sattr = {0};
4844         struct nfs_server *server = NFS_SERVER(inode);
4845         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4846         struct nfs_setattrargs arg = {
4847                 .fh             = NFS_FH(inode),
4848                 .iap            = &sattr,
4849                 .server         = server,
4850                 .bitmask        = bitmask,
4851                 .label          = ilabel,
4852         };
4853         struct nfs_setattrres res = {
4854                 .fattr          = fattr,
4855                 .label          = olabel,
4856                 .server         = server,
4857         };
4858         struct rpc_message msg = {
4859                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4860                 .rpc_argp       = &arg,
4861                 .rpc_resp       = &res,
4862         };
4863         int status;
4864
4865         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4866
4867         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4868         if (status)
4869                 dprintk("%s failed: %d\n", __func__, status);
4870
4871         return status;
4872 }
4873
4874 static int nfs4_do_set_security_label(struct inode *inode,
4875                 struct nfs4_label *ilabel,
4876                 struct nfs_fattr *fattr,
4877                 struct nfs4_label *olabel)
4878 {
4879         struct nfs4_exception exception = { };
4880         int err;
4881
4882         do {
4883                 err = _nfs4_do_set_security_label(inode, ilabel,
4884                                 fattr, olabel);
4885                 trace_nfs4_set_security_label(inode, err);
4886                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4887                                 &exception);
4888         } while (exception.retry);
4889         return err;
4890 }
4891
4892 static int
4893 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4894 {
4895         struct nfs4_label ilabel, *olabel = NULL;
4896         struct nfs_fattr fattr;
4897         struct rpc_cred *cred;
4898         struct inode *inode = d_inode(dentry);
4899         int status;
4900
4901         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4902                 return -EOPNOTSUPP;
4903
4904         nfs_fattr_init(&fattr);
4905
4906         ilabel.pi = 0;
4907         ilabel.lfs = 0;
4908         ilabel.label = (char *)buf;
4909         ilabel.len = buflen;
4910
4911         cred = rpc_lookup_cred();
4912         if (IS_ERR(cred))
4913                 return PTR_ERR(cred);
4914
4915         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4916         if (IS_ERR(olabel)) {
4917                 status = -PTR_ERR(olabel);
4918                 goto out;
4919         }
4920
4921         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4922         if (status == 0)
4923                 nfs_setsecurity(inode, &fattr, olabel);
4924
4925         nfs4_label_free(olabel);
4926 out:
4927         put_rpccred(cred);
4928         return status;
4929 }
4930 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4931
4932
4933 static int
4934 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4935                         struct nfs4_state *state, long *timeout)
4936 {
4937         struct nfs_client *clp = server->nfs_client;
4938
4939         if (task->tk_status >= 0)
4940                 return 0;
4941         switch(task->tk_status) {
4942                 case -NFS4ERR_DELEG_REVOKED:
4943                 case -NFS4ERR_ADMIN_REVOKED:
4944                 case -NFS4ERR_BAD_STATEID:
4945                 case -NFS4ERR_OPENMODE:
4946                         if (state == NULL)
4947                                 break;
4948                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4949                                 goto recovery_failed;
4950                         goto wait_on_recovery;
4951                 case -NFS4ERR_EXPIRED:
4952                         if (state != NULL) {
4953                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4954                                         goto recovery_failed;
4955                         }
4956                 case -NFS4ERR_STALE_STATEID:
4957                 case -NFS4ERR_STALE_CLIENTID:
4958                         nfs4_schedule_lease_recovery(clp);
4959                         goto wait_on_recovery;
4960                 case -NFS4ERR_MOVED:
4961                         if (nfs4_schedule_migration_recovery(server) < 0)
4962                                 goto recovery_failed;
4963                         goto wait_on_recovery;
4964                 case -NFS4ERR_LEASE_MOVED:
4965                         nfs4_schedule_lease_moved_recovery(clp);
4966                         goto wait_on_recovery;
4967 #if defined(CONFIG_NFS_V4_1)
4968                 case -NFS4ERR_BADSESSION:
4969                 case -NFS4ERR_BADSLOT:
4970                 case -NFS4ERR_BAD_HIGH_SLOT:
4971                 case -NFS4ERR_DEADSESSION:
4972                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4973                 case -NFS4ERR_SEQ_FALSE_RETRY:
4974                 case -NFS4ERR_SEQ_MISORDERED:
4975                         dprintk("%s ERROR %d, Reset session\n", __func__,
4976                                 task->tk_status);
4977                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4978                         goto wait_on_recovery;
4979 #endif /* CONFIG_NFS_V4_1 */
4980                 case -NFS4ERR_DELAY:
4981                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4982                         rpc_delay(task, nfs4_update_delay(timeout));
4983                         goto restart_call;
4984                 case -NFS4ERR_GRACE:
4985                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4986                 case -NFS4ERR_RETRY_UNCACHED_REP:
4987                 case -NFS4ERR_OLD_STATEID:
4988                         goto restart_call;
4989         }
4990         task->tk_status = nfs4_map_errors(task->tk_status);
4991         return 0;
4992 recovery_failed:
4993         task->tk_status = -EIO;
4994         return 0;
4995 wait_on_recovery:
4996         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4997         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4998                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4999         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
5000                 goto recovery_failed;
5001 restart_call:
5002         task->tk_status = 0;
5003         return -EAGAIN;
5004 }
5005
5006 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5007                                     nfs4_verifier *bootverf)
5008 {
5009         __be32 verf[2];
5010
5011         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5012                 /* An impossible timestamp guarantees this value
5013                  * will never match a generated boot time. */
5014                 verf[0] = 0;
5015                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5016         } else {
5017                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5018                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5019                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5020         }
5021         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5022 }
5023
5024 static int
5025 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5026 {
5027         int result;
5028         size_t len;
5029         char *str;
5030
5031         if (clp->cl_owner_id != NULL)
5032                 return 0;
5033
5034         rcu_read_lock();
5035         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5036                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5037                 1 +
5038                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5039                 1;
5040         rcu_read_unlock();
5041
5042         if (len > NFS4_OPAQUE_LIMIT + 1)
5043                 return -EINVAL;
5044
5045         /*
5046          * Since this string is allocated at mount time, and held until the
5047          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5048          * about a memory-reclaim deadlock.
5049          */
5050         str = kmalloc(len, GFP_KERNEL);
5051         if (!str)
5052                 return -ENOMEM;
5053
5054         rcu_read_lock();
5055         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5056                         clp->cl_ipaddr,
5057                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5058                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5059         rcu_read_unlock();
5060
5061         clp->cl_owner_id = str;
5062         return 0;
5063 }
5064
5065 static int
5066 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5067 {
5068         int result;
5069         size_t len;
5070         char *str;
5071
5072         len = 10 + 10 + 1 + 10 + 1 +
5073                 strlen(nfs4_client_id_uniquifier) + 1 +
5074                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5075
5076         if (len > NFS4_OPAQUE_LIMIT + 1)
5077                 return -EINVAL;
5078
5079         /*
5080          * Since this string is allocated at mount time, and held until the
5081          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5082          * about a memory-reclaim deadlock.
5083          */
5084         str = kmalloc(len, GFP_KERNEL);
5085         if (!str)
5086                 return -ENOMEM;
5087
5088         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5089                         clp->rpc_ops->version, clp->cl_minorversion,
5090                         nfs4_client_id_uniquifier,
5091                         clp->cl_rpcclient->cl_nodename);
5092         clp->cl_owner_id = str;
5093         return 0;
5094 }
5095
5096 static int
5097 nfs4_init_uniform_client_string(struct nfs_client *clp)
5098 {
5099         int result;
5100         size_t len;
5101         char *str;
5102
5103         if (clp->cl_owner_id != NULL)
5104                 return 0;
5105
5106         if (nfs4_client_id_uniquifier[0] != '\0')
5107                 return nfs4_init_uniquifier_client_string(clp);
5108
5109         len = 10 + 10 + 1 + 10 + 1 +
5110                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5111
5112         if (len > NFS4_OPAQUE_LIMIT + 1)
5113                 return -EINVAL;
5114
5115         /*
5116          * Since this string is allocated at mount time, and held until the
5117          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5118          * about a memory-reclaim deadlock.
5119          */
5120         str = kmalloc(len, GFP_KERNEL);
5121         if (!str)
5122                 return -ENOMEM;
5123
5124         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5125                         clp->rpc_ops->version, clp->cl_minorversion,
5126                         clp->cl_rpcclient->cl_nodename);
5127         clp->cl_owner_id = str;
5128         return 0;
5129 }
5130
5131 /*
5132  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5133  * services.  Advertise one based on the address family of the
5134  * clientaddr.
5135  */
5136 static unsigned int
5137 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5138 {
5139         if (strchr(clp->cl_ipaddr, ':') != NULL)
5140                 return scnprintf(buf, len, "tcp6");
5141         else
5142                 return scnprintf(buf, len, "tcp");
5143 }
5144
5145 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5146 {
5147         struct nfs4_setclientid *sc = calldata;
5148
5149         if (task->tk_status == 0)
5150                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5151 }
5152
5153 static const struct rpc_call_ops nfs4_setclientid_ops = {
5154         .rpc_call_done = nfs4_setclientid_done,
5155 };
5156
5157 /**
5158  * nfs4_proc_setclientid - Negotiate client ID
5159  * @clp: state data structure
5160  * @program: RPC program for NFSv4 callback service
5161  * @port: IP port number for NFS4 callback service
5162  * @cred: RPC credential to use for this call
5163  * @res: where to place the result
5164  *
5165  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5166  */
5167 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5168                 unsigned short port, struct rpc_cred *cred,
5169                 struct nfs4_setclientid_res *res)
5170 {
5171         nfs4_verifier sc_verifier;
5172         struct nfs4_setclientid setclientid = {
5173                 .sc_verifier = &sc_verifier,
5174                 .sc_prog = program,
5175                 .sc_clnt = clp,
5176         };
5177         struct rpc_message msg = {
5178                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5179                 .rpc_argp = &setclientid,
5180                 .rpc_resp = res,
5181                 .rpc_cred = cred,
5182         };
5183         struct rpc_task *task;
5184         struct rpc_task_setup task_setup_data = {
5185                 .rpc_client = clp->cl_rpcclient,
5186                 .rpc_message = &msg,
5187                 .callback_ops = &nfs4_setclientid_ops,
5188                 .callback_data = &setclientid,
5189                 .flags = RPC_TASK_TIMEOUT,
5190         };
5191         int status;
5192
5193         /* nfs_client_id4 */
5194         nfs4_init_boot_verifier(clp, &sc_verifier);
5195
5196         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5197                 status = nfs4_init_uniform_client_string(clp);
5198         else
5199                 status = nfs4_init_nonuniform_client_string(clp);
5200
5201         if (status)
5202                 goto out;
5203
5204         /* cb_client4 */
5205         setclientid.sc_netid_len =
5206                                 nfs4_init_callback_netid(clp,
5207                                                 setclientid.sc_netid,
5208                                                 sizeof(setclientid.sc_netid));
5209         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5210                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5211                                 clp->cl_ipaddr, port >> 8, port & 255);
5212
5213         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5214                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5215                 clp->cl_owner_id);
5216         task = rpc_run_task(&task_setup_data);
5217         if (IS_ERR(task)) {
5218                 status = PTR_ERR(task);
5219                 goto out;
5220         }
5221         status = task->tk_status;
5222         if (setclientid.sc_cred) {
5223                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5224                 put_rpccred(setclientid.sc_cred);
5225         }
5226         rpc_put_task(task);
5227 out:
5228         trace_nfs4_setclientid(clp, status);
5229         dprintk("NFS reply setclientid: %d\n", status);
5230         return status;
5231 }
5232
5233 /**
5234  * nfs4_proc_setclientid_confirm - Confirm client ID
5235  * @clp: state data structure
5236  * @res: result of a previous SETCLIENTID
5237  * @cred: RPC credential to use for this call
5238  *
5239  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5240  */
5241 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5242                 struct nfs4_setclientid_res *arg,
5243                 struct rpc_cred *cred)
5244 {
5245         struct rpc_message msg = {
5246                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5247                 .rpc_argp = arg,
5248                 .rpc_cred = cred,
5249         };
5250         int status;
5251
5252         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5253                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5254                 clp->cl_clientid);
5255         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5256         trace_nfs4_setclientid_confirm(clp, status);
5257         dprintk("NFS reply setclientid_confirm: %d\n", status);
5258         return status;
5259 }
5260
5261 struct nfs4_delegreturndata {
5262         struct nfs4_delegreturnargs args;
5263         struct nfs4_delegreturnres res;
5264         struct nfs_fh fh;
5265         nfs4_stateid stateid;
5266         unsigned long timestamp;
5267         struct nfs_fattr fattr;
5268         int rpc_status;
5269         struct inode *inode;
5270         bool roc;
5271         u32 roc_barrier;
5272 };
5273
5274 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5275 {
5276         struct nfs4_delegreturndata *data = calldata;
5277
5278         if (!nfs4_sequence_done(task, &data->res.seq_res))
5279                 return;
5280
5281         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5282         switch (task->tk_status) {
5283         case 0:
5284                 renew_lease(data->res.server, data->timestamp);
5285         case -NFS4ERR_ADMIN_REVOKED:
5286         case -NFS4ERR_DELEG_REVOKED:
5287         case -NFS4ERR_BAD_STATEID:
5288         case -NFS4ERR_OLD_STATEID:
5289         case -NFS4ERR_STALE_STATEID:
5290         case -NFS4ERR_EXPIRED:
5291                 task->tk_status = 0;
5292                 if (data->roc)
5293                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5294                 break;
5295         default:
5296                 if (nfs4_async_handle_error(task, data->res.server,
5297                                             NULL, NULL) == -EAGAIN) {
5298                         rpc_restart_call_prepare(task);
5299                         return;
5300                 }
5301         }
5302         data->rpc_status = task->tk_status;
5303 }
5304
5305 static void nfs4_delegreturn_release(void *calldata)
5306 {
5307         struct nfs4_delegreturndata *data = calldata;
5308         struct inode *inode = data->inode;
5309
5310         if (inode) {
5311                 if (data->roc)
5312                         pnfs_roc_release(inode);
5313                 nfs_iput_and_deactive(inode);
5314         }
5315         kfree(calldata);
5316 }
5317
5318 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5319 {
5320         struct nfs4_delegreturndata *d_data;
5321
5322         d_data = (struct nfs4_delegreturndata *)data;
5323
5324         if (d_data->roc)
5325                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5326
5327         nfs4_setup_sequence(d_data->res.server,
5328                         &d_data->args.seq_args,
5329                         &d_data->res.seq_res,
5330                         task);
5331 }
5332
5333 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5334         .rpc_call_prepare = nfs4_delegreturn_prepare,
5335         .rpc_call_done = nfs4_delegreturn_done,
5336         .rpc_release = nfs4_delegreturn_release,
5337 };
5338
5339 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5340 {
5341         struct nfs4_delegreturndata *data;
5342         struct nfs_server *server = NFS_SERVER(inode);
5343         struct rpc_task *task;
5344         struct rpc_message msg = {
5345                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5346                 .rpc_cred = cred,
5347         };
5348         struct rpc_task_setup task_setup_data = {
5349                 .rpc_client = server->client,
5350                 .rpc_message = &msg,
5351                 .callback_ops = &nfs4_delegreturn_ops,
5352                 .flags = RPC_TASK_ASYNC,
5353         };
5354         int status = 0;
5355
5356         data = kzalloc(sizeof(*data), GFP_NOFS);
5357         if (data == NULL)
5358                 return -ENOMEM;
5359         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5360         data->args.fhandle = &data->fh;
5361         data->args.stateid = &data->stateid;
5362         data->args.bitmask = server->cache_consistency_bitmask;
5363         nfs_copy_fh(&data->fh, NFS_FH(inode));
5364         nfs4_stateid_copy(&data->stateid, stateid);
5365         data->res.fattr = &data->fattr;
5366         data->res.server = server;
5367         nfs_fattr_init(data->res.fattr);
5368         data->timestamp = jiffies;
5369         data->rpc_status = 0;
5370         data->inode = nfs_igrab_and_active(inode);
5371         if (data->inode)
5372                 data->roc = nfs4_roc(inode);
5373
5374         task_setup_data.callback_data = data;
5375         msg.rpc_argp = &data->args;
5376         msg.rpc_resp = &data->res;
5377         task = rpc_run_task(&task_setup_data);
5378         if (IS_ERR(task))
5379                 return PTR_ERR(task);
5380         if (!issync)
5381                 goto out;
5382         status = nfs4_wait_for_completion_rpc_task(task);
5383         if (status != 0)
5384                 goto out;
5385         status = data->rpc_status;
5386         if (status == 0)
5387                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5388         else
5389                 nfs_refresh_inode(inode, &data->fattr);
5390 out:
5391         rpc_put_task(task);
5392         return status;
5393 }
5394
5395 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5396 {
5397         struct nfs_server *server = NFS_SERVER(inode);
5398         struct nfs4_exception exception = { };
5399         int err;
5400         do {
5401                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5402                 trace_nfs4_delegreturn(inode, err);
5403                 switch (err) {
5404                         case -NFS4ERR_STALE_STATEID:
5405                         case -NFS4ERR_EXPIRED:
5406                         case 0:
5407                                 return 0;
5408                 }
5409                 err = nfs4_handle_exception(server, err, &exception);
5410         } while (exception.retry);
5411         return err;
5412 }
5413
5414 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5415 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5416
5417 /*
5418  * sleep, with exponential backoff, and retry the LOCK operation.
