fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
  82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  86                             struct nfs_fattr *fattr, struct iattr *sattr,
  87                             struct nfs4_state *state, struct nfs4_label *ilabel,
  88                             struct nfs4_label *olabel);
  89 #ifdef CONFIG_NFS_V4_1
  90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  91                 struct rpc_cred *);
  92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  93                 struct rpc_cred *);
  94 #endif
  95
  96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  97 static inline struct nfs4_label *
  98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  99         struct iattr *sattr, struct nfs4_label *label)
 100 {
 101         int err;
 102
 103         if (label == NULL)
 104                 return NULL;
 105
 106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 107                 return NULL;
 108
 109         err = security_dentry_init_security(dentry, sattr->ia_mode,
 110                                 &dentry->d_name, (void **)&label->label, &label->len);
 111         if (err == 0)
 112                 return label;
 113
 114         return NULL;
 115 }
 116 static inline void
 117 nfs4_label_release_security(struct nfs4_label *label)
 118 {
 119         if (label)
 120                 security_release_secctx(label->label, label->len);
 121 }
 122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 123 {
 124         if (label)
 125                 return server->attr_bitmask;
 126
 127         return server->attr_bitmask_nl;
 128 }
 129 #else
 130 static inline struct nfs4_label *
 131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 132         struct iattr *sattr, struct nfs4_label *l)
 133 { return NULL; }
 134 static inline void
 135 nfs4_label_release_security(struct nfs4_label *label)
 136 { return; }
 137 static inline u32 *
 138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 139 { return server->attr_bitmask; }
 140 #endif
 141
 142 /* Prevent leaks of NFSv4 errors into userland */
 143 static int nfs4_map_errors(int err)
 144 {
 145         if (err >= -1000)
 146                 return err;
 147         switch (err) {
 148         case -NFS4ERR_RESOURCE:
 149         case -NFS4ERR_LAYOUTTRYLATER:
 150         case -NFS4ERR_RECALLCONFLICT:
 151                 return -EREMOTEIO;
 152         case -NFS4ERR_WRONGSEC:
 153         case -NFS4ERR_WRONG_CRED:
 154                 return -EPERM;
 155         case -NFS4ERR_BADOWNER:
 156         case -NFS4ERR_BADNAME:
 157                 return -EINVAL;
 158         case -NFS4ERR_SHARE_DENIED:
 159                 return -EACCES;
 160         case -NFS4ERR_MINOR_VERS_MISMATCH:
 161                 return -EPROTONOSUPPORT;
 162         case -NFS4ERR_FILE_OPEN:
 163                 return -EBUSY;
 164         default:
 165                 dprintk("%s could not handle NFSv4 error %d\n",
 166                                 __func__, -err);
 167                 break;
 168         }
 169         return -EIO;
 170 }
 171
 172 /*
 173  * This is our standard bitmap for GETATTR requests.
 174  */
 175 const u32 nfs4_fattr_bitmap[3] = {
 176         FATTR4_WORD0_TYPE
 177         | FATTR4_WORD0_CHANGE
 178         | FATTR4_WORD0_SIZE
 179         | FATTR4_WORD0_FSID
 180         | FATTR4_WORD0_FILEID,
 181         FATTR4_WORD1_MODE
 182         | FATTR4_WORD1_NUMLINKS
 183         | FATTR4_WORD1_OWNER
 184         | FATTR4_WORD1_OWNER_GROUP
 185         | FATTR4_WORD1_RAWDEV
 186         | FATTR4_WORD1_SPACE_USED
 187         | FATTR4_WORD1_TIME_ACCESS
 188         | FATTR4_WORD1_TIME_METADATA
 189         | FATTR4_WORD1_TIME_MODIFY
 190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 192         FATTR4_WORD2_SECURITY_LABEL
 193 #endif
 194 };
 195
 196 static const u32 nfs4_pnfs_open_bitmap[3] = {
 197         FATTR4_WORD0_TYPE
 198         | FATTR4_WORD0_CHANGE
 199         | FATTR4_WORD0_SIZE
 200         | FATTR4_WORD0_FSID
 201         | FATTR4_WORD0_FILEID,
 202         FATTR4_WORD1_MODE
 203         | FATTR4_WORD1_NUMLINKS
 204         | FATTR4_WORD1_OWNER
 205         | FATTR4_WORD1_OWNER_GROUP
 206         | FATTR4_WORD1_RAWDEV
 207         | FATTR4_WORD1_SPACE_USED
 208         | FATTR4_WORD1_TIME_ACCESS
 209         | FATTR4_WORD1_TIME_METADATA
 210         | FATTR4_WORD1_TIME_MODIFY,
 211         FATTR4_WORD2_MDSTHRESHOLD
 212 };
 213
 214 static const u32 nfs4_open_noattr_bitmap[3] = {
 215         FATTR4_WORD0_TYPE
 216         | FATTR4_WORD0_CHANGE
 217         | FATTR4_WORD0_FILEID,
 218 };
 219
 220 const u32 nfs4_statfs_bitmap[3] = {
 221         FATTR4_WORD0_FILES_AVAIL
 222         | FATTR4_WORD0_FILES_FREE
 223         | FATTR4_WORD0_FILES_TOTAL,
 224         FATTR4_WORD1_SPACE_AVAIL
 225         | FATTR4_WORD1_SPACE_FREE
 226         | FATTR4_WORD1_SPACE_TOTAL
 227 };
 228
 229 const u32 nfs4_pathconf_bitmap[3] = {
 230         FATTR4_WORD0_MAXLINK
 231         | FATTR4_WORD0_MAXNAME,
 232         0
 233 };
 234
 235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 236                         | FATTR4_WORD0_MAXREAD
 237                         | FATTR4_WORD0_MAXWRITE
 238                         | FATTR4_WORD0_LEASE_TIME,
 239                         FATTR4_WORD1_TIME_DELTA
 240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 241                         FATTR4_WORD2_LAYOUT_BLKSIZE
 242 };
 243
 244 const u32 nfs4_fs_locations_bitmap[3] = {
 245         FATTR4_WORD0_TYPE
 246         | FATTR4_WORD0_CHANGE
 247         | FATTR4_WORD0_SIZE
 248         | FATTR4_WORD0_FSID
 249         | FATTR4_WORD0_FILEID
 250         | FATTR4_WORD0_FS_LOCATIONS,
 251         FATTR4_WORD1_MODE
 252         | FATTR4_WORD1_NUMLINKS
 253         | FATTR4_WORD1_OWNER
 254         | FATTR4_WORD1_OWNER_GROUP
 255         | FATTR4_WORD1_RAWDEV
 256         | FATTR4_WORD1_SPACE_USED
 257         | FATTR4_WORD1_TIME_ACCESS
 258         | FATTR4_WORD1_TIME_METADATA
 259         | FATTR4_WORD1_TIME_MODIFY
 260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 261 };
 262
 263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 264                 struct nfs4_readdir_arg *readdir)
 265 {
 266         __be32 *start, *p;
 267
 268         if (cookie > 2) {
 269                 readdir->cookie = cookie;
 270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 271                 return;
 272         }
 273
 274         readdir->cookie = 0;
 275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 276         if (cookie == 2)
 277                 return;
 278
 279         /*
 280          * NFSv4 servers do not return entries for '.' and '..'
 281          * Therefore, we fake these entries here.  We let '.'
 282          * have cookie 0 and '..' have cookie 1.  Note that
 283          * when talking to the server, we always send cookie 0
 284          * instead of 1 or 2.
 285          */
 286         start = p = kmap_atomic(*readdir->pages);
 287
 288         if (cookie == 0) {
 289                 *p++ = xdr_one;                                  /* next */
 290                 *p++ = xdr_zero;                   /* cookie, first word */
 291                 *p++ = xdr_one;                   /* cookie, second word */
 292                 *p++ = xdr_one;                             /* entry len */
 293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 294                 p++;
 295                 *p++ = xdr_one;                         /* bitmap length */
 296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 297                 *p++ = htonl(8);              /* attribute buffer length */
 298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 299         }
 300
 301         *p++ = xdr_one;                                  /* next */
 302         *p++ = xdr_zero;                   /* cookie, first word */
 303         *p++ = xdr_two;                   /* cookie, second word */
 304         *p++ = xdr_two;                             /* entry len */
 305         memcpy(p, "..\0\0", 4);                         /* entry */
 306         p++;
 307         *p++ = xdr_one;                         /* bitmap length */
 308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309         *p++ = htonl(8);              /* attribute buffer length */
 310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 311
 312         readdir->pgbase = (char *)p - (char *)start;
 313         readdir->count -= readdir->pgbase;
 314         kunmap_atomic(start);
 315 }
 316
 317 static long nfs4_update_delay(long *timeout)
 318 {
 319         long ret;
 320         if (!timeout)
 321                 return NFS4_POLL_RETRY_MAX;
 322         if (*timeout <= 0)
 323                 *timeout = NFS4_POLL_RETRY_MIN;
 324         if (*timeout > NFS4_POLL_RETRY_MAX)
 325                 *timeout = NFS4_POLL_RETRY_MAX;
 326         ret = *timeout;
 327         *timeout <<= 1;
 328         return ret;
 329 }
 330
 331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 332 {
 333         int res = 0;
 334
 335         might_sleep();
 336
 337         freezable_schedule_timeout_killable_unsafe(
 338                 nfs4_update_delay(timeout));
 339         if (fatal_signal_pending(current))
 340                 res = -ERESTARTSYS;
 341         return res;
 342 }
 343
 344 /* This is the error handling routine for processes that are allowed
 345  * to sleep.