5419  */
5420 static unsigned long
5421 nfs4_set_lock_task_retry(unsigned long timeout)
5422 {
5423         freezable_schedule_timeout_killable_unsafe(timeout);
5424         timeout <<= 1;
5425         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5426                 return NFS4_LOCK_MAXTIMEOUT;
5427         return timeout;
5428 }
5429
5430 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5431 {
5432         struct inode *inode = state->inode;
5433         struct nfs_server *server = NFS_SERVER(inode);
5434         struct nfs_client *clp = server->nfs_client;
5435         struct nfs_lockt_args arg = {
5436                 .fh = NFS_FH(inode),
5437                 .fl = request,
5438         };
5439         struct nfs_lockt_res res = {
5440                 .denied = request,
5441         };
5442         struct rpc_message msg = {
5443                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5444                 .rpc_argp       = &arg,
5445                 .rpc_resp       = &res,
5446                 .rpc_cred       = state->owner->so_cred,
5447         };
5448         struct nfs4_lock_state *lsp;
5449         int status;
5450
5451         arg.lock_owner.clientid = clp->cl_clientid;
5452         status = nfs4_set_lock_state(state, request);
5453         if (status != 0)
5454                 goto out;
5455         lsp = request->fl_u.nfs4_fl.owner;
5456         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5457         arg.lock_owner.s_dev = server->s_dev;
5458         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5459         switch (status) {
5460                 case 0:
5461                         request->fl_type = F_UNLCK;
5462                         break;
5463                 case -NFS4ERR_DENIED:
5464                         status = 0;
5465         }
5466         request->fl_ops->fl_release_private(request);
5467         request->fl_ops = NULL;
5468 out:
5469         return status;
5470 }
5471
5472 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5473 {
5474         struct nfs4_exception exception = { };
5475         int err;
5476
5477         do {
5478                 err = _nfs4_proc_getlk(state, cmd, request);
5479                 trace_nfs4_get_lock(request, state, cmd, err);
5480                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5481                                 &exception);
5482         } while (exception.retry);
5483         return err;
5484 }
5485
5486 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5487 {
5488         int res = 0;
5489         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5490                 case FL_POSIX:
5491                         res = posix_lock_inode_wait(inode, fl);
5492                         break;
5493                 case FL_FLOCK:
5494                         res = flock_lock_inode_wait(inode, fl);
5495                         break;
5496                 default:
5497                         BUG();
5498         }
5499         return res;
5500 }
5501
5502 struct nfs4_unlockdata {
5503         struct nfs_locku_args arg;
5504         struct nfs_locku_res res;
5505         struct nfs4_lock_state *lsp;
5506         struct nfs_open_context *ctx;
5507         struct file_lock fl;
5508         const struct nfs_server *server;
5509         unsigned long timestamp;
5510 };
5511
5512 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5513                 struct nfs_open_context *ctx,
5514                 struct nfs4_lock_state *lsp,
5515                 struct nfs_seqid *seqid)
5516 {
5517         struct nfs4_unlockdata *p;
5518         struct inode *inode = lsp->ls_state->inode;
5519
5520         p = kzalloc(sizeof(*p), GFP_NOFS);
5521         if (p == NULL)
5522                 return NULL;
5523         p->arg.fh = NFS_FH(inode);
5524         p->arg.fl = &p->fl;
5525         p->arg.seqid = seqid;
5526         p->res.seqid = seqid;
5527         p->lsp = lsp;
5528         atomic_inc(&lsp->ls_count);
5529         /* Ensure we don't close file until we're done freeing locks! */
5530         p->ctx = get_nfs_open_context(ctx);
5531         memcpy(&p->fl, fl, sizeof(p->fl));
5532         p->server = NFS_SERVER(inode);
5533         return p;
5534 }
5535
5536 static void nfs4_locku_release_calldata(void *data)
5537 {
5538         struct nfs4_unlockdata *calldata = data;
5539         nfs_free_seqid(calldata->arg.seqid);
5540         nfs4_put_lock_state(calldata->lsp);
5541         put_nfs_open_context(calldata->ctx);
5542         kfree(calldata);
5543 }
5544
5545 static void nfs4_locku_done(struct rpc_task *task, void *data)
5546 {
5547         struct nfs4_unlockdata *calldata = data;
5548
5549         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5550                 return;
5551         switch (task->tk_status) {
5552                 case 0:
5553                         renew_lease(calldata->server, calldata->timestamp);
5554                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5555                         if (nfs4_update_lock_stateid(calldata->lsp,
5556                                         &calldata->res.stateid))
5557                                 break;
5558                 case -NFS4ERR_BAD_STATEID:
5559                 case -NFS4ERR_OLD_STATEID:
5560                 case -NFS4ERR_STALE_STATEID:
5561                 case -NFS4ERR_EXPIRED:
5562                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5563                                                 &calldata->lsp->ls_stateid))
5564                                 rpc_restart_call_prepare(task);
5565                         break;
5566                 default:
5567                         if (nfs4_async_handle_error(task, calldata->server,
5568                                                     NULL, NULL) == -EAGAIN)
5569                                 rpc_restart_call_prepare(task);
5570         }
5571         nfs_release_seqid(calldata->arg.seqid);
5572 }
5573
5574 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5575 {
5576         struct nfs4_unlockdata *calldata = data;
5577
5578         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5579                 goto out_wait;
5580         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5581         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5582                 /* Note: exit _without_ running nfs4_locku_done */
5583                 goto out_no_action;
5584         }
5585         calldata->timestamp = jiffies;
5586         if (nfs4_setup_sequence(calldata->server,
5587                                 &calldata->arg.seq_args,
5588                                 &calldata->res.seq_res,
5589                                 task) != 0)
5590                 nfs_release_seqid(calldata->arg.seqid);
5591         return;
5592 out_no_action:
5593         task->tk_action = NULL;
5594 out_wait:
5595         nfs4_sequence_done(task, &calldata->res.seq_res);
5596 }
5597
5598 static const struct rpc_call_ops nfs4_locku_ops = {
5599         .rpc_call_prepare = nfs4_locku_prepare,
5600         .rpc_call_done = nfs4_locku_done,
5601         .rpc_release = nfs4_locku_release_calldata,
5602 };
5603
5604 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5605                 struct nfs_open_context *ctx,
5606                 struct nfs4_lock_state *lsp,
5607                 struct nfs_seqid *seqid)
5608 {
5609         struct nfs4_unlockdata *data;
5610         struct rpc_message msg = {
5611                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5612                 .rpc_cred = ctx->cred,
5613         };
5614         struct rpc_task_setup task_setup_data = {
5615                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5616                 .rpc_message = &msg,
5617                 .callback_ops = &nfs4_locku_ops,
5618                 .workqueue = nfsiod_workqueue,
5619                 .flags = RPC_TASK_ASYNC,
5620         };
5621
5622         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5623                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5624
5625         /* Ensure this is an unlock - when canceling a lock, the
5626          * canceled lock is passed in, and it won't be an unlock.
5627          */
5628         fl->fl_type = F_UNLCK;
5629
5630         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5631         if (data == NULL) {
5632                 nfs_free_seqid(seqid);
5633                 return ERR_PTR(-ENOMEM);
5634         }
5635
5636         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5637         msg.rpc_argp = &data->arg;
5638         msg.rpc_resp = &data->res;
5639         task_setup_data.callback_data = data;
5640         return rpc_run_task(&task_setup_data);
5641 }
5642
5643 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5644 {
5645         struct inode *inode = state->inode;
5646         struct nfs4_state_owner *sp = state->owner;
5647         struct nfs_inode *nfsi = NFS_I(inode);
5648         struct nfs_seqid *seqid;
5649         struct nfs4_lock_state *lsp;
5650         struct rpc_task *task;
5651         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5652         int status = 0;
5653         unsigned char fl_flags = request->fl_flags;
5654
5655         status = nfs4_set_lock_state(state, request);
5656         /* Unlock _before_ we do the RPC call */
5657         request->fl_flags |= FL_EXISTS;
5658         /* Exclude nfs_delegation_claim_locks() */
5659         mutex_lock(&sp->so_delegreturn_mutex);
5660         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5661         down_read(&nfsi->rwsem);
5662         if (do_vfs_lock(inode, request) == -ENOENT) {
5663                 up_read(&nfsi->rwsem);
5664                 mutex_unlock(&sp->so_delegreturn_mutex);
5665                 goto out;
5666         }
5667         up_read(&nfsi->rwsem);
5668         mutex_unlock(&sp->so_delegreturn_mutex);
5669         if (status != 0)
5670                 goto out;
5671         /* Is this a delegated lock? */
5672         lsp = request->fl_u.nfs4_fl.owner;
5673         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5674                 goto out;
5675         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5676         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5677         status = -ENOMEM;
5678         if (IS_ERR(seqid))
5679                 goto out;
5680         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5681         status = PTR_ERR(task);
5682         if (IS_ERR(task))
5683                 goto out;
5684         status = nfs4_wait_for_completion_rpc_task(task);
5685         rpc_put_task(task);
5686 out:
5687         request->fl_flags = fl_flags;
5688         trace_nfs4_unlock(request, state, F_SETLK, status);
5689         return status;
5690 }
5691
5692 struct nfs4_lockdata {
5693         struct nfs_lock_args arg;
5694         struct nfs_lock_res res;
5695         struct nfs4_lock_state *lsp;
5696         struct nfs_open_context *ctx;
5697         struct file_lock fl;
5698         unsigned long timestamp;
5699         int rpc_status;
5700         int cancelled;
5701         struct nfs_server *server;
5702 };
5703
5704 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5705                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5706                 gfp_t gfp_mask)
5707 {
5708         struct nfs4_lockdata *p;
5709         struct inode *inode = lsp->ls_state->inode;
5710         struct nfs_server *server = NFS_SERVER(inode);
5711         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5712
5713         p = kzalloc(sizeof(*p), gfp_mask);
5714         if (p == NULL)
5715                 return NULL;
5716
5717         p->arg.fh = NFS_FH(inode);
5718         p->arg.fl = &p->fl;
5719         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5720         if (IS_ERR(p->arg.open_seqid))
5721                 goto out_free;
5722         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5723         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5724         if (IS_ERR(p->arg.lock_seqid))
5725                 goto out_free_seqid;
5726         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5727         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5728         p->arg.lock_owner.s_dev = server->s_dev;
5729         p->res.lock_seqid = p->arg.lock_seqid;
5730         p->lsp = lsp;
5731         p->server = server;
5732         atomic_inc(&lsp->ls_count);
5733         p->ctx = get_nfs_open_context(ctx);
5734         get_file(fl->fl_file);
5735         memcpy(&p->fl, fl, sizeof(p->fl));
5736         return p;
5737 out_free_seqid:
5738         nfs_free_seqid(p->arg.open_seqid);
5739 out_free:
5740         kfree(p);
5741         return NULL;
5742 }
5743
5744 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5745 {
5746         struct nfs4_lockdata *data = calldata;
5747         struct nfs4_state *state = data->lsp->ls_state;
5748
5749         dprintk("%s: begin!\n", __func__);
5750         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5751                 goto out_wait;
5752         /* Do we need to do an open_to_lock_owner? */
5753         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5754                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5755                         goto out_release_lock_seqid;
5756                 }
5757                 nfs4_stateid_copy(&data->arg.open_stateid,
5758                                 &state->open_stateid);
5759                 data->arg.new_lock_owner = 1;
5760                 data->res.open_seqid = data->arg.open_seqid;
5761         } else {
5762                 data->arg.new_lock_owner = 0;
5763                 nfs4_stateid_copy(&data->arg.lock_stateid,
5764                                 &data->lsp->ls_stateid);
5765         }
5766         if (!nfs4_valid_open_stateid(state)) {
5767                 data->rpc_status = -EBADF;
5768                 task->tk_action = NULL;
5769                 goto out_release_open_seqid;
5770         }
5771         data->timestamp = jiffies;
5772         if (nfs4_setup_sequence(data->server,
5773                                 &data->arg.seq_args,
5774                                 &data->res.seq_res,
5775                                 task) == 0)
5776                 return;
5777 out_release_open_seqid:
5778         nfs_release_seqid(data->arg.open_seqid);
5779 out_release_lock_seqid:
5780         nfs_release_seqid(data->arg.lock_seqid);
5781 out_wait:
5782         nfs4_sequence_done(task, &data->res.seq_res);
5783         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5784 }
5785
5786 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5787 {
5788         struct nfs4_lockdata *data = calldata;
5789         struct nfs4_lock_state *lsp = data->lsp;
5790
5791         dprintk("%s: begin!\n", __func__);
5792
5793         if (!nfs4_sequence_done(task, &data->res.seq_res))
5794                 return;
5795
5796         data->rpc_status = task->tk_status;
5797         switch (task->tk_status) {
5798         case 0:
5799                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5800                                 data->timestamp);
5801                 if (data->arg.new_lock) {
5802                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5803                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5804                                 rpc_restart_call_prepare(task);
5805                                 break;
5806                         }
5807                 }
5808                 if (data->arg.new_lock_owner != 0) {
5809                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5810                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5811                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5812                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5813                         rpc_restart_call_prepare(task);
5814                 break;
5815         case -NFS4ERR_BAD_STATEID:
5816         case -NFS4ERR_OLD_STATEID:
5817         case -NFS4ERR_STALE_STATEID:
5818         case -NFS4ERR_EXPIRED:
5819                 if (data->arg.new_lock_owner != 0) {
5820                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5821                                                 &lsp->ls_state->open_stateid))
5822                                 rpc_restart_call_prepare(task);
5823                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5824                                                 &lsp->ls_stateid))
5825                                 rpc_restart_call_prepare(task);
5826         }
5827         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5828 }
5829
5830 static void nfs4_lock_release(void *calldata)
5831 {
5832         struct nfs4_lockdata *data = calldata;
5833
5834         dprintk("%s: begin!\n", __func__);
5835         nfs_free_seqid(data->arg.open_seqid);
5836         if (data->cancelled != 0) {
5837                 struct rpc_task *task;
5838                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5839                                 data->arg.lock_seqid);
5840                 if (!IS_ERR(task))
5841                         rpc_put_task_async(task);
5842                 dprintk("%s: cancelling lock!\n", __func__);
5843         } else
5844                 nfs_free_seqid(data->arg.lock_seqid);
5845         nfs4_put_lock_state(data->lsp);
5846         put_nfs_open_context(data->ctx);
5847         fput(data->fl.fl_file);
5848         kfree(data);
5849         dprintk("%s: done!\n", __func__);
5850 }
5851
5852 static const struct rpc_call_ops nfs4_lock_ops = {
5853         .rpc_call_prepare = nfs4_lock_prepare,
5854         .rpc_call_done = nfs4_lock_done,
5855         .rpc_release = nfs4_lock_release,
5856 };
5857
5858 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5859 {
5860         switch (error) {
5861         case -NFS4ERR_ADMIN_REVOKED:
5862         case -NFS4ERR_BAD_STATEID:
5863                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5864                 if (new_lock_owner != 0 ||
5865                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5866                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5867                 break;
5868         case -NFS4ERR_STALE_STATEID:
5869                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5870         case -NFS4ERR_EXPIRED:
5871                 nfs4_schedule_lease_recovery(server->nfs_client);
5872         };
5873 }
5874
5875 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5876 {
5877         struct nfs4_lockdata *data;
5878         struct rpc_task *task;
5879         struct rpc_message msg = {
5880                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5881                 .rpc_cred = state->owner->so_cred,
5882         };
5883         struct rpc_task_setup task_setup_data = {
5884                 .rpc_client = NFS_CLIENT(state->inode),
5885                 .rpc_message = &msg,
5886                 .callback_ops = &nfs4_lock_ops,
5887                 .workqueue = nfsiod_workqueue,
5888                 .flags = RPC_TASK_ASYNC,
5889         };
5890         int ret;
5891
5892         dprintk("%s: begin!\n", __func__);
5893         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5894                         fl->fl_u.nfs4_fl.owner,
5895                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5896         if (data == NULL)
5897                 return -ENOMEM;
5898         if (IS_SETLKW(cmd))
5899                 data->arg.block = 1;
5900         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5901         msg.rpc_argp = &data->arg;
5902         msg.rpc_resp = &data->res;
5903         task_setup_data.callback_data = data;
5904         if (recovery_type > NFS_LOCK_NEW) {
5905                 if (recovery_type == NFS_LOCK_RECLAIM)
5906                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5907                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5908         } else
5909                 data->arg.new_lock = 1;
5910         task = rpc_run_task(&task_setup_data);
5911         if (IS_ERR(task))
5912                 return PTR_ERR(task);
5913         ret = nfs4_wait_for_completion_rpc_task(task);
5914         if (ret == 0) {
5915                 ret = data->rpc_status;
5916                 if (ret)
5917                         nfs4_handle_setlk_error(data->server, data->lsp,
5918                                         data->arg.new_lock_owner, ret);
5919         } else
5920                 data->cancelled = 1;
5921         rpc_put_task(task);
5922         dprintk("%s: done, ret = %d!\n", __func__, ret);
5923         return ret;
5924 }
5925
5926 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5927 {
5928         struct nfs_server *server = NFS_SERVER(state->inode);
5929         struct nfs4_exception exception = {
5930                 .inode = state->inode,
5931         };
5932         int err;
5933
5934         do {
5935                 /* Cache the lock if possible... */
5936                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5937                         return 0;
5938                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5939                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5940                 if (err != -NFS4ERR_DELAY)
5941                         break;
5942                 nfs4_handle_exception(server, err, &exception);
5943         } while (exception.retry);
5944         return err;
5945 }
5946
5947 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5948 {
5949         struct nfs_server *server = NFS_SERVER(state->inode);
5950         struct nfs4_exception exception = {
5951                 .inode = state->inode,
5952         };
5953         int err;
5954
5955         err = nfs4_set_lock_state(state, request);
5956         if (err != 0)
5957                 return err;
5958         if (!recover_lost_locks) {
5959                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5960                 return 0;
5961         }
5962         do {
5963                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5964                         return 0;
5965                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5966                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5967                 switch (err) {
5968                 default:
5969                         goto out;
5970                 case -NFS4ERR_GRACE:
5971                 case -NFS4ERR_DELAY:
5972                         nfs4_handle_exception(server, err, &exception);
5973                         err = 0;
5974                 }
5975         } while (exception.retry);
5976 out:
5977         return err;
5978 }
5979
5980 #if defined(CONFIG_NFS_V4_1)
5981 /**
5982  * nfs41_check_expired_locks - possibly free a lock stateid
5983  *
5984  * @state: NFSv4 state for an inode
5985  *
5986  * Returns NFS_OK if recovery for this stateid is now finished.
5987  * Otherwise a negative NFS4ERR value is returned.
5988  */
5989 static int nfs41_check_expired_locks(struct nfs4_state *state)
5990 {
5991         int status, ret = -NFS4ERR_BAD_STATEID;
5992         struct nfs4_lock_state *lsp;
5993         struct nfs_server *server = NFS_SERVER(state->inode);
5994
5995         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5996                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5997                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5998
5999                         status = nfs41_test_stateid(server,
6000                                         &lsp->ls_stateid,
6001                                         cred);
6002                         trace_nfs4_test_lock_stateid(state, lsp, status);
6003                         if (status != NFS_OK) {
6004                                 /* Free the stateid unless the server
6005                                  * informs us the stateid is unrecognized. */
6006                                 if (status != -NFS4ERR_BAD_STATEID)
6007                                         nfs41_free_stateid(server,
6008                                                         &lsp->ls_stateid,
6009                                                         cred);
6010                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6011                                 ret = status;
6012                         }
6013                 }
6014         };
6015
6016         return ret;
6017 }
6018
6019 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6020 {
6021         int status = NFS_OK;
6022
6023         if (test_bit(LK_STATE_IN_USE, &state->flags))
6024                 status = nfs41_check_expired_locks(state);
6025         if (status != NFS_OK)
6026                 status = nfs4_lock_expired(state, request);
6027         return status;
6028 }
6029 #endif
6030
6031 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6032 {
6033         struct nfs_inode *nfsi = NFS_I(state->inode);
6034         unsigned char fl_flags = request->fl_flags;
6035         int status = -ENOLCK;
6036
6037         if ((fl_flags & FL_POSIX) &&
6038                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6039                 goto out;
6040         /* Is this a delegated open? */
6041         status = nfs4_set_lock_state(state, request);
6042         if (status != 0)
6043                 goto out;
6044         request->fl_flags |= FL_ACCESS;
6045         status = do_vfs_lock(state->inode, request);
6046         if (status < 0)
6047                 goto out;
6048         down_read(&nfsi->rwsem);
6049         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6050                 /* Yes: cache locks! */
6051                 /* ...but avoid races with delegation recall... */
6052                 request->fl_flags = fl_flags & ~FL_SLEEP;
6053                 status = do_vfs_lock(state->inode, request);
6054                 up_read(&nfsi->rwsem);
6055                 goto out;
6056         }
6057         up_read(&nfsi->rwsem);
6058         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6059 out:
6060         request->fl_flags = fl_flags;
6061         return status;
6062 }
6063
6064 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6065 {
6066         struct nfs4_exception exception = {
6067                 .state = state,
6068                 .inode = state->inode,
6069         };
6070         int err;
6071
6072         do {
6073                 err = _nfs4_proc_setlk(state, cmd, request);
6074                 trace_nfs4_set_lock(request, state, cmd, err);
6075                 if (err == -NFS4ERR_DENIED)
6076                         err = -EAGAIN;
6077                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6078                                 err, &exception);
6079         } while (exception.retry);
6080         return err;
6081 }
6082
6083 static int
6084 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6085 {
6086         struct nfs_open_context *ctx;
6087         struct nfs4_state *state;
6088         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6089         int status;
6090
6091         /* verify open state */
6092         ctx = nfs_file_open_context(filp);
6093         state = ctx->state;
6094
6095         if (request->fl_start < 0 || request->fl_end < 0)
6096                 return -EINVAL;
6097
6098         if (IS_GETLK(cmd)) {
6099                 if (state != NULL)
6100                         return nfs4_proc_getlk(state, F_GETLK, request);
6101                 return 0;
6102         }
6103
6104         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6105                 return -EINVAL;
6106
6107         if (request->fl_type == F_UNLCK) {
6108                 if (state != NULL)
6109                         return nfs4_proc_unlck(state, cmd, request);
6110                 return 0;
6111         }
6112
6113         if (state == NULL)
6114                 return -ENOLCK;
6115         /*
6116          * Don't rely on the VFS having checked the file open mode,
6117          * since it won't do this for flock() locks.
6118          */
6119         switch (request->fl_type) {
6120         case F_RDLCK:
6121                 if (!(filp->f_mode & FMODE_READ))
6122                         return -EBADF;
6123                 break;
6124         case F_WRLCK:
6125                 if (!(filp->f_mode & FMODE_WRITE))
6126                         return -EBADF;
6127         }
6128
6129         do {
6130                 status = nfs4_proc_setlk(state, cmd, request);
6131                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6132                         break;
6133                 timeout = nfs4_set_lock_task_retry(timeout);
6134                 status = -ERESTARTSYS;
6135                 if (signalled())
6136                         break;
6137         } while(status < 0);
6138         return status;
6139 }
6140
6141 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6142 {
6143         struct nfs_server *server = NFS_SERVER(state->inode);
6144         int err;
6145
6146         err = nfs4_set_lock_state(state, fl);
6147         if (err != 0)
6148                 return err;
6149         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6150         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6151 }
6152
6153 struct nfs_release_lockowner_data {
6154         struct nfs4_lock_state *lsp;
6155         struct nfs_server *server;
6156         struct nfs_release_lockowner_args args;
6157         struct nfs_release_lockowner_res res;
6158         unsigned long timestamp;
6159 };
6160
6161 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6162 {
6163         struct nfs_release_lockowner_data *data = calldata;
6164         struct nfs_server *server = data->server;
6165         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6166                              &data->args.seq_args, &data->res.seq_res, task);
6167         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6168         data->timestamp = jiffies;
6169 }
6170
6171 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6172 {
6173         struct nfs_release_lockowner_data *data = calldata;
6174         struct nfs_server *server = data->server;
6175
6176         nfs40_sequence_done(task, &data->res.seq_res);
6177
6178         switch (task->tk_status) {
6179         case 0:
6180                 renew_lease(server, data->timestamp);
6181                 break;
6182         case -NFS4ERR_STALE_CLIENTID:
6183         case -NFS4ERR_EXPIRED:
6184                 nfs4_schedule_lease_recovery(server->nfs_client);
6185                 break;
6186         case -NFS4ERR_LEASE_MOVED:
6187         case -NFS4ERR_DELAY:
6188                 if (nfs4_async_handle_error(task, server,
6189                                             NULL, NULL) == -EAGAIN)
6190                         rpc_restart_call_prepare(task);
6191         }
6192 }
6193
6194 static void nfs4_release_lockowner_release(void *calldata)
6195 {
6196         struct nfs_release_lockowner_data *data = calldata;
6197         nfs4_free_lock_state(data->server, data->lsp);
6198         kfree(calldata);
6199 }
6200
6201 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6202         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6203         .rpc_call_done = nfs4_release_lockowner_done,
6204         .rpc_release = nfs4_release_lockowner_release,
6205 };
6206
6207 static void
6208 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6209 {
6210         struct nfs_release_lockowner_data *data;
6211         struct rpc_message msg = {
6212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6213         };
6214
6215         if (server->nfs_client->cl_mvops->minor_version != 0)
6216                 return;
6217
6218         data = kmalloc(sizeof(*data), GFP_NOFS);
6219         if (!data)
6220                 return;
6221         data->lsp = lsp;
6222         data->server = server;
6223         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6224         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6225         data->args.lock_owner.s_dev = server->s_dev;
6226
6227         msg.rpc_argp = &data->args;
6228         msg.rpc_resp = &data->res;
6229         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6230         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6231 }
6232
6233 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6234
6235 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6236                                    const void *buf, size_t buflen,
6237                                    int flags, int type)
6238 {
6239         if (strcmp(key, "") != 0)
6240                 return -EINVAL;
6241
6242         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6243 }
6244
6245 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6246                                    void *buf, size_t buflen, int type)
6247 {
6248         if (strcmp(key, "") != 0)
6249                 return -EINVAL;
6250
6251         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6252 }
6253
6254 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6255                                        size_t list_len, const char *name,
6256                                        size_t name_len, int type)
6257 {
6258         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6259
6260         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6261                 return 0;
6262
6263         if (list && len <= list_len)
6264                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6265         return len;
6266 }
6267
6268 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6269 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6270 {
6271         return server->caps & NFS_CAP_SECURITY_LABEL;
6272 }
6273
6274 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6275                                    const void *buf, size_t buflen,
6276                                    int flags, int type)
6277 {
6278         if (security_ismaclabel(key))
6279                 return nfs4_set_security_label(dentry, buf, buflen);
6280
6281         return -EOPNOTSUPP;
6282 }
6283
6284 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6285                                    void *buf, size_t buflen, int type)
6286 {
6287         if (security_ismaclabel(key))
6288                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6289         return -EOPNOTSUPP;
6290 }
6291
6292 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6293                                        size_t list_len, const char *name,
6294                                        size_t name_len, int type)
6295 {
6296         size_t len = 0;
6297
6298         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6299                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6300                 if (list && len <= list_len)
6301                         security_inode_listsecurity(d_inode(dentry), list, len);
6302         }
6303         return len;
6304 }
6305
6306 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6307         .prefix = XATTR_SECURITY_PREFIX,
6308         .list   = nfs4_xattr_list_nfs4_label,
6309         .get    = nfs4_xattr_get_nfs4_label,
6310         .set    = nfs4_xattr_set_nfs4_label,
6311 };
6312 #endif
6313
6314
6315 /*
6316  * nfs_fhget will use either the mounted_on_fileid or the fileid
6317  */
6318 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6319 {
6320         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6321                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6322               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6323               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6324                 return;
6325
6326         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6327                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6328         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6329         fattr->nlink = 2;
6330 }
6331
6332 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6333                                    const struct qstr *name,
6334                                    struct nfs4_fs_locations *fs_locations,
6335                                    struct page *page)
6336 {
6337         struct nfs_server *server = NFS_SERVER(dir);
6338         u32 bitmask[3] = {
6339                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6340         };
6341         struct nfs4_fs_locations_arg args = {
6342                 .dir_fh = NFS_FH(dir),
6343                 .name = name,
6344                 .page = page,
6345                 .bitmask = bitmask,
6346         };
6347         struct nfs4_fs_locations_res res = {
6348                 .fs_locations = fs_locations,
6349         };
6350         struct rpc_message msg = {
6351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6352                 .rpc_argp = &args,
6353                 .rpc_resp = &res,
6354         };
6355         int status;
6356
6357         dprintk("%s: start\n", __func__);
6358
6359         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6360          * is not supported */
6361         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6362                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6363         else
6364                 bitmask[0] |= FATTR4_WORD0_FILEID;
6365
6366         nfs_fattr_init(&fs_locations->fattr);
6367         fs_locations->server = server;
6368         fs_locations->nlocations = 0;
6369         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6370         dprintk("%s: returned status = %d\n", __func__, status);
6371         return status;
6372 }
6373
6374 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6375                            const struct qstr *name,
6376                            struct nfs4_fs_locations *fs_locations,
6377                            struct page *page)
6378 {
6379         struct nfs4_exception exception = { };
6380         int err;
6381         do {
6382                 err = _nfs4_proc_fs_locations(client, dir, name,
6383                                 fs_locations, page);
6384                 trace_nfs4_get_fs_locations(dir, name, err);
6385                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6386                                 &exception);
6387         } while (exception.retry);
6388         return err;
6389 }
6390
6391 /*
6392  * This operation also signals the server that this client is
6393  * performing migration recovery.  The server can stop returning
6394  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6395  * appended to this compound to identify the client ID which is
6396  * performing recovery.