 346  */
 347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 348 {
 349         struct nfs_client *clp = server->nfs_client;
 350         struct nfs4_state *state = exception->state;
 351         struct inode *inode = exception->inode;
 352         int ret = errorcode;
 353
 354         exception->retry = 0;
 355         switch(errorcode) {
 356                 case 0:
 357                         return 0;
 358                 case -NFS4ERR_OPENMODE:
 359                 case -NFS4ERR_DELEG_REVOKED:
 360                 case -NFS4ERR_ADMIN_REVOKED:
 361                 case -NFS4ERR_BAD_STATEID:
 362                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 363                                 nfs4_inode_return_delegation(inode);
 364                                 exception->retry = 1;
 365                                 return 0;
 366                         }
 367                         if (state == NULL)
 368                                 break;
 369                         ret = nfs4_schedule_stateid_recovery(server, state);
 370                         if (ret < 0)
 371                                 break;
 372                         goto wait_on_recovery;
 373                 case -NFS4ERR_EXPIRED:
 374                         if (state != NULL) {
 375                                 ret = nfs4_schedule_stateid_recovery(server, state);
 376                                 if (ret < 0)
 377                                         break;
 378                         }
 379                 case -NFS4ERR_STALE_STATEID:
 380                 case -NFS4ERR_STALE_CLIENTID:
 381                         nfs4_schedule_lease_recovery(clp);
 382                         goto wait_on_recovery;
 383                 case -NFS4ERR_MOVED:
 384                         ret = nfs4_schedule_migration_recovery(server);
 385                         if (ret < 0)
 386                                 break;
 387                         goto wait_on_recovery;
 388                 case -NFS4ERR_LEASE_MOVED:
 389                         nfs4_schedule_lease_moved_recovery(clp);
 390                         goto wait_on_recovery;
 391 #if defined(CONFIG_NFS_V4_1)
 392                 case -NFS4ERR_BADSESSION:
 393                 case -NFS4ERR_BADSLOT:
 394                 case -NFS4ERR_BAD_HIGH_SLOT:
 395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 396                 case -NFS4ERR_DEADSESSION:
 397                 case -NFS4ERR_SEQ_FALSE_RETRY:
 398                 case -NFS4ERR_SEQ_MISORDERED:
 399                         dprintk("%s ERROR: %d Reset session\n", __func__,
 400                                 errorcode);
 401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 402                         goto wait_on_recovery;
 403 #endif /* defined(CONFIG_NFS_V4_1) */
 404                 case -NFS4ERR_FILE_OPEN:
 405                         if (exception->timeout > HZ) {
 406                                 /* We have retried a decent amount, time to
 407                                  * fail
 408                                  */
 409                                 ret = -EBUSY;
 410                                 break;
 411                         }
 412                 case -NFS4ERR_GRACE:
 413                 case -NFS4ERR_DELAY:
 414                         ret = nfs4_delay(server->client, &exception->timeout);
 415                         if (ret != 0)
 416                                 break;
 417                 case -NFS4ERR_RETRY_UNCACHED_REP:
 418                 case -NFS4ERR_OLD_STATEID:
 419                         exception->retry = 1;
 420                         break;
 421                 case -NFS4ERR_BADOWNER:
 422                         /* The following works around a Linux server bug! */
 423                 case -NFS4ERR_BADNAME:
 424                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 425                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 426                                 exception->retry = 1;
 427                                 printk(KERN_WARNING "NFS: v4 server %s "
 428                                                 "does not accept raw "
 429                                                 "uid/gids. "
 430                                                 "Reenabling the idmapper.\n",
 431                                                 server->nfs_client->cl_hostname);
 432                         }
 433         }
 434         /* We failed to handle the error */
 435         return nfs4_map_errors(ret);
 436 wait_on_recovery:
 437         ret = nfs4_wait_clnt_recover(clp);
 438         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 439                 return -EIO;
 440         if (ret == 0)
 441                 exception->retry = 1;
 442         return ret;
 443 }
 444
 445 /*
 446  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 447  * or 'false' otherwise.
 448  */
 449 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 450 {
 451         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 452
 453         if (flavor == RPC_AUTH_GSS_KRB5I ||
 454             flavor == RPC_AUTH_GSS_KRB5P)
 455                 return true;
 456
 457         return false;
 458 }
 459
 460 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 461 {
 462         spin_lock(&clp->cl_lock);
 463         if (time_before(clp->cl_last_renewal,timestamp))
 464                 clp->cl_last_renewal = timestamp;
 465         spin_unlock(&clp->cl_lock);
 466 }
 467
 468 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 469 {
 470         struct nfs_client *clp = server->nfs_client;
 471
 472         if (!nfs4_has_session(clp))
 473                 do_renew_lease(clp, timestamp);
 474 }
 475
 476 struct nfs4_call_sync_data {
 477         const struct nfs_server *seq_server;
 478         struct nfs4_sequence_args *seq_args;
 479         struct nfs4_sequence_res *seq_res;
 480 };
 481
 482 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 483                         struct nfs4_sequence_res *res, int cache_reply)
 484 {
 485         args->sa_slot = NULL;
 486         args->sa_cache_this = cache_reply;
 487         args->sa_privileged = 0;
 488
 489         res->sr_slot = NULL;
 490 }
 491
 492 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 493 {
 494         args->sa_privileged = 1;
 495 }
 496
 497 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 498                          struct nfs4_sequence_args *args,
 499                          struct nfs4_sequence_res *res,
 500                          struct rpc_task *task)
 501 {
 502         struct nfs4_slot *slot;
 503
 504         /* slot already allocated? */
 505         if (res->sr_slot != NULL)
 506                 goto out_start;
 507
 508         spin_lock(&tbl->slot_tbl_lock);
 509         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 510                 goto out_sleep;
 511
 512         slot = nfs4_alloc_slot(tbl);
 513         if (IS_ERR(slot)) {
 514                 if (slot == ERR_PTR(-ENOMEM))
 515                         task->tk_timeout = HZ >> 2;
 516                 goto out_sleep;
 517         }
 518         spin_unlock(&tbl->slot_tbl_lock);
 519
 520         args->sa_slot = slot;
 521         res->sr_slot = slot;
 522
 523 out_start:
 524         rpc_call_start(task);
 525         return 0;
 526
 527 out_sleep:
 528         if (args->sa_privileged)
 529                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 530                                 NULL, RPC_PRIORITY_PRIVILEGED);
 531         else
 532                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 533         spin_unlock(&tbl->slot_tbl_lock);
 534         return -EAGAIN;
 535 }
 536 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 537
 538 static int nfs40_sequence_done(struct rpc_task *task,
 539                                struct nfs4_sequence_res *res)
 540 {
 541         struct nfs4_slot *slot = res->sr_slot;
 542         struct nfs4_slot_table *tbl;
 543
 544         if (slot == NULL)
 545                 goto out;
 546
 547         tbl = slot->table;
 548         spin_lock(&tbl->slot_tbl_lock);
 549         if (!nfs41_wake_and_assign_slot(tbl, slot))
 550                 nfs4_free_slot(tbl, slot);
 551         spin_unlock(&tbl->slot_tbl_lock);
 552
 553         res->sr_slot = NULL;
 554 out:
 555         return 1;
 556 }
 557
 558 #if defined(CONFIG_NFS_V4_1)
 559
 560 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 561 {
 562         struct nfs4_session *session;
 563         struct nfs4_slot_table *tbl;
 564         struct nfs4_slot *slot = res->sr_slot;
 565         bool send_new_highest_used_slotid = false;
 566
 567         tbl = slot->table;
 568         session = tbl->session;
 569
 570         spin_lock(&tbl->slot_tbl_lock);
 571         /* Be nice to the server: try to ensure that the last transmitted
 572          * value for highest_user_slotid <= target_highest_slotid
 573          */
 574         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 575                 send_new_highest_used_slotid = true;
 576
 577         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 578                 send_new_highest_used_slotid = false;
 579                 goto out_unlock;
 580         }
 581         nfs4_free_slot(tbl, slot);
 582
 583         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 584                 send_new_highest_used_slotid = false;
 585 out_unlock:
 586         spin_unlock(&tbl->slot_tbl_lock);
 587         res->sr_slot = NULL;
 588         if (send_new_highest_used_slotid)
 589                 nfs41_notify_server(session->clp);
 590 }
 591
 592 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 593 {
 594         struct nfs4_session *session;
 595         struct nfs4_slot *slot = res->sr_slot;
 596         struct nfs_client *clp;
 597         bool interrupted = false;
 598         int ret = 1;
 599
 600         if (slot == NULL)
 601                 goto out_noaction;
 602         /* don't increment the sequence number if the task wasn't sent */
 603         if (!RPC_WAS_SENT(task))
 604                 goto out;
 605
 606         session = slot->table->session;
 607
 608         if (slot->interrupted) {
 609                 slot->interrupted = 0;
 610                 interrupted = true;
 611         }
 612
 613         trace_nfs4_sequence_done(session, res);
 614         /* Check the SEQUENCE operation status */
 615         switch (res->sr_status) {
 616         case 0:
 617                 /* Update the slot's sequence and clientid lease timer */
 618                 ++slot->seq_nr;
 619                 clp = session->clp;
 620                 do_renew_lease(clp, res->sr_timestamp);
 621                 /* Check sequence flags */
 622                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 623                 nfs41_update_target_slotid(slot->table, slot, res);
 624                 break;
 625         case 1:
 626                 /*
 627                  * sr_status remains 1 if an RPC level error occurred.
 628                  * The server may or may not have processed the sequence
 629                  * operation..
 630                  * Mark the slot as having hosted an interrupted RPC call.
 631                  */
 632                 slot->interrupted = 1;
 633                 goto out;
 634         case -NFS4ERR_DELAY:
 635                 /* The server detected a resend of the RPC call and
 636                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 637                  * of RFC5661.
 638                  */
 639                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 640                         __func__,
 641                         slot->slot_nr,
 642                         slot->seq_nr);
 643                 goto out_retry;
 644         case -NFS4ERR_BADSLOT:
 645                 /*
 646                  * The slot id we used was probably retired. Try again
 647                  * using a different slot id.
 648                  */
 649                 goto retry_nowait;
 650         case -NFS4ERR_SEQ_MISORDERED:
 651                 /*
 652                  * Was the last operation on this sequence interrupted?
 653                  * If so, retry after bumping the sequence number.
 654                  */
 655                 if (interrupted) {
 656                         ++slot->seq_nr;
 657                         goto retry_nowait;
 658                 }
 659                 /*
 660                  * Could this slot have been previously retired?
 661                  * If so, then the server may be expecting seq_nr = 1!
 662                  */
 663                 if (slot->seq_nr != 1) {
 664                         slot->seq_nr = 1;
 665                         goto retry_nowait;
 666                 }
 667                 break;
 668         case -NFS4ERR_SEQ_FALSE_RETRY:
 669                 ++slot->seq_nr;
 670                 goto retry_nowait;
 671         default:
 672                 /* Just update the slot sequence no. */
 673                 ++slot->seq_nr;
 674         }
 675 out:
 676         /* The session may be reset by one of the error handlers. */
 677         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 678         nfs41_sequence_free_slot(res);
 679 out_noaction:
 680         return ret;
 681 retry_nowait:
 682         if (rpc_restart_call_prepare(task)) {
 683                 task->tk_status = 0;
 684                 ret = 0;
 685         }
 686         goto out;
 687 out_retry:
 688         if (!rpc_restart_call(task))
 689                 goto out;
 690         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 691         return 0;
 692 }
 693 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 694
 695 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 696 {
 697         if (res->sr_slot == NULL)
 698                 return 1;
 699         if (!res->sr_slot->table->session)
 700                 return nfs40_sequence_done(task, res);
 701         return nfs41_sequence_done(task, res);
 702 }
 703 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 704
 705 int nfs41_setup_sequence(struct nfs4_session *session,
 706                                 struct nfs4_sequence_args *args,
 707                                 struct nfs4_sequence_res *res,
 708                                 struct rpc_task *task)
 709 {
 710         struct nfs4_slot *slot;
 711         struct nfs4_slot_table *tbl;
 712
 713         dprintk("--> %s\n", __func__);
 714         /* slot already allocated? */
 715         if (res->sr_slot != NULL)
 716                 goto out_success;
 717
 718         tbl = &session->fc_slot_table;
 719
 720         task->tk_timeout = 0;
 721
 722         spin_lock(&tbl->slot_tbl_lock);
 723         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 724             !args->sa_privileged) {
 725                 /* The state manager will wait until the slot table is empty */
 726                 dprintk("%s session is draining\n", __func__);
 727                 goto out_sleep;
 728         }
 729
 730         slot = nfs4_alloc_slot(tbl);
 731         if (IS_ERR(slot)) {
 732                 /* If out of memory, try again in 1/4 second */
 733                 if (slot == ERR_PTR(-ENOMEM))
 734                         task->tk_timeout = HZ >> 2;
 735                 dprintk("<-- %s: no free slots\n", __func__);
 736                 goto out_sleep;
 737         }
 738         spin_unlock(&tbl->slot_tbl_lock);
 739
 740         args->sa_slot = slot;
 741
 742         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 743                         slot->slot_nr, slot->seq_nr);
 744
 745         res->sr_slot = slot;
 746         res->sr_timestamp = jiffies;
 747         res->sr_status_flags = 0;
 748         /*
 749          * sr_status is only set in decode_sequence, and so will remain
 750          * set to 1 if an rpc level failure occurs.
 751          */
 752         res->sr_status = 1;
 753         trace_nfs4_setup_sequence(session, args);
 754 out_success:
 755         rpc_call_start(task);
 756         return 0;
 757 out_sleep:
 758         /* Privileged tasks are queued with top priority */
 759         if (args->sa_privileged)
 760                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 761                                 NULL, RPC_PRIORITY_PRIVILEGED);
 762         else
 763                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 764         spin_unlock(&tbl->slot_tbl_lock);
 765         return -EAGAIN;
 766 }
 767 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 768
 769 static int nfs4_setup_sequence(const struct nfs_server *server,
 770                                struct nfs4_sequence_args *args,
 771                                struct nfs4_sequence_res *res,
 772                                struct rpc_task *task)
 773 {
 774         struct nfs4_session *session = nfs4_get_session(server);
 775         int ret = 0;
 776
 777         if (!session)
 778                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 779                                             args, res, task);
 780
 781         dprintk("--> %s clp %p session %p sr_slot %u\n",
 782                 __func__, session->clp, session, res->sr_slot ?
 783                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 784
 785         ret = nfs41_setup_sequence(session, args, res, task);
 786
 787         dprintk("<-- %s status=%d\n", __func__, ret);
 788         return ret;
 789 }
 790
 791 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 792 {
 793         struct nfs4_call_sync_data *data = calldata;
 794         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 795
 796         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 797
 798         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 799 }
 800
 801 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 802 {
 803         struct nfs4_call_sync_data *data = calldata;
 804
 805         nfs41_sequence_done(task, data->seq_res);
 806 }
 807
 808 static const struct rpc_call_ops nfs41_call_sync_ops = {
 809         .rpc_call_prepare = nfs41_call_sync_prepare,
 810         .rpc_call_done = nfs41_call_sync_done,
 811 };
 812
 813 #else   /* !CONFIG_NFS_V4_1 */
 814
 815 static int nfs4_setup_sequence(const struct nfs_server *server,
 816                                struct nfs4_sequence_args *args,
 817                                struct nfs4_sequence_res *res,
 818                                struct rpc_task *task)
 819 {
 820         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 821                                     args, res, task);
 822 }
 823
 824 int nfs4_sequence_done(struct rpc_task *task,
 825                        struct nfs4_sequence_res *res)
 826 {
 827         return nfs40_sequence_done(task, res);
 828 }
 829 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 830
 831 #endif  /* !CONFIG_NFS_V4_1 */
 832
 833 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 834 {
 835         struct nfs4_call_sync_data *data = calldata;
 836         nfs4_setup_sequence(data->seq_server,
 837                                 data->seq_args, data->seq_res, task);
 838 }
 839
 840 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 841 {
 842         struct nfs4_call_sync_data *data = calldata;
 843         nfs4_sequence_done(task, data->seq_res);
 844 }
 845
 846 static const struct rpc_call_ops nfs40_call_sync_ops = {
 847         .rpc_call_prepare = nfs40_call_sync_prepare,
 848         .rpc_call_done = nfs40_call_sync_done,
 849 };
 850
 851 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 852                                    struct nfs_server *server,
 853                                    struct rpc_message *msg,
 854                                    struct nfs4_sequence_args *args,
 855                                    struct nfs4_sequence_res *res)
 856 {
 857         int ret;
 858         struct rpc_task *task;
 859         struct nfs_client *clp = server->nfs_client;
 860         struct nfs4_call_sync_data data = {
 861                 .seq_server = server,
 862                 .seq_args = args,
 863                 .seq_res = res,
 864         };
 865         struct rpc_task_setup task_setup = {
 866                 .rpc_client = clnt,
 867                 .rpc_message = msg,
 868                 .callback_ops = clp->cl_mvops->call_sync_ops,
 869                 .callback_data = &data
 870         };
 871
 872         task = rpc_run_task(&task_setup);
 873         if (IS_ERR(task))
 874                 ret = PTR_ERR(task);
 875         else {
 876                 ret = task->tk_status;
 877                 rpc_put_task(task);
 878         }
 879         return ret;
 880 }
 881
 882 int nfs4_call_sync(struct rpc_clnt *clnt,
 883                    struct nfs_server *server,
 884                    struct rpc_message *msg,
 885                    struct nfs4_sequence_args *args,
 886                    struct nfs4_sequence_res *res,
 887                    int cache_reply)
 888 {
 889         nfs4_init_sequence(args, res, cache_reply);
 890         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 891 }
 892
 893 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 894 {
 895         struct nfs_inode *nfsi = NFS_I(dir);
 896
 897         spin_lock(&dir->i_lock);
 898         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 899         if (!cinfo->atomic || cinfo->before != dir->i_version)
 900                 nfs_force_lookup_revalidate(dir);
 901         dir->i_version = cinfo->after;
 902         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 903         nfs_fscache_invalidate(dir);
 904         spin_unlock(&dir->i_lock);
 905 }
 906
 907 struct nfs4_opendata {
 908         struct kref kref;
 909         struct nfs_openargs o_arg;
 910         struct nfs_openres o_res;
 911         struct nfs_open_confirmargs c_arg;
 912         struct nfs_open_confirmres c_res;
 913         struct nfs4_string owner_name;
 914         struct nfs4_string group_name;
 915         struct nfs4_label *a_label;
 916         struct nfs_fattr f_attr;
 917         struct nfs4_label *f_label;
 918         struct dentry *dir;
 919         struct dentry *dentry;
 920         struct nfs4_state_owner *owner;
 921         struct nfs4_state *state;
 922         struct iattr attrs;
 923         unsigned long timestamp;
 924         unsigned int rpc_done : 1;
 925         unsigned int file_created : 1;
 926         unsigned int is_recover : 1;
 927         int rpc_status;
 928         int cancelled;
 929 };
 930
 931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 932                 int err, struct nfs4_exception *exception)
 933 {
 934         if (err != -EINVAL)
 935                 return false;
 936         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 937                 return false;
 938         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 939         exception->retry = 1;
 940         return true;
 941 }
 942
 943 static u32
 944 nfs4_map_atomic_open_share(struct nfs_server *server,
 945                 fmode_t fmode, int openflags)
 946 {
 947         u32 res = 0;
 948
 949         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
 950         case FMODE_READ:
 951                 res = NFS4_SHARE_ACCESS_READ;
 952                 break;
 953         case FMODE_WRITE:
 954                 res = NFS4_SHARE_ACCESS_WRITE;
 955                 break;
 956         case FMODE_READ|FMODE_WRITE:
 957                 res = NFS4_SHARE_ACCESS_BOTH;
 958         }
 959         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 960                 goto out;
 961         /* Want no delegation if we're using O_DIRECT */
 962         if (openflags & O_DIRECT)
 963                 res |= NFS4_SHARE_WANT_NO_DELEG;
 964 out:
 965         return res;
 966 }
 967
 968 static enum open_claim_type4
 969 nfs4_map_atomic_open_claim(struct nfs_server *server,
 970                 enum open_claim_type4 claim)
 971 {
 972         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 973                 return claim;
 974         switch (claim) {
 975         default:
 976                 return claim;
 977         case NFS4_OPEN_CLAIM_FH:
 978                 return NFS4_OPEN_CLAIM_NULL;
 979         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 980                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 981         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 982                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 983         }
 984 }
 985
 986 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 987 {
 988         p->o_res.f_attr = &p->f_attr;
 989         p->o_res.f_label = p->f_label;
 990         p->o_res.seqid = p->o_arg.seqid;
 991         p->c_res.seqid = p->c_arg.seqid;
 992         p->o_res.server = p->o_arg.server;
 993         p->o_res.access_request = p->o_arg.access;
 994         nfs_fattr_init(&p->f_attr);
 995         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 996 }
 997
 998 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 999                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1000                 const struct iattr *attrs,
1001                 struct nfs4_label *label,
1002                 enum open_claim_type4 claim,
1003                 gfp_t gfp_mask)
1004 {
1005         struct dentry *parent = dget_parent(dentry);
1006         struct inode *dir = d_inode(parent);
1007         struct nfs_server *server = NFS_SERVER(dir);
1008         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1009         struct nfs4_opendata *p;
1010
1011         p = kzalloc(sizeof(*p), gfp_mask);
1012         if (p == NULL)
1013                 goto err;
1014
1015         p->f_label = nfs4_label_alloc(server, gfp_mask);
1016         if (IS_ERR(p->f_label))
1017                 goto err_free_p;
1018
1019         p->a_label = nfs4_label_alloc(server, gfp_mask);
1020         if (IS_ERR(p->a_label))
1021                 goto err_free_f;
1022
1023         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1024         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1025         if (IS_ERR(p->o_arg.seqid))
1026                 goto err_free_label;
1027         nfs_sb_active(dentry->d_sb);
1028         p->dentry = dget(dentry);
1029         p->dir = parent;
1030         p->owner = sp;
1031         atomic_inc(&sp->so_count);
1032         p->o_arg.open_flags = flags;
1033         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1034         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1035                         fmode, flags);
1036         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1037          * will return permission denied for all bits until close */
1038         if (!(flags & O_EXCL)) {
1039                 /* ask server to check for all possible rights as results
1040                  * are cached */
1041                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1042                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1043         }
1044         p->o_arg.clientid = server->nfs_client->cl_clientid;
1045         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1046         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1047         p->o_arg.name = &dentry->d_name;
1048         p->o_arg.server = server;
1049         p->o_arg.bitmask = nfs4_bitmask(server, label);
1050         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1051         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1052         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1053         switch (p->o_arg.claim) {
1054         case NFS4_OPEN_CLAIM_NULL:
1055         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1056         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1057                 p->o_arg.fh = NFS_FH(dir);
1058                 break;
1059         case NFS4_OPEN_CLAIM_PREVIOUS:
1060         case NFS4_OPEN_CLAIM_FH:
1061         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1062         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1063                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1064         }
1065         if (attrs != NULL && attrs->ia_valid != 0) {
1066                 __u32 verf[2];
1067
1068                 p->o_arg.u.attrs = &p->attrs;
1069                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1070
1071                 verf[0] = jiffies;
1072                 verf[1] = current->pid;
1073                 memcpy(p->o_arg.u.verifier.data, verf,
1074                                 sizeof(p->o_arg.u.verifier.data));
1075         }
1076         p->c_arg.fh = &p->o_res.fh;
1077         p->c_arg.stateid = &p->o_res.stateid;
1078         p->c_arg.seqid = p->o_arg.seqid;
1079         nfs4_init_opendata_res(p);
1080         kref_init(&p->kref);
1081         return p;
1082
1083 err_free_label:
1084         nfs4_label_free(p->a_label);
1085 err_free_f:
1086         nfs4_label_free(p->f_label);
1087 err_free_p:
1088         kfree(p);
1089 err:
1090         dput(parent);
1091         return NULL;
1092 }
1093
1094 static void nfs4_opendata_free(struct kref *kref)
1095 {
1096         struct nfs4_opendata *p = container_of(kref,
1097                         struct nfs4_opendata, kref);
1098         struct super_block *sb = p->dentry->d_sb;
1099
1100         nfs_free_seqid(p->o_arg.seqid);
1101         if (p->state != NULL)
1102                 nfs4_put_open_state(p->state);
1103         nfs4_put_state_owner(p->owner);
1104
1105         nfs4_label_free(p->a_label);
1106         nfs4_label_free(p->f_label);
1107
1108         dput(p->dir);