6397  */
6398 static int _nfs40_proc_get_locations(struct inode *inode,
6399                                      struct nfs4_fs_locations *locations,
6400                                      struct page *page, struct rpc_cred *cred)
6401 {
6402         struct nfs_server *server = NFS_SERVER(inode);
6403         struct rpc_clnt *clnt = server->client;
6404         u32 bitmask[2] = {
6405                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6406         };
6407         struct nfs4_fs_locations_arg args = {
6408                 .clientid       = server->nfs_client->cl_clientid,
6409                 .fh             = NFS_FH(inode),
6410                 .page           = page,
6411                 .bitmask        = bitmask,
6412                 .migration      = 1,            /* skip LOOKUP */
6413                 .renew          = 1,            /* append RENEW */
6414         };
6415         struct nfs4_fs_locations_res res = {
6416                 .fs_locations   = locations,
6417                 .migration      = 1,
6418                 .renew          = 1,
6419         };
6420         struct rpc_message msg = {
6421                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6422                 .rpc_argp       = &args,
6423                 .rpc_resp       = &res,
6424                 .rpc_cred       = cred,
6425         };
6426         unsigned long now = jiffies;
6427         int status;
6428
6429         nfs_fattr_init(&locations->fattr);
6430         locations->server = server;
6431         locations->nlocations = 0;
6432
6433         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6434         nfs4_set_sequence_privileged(&args.seq_args);
6435         status = nfs4_call_sync_sequence(clnt, server, &msg,
6436                                         &args.seq_args, &res.seq_res);
6437         if (status)
6438                 return status;
6439
6440         renew_lease(server, now);
6441         return 0;
6442 }
6443
6444 #ifdef CONFIG_NFS_V4_1
6445
6446 /*
6447  * This operation also signals the server that this client is
6448  * performing migration recovery.  The server can stop asserting
6449  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6450  * performing this operation is identified in the SEQUENCE
6451  * operation in this compound.
6452  *
6453  * When the client supports GETATTR(fs_locations_info), it can
6454  * be plumbed in here.
6455  */
6456 static int _nfs41_proc_get_locations(struct inode *inode,
6457                                      struct nfs4_fs_locations *locations,
6458                                      struct page *page, struct rpc_cred *cred)
6459 {
6460         struct nfs_server *server = NFS_SERVER(inode);
6461         struct rpc_clnt *clnt = server->client;
6462         u32 bitmask[2] = {
6463                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6464         };
6465         struct nfs4_fs_locations_arg args = {
6466                 .fh             = NFS_FH(inode),
6467                 .page           = page,
6468                 .bitmask        = bitmask,
6469                 .migration      = 1,            /* skip LOOKUP */
6470         };
6471         struct nfs4_fs_locations_res res = {
6472                 .fs_locations   = locations,
6473                 .migration      = 1,
6474         };
6475         struct rpc_message msg = {
6476                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6477                 .rpc_argp       = &args,
6478                 .rpc_resp       = &res,
6479                 .rpc_cred       = cred,
6480         };
6481         int status;
6482
6483         nfs_fattr_init(&locations->fattr);
6484         locations->server = server;
6485         locations->nlocations = 0;
6486
6487         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6488         nfs4_set_sequence_privileged(&args.seq_args);
6489         status = nfs4_call_sync_sequence(clnt, server, &msg,
6490                                         &args.seq_args, &res.seq_res);
6491         if (status == NFS4_OK &&
6492             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6493                 status = -NFS4ERR_LEASE_MOVED;
6494         return status;
6495 }
6496
6497 #endif  /* CONFIG_NFS_V4_1 */
6498
6499 /**
6500  * nfs4_proc_get_locations - discover locations for a migrated FSID
6501  * @inode: inode on FSID that is migrating
6502  * @locations: result of query
6503  * @page: buffer
6504  * @cred: credential to use for this operation
6505  *
6506  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6507  * operation failed, or a negative errno if a local error occurred.
6508  *
6509  * On success, "locations" is filled in, but if the server has
6510  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6511  * asserted.
6512  *
6513  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6514  * from this client that require migration recovery.
6515  */
6516 int nfs4_proc_get_locations(struct inode *inode,
6517                             struct nfs4_fs_locations *locations,
6518                             struct page *page, struct rpc_cred *cred)
6519 {
6520         struct nfs_server *server = NFS_SERVER(inode);
6521         struct nfs_client *clp = server->nfs_client;
6522         const struct nfs4_mig_recovery_ops *ops =
6523                                         clp->cl_mvops->mig_recovery_ops;
6524         struct nfs4_exception exception = { };
6525         int status;
6526
6527         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6528                 (unsigned long long)server->fsid.major,
6529                 (unsigned long long)server->fsid.minor,
6530                 clp->cl_hostname);
6531         nfs_display_fhandle(NFS_FH(inode), __func__);
6532
6533         do {
6534                 status = ops->get_locations(inode, locations, page, cred);
6535                 if (status != -NFS4ERR_DELAY)
6536                         break;
6537                 nfs4_handle_exception(server, status, &exception);
6538         } while (exception.retry);
6539         return status;
6540 }
6541
6542 /*
6543  * This operation also signals the server that this client is
6544  * performing "lease moved" recovery.  The server can stop
6545  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6546  * is appended to this compound to identify the client ID which is
6547  * performing recovery.
6548  */
6549 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6550 {
6551         struct nfs_server *server = NFS_SERVER(inode);
6552         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6553         struct rpc_clnt *clnt = server->client;
6554         struct nfs4_fsid_present_arg args = {
6555                 .fh             = NFS_FH(inode),
6556                 .clientid       = clp->cl_clientid,
6557                 .renew          = 1,            /* append RENEW */
6558         };
6559         struct nfs4_fsid_present_res res = {
6560                 .renew          = 1,
6561         };
6562         struct rpc_message msg = {
6563                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6564                 .rpc_argp       = &args,
6565                 .rpc_resp       = &res,
6566                 .rpc_cred       = cred,
6567         };
6568         unsigned long now = jiffies;
6569         int status;
6570
6571         res.fh = nfs_alloc_fhandle();
6572         if (res.fh == NULL)
6573                 return -ENOMEM;
6574
6575         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6576         nfs4_set_sequence_privileged(&args.seq_args);
6577         status = nfs4_call_sync_sequence(clnt, server, &msg,
6578                                                 &args.seq_args, &res.seq_res);
6579         nfs_free_fhandle(res.fh);
6580         if (status)
6581                 return status;
6582
6583         do_renew_lease(clp, now);
6584         return 0;
6585 }
6586
6587 #ifdef CONFIG_NFS_V4_1
6588
6589 /*
6590  * This operation also signals the server that this client is
6591  * performing "lease moved" recovery.  The server can stop asserting
6592  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6593  * this operation is identified in the SEQUENCE operation in this
6594  * compound.
6595  */
6596 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6597 {
6598         struct nfs_server *server = NFS_SERVER(inode);
6599         struct rpc_clnt *clnt = server->client;
6600         struct nfs4_fsid_present_arg args = {
6601                 .fh             = NFS_FH(inode),
6602         };
6603         struct nfs4_fsid_present_res res = {
6604         };
6605         struct rpc_message msg = {
6606                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6607                 .rpc_argp       = &args,
6608                 .rpc_resp       = &res,
6609                 .rpc_cred       = cred,
6610         };
6611         int status;
6612
6613         res.fh = nfs_alloc_fhandle();
6614         if (res.fh == NULL)
6615                 return -ENOMEM;
6616
6617         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6618         nfs4_set_sequence_privileged(&args.seq_args);
6619         status = nfs4_call_sync_sequence(clnt, server, &msg,
6620                                                 &args.seq_args, &res.seq_res);
6621         nfs_free_fhandle(res.fh);
6622         if (status == NFS4_OK &&
6623             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6624                 status = -NFS4ERR_LEASE_MOVED;
6625         return status;
6626 }
6627
6628 #endif  /* CONFIG_NFS_V4_1 */
6629
6630 /**
6631  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6632  * @inode: inode on FSID to check
6633  * @cred: credential to use for this operation
6634  *
6635  * Server indicates whether the FSID is present, moved, or not
6636  * recognized.  This operation is necessary to clear a LEASE_MOVED
6637  * condition for this client ID.
6638  *
6639  * Returns NFS4_OK if the FSID is present on this server,
6640  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6641  *  NFS4ERR code if some error occurred on the server, or a
6642  *  negative errno if a local failure occurred.
6643  */
6644 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6645 {
6646         struct nfs_server *server = NFS_SERVER(inode);
6647         struct nfs_client *clp = server->nfs_client;
6648         const struct nfs4_mig_recovery_ops *ops =
6649                                         clp->cl_mvops->mig_recovery_ops;
6650         struct nfs4_exception exception = { };
6651         int status;
6652
6653         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6654                 (unsigned long long)server->fsid.major,
6655                 (unsigned long long)server->fsid.minor,
6656                 clp->cl_hostname);
6657         nfs_display_fhandle(NFS_FH(inode), __func__);
6658
6659         do {
6660                 status = ops->fsid_present(inode, cred);
6661                 if (status != -NFS4ERR_DELAY)
6662                         break;
6663                 nfs4_handle_exception(server, status, &exception);
6664         } while (exception.retry);
6665         return status;
6666 }
6667
6668 /**
6669  * If 'use_integrity' is true and the state managment nfs_client
6670  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6671  * and the machine credential as per RFC3530bis and RFC5661 Security
6672  * Considerations sections. Otherwise, just use the user cred with the
6673  * filesystem's rpc_client.
6674  */
6675 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6676 {
6677         int status;
6678         struct nfs4_secinfo_arg args = {
6679                 .dir_fh = NFS_FH(dir),
6680                 .name   = name,
6681         };
6682         struct nfs4_secinfo_res res = {
6683                 .flavors     = flavors,
6684         };
6685         struct rpc_message msg = {
6686                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6687                 .rpc_argp = &args,
6688                 .rpc_resp = &res,
6689         };
6690         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6691         struct rpc_cred *cred = NULL;
6692
6693         if (use_integrity) {
6694                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6695                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6696                 msg.rpc_cred = cred;
6697         }
6698
6699         dprintk("NFS call  secinfo %s\n", name->name);
6700
6701         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6702                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6703
6704         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6705                                 &res.seq_res, 0);
6706         dprintk("NFS reply  secinfo: %d\n", status);
6707
6708         if (cred)
6709                 put_rpccred(cred);
6710
6711         return status;
6712 }
6713
6714 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6715                       struct nfs4_secinfo_flavors *flavors)
6716 {
6717         struct nfs4_exception exception = { };
6718         int err;
6719         do {
6720                 err = -NFS4ERR_WRONGSEC;
6721
6722                 /* try to use integrity protection with machine cred */
6723                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6724                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6725
6726                 /*
6727                  * if unable to use integrity protection, or SECINFO with
6728                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6729                  * disallowed by spec, but exists in deployed servers) use
6730                  * the current filesystem's rpc_client and the user cred.
6731                  */
6732                 if (err == -NFS4ERR_WRONGSEC)
6733                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6734
6735                 trace_nfs4_secinfo(dir, name, err);
6736                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6737                                 &exception);
6738         } while (exception.retry);
6739         return err;
6740 }
6741
6742 #ifdef CONFIG_NFS_V4_1
6743 /*
6744  * Check the exchange flags returned by the server for invalid flags, having
6745  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6746  * DS flags set.
6747  */
6748 static int nfs4_check_cl_exchange_flags(u32 flags)
6749 {
6750         if (flags & ~EXCHGID4_FLAG_MASK_R)
6751                 goto out_inval;
6752         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6753             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6754                 goto out_inval;
6755         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6756                 goto out_inval;
6757         return NFS_OK;
6758 out_inval:
6759         return -NFS4ERR_INVAL;
6760 }
6761
6762 static bool
6763 nfs41_same_server_scope(struct nfs41_server_scope *a,
6764                         struct nfs41_server_scope *b)
6765 {
6766         if (a->server_scope_sz == b->server_scope_sz &&
6767             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6768                 return true;
6769
6770         return false;
6771 }
6772
6773 /*
6774  * nfs4_proc_bind_conn_to_session()
6775  *
6776  * The 4.1 client currently uses the same TCP connection for the
6777  * fore and backchannel.
6778  */
6779 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6780 {
6781         int status;
6782         struct nfs41_bind_conn_to_session_args args = {
6783                 .client = clp,
6784                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6785         };
6786         struct nfs41_bind_conn_to_session_res res;
6787         struct rpc_message msg = {
6788                 .rpc_proc =
6789                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6790                 .rpc_argp = &args,
6791                 .rpc_resp = &res,
6792                 .rpc_cred = cred,
6793         };
6794
6795         dprintk("--> %s\n", __func__);
6796
6797         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6798         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6799                 args.dir = NFS4_CDFC4_FORE;
6800
6801         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6802         trace_nfs4_bind_conn_to_session(clp, status);
6803         if (status == 0) {
6804                 if (memcmp(res.sessionid.data,
6805                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6806                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6807                         status = -EIO;
6808                         goto out;
6809                 }
6810                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6811                         dprintk("NFS: %s: Unexpected direction from server\n",
6812                                 __func__);
6813                         status = -EIO;
6814                         goto out;
6815                 }
6816                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6817                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6818                                 __func__);
6819                         status = -EIO;
6820                         goto out;
6821                 }
6822         }
6823 out:
6824         dprintk("<-- %s status= %d\n", __func__, status);
6825         return status;
6826 }
6827
6828 /*
6829  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6830  * and operations we'd like to see to enable certain features in the allow map
6831  */
6832 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6833         .how = SP4_MACH_CRED,
6834         .enforce.u.words = {
6835                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6836                       1 << (OP_EXCHANGE_ID - 32) |
6837                       1 << (OP_CREATE_SESSION - 32) |
6838                       1 << (OP_DESTROY_SESSION - 32) |
6839                       1 << (OP_DESTROY_CLIENTID - 32)
6840         },
6841         .allow.u.words = {
6842                 [0] = 1 << (OP_CLOSE) |
6843                       1 << (OP_LOCKU) |
6844                       1 << (OP_COMMIT),
6845                 [1] = 1 << (OP_SECINFO - 32) |
6846                       1 << (OP_SECINFO_NO_NAME - 32) |
6847                       1 << (OP_TEST_STATEID - 32) |
6848                       1 << (OP_FREE_STATEID - 32) |
6849                       1 << (OP_WRITE - 32)
6850         }
6851 };
6852
6853 /*
6854  * Select the state protection mode for client `clp' given the server results
6855  * from exchange_id in `sp'.