1109         dput(p->dentry);
1110         nfs_sb_deactive(sb);
1111         nfs_fattr_free_names(&p->f_attr);
1112         kfree(p->f_attr.mdsthreshold);
1113         kfree(p);
1114 }
1115
1116 static void nfs4_opendata_put(struct nfs4_opendata *p)
1117 {
1118         if (p != NULL)
1119                 kref_put(&p->kref, nfs4_opendata_free);
1120 }
1121
1122 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1123 {
1124         int ret;
1125
1126         ret = rpc_wait_for_completion_task(task);
1127         return ret;
1128 }
1129
1130 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1131 {
1132         int ret = 0;
1133
1134         if (open_mode & (O_EXCL|O_TRUNC))
1135                 goto out;
1136         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1137                 case FMODE_READ:
1138                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1139                                 && state->n_rdonly != 0;
1140                         break;
1141                 case FMODE_WRITE:
1142                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1143                                 && state->n_wronly != 0;
1144                         break;
1145                 case FMODE_READ|FMODE_WRITE:
1146                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1147                                 && state->n_rdwr != 0;
1148         }
1149 out:
1150         return ret;
1151 }
1152
1153 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1154                 enum open_claim_type4 claim)
1155 {
1156         if (delegation == NULL)
1157                 return 0;
1158         if ((delegation->type & fmode) != fmode)
1159                 return 0;
1160         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1161                 return 0;
1162         switch (claim) {
1163         case NFS4_OPEN_CLAIM_NULL:
1164         case NFS4_OPEN_CLAIM_FH:
1165                 break;
1166         case NFS4_OPEN_CLAIM_PREVIOUS:
1167                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1168                         break;
1169         default:
1170                 return 0;
1171         }
1172         nfs_mark_delegation_referenced(delegation);
1173         return 1;
1174 }
1175
1176 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1177 {
1178         switch (fmode) {
1179                 case FMODE_WRITE:
1180                         state->n_wronly++;
1181                         break;
1182                 case FMODE_READ:
1183                         state->n_rdonly++;
1184                         break;
1185                 case FMODE_READ|FMODE_WRITE:
1186                         state->n_rdwr++;
1187         }
1188         nfs4_state_set_mode_locked(state, state->state | fmode);
1189 }
1190
1191 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1192 {
1193         struct nfs_client *clp = state->owner->so_server->nfs_client;
1194         bool need_recover = false;
1195
1196         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1197                 need_recover = true;
1198         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1199                 need_recover = true;
1200         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1201                 need_recover = true;
1202         if (need_recover)
1203                 nfs4_state_mark_reclaim_nograce(clp, state);
1204 }
1205
1206 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1207                 nfs4_stateid *stateid)
1208 {
1209         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1210                 return true;
1211         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1212                 nfs_test_and_clear_all_open_stateid(state);
1213                 return true;
1214         }
1215         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1216                 return true;
1217         return false;
1218 }
1219
1220 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1221 {
1222         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1223                 return;
1224         if (state->n_wronly)
1225                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1226         if (state->n_rdonly)
1227                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1228         if (state->n_rdwr)
1229                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1230         set_bit(NFS_OPEN_STATE, &state->flags);
1231 }
1232
1233 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1234                 nfs4_stateid *stateid, fmode_t fmode)
1235 {
1236         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1237         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1238         case FMODE_WRITE:
1239                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1240                 break;
1241         case FMODE_READ:
1242                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1243                 break;
1244         case 0:
1245                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1246                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1247                 clear_bit(NFS_OPEN_STATE, &state->flags);
1248         }
1249         if (stateid == NULL)
1250                 return;
1251         /* Handle races with OPEN */
1252         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1253             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1254                 nfs_resync_open_stateid_locked(state);
1255                 return;
1256         }
1257         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1258                 nfs4_stateid_copy(&state->stateid, stateid);
1259         nfs4_stateid_copy(&state->open_stateid, stateid);
1260 }
1261
1262 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1263 {
1264         write_seqlock(&state->seqlock);
1265         nfs_clear_open_stateid_locked(state, stateid, fmode);
1266         write_sequnlock(&state->seqlock);
1267         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1268                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1269 }
1270
1271 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1272 {
1273         switch (fmode) {
1274                 case FMODE_READ:
1275                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1276                         break;
1277                 case FMODE_WRITE:
1278                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1279                         break;
1280                 case FMODE_READ|FMODE_WRITE:
1281                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1282         }
1283         if (!nfs_need_update_open_stateid(state, stateid))
1284                 return;
1285         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1286                 nfs4_stateid_copy(&state->stateid, stateid);
1287         nfs4_stateid_copy(&state->open_stateid, stateid);
1288 }
1289
1290 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1291 {
1292         /*
1293          * Protect the call to nfs4_state_set_mode_locked and
1294          * serialise the stateid update
1295          */
1296         write_seqlock(&state->seqlock);
1297         if (deleg_stateid != NULL) {
1298                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1299                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1300         }
1301         if (open_stateid != NULL)
1302                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1303         write_sequnlock(&state->seqlock);
1304         spin_lock(&state->owner->so_lock);
1305         update_open_stateflags(state, fmode);
1306         spin_unlock(&state->owner->so_lock);
1307 }
1308
1309 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1310 {
1311         struct nfs_inode *nfsi = NFS_I(state->inode);
1312         struct nfs_delegation *deleg_cur;
1313         int ret = 0;
1314
1315         fmode &= (FMODE_READ|FMODE_WRITE);
1316
1317         rcu_read_lock();
1318         deleg_cur = rcu_dereference(nfsi->delegation);
1319         if (deleg_cur == NULL)
1320                 goto no_delegation;
1321
1322         spin_lock(&deleg_cur->lock);
1323         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1324            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1325             (deleg_cur->type & fmode) != fmode)
1326                 goto no_delegation_unlock;
1327
1328         if (delegation == NULL)
1329                 delegation = &deleg_cur->stateid;
1330         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1331                 goto no_delegation_unlock;
1332
1333         nfs_mark_delegation_referenced(deleg_cur);
1334         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1335         ret = 1;
1336 no_delegation_unlock:
1337         spin_unlock(&deleg_cur->lock);
1338 no_delegation:
1339         rcu_read_unlock();
1340
1341         if (!ret && open_stateid != NULL) {
1342                 __update_open_stateid(state, open_stateid, NULL, fmode);
1343                 ret = 1;
1344         }
1345         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1346                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1347
1348         return ret;
1349 }
1350
1351 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1352                 const nfs4_stateid *stateid)
1353 {
1354         struct nfs4_state *state = lsp->ls_state;
1355         bool ret = false;
1356
1357         spin_lock(&state->state_lock);
1358         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1359                 goto out_noupdate;
1360         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1361                 goto out_noupdate;
1362         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1363         ret = true;
1364 out_noupdate:
1365         spin_unlock(&state->state_lock);
1366         return ret;
1367 }
1368
1369 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1370 {
1371         struct nfs_delegation *delegation;
1372
1373         rcu_read_lock();
1374         delegation = rcu_dereference(NFS_I(inode)->delegation);
1375         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1376                 rcu_read_unlock();
1377                 return;
1378         }
1379         rcu_read_unlock();
1380         nfs4_inode_return_delegation(inode);
1381 }
1382
1383 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1384 {
1385         struct nfs4_state *state = opendata->state;
1386         struct nfs_inode *nfsi = NFS_I(state->inode);
1387         struct nfs_delegation *delegation;
1388         int open_mode = opendata->o_arg.open_flags;
1389         fmode_t fmode = opendata->o_arg.fmode;
1390         enum open_claim_type4 claim = opendata->o_arg.claim;
1391         nfs4_stateid stateid;
1392         int ret = -EAGAIN;
1393
1394         for (;;) {
1395                 spin_lock(&state->owner->so_lock);
1396                 if (can_open_cached(state, fmode, open_mode)) {
1397                         update_open_stateflags(state, fmode);
1398                         spin_unlock(&state->owner->so_lock);
1399                         goto out_return_state;
1400                 }
1401                 spin_unlock(&state->owner->so_lock);
1402                 rcu_read_lock();
1403                 delegation = rcu_dereference(nfsi->delegation);
1404                 if (!can_open_delegated(delegation, fmode, claim)) {
1405                         rcu_read_unlock();
1406                         break;
1407                 }
1408                 /* Save the delegation */
1409                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1410                 rcu_read_unlock();
1411                 nfs_release_seqid(opendata->o_arg.seqid);
1412                 if (!opendata->is_recover) {
1413                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1414                         if (ret != 0)
1415                                 goto out;
1416                 }
1417                 ret = -EAGAIN;
1418
1419                 /* Try to update the stateid using the delegation */
1420                 if (update_open_stateid(state, NULL, &stateid, fmode))
1421                         goto out_return_state;
1422         }
1423 out:
1424         return ERR_PTR(ret);
1425 out_return_state:
1426         atomic_inc(&state->count);
1427         return state;
1428 }
1429
1430 static void
1431 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1432 {
1433         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1434         struct nfs_delegation *delegation;
1435         int delegation_flags = 0;
1436
1437         rcu_read_lock();
1438         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1439         if (delegation)
1440                 delegation_flags = delegation->flags;
1441         rcu_read_unlock();
1442         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1443                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1444                                    "returning a delegation for "
1445                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1446                                    clp->cl_hostname);
1447         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1448                 nfs_inode_set_delegation(state->inode,
1449                                          data->owner->so_cred,
1450                                          &data->o_res);
1451         else
1452                 nfs_inode_reclaim_delegation(state->inode,
1453                                              data->owner->so_cred,
1454                                              &data->o_res);
1455 }
1456
1457 /*
1458  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1459  * and update the nfs4_state.
1460  */
1461 static struct nfs4_state *
1462 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1463 {
1464         struct inode *inode = data->state->inode;
1465         struct nfs4_state *state = data->state;
1466         int ret;
1467
1468         if (!data->rpc_done) {
1469                 if (data->rpc_status) {
1470                         ret = data->rpc_status;
1471                         goto err;
1472                 }
1473                 /* cached opens have already been processed */
1474                 goto update;
1475         }
1476
1477         ret = nfs_refresh_inode(inode, &data->f_attr);
1478         if (ret)
1479                 goto err;
1480
1481         if (data->o_res.delegation_type != 0)
1482                 nfs4_opendata_check_deleg(data, state);
1483 update:
1484         update_open_stateid(state, &data->o_res.stateid, NULL,
1485                             data->o_arg.fmode);
1486         atomic_inc(&state->count);
1487
1488         return state;
1489 err:
1490         return ERR_PTR(ret);
1491
1492 }
1493
1494 static struct nfs4_state *
1495 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1496 {
1497         struct inode *inode;
1498         struct nfs4_state *state = NULL;
1499         int ret;
1500
1501         if (!data->rpc_done) {
1502                 state = nfs4_try_open_cached(data);
1503                 goto out;
1504         }
1505
1506         ret = -EAGAIN;
1507         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1508                 goto err;
1509         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1510         ret = PTR_ERR(inode);
1511         if (IS_ERR(inode))
1512                 goto err;
1513         ret = -ENOMEM;
1514         state = nfs4_get_open_state(inode, data->owner);
1515         if (state == NULL)
1516                 goto err_put_inode;
1517         if (data->o_res.delegation_type != 0)
1518                 nfs4_opendata_check_deleg(data, state);
1519         update_open_stateid(state, &data->o_res.stateid, NULL,
1520                         data->o_arg.fmode);
1521         iput(inode);
1522 out:
1523         nfs_release_seqid(data->o_arg.seqid);
1524         return state;
1525 err_put_inode:
1526         iput(inode);
1527 err:
1528         return ERR_PTR(ret);
1529 }
1530
1531 static struct nfs4_state *
1532 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1533 {
1534         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1535                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1536         return _nfs4_opendata_to_nfs4_state(data);
1537 }
1538
1539 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1540 {
1541         struct nfs_inode *nfsi = NFS_I(state->inode);
1542         struct nfs_open_context *ctx;
1543
1544         spin_lock(&state->inode->i_lock);
1545         list_for_each_entry(ctx, &nfsi->open_files, list) {
1546                 if (ctx->state != state)
1547                         continue;
1548                 get_nfs_open_context(ctx);
1549                 spin_unlock(&state->inode->i_lock);
1550                 return ctx;
1551         }
1552         spin_unlock(&state->inode->i_lock);
1553         return ERR_PTR(-ENOENT);
1554 }
1555
1556 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1557                 struct nfs4_state *state, enum open_claim_type4 claim)
1558 {
1559         struct nfs4_opendata *opendata;
1560
1561         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1562                         NULL, NULL, claim, GFP_NOFS);
1563         if (opendata == NULL)
1564                 return ERR_PTR(-ENOMEM);
1565         opendata->state = state;
1566         atomic_inc(&state->count);
1567         return opendata;
1568 }
1569
1570 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1571 {
1572         struct nfs4_state *newstate;
1573         int ret;
1574
1575         if ((opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
1576              opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEG_CUR_FH) &&
1577             (opendata->o_arg.u.delegation_type & fmode) != fmode)
1578                 /* This mode can't have been delegated, so we must have
1579                  * a valid open_stateid to cover it - not need to reclaim.
1580                  */
1581                 return 0;
1582         opendata->o_arg.open_flags = 0;
1583         opendata->o_arg.fmode = fmode;
1584         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1585                         NFS_SB(opendata->dentry->d_sb),
1586                         fmode, 0);
1587         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1588         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1589         nfs4_init_opendata_res(opendata);
1590         ret = _nfs4_recover_proc_open(opendata);
1591         if (ret != 0)
1592                 return ret;
1593         newstate = nfs4_opendata_to_nfs4_state(opendata);
1594         if (IS_ERR(newstate))
1595                 return PTR_ERR(newstate);
1596         nfs4_close_state(newstate, fmode);
1597         *res = newstate;
1598         return 0;
1599 }
1600
1601 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1602 {
1603         struct nfs4_state *newstate;
1604         int ret;
1605
1606         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1607         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1608         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1609         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1610         /* memory barrier prior to reading state->n_* */
1611         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1612         clear_bit(NFS_OPEN_STATE, &state->flags);
1613         smp_rmb();
1614         if (state->n_rdwr != 0) {
1615                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1616                 if (ret != 0)
1617                         return ret;
1618                 if (newstate != state)
1619                         return -ESTALE;
1620         }
1621         if (state->n_wronly != 0) {
1622                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1623                 if (ret != 0)
1624                         return ret;
1625                 if (newstate != state)
1626                         return -ESTALE;
1627         }
1628         if (state->n_rdonly != 0) {
1629                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1630                 if (ret != 0)
1631                         return ret;
1632                 if (newstate != state)
1633                         return -ESTALE;
1634         }
1635         /*
1636          * We may have performed cached opens for all three recoveries.
1637          * Check if we need to update the current stateid.
1638          */
1639         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1640             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1641                 write_seqlock(&state->seqlock);
1642                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1643                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1644                 write_sequnlock(&state->seqlock);
1645         }
1646         return 0;
1647 }
1648
1649 /*
1650  * OPEN_RECLAIM:
1651  *      reclaim state on the server after a reboot.
1652  */
1653 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1654 {
1655         struct nfs_delegation *delegation;
1656         struct nfs4_opendata *opendata;
1657         fmode_t delegation_type = 0;
1658         int status;
1659
1660         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1661                         NFS4_OPEN_CLAIM_PREVIOUS);
1662         if (IS_ERR(opendata))
1663                 return PTR_ERR(opendata);
1664         rcu_read_lock();
1665         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1666         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1667                 delegation_type = delegation->type;
1668         rcu_read_unlock();
1669         opendata->o_arg.u.delegation_type = delegation_type;
1670         status = nfs4_open_recover(opendata, state);
1671         nfs4_opendata_put(opendata);
1672         return status;
1673 }
1674
1675 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1676 {
1677         struct nfs_server *server = NFS_SERVER(state->inode);
1678         struct nfs4_exception exception = { };
1679         int err;
1680         do {
1681                 err = _nfs4_do_open_reclaim(ctx, state);
1682                 trace_nfs4_open_reclaim(ctx, 0, err);
1683                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1684                         continue;
1685                 if (err != -NFS4ERR_DELAY)
1686                         break;
1687                 nfs4_handle_exception(server, err, &exception);
1688         } while (exception.retry);
1689         return err;
1690 }
1691
1692 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1693 {
1694         struct nfs_open_context *ctx;
1695         int ret;
1696
1697         ctx = nfs4_state_find_open_context(state);
1698         if (IS_ERR(ctx))
1699                 return -EAGAIN;
1700         ret = nfs4_do_open_reclaim(ctx, state);
1701         put_nfs_open_context(ctx);
1702         return ret;
1703 }
1704
1705 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1706 {
1707         switch (err) {
1708                 default:
1709                         printk(KERN_ERR "NFS: %s: unhandled error "
1710                                         "%d.\n", __func__, err);
1711                 case 0:
1712                 case -ENOENT:
1713                 case -EAGAIN:
1714                 case -ESTALE:
1715                         break;
1716                 case -NFS4ERR_BADSESSION:
1717                 case -NFS4ERR_BADSLOT:
1718                 case -NFS4ERR_BAD_HIGH_SLOT:
1719                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1720                 case -NFS4ERR_DEADSESSION:
1721                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1722                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1723                         return -EAGAIN;
1724                 case -NFS4ERR_STALE_CLIENTID:
1725                 case -NFS4ERR_STALE_STATEID:
1726                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1727                 case -NFS4ERR_EXPIRED:
1728                         /* Don't recall a delegation if it was lost */
1729                         nfs4_schedule_lease_recovery(server->nfs_client);
1730                         return -EAGAIN;
1731                 case -NFS4ERR_MOVED:
1732                         nfs4_schedule_migration_recovery(server);
1733                         return -EAGAIN;
1734                 case -NFS4ERR_LEASE_MOVED:
1735                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1736                         return -EAGAIN;
1737                 case -NFS4ERR_DELEG_REVOKED:
1738                 case -NFS4ERR_ADMIN_REVOKED:
1739                 case -NFS4ERR_BAD_STATEID:
1740                 case -NFS4ERR_OPENMODE:
1741                         nfs_inode_find_state_and_recover(state->inode,
1742                                         stateid);
1743                         nfs4_schedule_stateid_recovery(server, state);
1744                         return -EAGAIN;
1745                 case -NFS4ERR_DELAY:
1746                 case -NFS4ERR_GRACE:
1747                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1748                         ssleep(1);
1749                         return -EAGAIN;
1750                 case -ENOMEM:
1751                 case -NFS4ERR_DENIED:
1752                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1753                         return 0;
1754         }
1755         return err;
1756 }
1757
1758 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1759 {
1760         struct nfs_server *server = NFS_SERVER(state->inode);
1761         struct nfs4_opendata *opendata;
1762         int err;
1763
1764         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1765                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1766         if (IS_ERR(opendata))
1767                 return PTR_ERR(opendata);
1768         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1769         err = nfs4_open_recover(opendata, state);
1770         nfs4_opendata_put(opendata);
1771         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1772 }
1773
1774 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1775 {
1776         struct nfs4_opendata *data = calldata;
1777
1778         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1779                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1780 }
1781
1782 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1783 {
1784         struct nfs4_opendata *data = calldata;
1785
1786         nfs40_sequence_done(task, &data->c_res.seq_res);
1787
1788         data->rpc_status = task->tk_status;
1789         if (data->rpc_status == 0) {
1790                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1791                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1792                 renew_lease(data->o_res.server, data->timestamp);
1793                 data->rpc_done = 1;
1794         }
1795 }
1796
1797 static void nfs4_open_confirm_release(void *calldata)
1798 {
1799         struct nfs4_opendata *data = calldata;
1800         struct nfs4_state *state = NULL;
1801
1802         /* If this request hasn't been cancelled, do nothing */
1803         if (data->cancelled == 0)
1804                 goto out_free;
1805         /* In case of error, no cleanup! */
1806         if (!data->rpc_done)
1807                 goto out_free;
1808         state = nfs4_opendata_to_nfs4_state(data);
1809         if (!IS_ERR(state))
1810                 nfs4_close_state(state, data->o_arg.fmode);
1811 out_free:
1812         nfs4_opendata_put(data);
1813 }
1814
1815 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1816         .rpc_call_prepare = nfs4_open_confirm_prepare,
1817         .rpc_call_done = nfs4_open_confirm_done,
1818         .rpc_release = nfs4_open_confirm_release,
1819 };
1820
1821 /*
1822  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1823  */
1824 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1825 {
1826         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1827         struct rpc_task *task;
1828         struct  rpc_message msg = {
1829                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1830                 .rpc_argp = &data->c_arg,
1831                 .rpc_resp = &data->c_res,
1832                 .rpc_cred = data->owner->so_cred,
1833         };
1834         struct rpc_task_setup task_setup_data = {
1835                 .rpc_client = server->client,
1836                 .rpc_message = &msg,
1837                 .callback_ops = &nfs4_open_confirm_ops,
1838                 .callback_data = data,
1839                 .workqueue = nfsiod_workqueue,
1840                 .flags = RPC_TASK_ASYNC,
1841         };
1842         int status;
1843
1844         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1845         kref_get(&data->kref);
1846         data->rpc_done = 0;
1847         data->rpc_status = 0;
1848         data->timestamp = jiffies;
1849         task = rpc_run_task(&task_setup_data);
1850         if (IS_ERR(task))
1851                 return PTR_ERR(task);
1852         status = nfs4_wait_for_completion_rpc_task(task);
1853         if (status != 0) {
1854                 data->cancelled = 1;
1855                 smp_wmb();
1856         } else
1857                 status = data->rpc_status;
1858         rpc_put_task(task);
1859         return status;
1860 }
1861
1862 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1863 {
1864         struct nfs4_opendata *data = calldata;
1865         struct nfs4_state_owner *sp = data->owner;
1866         struct nfs_client *clp = sp->so_server->nfs_client;
1867         enum open_claim_type4 claim = data->o_arg.claim;
1868
1869         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1870                 goto out_wait;
1871         /*
1872          * Check if we still need to send an OPEN call, or if we can use
1873          * a delegation instead.