6856  *
6857  * Returns 0 on success, negative errno otherwise.
6858  */
6859 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6860                                  struct nfs41_state_protection *sp)
6861 {
6862         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6863                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6864                       1 << (OP_EXCHANGE_ID - 32) |
6865                       1 << (OP_CREATE_SESSION - 32) |
6866                       1 << (OP_DESTROY_SESSION - 32) |
6867                       1 << (OP_DESTROY_CLIENTID - 32)
6868         };
6869         unsigned int i;
6870
6871         if (sp->how == SP4_MACH_CRED) {
6872                 /* Print state protect result */
6873                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6874                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6875                         if (test_bit(i, sp->enforce.u.longs))
6876                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6877                         if (test_bit(i, sp->allow.u.longs))
6878                                 dfprintk(MOUNT, "  allow op %d\n", i);
6879                 }
6880
6881                 /* make sure nothing is on enforce list that isn't supported */
6882                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6883                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6884                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6885                                 return -EINVAL;
6886                         }
6887                 }
6888
6889                 /*
6890                  * Minimal mode - state operations are allowed to use machine
6891                  * credential.  Note this already happens by default, so the
6892                  * client doesn't have to do anything more than the negotiation.
6893                  *
6894                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6895                  *       we're already using the machine cred for exchange_id
6896                  *       and will never use a different cred.
6897                  */
6898                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6899                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6900                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6901                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6902                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6903                         dfprintk(MOUNT, "  minimal mode enabled\n");
6904                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6905                 } else {
6906                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6907                         return -EINVAL;
6908                 }
6909
6910                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6911                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6912                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6913                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6914                 }
6915
6916                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6917                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6918                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6919                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6920                 }
6921
6922                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6923                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6924                         dfprintk(MOUNT, "  stateid mode enabled\n");
6925                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6926                 }
6927
6928                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6929                         dfprintk(MOUNT, "  write mode enabled\n");
6930                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6931                 }
6932
6933                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6934                         dfprintk(MOUNT, "  commit mode enabled\n");
6935                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6936                 }
6937         }
6938
6939         return 0;
6940 }
6941
6942 /*
6943  * _nfs4_proc_exchange_id()
6944  *
6945  * Wrapper for EXCHANGE_ID operation.
6946  */
6947 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6948         u32 sp4_how)
6949 {
6950         nfs4_verifier verifier;
6951         struct nfs41_exchange_id_args args = {
6952                 .verifier = &verifier,
6953                 .client = clp,
6954 #ifdef CONFIG_NFS_V4_1_MIGRATION
6955                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6956                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6957                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6958 #else
6959                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6960                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6961 #endif
6962         };
6963         struct nfs41_exchange_id_res res = {
6964                 0
6965         };
6966         int status;
6967         struct rpc_message msg = {
6968                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6969                 .rpc_argp = &args,
6970                 .rpc_resp = &res,
6971                 .rpc_cred = cred,
6972         };
6973
6974         nfs4_init_boot_verifier(clp, &verifier);
6975
6976         status = nfs4_init_uniform_client_string(clp);
6977         if (status)
6978                 goto out;
6979
6980         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6981                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6982                 clp->cl_owner_id);
6983
6984         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6985                                         GFP_NOFS);
6986         if (unlikely(res.server_owner == NULL)) {
6987                 status = -ENOMEM;
6988                 goto out;
6989         }
6990
6991         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6992                                         GFP_NOFS);
6993         if (unlikely(res.server_scope == NULL)) {
6994                 status = -ENOMEM;
6995                 goto out_server_owner;
6996         }
6997
6998         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6999         if (unlikely(res.impl_id == NULL)) {
7000                 status = -ENOMEM;
7001                 goto out_server_scope;
7002         }
7003
7004         switch (sp4_how) {
7005         case SP4_NONE:
7006                 args.state_protect.how = SP4_NONE;
7007                 break;
7008
7009         case SP4_MACH_CRED:
7010                 args.state_protect = nfs4_sp4_mach_cred_request;
7011                 break;
7012
7013         default:
7014                 /* unsupported! */
7015                 WARN_ON_ONCE(1);
7016                 status = -EINVAL;
7017                 goto out_impl_id;
7018         }
7019
7020         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7021         trace_nfs4_exchange_id(clp, status);
7022         if (status == 0)
7023                 status = nfs4_check_cl_exchange_flags(res.flags);
7024
7025         if (status == 0)
7026                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7027
7028         if (status == 0) {
7029                 clp->cl_clientid = res.clientid;
7030                 clp->cl_exchange_flags = res.flags;
7031                 /* Client ID is not confirmed */
7032                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7033                         clear_bit(NFS4_SESSION_ESTABLISHED,
7034                                         &clp->cl_session->session_state);
7035                         clp->cl_seqid = res.seqid;
7036                 }
7037
7038                 kfree(clp->cl_serverowner);
7039                 clp->cl_serverowner = res.server_owner;
7040                 res.server_owner = NULL;
7041
7042                 /* use the most recent implementation id */
7043                 kfree(clp->cl_implid);
7044                 clp->cl_implid = res.impl_id;
7045                 res.impl_id = NULL;
7046
7047                 if (clp->cl_serverscope != NULL &&
7048                     !nfs41_same_server_scope(clp->cl_serverscope,
7049                                              res.server_scope)) {
7050                         dprintk("%s: server_scope mismatch detected\n",
7051                                 __func__);
7052                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7053                         kfree(clp->cl_serverscope);
7054                         clp->cl_serverscope = NULL;
7055                 }
7056
7057                 if (clp->cl_serverscope == NULL) {
7058                         clp->cl_serverscope = res.server_scope;
7059                         res.server_scope = NULL;
7060                 }
7061         }
7062
7063 out_impl_id:
7064         kfree(res.impl_id);
7065 out_server_scope:
7066         kfree(res.server_scope);
7067 out_server_owner:
7068         kfree(res.server_owner);
7069 out:
7070         if (clp->cl_implid != NULL)
7071                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7072                         "domain: %s, name: %s, date: %llu,%u\n",
7073                         clp->cl_implid->domain, clp->cl_implid->name,
7074                         clp->cl_implid->date.seconds,
7075                         clp->cl_implid->date.nseconds);
7076         dprintk("NFS reply exchange_id: %d\n", status);
7077         return status;
7078 }
7079
7080 /*
7081  * nfs4_proc_exchange_id()
7082  *
7083  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7084  *
7085  * Since the clientid has expired, all compounds using sessions
7086  * associated with the stale clientid will be returning
7087  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7088  * be in some phase of session reset.
7089  *
7090  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7091  */
7092 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7093 {
7094         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7095         int status;
7096
7097         /* try SP4_MACH_CRED if krb5i/p */
7098         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7099             authflavor == RPC_AUTH_GSS_KRB5P) {
7100                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7101                 if (!status)
7102                         return 0;
7103         }
7104
7105         /* try SP4_NONE */
7106         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7107 }
7108
7109 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7110                 struct rpc_cred *cred)
7111 {
7112         struct rpc_message msg = {
7113                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7114                 .rpc_argp = clp,
7115                 .rpc_cred = cred,
7116         };
7117         int status;
7118
7119         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7120         trace_nfs4_destroy_clientid(clp, status);
7121         if (status)
7122                 dprintk("NFS: Got error %d from the server %s on "
7123                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7124         return status;
7125 }
7126
7127 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7128                 struct rpc_cred *cred)
7129 {
7130         unsigned int loop;
7131         int ret;
7132
7133         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7134                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7135                 switch (ret) {
7136                 case -NFS4ERR_DELAY:
7137                 case -NFS4ERR_CLIENTID_BUSY:
7138                         ssleep(1);
7139                         break;
7140                 default:
7141                         return ret;
7142                 }
7143         }
7144         return 0;
7145 }
7146
7147 int nfs4_destroy_clientid(struct nfs_client *clp)
7148 {
7149         struct rpc_cred *cred;
7150         int ret = 0;
7151
7152         if (clp->cl_mvops->minor_version < 1)
7153                 goto out;
7154         if (clp->cl_exchange_flags == 0)
7155                 goto out;
7156         if (clp->cl_preserve_clid)
7157                 goto out;
7158         cred = nfs4_get_clid_cred(clp);
7159         ret = nfs4_proc_destroy_clientid(clp, cred);
7160         if (cred)
7161                 put_rpccred(cred);
7162         switch (ret) {
7163         case 0:
7164         case -NFS4ERR_STALE_CLIENTID:
7165                 clp->cl_exchange_flags = 0;
7166         }
7167 out:
7168         return ret;
7169 }
7170
7171 struct nfs4_get_lease_time_data {
7172         struct nfs4_get_lease_time_args *args;
7173         struct nfs4_get_lease_time_res *res;
7174         struct nfs_client *clp;
7175 };
7176
7177 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7178                                         void *calldata)
7179 {
7180         struct nfs4_get_lease_time_data *data =
7181                         (struct nfs4_get_lease_time_data *)calldata;
7182
7183         dprintk("--> %s\n", __func__);
7184         /* just setup sequence, do not trigger session recovery
7185            since we're invoked within one */
7186         nfs41_setup_sequence(data->clp->cl_session,
7187                         &data->args->la_seq_args,
7188                         &data->res->lr_seq_res,
7189                         task);
7190         dprintk("<-- %s\n", __func__);
7191 }
7192
7193 /*
7194  * Called from nfs4_state_manager thread for session setup, so don't recover
7195  * from sequence operation or clientid errors.
7196  */
7197 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7198 {
7199         struct nfs4_get_lease_time_data *data =
7200                         (struct nfs4_get_lease_time_data *)calldata;
7201
7202         dprintk("--> %s\n", __func__);
7203         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7204                 return;
7205         switch (task->tk_status) {
7206         case -NFS4ERR_DELAY:
7207         case -NFS4ERR_GRACE:
7208                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7209                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7210                 task->tk_status = 0;
7211                 /* fall through */
7212         case -NFS4ERR_RETRY_UNCACHED_REP:
7213                 rpc_restart_call_prepare(task);
7214                 return;
7215         }
7216         dprintk("<-- %s\n", __func__);
7217 }
7218
7219 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7220         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7221         .rpc_call_done = nfs4_get_lease_time_done,
7222 };
7223
7224 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7225 {
7226         struct rpc_task *task;
7227         struct nfs4_get_lease_time_args args;
7228         struct nfs4_get_lease_time_res res = {
7229                 .lr_fsinfo = fsinfo,
7230         };
7231         struct nfs4_get_lease_time_data data = {
7232                 .args = &args,
7233                 .res = &res,
7234                 .clp = clp,
7235         };
7236         struct rpc_message msg = {
7237                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7238                 .rpc_argp = &args,
7239                 .rpc_resp = &res,
7240         };
7241         struct rpc_task_setup task_setup = {
7242                 .rpc_client = clp->cl_rpcclient,
7243                 .rpc_message = &msg,
7244                 .callback_ops = &nfs4_get_lease_time_ops,
7245                 .callback_data = &data,
7246                 .flags = RPC_TASK_TIMEOUT,
7247         };
7248         int status;
7249
7250         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7251         nfs4_set_sequence_privileged(&args.la_seq_args);
7252         dprintk("--> %s\n", __func__);
7253         task = rpc_run_task(&task_setup);
7254
7255         if (IS_ERR(task))
7256                 status = PTR_ERR(task);
7257         else {
7258                 status = task->tk_status;
7259                 rpc_put_task(task);
7260         }
7261         dprintk("<-- %s return %d\n", __func__, status);
7262
7263         return status;
7264 }
7265
7266 /*
7267  * Initialize the values to be used by the client in CREATE_SESSION
7268  * If nfs4_init_session set the fore channel request and response sizes,
7269  * use them.
7270  *
7271  * Set the back channel max_resp_sz_cached to zero to force the client to
7272  * always set csa_cachethis to FALSE because the current implementation
7273  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7274  */
7275 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7276 {
7277         unsigned int max_rqst_sz, max_resp_sz;
7278
7279         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7280         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7281
7282         /* Fore channel attributes */
7283         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7284         args->fc_attrs.max_resp_sz = max_resp_sz;
7285         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7286         args->fc_attrs.max_reqs = max_session_slots;
7287
7288         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7289                 "max_ops=%u max_reqs=%u\n",
7290                 __func__,
7291                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7292                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7293
7294         /* Back channel attributes */
7295         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7296         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7297         args->bc_attrs.max_resp_sz_cached = 0;
7298         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7299         args->bc_attrs.max_reqs = 1;
7300
7301         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7302                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7303                 __func__,
7304                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7305                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7306                 args->bc_attrs.max_reqs);
7307 }
7308
7309 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7310                 struct nfs41_create_session_res *res)
7311 {
7312         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7313         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7314
7315         if (rcvd->max_resp_sz > sent->max_resp_sz)
7316                 return -EINVAL;
7317         /*
7318          * Our requested max_ops is the minimum we need; we're not
7319          * prepared to break up compounds into smaller pieces than that.
7320          * So, no point even trying to continue if the server won't
7321          * cooperate:
7322          */
7323         if (rcvd->max_ops < sent->max_ops)
7324                 return -EINVAL;
7325         if (rcvd->max_reqs == 0)
7326                 return -EINVAL;
7327         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7328                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7329         return 0;
7330 }
7331
7332 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7333                 struct nfs41_create_session_res *res)
7334 {
7335         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7336         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7337
7338         if (!(res->flags & SESSION4_BACK_CHAN))
7339                 goto out;
7340         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7341                 return -EINVAL;
7342         if (rcvd->max_resp_sz < sent->max_resp_sz)
7343                 return -EINVAL;
7344         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7345                 return -EINVAL;
7346         /* These would render the backchannel useless: */
7347         if (rcvd->max_ops != sent->max_ops)
7348                 return -EINVAL;
7349         if (rcvd->max_reqs != sent->max_reqs)
7350                 return -EINVAL;
7351 out:
7352         return 0;
7353 }
7354
7355 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7356                                      struct nfs41_create_session_res *res)
7357 {
7358         int ret;
7359
7360         ret = nfs4_verify_fore_channel_attrs(args, res);
7361         if (ret)
7362                 return ret;
7363         return nfs4_verify_back_channel_attrs(args, res);
7364 }
7365
7366 static void nfs4_update_session(struct nfs4_session *session,
7367                 struct nfs41_create_session_res *res)
7368 {
7369         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7370         /* Mark client id and session as being confirmed */
7371         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7372         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7373         session->flags = res->flags;
7374         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7375         if (res->flags & SESSION4_BACK_CHAN)
7376                 memcpy(&session->bc_attrs, &res->bc_attrs,
7377                                 sizeof(session->bc_attrs));
7378 }
7379
7380 static int _nfs4_proc_create_session(struct nfs_client *clp,
7381                 struct rpc_cred *cred)
7382 {
7383         struct nfs4_session *session = clp->cl_session;
7384         struct nfs41_create_session_args args = {
7385                 .client = clp,
7386                 .clientid = clp->cl_clientid,
7387                 .seqid = clp->cl_seqid,
7388                 .cb_program = NFS4_CALLBACK,
7389         };
7390         struct nfs41_create_session_res res;
7391
7392         struct rpc_message msg = {
7393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7394                 .rpc_argp = &args,
7395                 .rpc_resp = &res,
7396                 .rpc_cred = cred,
7397         };
7398         int status;
7399
7400         nfs4_init_channel_attrs(&args);
7401         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7402
7403         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7404         trace_nfs4_create_session(clp, status);
7405
7406         if (!status) {
7407                 /* Verify the session's negotiated channel_attrs values */
7408                 status = nfs4_verify_channel_attrs(&args, &res);
7409                 /* Increment the clientid slot sequence id */
7410                 if (clp->cl_seqid == res.seqid)
7411                         clp->cl_seqid++;
7412                 if (status)
7413                         goto out;
7414                 nfs4_update_session(session, &res);
7415         }
7416 out:
7417         return status;
7418 }
7419
7420 /*
7421  * Issues a CREATE_SESSION operation to the server.