1874          */
1875         if (data->state != NULL) {
1876                 struct nfs_delegation *delegation;
1877
1878                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1879                         goto out_no_action;
1880                 rcu_read_lock();
1881                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1882                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1883                         goto unlock_no_action;
1884                 rcu_read_unlock();
1885         }
1886         /* Update client id. */
1887         data->o_arg.clientid = clp->cl_clientid;
1888         switch (claim) {
1889         default:
1890                 break;
1891         case NFS4_OPEN_CLAIM_PREVIOUS:
1892         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1893         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1894                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1895         case NFS4_OPEN_CLAIM_FH:
1896                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1897                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1898         }
1899         data->timestamp = jiffies;
1900         if (nfs4_setup_sequence(data->o_arg.server,
1901                                 &data->o_arg.seq_args,
1902                                 &data->o_res.seq_res,
1903                                 task) != 0)
1904                 nfs_release_seqid(data->o_arg.seqid);
1905
1906         /* Set the create mode (note dependency on the session type) */
1907         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1908         if (data->o_arg.open_flags & O_EXCL) {
1909                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1910                 if (nfs4_has_persistent_session(clp))
1911                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1912                 else if (clp->cl_mvops->minor_version > 0)
1913                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1914         }
1915         return;
1916 unlock_no_action:
1917         rcu_read_unlock();
1918 out_no_action:
1919         task->tk_action = NULL;
1920 out_wait:
1921         nfs4_sequence_done(task, &data->o_res.seq_res);
1922 }
1923
1924 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1925 {
1926         struct nfs4_opendata *data = calldata;
1927
1928         data->rpc_status = task->tk_status;
1929
1930         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1931                 return;
1932
1933         if (task->tk_status == 0) {
1934                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1935                         switch (data->o_res.f_attr->mode & S_IFMT) {
1936                         case S_IFREG:
1937                                 break;
1938                         case S_IFLNK:
1939                                 data->rpc_status = -ELOOP;
1940                                 break;
1941                         case S_IFDIR:
1942                                 data->rpc_status = -EISDIR;
1943                                 break;
1944                         default:
1945                                 data->rpc_status = -ENOTDIR;
1946                         }
1947                 }
1948                 renew_lease(data->o_res.server, data->timestamp);
1949                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1950                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1951         }
1952         data->rpc_done = 1;
1953 }
1954
1955 static void nfs4_open_release(void *calldata)
1956 {
1957         struct nfs4_opendata *data = calldata;
1958         struct nfs4_state *state = NULL;
1959
1960         /* If this request hasn't been cancelled, do nothing */
1961         if (data->cancelled == 0)
1962                 goto out_free;
1963         /* In case of error, no cleanup! */
1964         if (data->rpc_status != 0 || !data->rpc_done)
1965                 goto out_free;
1966         /* In case we need an open_confirm, no cleanup! */
1967         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1968                 goto out_free;
1969         state = nfs4_opendata_to_nfs4_state(data);
1970         if (!IS_ERR(state))
1971                 nfs4_close_state(state, data->o_arg.fmode);
1972 out_free:
1973         nfs4_opendata_put(data);
1974 }
1975
1976 static const struct rpc_call_ops nfs4_open_ops = {
1977         .rpc_call_prepare = nfs4_open_prepare,
1978         .rpc_call_done = nfs4_open_done,
1979         .rpc_release = nfs4_open_release,
1980 };
1981
1982 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1983 {
1984         struct inode *dir = d_inode(data->dir);
1985         struct nfs_server *server = NFS_SERVER(dir);
1986         struct nfs_openargs *o_arg = &data->o_arg;
1987         struct nfs_openres *o_res = &data->o_res;
1988         struct rpc_task *task;
1989         struct rpc_message msg = {
1990                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1991                 .rpc_argp = o_arg,
1992                 .rpc_resp = o_res,
1993                 .rpc_cred = data->owner->so_cred,
1994         };
1995         struct rpc_task_setup task_setup_data = {
1996                 .rpc_client = server->client,
1997                 .rpc_message = &msg,
1998                 .callback_ops = &nfs4_open_ops,
1999                 .callback_data = data,
2000                 .workqueue = nfsiod_workqueue,
2001                 .flags = RPC_TASK_ASYNC,
2002         };
2003         int status;
2004
2005         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2006         kref_get(&data->kref);
2007         data->rpc_done = 0;
2008         data->rpc_status = 0;
2009         data->cancelled = 0;
2010         data->is_recover = 0;
2011         if (isrecover) {
2012                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2013                 data->is_recover = 1;
2014         }
2015         task = rpc_run_task(&task_setup_data);
2016         if (IS_ERR(task))
2017                 return PTR_ERR(task);
2018         status = nfs4_wait_for_completion_rpc_task(task);
2019         if (status != 0) {
2020                 data->cancelled = 1;
2021                 smp_wmb();
2022         } else
2023                 status = data->rpc_status;
2024         rpc_put_task(task);
2025
2026         return status;
2027 }
2028
2029 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2030 {
2031         struct inode *dir = d_inode(data->dir);
2032         struct nfs_openres *o_res = &data->o_res;
2033         int status;
2034
2035         status = nfs4_run_open_task(data, 1);
2036         if (status != 0 || !data->rpc_done)
2037                 return status;
2038
2039         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2040
2041         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2042                 status = _nfs4_proc_open_confirm(data);
2043                 if (status != 0)
2044                         return status;
2045         }
2046
2047         return status;
2048 }
2049
2050 /*
2051  * Additional permission checks in order to distinguish between an
2052  * open for read, and an open for execute. This works around the
2053  * fact that NFSv4 OPEN treats read and execute permissions as being
2054  * the same.
2055  * Note that in the non-execute case, we want to turn off permission
2056  * checking if we just created a new file (POSIX open() semantics).
2057  */
2058 static int nfs4_opendata_access(struct rpc_cred *cred,
2059                                 struct nfs4_opendata *opendata,
2060                                 struct nfs4_state *state, fmode_t fmode,
2061                                 int openflags)
2062 {
2063         struct nfs_access_entry cache;
2064         u32 mask;
2065
2066         /* access call failed or for some reason the server doesn't
2067          * support any access modes -- defer access call until later */
2068         if (opendata->o_res.access_supported == 0)
2069                 return 0;
2070
2071         mask = 0;
2072         /*
2073          * Use openflags to check for exec, because fmode won't
2074          * always have FMODE_EXEC set when file open for exec.
2075          */
2076         if (openflags & __FMODE_EXEC) {
2077                 /* ONLY check for exec rights */
2078                 mask = MAY_EXEC;
2079         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2080                 mask = MAY_READ;
2081
2082         cache.cred = cred;
2083         cache.jiffies = jiffies;
2084         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2085         nfs_access_add_cache(state->inode, &cache);
2086
2087         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2088                 return 0;
2089
2090         /* even though OPEN succeeded, access is denied. Close the file */
2091         nfs4_close_state(state, fmode);
2092         return -EACCES;
2093 }
2094
2095 /*
2096  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2097  */
2098 static int _nfs4_proc_open(struct nfs4_opendata *data)
2099 {
2100         struct inode *dir = d_inode(data->dir);
2101         struct nfs_server *server = NFS_SERVER(dir);
2102         struct nfs_openargs *o_arg = &data->o_arg;
2103         struct nfs_openres *o_res = &data->o_res;
2104         int status;
2105
2106         status = nfs4_run_open_task(data, 0);
2107         if (!data->rpc_done)
2108                 return status;
2109         if (status != 0) {
2110                 if (status == -NFS4ERR_BADNAME &&
2111                                 !(o_arg->open_flags & O_CREAT))
2112                         return -ENOENT;
2113                 return status;
2114         }
2115
2116         nfs_fattr_map_and_free_names(server, &data->f_attr);
2117
2118         if (o_arg->open_flags & O_CREAT) {
2119                 update_changeattr(dir, &o_res->cinfo);
2120                 if (o_arg->open_flags & O_EXCL)
2121                         data->file_created = 1;
2122                 else if (o_res->cinfo.before != o_res->cinfo.after)
2123                         data->file_created = 1;
2124         }
2125         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2126                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2127         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2128                 status = _nfs4_proc_open_confirm(data);
2129                 if (status != 0)
2130                         return status;
2131         }
2132         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2133                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2134         return 0;
2135 }
2136
2137 static int nfs4_recover_expired_lease(struct nfs_server *server)
2138 {
2139         return nfs4_client_recover_expired_lease(server->nfs_client);
2140 }
2141
2142 /*
2143  * OPEN_EXPIRED:
2144  *      reclaim state on the server after a network partition.
2145  *      Assumes caller holds the appropriate lock
2146  */
2147 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2148 {
2149         struct nfs4_opendata *opendata;
2150         int ret;
2151
2152         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2153                         NFS4_OPEN_CLAIM_FH);
2154         if (IS_ERR(opendata))
2155                 return PTR_ERR(opendata);
2156         ret = nfs4_open_recover(opendata, state);
2157         if (ret == -ESTALE)
2158                 d_drop(ctx->dentry);
2159         nfs4_opendata_put(opendata);
2160         return ret;
2161 }
2162
2163 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2164 {
2165         struct nfs_server *server = NFS_SERVER(state->inode);
2166         struct nfs4_exception exception = { };
2167         int err;
2168
2169         do {
2170                 err = _nfs4_open_expired(ctx, state);
2171                 trace_nfs4_open_expired(ctx, 0, err);
2172                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2173                         continue;
2174                 switch (err) {
2175                 default:
2176                         goto out;
2177                 case -NFS4ERR_GRACE:
2178                 case -NFS4ERR_DELAY:
2179                         nfs4_handle_exception(server, err, &exception);
2180                         err = 0;
2181                 }
2182         } while (exception.retry);
2183 out:
2184         return err;
2185 }
2186
2187 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2188 {
2189         struct nfs_open_context *ctx;
2190         int ret;
2191
2192         ctx = nfs4_state_find_open_context(state);
2193         if (IS_ERR(ctx))
2194                 return -EAGAIN;
2195         ret = nfs4_do_open_expired(ctx, state);
2196         put_nfs_open_context(ctx);
2197         return ret;
2198 }
2199
2200 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2201 {
2202         nfs_remove_bad_delegation(state->inode);
2203         write_seqlock(&state->seqlock);
2204         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2205         write_sequnlock(&state->seqlock);
2206         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2207 }
2208
2209 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2210 {
2211         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2212                 nfs_finish_clear_delegation_stateid(state);
2213 }
2214
2215 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2216 {
2217         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2218         nfs40_clear_delegation_stateid(state);
2219         return nfs4_open_expired(sp, state);
2220 }
2221
2222 #if defined(CONFIG_NFS_V4_1)
2223 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2224 {
2225         struct nfs_server *server = NFS_SERVER(state->inode);
2226         nfs4_stateid stateid;
2227         struct nfs_delegation *delegation;
2228         struct rpc_cred *cred;
2229         int status;
2230
2231         /* Get the delegation credential for use by test/free_stateid */
2232         rcu_read_lock();
2233         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2234         if (delegation == NULL) {
2235                 rcu_read_unlock();
2236                 return;
2237         }
2238
2239         nfs4_stateid_copy(&stateid, &delegation->stateid);
2240         cred = get_rpccred(delegation->cred);
2241         rcu_read_unlock();
2242         status = nfs41_test_stateid(server, &stateid, cred);
2243         trace_nfs4_test_delegation_stateid(state, NULL, status);
2244
2245         if (status != NFS_OK) {
2246                 /* Free the stateid unless the server explicitly
2247                  * informs us the stateid is unrecognized. */
2248                 if (status != -NFS4ERR_BAD_STATEID)
2249                         nfs41_free_stateid(server, &stateid, cred);
2250                 nfs_finish_clear_delegation_stateid(state);
2251         }
2252
2253         put_rpccred(cred);
2254 }
2255
2256 /**
2257  * nfs41_check_open_stateid - possibly free an open stateid
2258  *
2259  * @state: NFSv4 state for an inode
2260  *
2261  * Returns NFS_OK if recovery for this stateid is now finished.
2262  * Otherwise a negative NFS4ERR value is returned.
2263  */
2264 static int nfs41_check_open_stateid(struct nfs4_state *state)
2265 {
2266         struct nfs_server *server = NFS_SERVER(state->inode);
2267         nfs4_stateid *stateid = &state->open_stateid;
2268         struct rpc_cred *cred = state->owner->so_cred;
2269         int status;
2270
2271         /* If a state reset has been done, test_stateid is unneeded */
2272         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2273             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2274             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2275                 return -NFS4ERR_BAD_STATEID;
2276
2277         status = nfs41_test_stateid(server, stateid, cred);
2278         trace_nfs4_test_open_stateid(state, NULL, status);
2279         if (status != NFS_OK) {
2280                 /* Free the stateid unless the server explicitly
2281                  * informs us the stateid is unrecognized. */
2282                 if (status != -NFS4ERR_BAD_STATEID)
2283                         nfs41_free_stateid(server, stateid, cred);
2284
2285                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2286                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2287                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2288                 clear_bit(NFS_OPEN_STATE, &state->flags);
2289         }
2290         return status;
2291 }
2292
2293 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2294 {
2295         int status;
2296
2297         nfs41_check_delegation_stateid(state);
2298         status = nfs41_check_open_stateid(state);
2299         if (status != NFS_OK)
2300                 status = nfs4_open_expired(sp, state);
2301         return status;
2302 }
2303 #endif
2304
2305 /*
2306  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2307  * fields corresponding to attributes that were used to store the verifier.
2308  * Make sure we clobber those fields in the later setattr call
2309  */
2310 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2311 {
2312         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2313             !(sattr->ia_valid & ATTR_ATIME_SET))
2314                 sattr->ia_valid |= ATTR_ATIME;
2315
2316         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2317             !(sattr->ia_valid & ATTR_MTIME_SET))
2318                 sattr->ia_valid |= ATTR_MTIME;
2319 }
2320
2321 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2322                 fmode_t fmode,
2323                 int flags,
2324                 struct nfs_open_context *ctx)
2325 {
2326         struct nfs4_state_owner *sp = opendata->owner;
2327         struct nfs_server *server = sp->so_server;
2328         struct dentry *dentry;
2329         struct nfs4_state *state;
2330         unsigned int seq;
2331         int ret;
2332
2333         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2334
2335         ret = _nfs4_proc_open(opendata);
2336         if (ret != 0)
2337                 goto out;
2338
2339         state = nfs4_opendata_to_nfs4_state(opendata);
2340         ret = PTR_ERR(state);
2341         if (IS_ERR(state))
2342                 goto out;
2343         if (server->caps & NFS_CAP_POSIX_LOCK)
2344                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2345
2346         dentry = opendata->dentry;
2347         if (d_really_is_negative(dentry)) {
2348                 /* FIXME: Is this d_drop() ever needed? */
2349                 d_drop(dentry);
2350                 dentry = d_add_unique(dentry, igrab(state->inode));
2351                 if (dentry == NULL) {
2352                         dentry = opendata->dentry;
2353                 } else if (dentry != ctx->dentry) {
2354                         dput(ctx->dentry);
2355                         ctx->dentry = dget(dentry);
2356                 }
2357                 nfs_set_verifier(dentry,
2358                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2359         }
2360
2361         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2362         if (ret != 0)
2363                 goto out;
2364
2365         ctx->state = state;
2366         if (d_inode(dentry) == state->inode) {
2367                 nfs_inode_attach_open_context(ctx);
2368                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2369                         nfs4_schedule_stateid_recovery(server, state);
2370         }
2371 out:
2372         return ret;
2373 }
2374
2375 /*
2376  * Returns a referenced nfs4_state
2377  */
2378 static int _nfs4_do_open(struct inode *dir,
2379                         struct nfs_open_context *ctx,
2380                         int flags,
2381                         struct iattr *sattr,
2382                         struct nfs4_label *label,
2383                         int *opened)
2384 {
2385         struct nfs4_state_owner  *sp;
2386         struct nfs4_state     *state = NULL;
2387         struct nfs_server       *server = NFS_SERVER(dir);
2388         struct nfs4_opendata *opendata;
2389         struct dentry *dentry = ctx->dentry;
2390         struct rpc_cred *cred = ctx->cred;
2391         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2392         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2393         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2394         struct nfs4_label *olabel = NULL;
2395         int status;
2396
2397         /* Protect against reboot recovery conflicts */
2398         status = -ENOMEM;
2399         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2400         if (sp == NULL) {
2401                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2402                 goto out_err;
2403         }
2404         status = nfs4_recover_expired_lease(server);
2405         if (status != 0)
2406                 goto err_put_state_owner;
2407         if (d_really_is_positive(dentry))
2408                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2409         status = -ENOMEM;
2410         if (d_really_is_positive(dentry))
2411                 claim = NFS4_OPEN_CLAIM_FH;
2412         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2413                         label, claim, GFP_KERNEL);
2414         if (opendata == NULL)
2415                 goto err_put_state_owner;
2416
2417         if (label) {
2418                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2419                 if (IS_ERR(olabel)) {
2420                         status = PTR_ERR(olabel);
2421                         goto err_opendata_put;
2422                 }
2423         }
2424
2425         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2426                 if (!opendata->f_attr.mdsthreshold) {
2427                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2428                         if (!opendata->f_attr.mdsthreshold)
2429                                 goto err_free_label;
2430                 }
2431                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2432         }
2433         if (d_really_is_positive(dentry))
2434                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2435
2436         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2437         if (status != 0)
2438                 goto err_free_label;
2439         state = ctx->state;
2440
2441         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2442             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2443                 nfs4_exclusive_attrset(opendata, sattr);
2444
2445                 nfs_fattr_init(opendata->o_res.f_attr);
2446                 status = nfs4_do_setattr(state->inode, cred,
2447                                 opendata->o_res.f_attr, sattr,
2448                                 state, label, olabel);
2449                 if (status == 0) {
2450                         nfs_setattr_update_inode(state->inode, sattr,
2451                                         opendata->o_res.f_attr);
2452                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2453                 }
2454         }
2455         if (opened && opendata->file_created)
2456                 *opened |= FILE_CREATED;
2457
2458         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2459                 *ctx_th = opendata->f_attr.mdsthreshold;
2460                 opendata->f_attr.mdsthreshold = NULL;
2461         }
2462
2463         nfs4_label_free(olabel);
2464
2465         nfs4_opendata_put(opendata);
2466         nfs4_put_state_owner(sp);
2467         return 0;
2468 err_free_label:
2469         nfs4_label_free(olabel);
2470 err_opendata_put:
2471         nfs4_opendata_put(opendata);
2472 err_put_state_owner:
2473         nfs4_put_state_owner(sp);
2474 out_err:
2475         return status;
2476 }
2477
2478
2479 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2480                                         struct nfs_open_context *ctx,
2481                                         int flags,
2482                                         struct iattr *sattr,
2483                                         struct nfs4_label *label,
2484                                         int *opened)
2485 {
2486         struct nfs_server *server = NFS_SERVER(dir);
2487         struct nfs4_exception exception = { };
2488         struct nfs4_state *res;
2489         int status;
2490
2491         do {
2492                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2493                 res = ctx->state;
2494                 trace_nfs4_open_file(ctx, flags, status);
2495                 if (status == 0)
2496                         break;
2497                 /* NOTE: BAD_SEQID means the server and client disagree about the
2498                  * book-keeping w.r.t. state-changing operations
2499                  * (OPEN/CLOSE/LOCK/LOCKU...)
2500                  * It is actually a sign of a bug on the client or on the server.
2501                  *
2502                  * If we receive a BAD_SEQID error in the particular case of
2503                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2504                  * have unhashed the old state_owner for us, and that we can
2505                  * therefore safely retry using a new one. We should still warn
2506                  * the user though...
2507                  */
2508                 if (status == -NFS4ERR_BAD_SEQID) {
2509                         pr_warn_ratelimited("NFS: v4 server %s "
2510                                         " returned a bad sequence-id error!\n",
2511                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2512                         exception.retry = 1;
2513                         continue;
2514                 }
2515                 /*
2516                  * BAD_STATEID on OPEN means that the server cancelled our
2517                  * state before it received the OPEN_CONFIRM.
2518                  * Recover by retrying the request as per the discussion
2519                  * on Page 181 of RFC3530.