7422  * It is the responsibility of the caller to verify the session is
7423  * expired before calling this routine.
7424  */
7425 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7426 {
7427         int status;
7428         unsigned *ptr;
7429         struct nfs4_session *session = clp->cl_session;
7430
7431         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7432
7433         status = _nfs4_proc_create_session(clp, cred);
7434         if (status)
7435                 goto out;
7436
7437         /* Init or reset the session slot tables */
7438         status = nfs4_setup_session_slot_tables(session);
7439         dprintk("slot table setup returned %d\n", status);
7440         if (status)
7441                 goto out;
7442
7443         ptr = (unsigned *)&session->sess_id.data[0];
7444         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7445                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7446 out:
7447         dprintk("<-- %s\n", __func__);
7448         return status;
7449 }
7450
7451 /*
7452  * Issue the over-the-wire RPC DESTROY_SESSION.
7453  * The caller must serialize access to this routine.
7454  */
7455 int nfs4_proc_destroy_session(struct nfs4_session *session,
7456                 struct rpc_cred *cred)
7457 {
7458         struct rpc_message msg = {
7459                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7460                 .rpc_argp = session,
7461                 .rpc_cred = cred,
7462         };
7463         int status = 0;
7464
7465         dprintk("--> nfs4_proc_destroy_session\n");
7466
7467         /* session is still being setup */
7468         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7469                 return 0;
7470
7471         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7472         trace_nfs4_destroy_session(session->clp, status);
7473
7474         if (status)
7475                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7476                         "Session has been destroyed regardless...\n", status);
7477
7478         dprintk("<-- nfs4_proc_destroy_session\n");
7479         return status;
7480 }
7481
7482 /*
7483  * Renew the cl_session lease.
7484  */
7485 struct nfs4_sequence_data {
7486         struct nfs_client *clp;
7487         struct nfs4_sequence_args args;
7488         struct nfs4_sequence_res res;
7489 };
7490
7491 static void nfs41_sequence_release(void *data)
7492 {
7493         struct nfs4_sequence_data *calldata = data;
7494         struct nfs_client *clp = calldata->clp;
7495
7496         if (atomic_read(&clp->cl_count) > 1)
7497                 nfs4_schedule_state_renewal(clp);
7498         nfs_put_client(clp);
7499         kfree(calldata);
7500 }
7501
7502 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7503 {
7504         switch(task->tk_status) {
7505         case -NFS4ERR_DELAY:
7506                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7507                 return -EAGAIN;
7508         default:
7509                 nfs4_schedule_lease_recovery(clp);
7510         }
7511         return 0;
7512 }
7513
7514 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7515 {
7516         struct nfs4_sequence_data *calldata = data;
7517         struct nfs_client *clp = calldata->clp;
7518
7519         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7520                 return;
7521
7522         trace_nfs4_sequence(clp, task->tk_status);
7523         if (task->tk_status < 0) {
7524                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7525                 if (atomic_read(&clp->cl_count) == 1)
7526                         goto out;
7527
7528                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7529                         rpc_restart_call_prepare(task);
7530                         return;
7531                 }
7532         }
7533         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7534 out:
7535         dprintk("<-- %s\n", __func__);
7536 }
7537
7538 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7539 {
7540         struct nfs4_sequence_data *calldata = data;
7541         struct nfs_client *clp = calldata->clp;
7542         struct nfs4_sequence_args *args;
7543         struct nfs4_sequence_res *res;
7544
7545         args = task->tk_msg.rpc_argp;
7546         res = task->tk_msg.rpc_resp;
7547
7548         nfs41_setup_sequence(clp->cl_session, args, res, task);
7549 }
7550
7551 static const struct rpc_call_ops nfs41_sequence_ops = {
7552         .rpc_call_done = nfs41_sequence_call_done,
7553         .rpc_call_prepare = nfs41_sequence_prepare,
7554         .rpc_release = nfs41_sequence_release,
7555 };
7556
7557 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7558                 struct rpc_cred *cred,
7559                 bool is_privileged)
7560 {
7561         struct nfs4_sequence_data *calldata;
7562         struct rpc_message msg = {
7563                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7564                 .rpc_cred = cred,
7565         };
7566         struct rpc_task_setup task_setup_data = {
7567                 .rpc_client = clp->cl_rpcclient,
7568                 .rpc_message = &msg,
7569                 .callback_ops = &nfs41_sequence_ops,
7570                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7571         };
7572
7573         if (!atomic_inc_not_zero(&clp->cl_count))
7574                 return ERR_PTR(-EIO);
7575         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7576         if (calldata == NULL) {
7577                 nfs_put_client(clp);
7578                 return ERR_PTR(-ENOMEM);
7579         }
7580         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7581         if (is_privileged)
7582                 nfs4_set_sequence_privileged(&calldata->args);
7583         msg.rpc_argp = &calldata->args;
7584         msg.rpc_resp = &calldata->res;
7585         calldata->clp = clp;
7586         task_setup_data.callback_data = calldata;
7587
7588         return rpc_run_task(&task_setup_data);
7589 }
7590
7591 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7592 {
7593         struct rpc_task *task;
7594         int ret = 0;
7595
7596         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7597                 return -EAGAIN;
7598         task = _nfs41_proc_sequence(clp, cred, false);
7599         if (IS_ERR(task))
7600                 ret = PTR_ERR(task);
7601         else
7602                 rpc_put_task_async(task);
7603         dprintk("<-- %s status=%d\n", __func__, ret);
7604         return ret;
7605 }
7606
7607 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7608 {
7609         struct rpc_task *task;
7610         int ret;
7611
7612         task = _nfs41_proc_sequence(clp, cred, true);
7613         if (IS_ERR(task)) {
7614                 ret = PTR_ERR(task);
7615                 goto out;
7616         }
7617         ret = rpc_wait_for_completion_task(task);
7618         if (!ret)
7619                 ret = task->tk_status;
7620         rpc_put_task(task);
7621 out:
7622         dprintk("<-- %s status=%d\n", __func__, ret);
7623         return ret;
7624 }
7625
7626 struct nfs4_reclaim_complete_data {
7627         struct nfs_client *clp;
7628         struct nfs41_reclaim_complete_args arg;
7629         struct nfs41_reclaim_complete_res res;
7630 };
7631
7632 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7633 {
7634         struct nfs4_reclaim_complete_data *calldata = data;
7635
7636         nfs41_setup_sequence(calldata->clp->cl_session,
7637                         &calldata->arg.seq_args,
7638                         &calldata->res.seq_res,
7639                         task);
7640 }
7641
7642 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7643 {
7644         switch(task->tk_status) {
7645         case 0:
7646         case -NFS4ERR_COMPLETE_ALREADY:
7647         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7648                 break;
7649         case -NFS4ERR_DELAY:
7650                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7651                 /* fall through */
7652         case -NFS4ERR_RETRY_UNCACHED_REP:
7653                 return -EAGAIN;
7654         default:
7655                 nfs4_schedule_lease_recovery(clp);
7656         }
7657         return 0;
7658 }
7659
7660 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7661 {
7662         struct nfs4_reclaim_complete_data *calldata = data;
7663         struct nfs_client *clp = calldata->clp;
7664         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7665
7666         dprintk("--> %s\n", __func__);
7667         if (!nfs41_sequence_done(task, res))
7668                 return;
7669
7670         trace_nfs4_reclaim_complete(clp, task->tk_status);
7671         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7672                 rpc_restart_call_prepare(task);
7673                 return;
7674         }
7675         dprintk("<-- %s\n", __func__);
7676 }
7677
7678 static void nfs4_free_reclaim_complete_data(void *data)
7679 {
7680         struct nfs4_reclaim_complete_data *calldata = data;
7681
7682         kfree(calldata);
7683 }
7684
7685 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7686         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7687         .rpc_call_done = nfs4_reclaim_complete_done,
7688         .rpc_release = nfs4_free_reclaim_complete_data,
7689 };
7690
7691 /*
7692  * Issue a global reclaim complete.
7693  */
7694 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7695                 struct rpc_cred *cred)
7696 {
7697         struct nfs4_reclaim_complete_data *calldata;
7698         struct rpc_task *task;
7699         struct rpc_message msg = {
7700                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7701                 .rpc_cred = cred,
7702         };
7703         struct rpc_task_setup task_setup_data = {
7704                 .rpc_client = clp->cl_rpcclient,
7705                 .rpc_message = &msg,
7706                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7707                 .flags = RPC_TASK_ASYNC,
7708         };
7709         int status = -ENOMEM;
7710
7711         dprintk("--> %s\n", __func__);
7712         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7713         if (calldata == NULL)
7714                 goto out;
7715         calldata->clp = clp;
7716         calldata->arg.one_fs = 0;
7717
7718         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7719         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7720         msg.rpc_argp = &calldata->arg;
7721         msg.rpc_resp = &calldata->res;
7722         task_setup_data.callback_data = calldata;
7723         task = rpc_run_task(&task_setup_data);
7724         if (IS_ERR(task)) {
7725                 status = PTR_ERR(task);
7726                 goto out;
7727         }
7728         status = nfs4_wait_for_completion_rpc_task(task);
7729         if (status == 0)
7730                 status = task->tk_status;
7731         rpc_put_task(task);
7732         return 0;
7733 out:
7734         dprintk("<-- %s status=%d\n", __func__, status);
7735         return status;
7736 }
7737
7738 static void
7739 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7740 {
7741         struct nfs4_layoutget *lgp = calldata;
7742         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7743         struct nfs4_session *session = nfs4_get_session(server);
7744
7745         dprintk("--> %s\n", __func__);
7746         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7747          * right now covering the LAYOUTGET we are about to send.
7748          * However, that is not so catastrophic, and there seems
7749          * to be no way to prevent it completely.
7750          */
7751         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7752                                 &lgp->res.seq_res, task))
7753                 return;
7754         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7755                                           NFS_I(lgp->args.inode)->layout,
7756                                           &lgp->args.range,
7757                                           lgp->args.ctx->state)) {
7758                 rpc_exit(task, NFS4_OK);
7759         }
7760 }
7761
7762 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7763 {
7764         struct nfs4_layoutget *lgp = calldata;
7765         struct inode *inode = lgp->args.inode;
7766         struct nfs_server *server = NFS_SERVER(inode);
7767         struct pnfs_layout_hdr *lo;
7768         struct nfs4_state *state = NULL;
7769         unsigned long timeo, now, giveup;
7770
7771         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7772
7773         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7774                 goto out;
7775
7776         switch (task->tk_status) {
7777         case 0:
7778                 goto out;
7779         /*
7780          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7781          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7782          */
7783         case -NFS4ERR_BADLAYOUT:
7784                 goto out_overflow;
7785         /*
7786          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7787          * (or clients) writing to the same RAID stripe except when
7788          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7789          */
7790         case -NFS4ERR_LAYOUTTRYLATER:
7791                 if (lgp->args.minlength == 0)
7792                         goto out_overflow;
7793         /*
7794          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7795          * existing layout before getting a new one).
7796          */
7797         case -NFS4ERR_RECALLCONFLICT:
7798                 timeo = rpc_get_timeout(task->tk_client);
7799                 giveup = lgp->args.timestamp + timeo;
7800                 now = jiffies;
7801                 if (time_after(giveup, now)) {
7802                         unsigned long delay;
7803
7804                         /* Delay for:
7805                          * - Not less then NFS4_POLL_RETRY_MIN.
7806                          * - One last time a jiffie before we give up
7807                          * - exponential backoff (time_now minus start_attempt)
7808                          */
7809                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7810                                     min((giveup - now - 1),
7811                                         now - lgp->args.timestamp));
7812
7813                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7814                                 __func__, delay);
7815                         rpc_delay(task, delay);
7816                         /* Do not call nfs4_async_handle_error() */
7817                         goto out_restart;
7818                 }
7819                 break;
7820         case -NFS4ERR_EXPIRED:
7821         case -NFS4ERR_BAD_STATEID:
7822                 spin_lock(&inode->i_lock);
7823                 if (nfs4_stateid_match(&lgp->args.stateid,
7824                                         &lgp->args.ctx->state->stateid)) {
7825                         spin_unlock(&inode->i_lock);
7826                         /* If the open stateid was bad, then recover it. */
7827                         state = lgp->args.ctx->state;
7828                         break;
7829                 }
7830                 lo = NFS_I(inode)->layout;
7831                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7832                                         &lo->plh_stateid)) {
7833                         LIST_HEAD(head);
7834
7835                         /*
7836                          * Mark the bad layout state as invalid, then retry
7837                          * with the current stateid.
7838                          */
7839                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7840                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7841                         spin_unlock(&inode->i_lock);
7842                         pnfs_free_lseg_list(&head);
7843                 } else
7844                         spin_unlock(&inode->i_lock);
7845                 goto out_restart;
7846         }
7847         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7848                 goto out_restart;
7849 out:
7850         dprintk("<-- %s\n", __func__);
7851         return;
7852 out_restart:
7853         task->tk_status = 0;
7854         rpc_restart_call_prepare(task);
7855         return;
7856 out_overflow:
7857         task->tk_status = -EOVERFLOW;
7858         goto out;
7859 }
7860
7861 static size_t max_response_pages(struct nfs_server *server)
7862 {
7863         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7864         return nfs_page_array_len(0, max_resp_sz);
7865 }
7866
7867 static void nfs4_free_pages(struct page **pages, size_t size)
7868 {
7869         int i;
7870
7871         if (!pages)
7872                 return;
7873
7874         for (i = 0; i < size; i++) {
7875                 if (!pages[i])
7876                         break;
7877                 __free_page(pages[i]);
7878         }
7879         kfree(pages);
7880 }
7881
7882 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7883 {
7884         struct page **pages;
7885         int i;
7886
7887         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7888         if (!pages) {
7889                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7890                 return NULL;
7891         }
7892
7893         for (i = 0; i < size; i++) {
7894                 pages[i] = alloc_page(gfp_flags);
7895                 if (!pages[i]) {
7896                         dprintk("%s: failed to allocate page\n", __func__);
7897                         nfs4_free_pages(pages, size);
7898                         return NULL;
7899                 }
7900         }
7901
7902         return pages;
7903 }
7904
7905 static void nfs4_layoutget_release(void *calldata)
7906 {
7907         struct nfs4_layoutget *lgp = calldata;
7908         struct inode *inode = lgp->args.inode;
7909         struct nfs_server *server = NFS_SERVER(inode);
7910         size_t max_pages = max_response_pages(server);
7911
7912         dprintk("--> %s\n", __func__);
7913         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7914         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7915         put_nfs_open_context(lgp->args.ctx);
7916         kfree(calldata);
7917         dprintk("<-- %s\n", __func__);
7918 }
7919
7920 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7921         .rpc_call_prepare = nfs4_layoutget_prepare,
7922         .rpc_call_done = nfs4_layoutget_done,
7923         .rpc_release = nfs4_layoutget_release,
7924 };
7925
7926 struct pnfs_layout_segment *
7927 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7928 {
7929         struct inode *inode = lgp->args.inode;
7930         struct nfs_server *server = NFS_SERVER(inode);
7931         size_t max_pages = max_response_pages(server);
7932         struct rpc_task *task;
7933         struct rpc_message msg = {
7934                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7935                 .rpc_argp = &lgp->args,
7936                 .rpc_resp = &lgp->res,
7937                 .rpc_cred = lgp->cred,
7938         };
7939         struct rpc_task_setup task_setup_data = {
7940                 .rpc_client = server->client,
7941                 .rpc_message = &msg,
7942                 .callback_ops = &nfs4_layoutget_call_ops,
7943                 .callback_data = lgp,
7944                 .flags = RPC_TASK_ASYNC,
7945         };
7946         struct pnfs_layout_segment *lseg = NULL;
7947         int status = 0;
7948
7949         dprintk("--> %s\n", __func__);
7950
7951         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7952         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7953
7954         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7955         if (!lgp->args.layout.pages) {
7956                 nfs4_layoutget_release(lgp);
7957                 return ERR_PTR(-ENOMEM);
7958         }
7959         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7960         lgp->args.timestamp = jiffies;
7961
7962         lgp->res.layoutp = &lgp->args.layout;
7963         lgp->res.seq_res.sr_slot = NULL;
7964         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7965
7966         task = rpc_run_task(&task_setup_data);
7967         if (IS_ERR(task))
7968                 return ERR_CAST(task);
7969         status = nfs4_wait_for_completion_rpc_task(task);
7970         if (status == 0)
7971                 status = task->tk_status;
7972         trace_nfs4_layoutget(lgp->args.ctx,
7973                         &lgp->args.range,
7974                         &lgp->res.range,
7975                         status);
7976         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7977         if (status == 0 && lgp->res.layoutp->len)
7978                 lseg = pnfs_layout_process(lgp);
7979         rpc_put_task(task);
7980         dprintk("<-- %s status=%d\n", __func__, status);
7981         if (status)
7982                 return ERR_PTR(status);
7983         return lseg;
7984 }
7985
7986 static void
7987 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7988 {
7989         struct nfs4_layoutreturn *lrp = calldata;
7990
7991         dprintk("--> %s\n", __func__);
7992         nfs41_setup_sequence(lrp->clp->cl_session,
7993                         &lrp->args.seq_args,
7994                         &lrp->res.seq_res,
7995                         task);
7996 }
7997
7998 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7999 {
8000         struct nfs4_layoutreturn *lrp = calldata;
8001         struct nfs_server *server;
8002
8003         dprintk("--> %s\n", __func__);
8004
8005         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8006                 return;
8007
8008         server = NFS_SERVER(lrp->args.inode);
8009         switch (task->tk_status) {
8010         default:
8011                 task->tk_status = 0;
8012         case 0:
8013                 break;
8014         case -NFS4ERR_DELAY:
8015                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8016                         break;
8017                 rpc_restart_call_prepare(task);
8018                 return;
8019         }
8020         dprintk("<-- %s\n", __func__);
8021 }
8022
8023 static void nfs4_layoutreturn_release(void *calldata)
8024 {
8025         struct nfs4_layoutreturn *lrp = calldata;
8026         struct pnfs_layout_hdr *lo = lrp->args.layout;
8027         LIST_HEAD(freeme);
8028
8029         dprintk("--> %s\n", __func__);
8030         spin_lock(&lo->plh_inode->i_lock);
8031         if (lrp->res.lrs_present)
8032                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8033         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8034         pnfs_clear_layoutreturn_waitbit(lo);
8035         lo->plh_block_lgets--;
8036         spin_unlock(&lo->plh_inode->i_lock);
8037         pnfs_free_lseg_list(&freeme);
8038         pnfs_put_layout_hdr(lrp->args.layout);
8039         nfs_iput_and_deactive(lrp->inode);
8040         kfree(calldata);
8041         dprintk("<-- %s\n", __func__);
8042 }
8043
8044 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8045         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8046         .rpc_call_done = nfs4_layoutreturn_done,
8047         .rpc_release = nfs4_layoutreturn_release,
8048 };
8049
8050 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8051 {
8052         struct rpc_task *task;
8053         struct rpc_message msg = {
8054                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8055                 .rpc_argp = &lrp->args,
8056                 .rpc_resp = &lrp->res,
8057                 .rpc_cred = lrp->cred,
8058         };
8059         struct rpc_task_setup task_setup_data = {
8060                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8061                 .rpc_message = &msg,
8062                 .callback_ops = &nfs4_layoutreturn_call_ops,
8063                 .callback_data = lrp,
8064         };
8065         int status = 0;
8066
8067         dprintk("--> %s\n", __func__);
8068         if (!sync) {
8069                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8070                 if (!lrp->inode) {
8071                         nfs4_layoutreturn_release(lrp);
8072                         return -EAGAIN;
8073                 }
8074                 task_setup_data.flags |= RPC_TASK_ASYNC;
8075         }
8076         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8077         task = rpc_run_task(&task_setup_data);
8078         if (IS_ERR(task))
8079                 return PTR_ERR(task);
8080         if (sync)
8081                 status = task->tk_status;
8082         trace_nfs4_layoutreturn(lrp->args.inode, status);
8083         dprintk("<-- %s status=%d\n", __func__, status);
8084         rpc_put_task(task);
8085         return status;
8086 }
8087
8088 static int
8089 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8090                 struct pnfs_device *pdev,
8091                 struct rpc_cred *cred)
8092 {
8093         struct nfs4_getdeviceinfo_args args = {
8094                 .pdev = pdev,
8095                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8096                         NOTIFY_DEVICEID4_DELETE,
8097         };
8098         struct nfs4_getdeviceinfo_res res = {
8099                 .pdev = pdev,
8100         };
8101         struct rpc_message msg = {
8102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8103                 .rpc_argp = &args,
8104                 .rpc_resp = &res,
8105                 .rpc_cred = cred,
8106         };
8107         int status;
8108
8109         dprintk("--> %s\n", __func__);
8110         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8111         if (res.notification & ~args.notify_types)
8112                 dprintk("%s: unsupported notification\n", __func__);
8113         if (res.notification != args.notify_types)
8114                 pdev->nocache = 1;
8115
8116         dprintk("<-- %s status=%d\n", __func__, status);
8117
8118         return status;
8119 }
8120
8121 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8122                 struct pnfs_device *pdev,
8123                 struct rpc_cred *cred)
8124 {
8125         struct nfs4_exception exception = { };
8126         int err;
8127
8128         do {
8129                 err = nfs4_handle_exception(server,
8130                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8131                                         &exception);
8132         } while (exception.retry);
8133         return err;
8134 }
8135 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8136
8137 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8138 {
8139         struct nfs4_layoutcommit_data *data = calldata;
8140         struct nfs_server *server = NFS_SERVER(data->args.inode);
8141         struct nfs4_session *session = nfs4_get_session(server);
8142
8143         nfs41_setup_sequence(session,
8144                         &data->args.seq_args,
8145                         &data->res.seq_res,
8146                         task);
8147 }
8148
8149 static void
8150 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8151 {
8152         struct nfs4_layoutcommit_data *data = calldata;
8153         struct nfs_server *server = NFS_SERVER(data->args.inode);
8154
8155         if (!nfs41_sequence_done(task, &data->res.seq_res))
8156                 return;
8157
8158         switch (task->tk_status) { /* Just ignore these failures */
8159         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8160         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8161         case -NFS4ERR_BADLAYOUT:     /* no layout */
8162         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8163                 task->tk_status = 0;
8164         case 0:
8165                 break;
8166         default:
8167                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8168                         rpc_restart_call_prepare(task);
8169                         return;
8170                 }
8171         }
8172 }
8173
8174 static void nfs4_layoutcommit_release(void *calldata)
8175 {
8176         struct nfs4_layoutcommit_data *data = calldata;
8177
8178         pnfs_cleanup_layoutcommit(data);
8179         nfs_post_op_update_inode_force_wcc(data->args.inode,
8180                                            data->res.fattr);
8181         put_rpccred(data->cred);
8182         nfs_iput_and_deactive(data->inode);
8183         kfree(data);
8184 }
8185
8186 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8187         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8188         .rpc_call_done = nfs4_layoutcommit_done,
8189         .rpc_release = nfs4_layoutcommit_release,
8190 };
8191
8192 int
8193 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8194 {
8195         struct rpc_message msg = {
8196                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8197                 .rpc_argp = &data->args,
8198                 .rpc_resp = &data->res,
8199                 .rpc_cred = data->cred,
8200         };
8201         struct rpc_task_setup task_setup_data = {
8202                 .task = &data->task,
8203                 .rpc_client = NFS_CLIENT(data->args.inode),
8204                 .rpc_message = &msg,
8205                 .callback_ops = &nfs4_layoutcommit_ops,
8206                 .callback_data = data,
8207         };
8208         struct rpc_task *task;
8209         int status = 0;
8210
8211         dprintk("NFS: initiating layoutcommit call. sync %d "
8212                 "lbw: %llu inode %lu\n", sync,
8213                 data->args.lastbytewritten,
8214                 data->args.inode->i_ino);
8215
8216         if (!sync) {
8217                 data->inode = nfs_igrab_and_active(data->args.inode);
8218                 if (data->inode == NULL) {
8219                         nfs4_layoutcommit_release(data);
8220                         return -EAGAIN;
8221                 }
8222                 task_setup_data.flags = RPC_TASK_ASYNC;
8223         }
8224         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8225         task = rpc_run_task(&task_setup_data);
8226         if (IS_ERR(task))
8227                 return PTR_ERR(task);
8228         if (sync)
8229                 status = task->tk_status;
8230         trace_nfs4_layoutcommit(data->args.inode, status);
8231         dprintk("%s: status %d\n", __func__, status);
8232         rpc_put_task(task);
8233         return status;
8234 }
8235
8236 /**
8237  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8238  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8239  */
8240 static int
8241 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8242                     struct nfs_fsinfo *info,
8243                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8244 {
8245         struct nfs41_secinfo_no_name_args args = {
8246                 .style = SECINFO_STYLE_CURRENT_FH,
8247         };
8248         struct nfs4_secinfo_res res = {
8249                 .flavors = flavors,
8250         };
8251         struct rpc_message msg = {
8252                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8253                 .rpc_argp = &args,
8254                 .rpc_resp = &res,
8255         };
8256         struct rpc_clnt *clnt = server->client;
8257         struct rpc_cred *cred = NULL;
8258         int status;
8259
8260         if (use_integrity) {
8261                 clnt = server->nfs_client->cl_rpcclient;
8262                 cred = nfs4_get_clid_cred(server->nfs_client);
8263                 msg.rpc_cred = cred;
8264         }
8265
8266         dprintk("--> %s\n", __func__);
8267         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8268                                 &res.seq_res, 0);
8269         dprintk("<-- %s status=%d\n", __func__, status);
8270
8271         if (cred)
8272                 put_rpccred(cred);
8273
8274         return status;
8275 }
8276
8277 static int
8278 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8279                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8280 {
8281         struct nfs4_exception exception = { };
8282         int err;
8283         do {
8284                 /* first try using integrity protection */
8285                 err = -NFS4ERR_WRONGSEC;
8286
8287                 /* try to use integrity protection with machine cred */
8288                 if (_nfs4_is_integrity_protected(server->nfs_client))
8289                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8290                                                           flavors, true);
8291
8292                 /*
8293                  * if unable to use integrity protection, or SECINFO with
8294                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8295                  * disallowed by spec, but exists in deployed servers) use
8296                  * the current filesystem's rpc_client and the user cred.
8297                  */
8298                 if (err == -NFS4ERR_WRONGSEC)
8299                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8300                                                           flavors, false);
8301
8302                 switch (err) {
8303                 case 0:
8304                 case -NFS4ERR_WRONGSEC:
8305                 case -ENOTSUPP:
8306                         goto out;
8307                 default:
8308                         err = nfs4_handle_exception(server, err, &exception);
8309                 }
8310         } while (exception.retry);
8311 out:
8312         return err;
8313 }
8314
8315 static int
8316 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8317                     struct nfs_fsinfo *info)
8318 {
8319         int err;
8320         struct page *page;
8321         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8322         struct nfs4_secinfo_flavors *flavors;
8323         struct nfs4_secinfo4 *secinfo;
8324         int i;
8325
8326         page = alloc_page(GFP_KERNEL);
8327         if (!page) {
8328                 err = -ENOMEM;
8329                 goto out;
8330         }
8331
8332         flavors = page_address(page);
8333         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8334
8335         /*
8336          * Fall back on "guess and check" method if
8337          * the server doesn't support SECINFO_NO_NAME
8338          */
8339         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8340                 err = nfs4_find_root_sec(server, fhandle, info);
8341                 goto out_freepage;
8342         }
8343         if (err)
8344                 goto out_freepage;
8345
8346         for (i = 0; i < flavors->num_flavors; i++) {
8347                 secinfo = &flavors->flavors[i];
8348
8349                 switch (secinfo->flavor) {
8350                 case RPC_AUTH_NULL:
8351                 case RPC_AUTH_UNIX:
8352                 case RPC_AUTH_GSS:
8353                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8354                                         &secinfo->flavor_info);
8355                         break;
8356                 default:
8357                         flavor = RPC_AUTH_MAXFLAVOR;
8358                         break;
8359                 }
8360
8361                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8362                         flavor = RPC_AUTH_MAXFLAVOR;
8363
8364                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8365                         err = nfs4_lookup_root_sec(server, fhandle,
8366                                                    info, flavor);
8367                         if (!err)
8368                                 break;
8369                 }
8370         }
8371
8372         if (flavor == RPC_AUTH_MAXFLAVOR)
8373                 err = -EPERM;
8374
8375 out_freepage:
8376         put_page(page);
8377         if (err == -EACCES)
8378                 return -EPERM;
8379 out:
8380         return err;
8381 }
8382
8383 static int _nfs41_test_stateid(struct nfs_server *server,
8384                 nfs4_stateid *stateid,
8385                 struct rpc_cred *cred)
8386 {
8387         int status;
8388         struct nfs41_test_stateid_args args = {
8389                 .stateid = stateid,
8390         };
8391         struct nfs41_test_stateid_res res;
8392         struct rpc_message msg = {
8393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8394                 .rpc_argp = &args,
8395                 .rpc_resp = &res,
8396                 .rpc_cred = cred,
8397         };
8398         struct rpc_clnt *rpc_client = server->client;
8399
8400         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8401                 &rpc_client, &msg);
8402
8403         dprintk("NFS call  test_stateid %p\n", stateid);
8404         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8405         nfs4_set_sequence_privileged(&args.seq_args);
8406         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8407                         &args.seq_args, &res.seq_res);
8408         if (status != NFS_OK) {
8409                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8410                 return status;
8411         }
8412         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8413         return -res.status;
8414 }
8415
8416 /**
8417  * nfs41_test_stateid - perform a TEST_STATEID operation
8418  *
8419  * @server: server / transport on which to perform the operation
8420  * @stateid: state ID to test
8421  * @cred: credential
8422  *
8423  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8424  * Otherwise a negative NFS4ERR value is returned if the operation
8425  * failed or the state ID is not currently valid.
8426  */
8427 static int nfs41_test_stateid(struct nfs_server *server,
8428                 nfs4_stateid *stateid,
8429                 struct rpc_cred *cred)
8430 {
8431         struct nfs4_exception exception = { };
8432         int err;
8433         do {
8434                 err = _nfs41_test_stateid(server, stateid, cred);
8435                 if (err != -NFS4ERR_DELAY)
8436                         break;
8437                 nfs4_handle_exception(server, err, &exception);
8438         } while (exception.retry);
8439         return err;
8440 }
8441
8442 struct nfs_free_stateid_data {
8443         struct nfs_server *server;
8444         struct nfs41_free_stateid_args args;
8445         struct nfs41_free_stateid_res res;
8446 };
8447
8448 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8449 {
8450         struct nfs_free_stateid_data *data = calldata;
8451         nfs41_setup_sequence(nfs4_get_session(data->server),
8452                         &data->args.seq_args,
8453                         &data->res.seq_res,
8454                         task);
8455 }
8456
8457 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8458 {
8459         struct nfs_free_stateid_data *data = calldata;
8460
8461         nfs41_sequence_done(task, &data->res.seq_res);
8462
8463         switch (task->tk_status) {
8464         case -NFS4ERR_DELAY:
8465                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8466                         rpc_restart_call_prepare(task);
8467         }
8468 }
8469
8470 static void nfs41_free_stateid_release(void *calldata)
8471 {
8472         kfree(calldata);
8473 }
8474
8475 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8476         .rpc_call_prepare = nfs41_free_stateid_prepare,
8477         .rpc_call_done = nfs41_free_stateid_done,
8478         .rpc_release = nfs41_free_stateid_release,
8479 };
8480
8481 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8482                 nfs4_stateid *stateid,
8483                 struct rpc_cred *cred,
8484                 bool privileged)
8485 {
8486         struct rpc_message msg = {
8487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8488                 .rpc_cred = cred,
8489         };
8490         struct rpc_task_setup task_setup = {
8491                 .rpc_client = server->client,
8492                 .rpc_message = &msg,
8493                 .callback_ops = &nfs41_free_stateid_ops,
8494                 .flags = RPC_TASK_ASYNC,
8495         };
8496         struct nfs_free_stateid_data *data;
8497
8498         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8499                 &task_setup.rpc_client, &msg);
8500
8501         dprintk("NFS call  free_stateid %p\n", stateid);
8502         data = kmalloc(sizeof(*data), GFP_NOFS);
8503         if (!data)
8504                 return ERR_PTR(-ENOMEM);
8505         data->server = server;
8506         nfs4_stateid_copy(&data->args.stateid, stateid);
8507
8508         task_setup.callback_data = data;
8509
8510         msg.rpc_argp = &data->args;
8511         msg.rpc_resp = &data->res;
8512         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8513         if (privileged)
8514                 nfs4_set_sequence_privileged(&data->args.seq_args);
8515
8516         return rpc_run_task(&task_setup);
8517 }
8518
8519 /**
8520  * nfs41_free_stateid - perform a FREE_STATEID operation
8521  *
8522  * @server: server / transport on which to perform the operation
8523  * @stateid: state ID to release
8524  * @cred: credential
8525  *
8526  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8527  * negative NFS4ERR value is returned.
8528  */
8529 static int nfs41_free_stateid(struct nfs_server *server,
8530                 nfs4_stateid *stateid,
8531                 struct rpc_cred *cred)
8532 {
8533         struct rpc_task *task;
8534         int ret;
8535
8536         task = _nfs41_free_stateid(server, stateid, cred, true);
8537         if (IS_ERR(task))
8538                 return PTR_ERR(task);
8539         ret = rpc_wait_for_completion_task(task);
8540         if (!ret)
8541                 ret = task->tk_status;
8542         rpc_put_task(task);
8543         return ret;
8544 }
8545
8546 static void
8547 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8548 {
8549         struct rpc_task *task;
8550         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8551
8552         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8553         nfs4_free_lock_state(server, lsp);
8554         if (IS_ERR(task))
8555                 return;
8556         rpc_put_task(task);
8557 }
8558
8559 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8560                 const nfs4_stateid *s2)
8561 {
8562         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8563                 return false;
8564
8565         if (s1->seqid == s2->seqid)
8566                 return true;
8567         if (s1->seqid == 0 || s2->seqid == 0)
8568                 return true;
8569
8570         return false;
8571 }
8572
8573 #endif /* CONFIG_NFS_V4_1 */
8574
8575 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8576                 const nfs4_stateid *s2)
8577 {
8578         return nfs4_stateid_match(s1, s2);
8579 }
8580
8581
8582 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8583         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8584         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8585         .recover_open   = nfs4_open_reclaim,
8586         .recover_lock   = nfs4_lock_reclaim,
8587         .establish_clid = nfs4_init_clientid,
8588         .detect_trunking = nfs40_discover_server_trunking,
8589 };
8590
8591 #if defined(CONFIG_NFS_V4_1)
8592 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8593         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8594         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8595         .recover_open   = nfs4_open_reclaim,
8596         .recover_lock   = nfs4_lock_reclaim,
8597         .establish_clid = nfs41_init_clientid,
8598         .reclaim_complete = nfs41_proc_reclaim_complete,
8599         .detect_trunking = nfs41_discover_server_trunking,
8600 };
8601 #endif /* CONFIG_NFS_V4_1 */
8602
8603 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8604         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8605         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8606         .recover_open   = nfs40_open_expired,
8607         .recover_lock   = nfs4_lock_expired,
8608         .establish_clid = nfs4_init_clientid,
8609 };
8610
8611 #if defined(CONFIG_NFS_V4_1)
8612 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8613         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8614         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8615         .recover_open   = nfs41_open_expired,
8616         .recover_lock   = nfs41_lock_expired,
8617         .establish_clid = nfs41_init_clientid,
8618 };
8619 #endif /* CONFIG_NFS_V4_1 */
8620
8621 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8622         .sched_state_renewal = nfs4_proc_async_renew,
8623         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8624         .renew_lease = nfs4_proc_renew,
8625 };
8626
8627 #if defined(CONFIG_NFS_V4_1)
8628 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8629         .sched_state_renewal = nfs41_proc_async_sequence,
8630         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8631         .renew_lease = nfs4_proc_sequence,
8632 };
8633 #endif
8634
8635 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8636         .get_locations = _nfs40_proc_get_locations,
8637         .fsid_present = _nfs40_proc_fsid_present,
8638 };
8639
8640 #if defined(CONFIG_NFS_V4_1)
8641 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8642         .get_locations = _nfs41_proc_get_locations,
8643         .fsid_present = _nfs41_proc_fsid_present,
8644 };
8645 #endif  /* CONFIG_NFS_V4_1 */
8646
8647 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8648         .minor_version = 0,
8649         .init_caps = NFS_CAP_READDIRPLUS
8650                 | NFS_CAP_ATOMIC_OPEN
8651                 | NFS_CAP_POSIX_LOCK,
8652         .init_client = nfs40_init_client,
8653         .shutdown_client = nfs40_shutdown_client,
8654         .match_stateid = nfs4_match_stateid,
8655         .find_root_sec = nfs4_find_root_sec,
8656         .free_lock_state = nfs4_release_lockowner,
8657         .alloc_seqid = nfs_alloc_seqid,
8658         .call_sync_ops = &nfs40_call_sync_ops,
8659         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8660         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8661         .state_renewal_ops = &nfs40_state_renewal_ops,
8662         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8663 };
8664
8665 #if defined(CONFIG_NFS_V4_1)
8666 static struct nfs_seqid *
8667 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8668 {
8669         return NULL;
8670 }
8671
8672 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8673         .minor_version = 1,
8674         .init_caps = NFS_CAP_READDIRPLUS
8675                 | NFS_CAP_ATOMIC_OPEN
8676                 | NFS_CAP_POSIX_LOCK
8677                 | NFS_CAP_STATEID_NFSV41
8678                 | NFS_CAP_ATOMIC_OPEN_V1,
8679         .init_client = nfs41_init_client,
8680         .shutdown_client = nfs41_shutdown_client,
8681         .match_stateid = nfs41_match_stateid,
8682         .find_root_sec = nfs41_find_root_sec,
8683         .free_lock_state = nfs41_free_lock_state,
8684         .alloc_seqid = nfs_alloc_no_seqid,
8685         .call_sync_ops = &nfs41_call_sync_ops,
8686         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8687         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8688         .state_renewal_ops = &nfs41_state_renewal_ops,
8689         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8690 };
8691 #endif
8692
8693 #if defined(CONFIG_NFS_V4_2)
8694 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8695         .minor_version = 2,
8696         .init_caps = NFS_CAP_READDIRPLUS
8697                 | NFS_CAP_ATOMIC_OPEN
8698                 | NFS_CAP_POSIX_LOCK
8699                 | NFS_CAP_STATEID_NFSV41
8700                 | NFS_CAP_ATOMIC_OPEN_V1
8701                 | NFS_CAP_ALLOCATE
8702                 | NFS_CAP_DEALLOCATE
8703                 | NFS_CAP_SEEK
8704                 | NFS_CAP_LAYOUTSTATS,
8705         .init_client = nfs41_init_client,
8706         .shutdown_client = nfs41_shutdown_client,
8707         .match_stateid = nfs41_match_stateid,
8708         .find_root_sec = nfs41_find_root_sec,
8709         .free_lock_state = nfs41_free_lock_state,
8710         .call_sync_ops = &nfs41_call_sync_ops,
8711         .alloc_seqid = nfs_alloc_no_seqid,
8712         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8713         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8714         .state_renewal_ops = &nfs41_state_renewal_ops,
8715         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8716 };
8717 #endif
8718
8719 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8720         [0] = &nfs_v4_0_minor_ops,
8721 #if defined(CONFIG_NFS_V4_1)
8722         [1] = &nfs_v4_1_minor_ops,
8723 #endif
8724 #if defined(CONFIG_NFS_V4_2)
8725         [2] = &nfs_v4_2_minor_ops,
8726 #endif
8727 };
8728
8729 static const struct inode_operations nfs4_dir_inode_operations = {
8730         .create         = nfs_create,
8731         .lookup         = nfs_lookup,
8732         .atomic_open    = nfs_atomic_open,
8733         .link           = nfs_link,
8734         .unlink         = nfs_unlink,
8735         .symlink        = nfs_symlink,
8736         .mkdir          = nfs_mkdir,
8737         .rmdir          = nfs_rmdir,
8738         .mknod          = nfs_mknod,
8739         .rename         = nfs_rename,
8740         .permission     = nfs_permission,
8741         .getattr        = nfs_getattr,
8742         .setattr        = nfs_setattr,
8743         .getxattr       = generic_getxattr,
8744         .setxattr       = generic_setxattr,
8745         .listxattr      = generic_listxattr,
8746         .removexattr    = generic_removexattr,
8747 };
8748
8749 static const struct inode_operations nfs4_file_inode_operations = {
8750         .permission     = nfs_permission,
8751         .getattr        = nfs_getattr,
8752         .setattr        = nfs_setattr,
8753         .getxattr       = generic_getxattr,
8754         .setxattr       = generic_setxattr,
8755         .listxattr      = generic_listxattr,
8756         .removexattr    = generic_removexattr,
8757 };
8758
8759 const struct nfs_rpc_ops nfs_v4_clientops = {
8760         .version        = 4,                    /* protocol version */
8761         .dentry_ops     = &nfs4_dentry_operations,
8762         .dir_inode_ops  = &nfs4_dir_inode_operations,
8763         .file_inode_ops = &nfs4_file_inode_operations,
8764         .file_ops       = &nfs4_file_operations,
8765         .getroot        = nfs4_proc_get_root,
8766         .submount       = nfs4_submount,
8767         .try_mount      = nfs4_try_mount,
8768         .getattr        = nfs4_proc_getattr,
8769         .setattr        = nfs4_proc_setattr,
8770         .lookup         = nfs4_proc_lookup,
8771         .access         = nfs4_proc_access,
8772         .readlink       = nfs4_proc_readlink,
8773         .create         = nfs4_proc_create,
8774         .remove         = nfs4_proc_remove,
8775         .unlink_setup   = nfs4_proc_unlink_setup,
8776         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8777         .unlink_done    = nfs4_proc_unlink_done,
8778         .rename_setup   = nfs4_proc_rename_setup,
8779         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8780         .rename_done    = nfs4_proc_rename_done,
8781         .link           = nfs4_proc_link,
8782         .symlink        = nfs4_proc_symlink,
8783         .mkdir          = nfs4_proc_mkdir,
8784         .rmdir          = nfs4_proc_remove,
8785         .readdir        = nfs4_proc_readdir,
8786         .mknod          = nfs4_proc_mknod,
8787         .statfs         = nfs4_proc_statfs,
8788         .fsinfo         = nfs4_proc_fsinfo,
8789         .pathconf       = nfs4_proc_pathconf,
8790         .set_capabilities = nfs4_server_capabilities,
8791         .decode_dirent  = nfs4_decode_dirent,
8792         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8793         .read_setup     = nfs4_proc_read_setup,
8794         .read_done      = nfs4_read_done,
8795         .write_setup    = nfs4_proc_write_setup,
8796         .write_done     = nfs4_write_done,
8797         .commit_setup   = nfs4_proc_commit_setup,
8798         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8799         .commit_done    = nfs4_commit_done,
8800         .lock           = nfs4_proc_lock,
8801         .clear_acl_cache = nfs4_zap_acl_attr,
8802         .close_context  = nfs4_close_context,
8803         .open_context   = nfs4_atomic_open,
8804         .have_delegation = nfs4_have_delegation,
8805         .return_delegation = nfs4_inode_return_delegation,
8806         .alloc_client   = nfs4_alloc_client,
8807         .init_client    = nfs4_init_client,
8808         .free_client    = nfs4_free_client,
8809         .create_server  = nfs4_create_server,
8810         .clone_server   = nfs_clone_server,
8811 };
8812
8813 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8814         .prefix = XATTR_NAME_NFSV4_ACL,
8815         .list   = nfs4_xattr_list_nfs4_acl,
8816         .get    = nfs4_xattr_get_nfs4_acl,
8817         .set    = nfs4_xattr_set_nfs4_acl,
8818 };
8819
8820 const struct xattr_handler *nfs4_xattr_handlers[] = {
8821         &nfs4_xattr_nfs4_acl_handler,
8822 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8823         &nfs4_xattr_nfs4_label_handler,
8824 #endif
8825         NULL
8826 };
8827
8828 /*
8829  * Local variables:
8830  *  c-basic-offset: 8
8831  * End:
8832  */