2520                  */
2521                 if (status == -NFS4ERR_BAD_STATEID) {
2522                         exception.retry = 1;
2523                         continue;
2524                 }
2525                 if (status == -EAGAIN) {
2526                         /* We must have found a delegation */
2527                         exception.retry = 1;
2528                         continue;
2529                 }
2530                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2531                         continue;
2532                 res = ERR_PTR(nfs4_handle_exception(server,
2533                                         status, &exception));
2534         } while (exception.retry);
2535         return res;
2536 }
2537
2538 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2539                             struct nfs_fattr *fattr, struct iattr *sattr,
2540                             struct nfs4_state *state, struct nfs4_label *ilabel,
2541                             struct nfs4_label *olabel)
2542 {
2543         struct nfs_server *server = NFS_SERVER(inode);
2544         struct nfs_setattrargs  arg = {
2545                 .fh             = NFS_FH(inode),
2546                 .iap            = sattr,
2547                 .server         = server,
2548                 .bitmask = server->attr_bitmask,
2549                 .label          = ilabel,
2550         };
2551         struct nfs_setattrres  res = {
2552                 .fattr          = fattr,
2553                 .label          = olabel,
2554                 .server         = server,
2555         };
2556         struct rpc_message msg = {
2557                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2558                 .rpc_argp       = &arg,
2559                 .rpc_resp       = &res,
2560                 .rpc_cred       = cred,
2561         };
2562         unsigned long timestamp = jiffies;
2563         fmode_t fmode;
2564         bool truncate;
2565         int status;
2566
2567         arg.bitmask = nfs4_bitmask(server, ilabel);
2568         if (ilabel)
2569                 arg.bitmask = nfs4_bitmask(server, olabel);
2570
2571         nfs_fattr_init(fattr);
2572
2573         /* Servers should only apply open mode checks for file size changes */
2574         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2575         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2576
2577         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2578                 /* Use that stateid */
2579         } else if (truncate && state != NULL) {
2580                 struct nfs_lockowner lockowner = {
2581                         .l_owner = current->files,
2582                         .l_pid = current->tgid,
2583                 };
2584                 if (!nfs4_valid_open_stateid(state))
2585                         return -EBADF;
2586                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2587                                 &lockowner) == -EIO)
2588                         return -EBADF;
2589         } else
2590                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2591
2592         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2593         if (status == 0 && state != NULL)
2594                 renew_lease(server, timestamp);
2595         return status;
2596 }
2597
2598 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2599                            struct nfs_fattr *fattr, struct iattr *sattr,
2600                            struct nfs4_state *state, struct nfs4_label *ilabel,
2601                            struct nfs4_label *olabel)
2602 {
2603         struct nfs_server *server = NFS_SERVER(inode);
2604         struct nfs4_exception exception = {
2605                 .state = state,
2606                 .inode = inode,
2607         };
2608         int err;
2609         do {
2610                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2611                 trace_nfs4_setattr(inode, err);
2612                 switch (err) {
2613                 case -NFS4ERR_OPENMODE:
2614                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2615                                 pr_warn_once("NFSv4: server %s is incorrectly "
2616                                                 "applying open mode checks to "
2617                                                 "a SETATTR that is not "
2618                                                 "changing file size.\n",
2619                                                 server->nfs_client->cl_hostname);
2620                         }
2621                         if (state && !(state->state & FMODE_WRITE)) {
2622                                 err = -EBADF;
2623                                 if (sattr->ia_valid & ATTR_OPEN)
2624                                         err = -EACCES;
2625                                 goto out;
2626                         }
2627                 }
2628                 err = nfs4_handle_exception(server, err, &exception);
2629         } while (exception.retry);
2630 out:
2631         return err;
2632 }
2633
2634 struct nfs4_closedata {
2635         struct inode *inode;
2636         struct nfs4_state *state;
2637         struct nfs_closeargs arg;
2638         struct nfs_closeres res;
2639         struct nfs_fattr fattr;
2640         unsigned long timestamp;
2641         bool roc;
2642         u32 roc_barrier;
2643 };
2644
2645 static void nfs4_free_closedata(void *data)
2646 {
2647         struct nfs4_closedata *calldata = data;
2648         struct nfs4_state_owner *sp = calldata->state->owner;
2649         struct super_block *sb = calldata->state->inode->i_sb;
2650
2651         if (calldata->roc)
2652                 pnfs_roc_release(calldata->state->inode);
2653         nfs4_put_open_state(calldata->state);
2654         nfs_free_seqid(calldata->arg.seqid);
2655         nfs4_put_state_owner(sp);
2656         nfs_sb_deactive(sb);
2657         kfree(calldata);
2658 }
2659
2660 static void nfs4_close_done(struct rpc_task *task, void *data)
2661 {
2662         struct nfs4_closedata *calldata = data;
2663         struct nfs4_state *state = calldata->state;
2664         struct nfs_server *server = NFS_SERVER(calldata->inode);
2665         nfs4_stateid *res_stateid = NULL;
2666
2667         dprintk("%s: begin!\n", __func__);
2668         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2669                 return;
2670         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2671         /* hmm. we are done with the inode, and in the process of freeing
2672          * the state_owner. we keep this around to process errors
2673          */
2674         switch (task->tk_status) {
2675                 case 0:
2676                         res_stateid = &calldata->res.stateid;
2677                         if (calldata->roc)
2678                                 pnfs_roc_set_barrier(state->inode,
2679                                                      calldata->roc_barrier);
2680                         renew_lease(server, calldata->timestamp);
2681                         break;
2682                 case -NFS4ERR_ADMIN_REVOKED:
2683                 case -NFS4ERR_STALE_STATEID:
2684                 case -NFS4ERR_OLD_STATEID:
2685                 case -NFS4ERR_BAD_STATEID:
2686                 case -NFS4ERR_EXPIRED:
2687                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2688                                                 &state->open_stateid)) {
2689                                 rpc_restart_call_prepare(task);
2690                                 goto out_release;
2691                         }
2692                         if (calldata->arg.fmode == 0)
2693                                 break;
2694                 default:
2695                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2696                                 rpc_restart_call_prepare(task);
2697                                 goto out_release;
2698                         }
2699         }
2700         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2701 out_release:
2702         nfs_release_seqid(calldata->arg.seqid);
2703         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2704         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2705 }
2706
2707 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2708 {
2709         struct nfs4_closedata *calldata = data;
2710         struct nfs4_state *state = calldata->state;
2711         struct inode *inode = calldata->inode;
2712         bool is_rdonly, is_wronly, is_rdwr;
2713         int call_close = 0;
2714
2715         dprintk("%s: begin!\n", __func__);
2716         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2717                 goto out_wait;
2718
2719         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2720         spin_lock(&state->owner->so_lock);
2721         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2722         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2723         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2724         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2725         /* Calculate the change in open mode */
2726         calldata->arg.fmode = 0;
2727         if (state->n_rdwr == 0) {
2728                 if (state->n_rdonly == 0)
2729                         call_close |= is_rdonly;
2730                 else if (is_rdonly)
2731                         calldata->arg.fmode |= FMODE_READ;
2732                 if (state->n_wronly == 0)
2733                         call_close |= is_wronly;
2734                 else if (is_wronly)
2735                         calldata->arg.fmode |= FMODE_WRITE;
2736         } else if (is_rdwr)
2737                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2738
2739         if (calldata->arg.fmode == 0)
2740                 call_close |= is_rdwr;
2741
2742         if (!nfs4_valid_open_stateid(state))
2743                 call_close = 0;
2744         spin_unlock(&state->owner->so_lock);
2745
2746         if (!call_close) {
2747                 /* Note: exit _without_ calling nfs4_close_done */
2748                 goto out_no_action;
2749         }
2750
2751         if (calldata->arg.fmode == 0)
2752                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2753         if (calldata->roc)
2754                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2755
2756         calldata->arg.share_access =
2757                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2758                                 calldata->arg.fmode, 0);
2759
2760         nfs_fattr_init(calldata->res.fattr);
2761         calldata->timestamp = jiffies;
2762         if (nfs4_setup_sequence(NFS_SERVER(inode),
2763                                 &calldata->arg.seq_args,
2764                                 &calldata->res.seq_res,
2765                                 task) != 0)
2766                 nfs_release_seqid(calldata->arg.seqid);
2767         dprintk("%s: done!\n", __func__);
2768         return;
2769 out_no_action:
2770         task->tk_action = NULL;
2771 out_wait:
2772         nfs4_sequence_done(task, &calldata->res.seq_res);
2773 }
2774
2775 static const struct rpc_call_ops nfs4_close_ops = {
2776         .rpc_call_prepare = nfs4_close_prepare,
2777         .rpc_call_done = nfs4_close_done,
2778         .rpc_release = nfs4_free_closedata,
2779 };
2780
2781 static bool nfs4_roc(struct inode *inode)
2782 {
2783         if (!nfs_have_layout(inode))
2784                 return false;
2785         return pnfs_roc(inode);
2786 }
2787
2788 /*
2789  * It is possible for data to be read/written from a mem-mapped file
2790  * after the sys_close call (which hits the vfs layer as a flush).
2791  * This means that we can't safely call nfsv4 close on a file until
2792  * the inode is cleared. This in turn means that we are not good
2793  * NFSv4 citizens - we do not indicate to the server to update the file's
2794  * share state even when we are done with one of the three share
2795  * stateid's in the inode.
2796  *
2797  * NOTE: Caller must be holding the sp->so_owner semaphore!
2798  */
2799 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2800 {
2801         struct nfs_server *server = NFS_SERVER(state->inode);
2802         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2803         struct nfs4_closedata *calldata;
2804         struct nfs4_state_owner *sp = state->owner;
2805         struct rpc_task *task;
2806         struct rpc_message msg = {
2807                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2808                 .rpc_cred = state->owner->so_cred,
2809         };
2810         struct rpc_task_setup task_setup_data = {
2811                 .rpc_client = server->client,
2812                 .rpc_message = &msg,
2813                 .callback_ops = &nfs4_close_ops,
2814                 .workqueue = nfsiod_workqueue,
2815                 .flags = RPC_TASK_ASYNC,
2816         };
2817         int status = -ENOMEM;
2818
2819         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2820                 &task_setup_data.rpc_client, &msg);
2821
2822         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2823         if (calldata == NULL)
2824                 goto out;
2825         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2826         calldata->inode = state->inode;
2827         calldata->state = state;
2828         calldata->arg.fh = NFS_FH(state->inode);
2829         /* Serialization for the sequence id */
2830         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2831         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2832         if (IS_ERR(calldata->arg.seqid))
2833                 goto out_free_calldata;
2834         calldata->arg.fmode = 0;
2835         calldata->arg.bitmask = server->cache_consistency_bitmask;
2836         calldata->res.fattr = &calldata->fattr;
2837         calldata->res.seqid = calldata->arg.seqid;
2838         calldata->res.server = server;
2839         calldata->roc = nfs4_roc(state->inode);
2840         nfs_sb_active(calldata->inode->i_sb);
2841
2842         msg.rpc_argp = &calldata->arg;
2843         msg.rpc_resp = &calldata->res;
2844         task_setup_data.callback_data = calldata;
2845         task = rpc_run_task(&task_setup_data);
2846         if (IS_ERR(task))
2847                 return PTR_ERR(task);
2848         status = 0;
2849         if (wait)
2850                 status = rpc_wait_for_completion_task(task);
2851         rpc_put_task(task);
2852         return status;
2853 out_free_calldata:
2854         kfree(calldata);
2855 out:
2856         nfs4_put_open_state(state);
2857         nfs4_put_state_owner(sp);
2858         return status;
2859 }
2860
2861 static struct inode *
2862 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2863                 int open_flags, struct iattr *attr, int *opened)
2864 {
2865         struct nfs4_state *state;
2866         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2867
2868         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2869
2870         /* Protect against concurrent sillydeletes */
2871         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2872
2873         nfs4_label_release_security(label);
2874
2875         if (IS_ERR(state))
2876                 return ERR_CAST(state);
2877         return state->inode;
2878 }
2879
2880 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2881 {
2882         if (ctx->state == NULL)
2883                 return;
2884         if (is_sync)
2885                 nfs4_close_sync(ctx->state, ctx->mode);
2886         else
2887                 nfs4_close_state(ctx->state, ctx->mode);
2888 }
2889
2890 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2891 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2892 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2893
2894 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2895 {
2896         struct nfs4_server_caps_arg args = {
2897                 .fhandle = fhandle,
2898         };
2899         struct nfs4_server_caps_res res = {};
2900         struct rpc_message msg = {
2901                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2902                 .rpc_argp = &args,
2903                 .rpc_resp = &res,
2904         };
2905         int status;
2906
2907         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2908         if (status == 0) {
2909                 /* Sanity check the server answers */
2910                 switch (server->nfs_client->cl_minorversion) {
2911                 case 0:
2912                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2913                         res.attr_bitmask[2] = 0;
2914                         break;
2915                 case 1:
2916                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2917                         break;
2918                 case 2:
2919                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2920                 }
2921                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2922                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2923                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2924                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2925                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2926                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2927                                 NFS_CAP_SECURITY_LABEL);
2928                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2929                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2930                         server->caps |= NFS_CAP_ACLS;
2931                 if (res.has_links != 0)
2932                         server->caps |= NFS_CAP_HARDLINKS;
2933                 if (res.has_symlinks != 0)
2934                         server->caps |= NFS_CAP_SYMLINKS;
2935                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2936                         server->caps |= NFS_CAP_FILEID;
2937                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2938                         server->caps |= NFS_CAP_MODE;
2939                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2940                         server->caps |= NFS_CAP_NLINK;
2941                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2942                         server->caps |= NFS_CAP_OWNER;
2943                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2944                         server->caps |= NFS_CAP_OWNER_GROUP;
2945                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2946                         server->caps |= NFS_CAP_ATIME;
2947                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2948                         server->caps |= NFS_CAP_CTIME;
2949                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2950                         server->caps |= NFS_CAP_MTIME;
2951 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2952                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2953                         server->caps |= NFS_CAP_SECURITY_LABEL;
2954 #endif
2955                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2956                                 sizeof(server->attr_bitmask));
2957                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2958
2959                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2960                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2961                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2962                 server->cache_consistency_bitmask[2] = 0;
2963                 server->acl_bitmask = res.acl_bitmask;
2964                 server->fh_expire_type = res.fh_expire_type;
2965         }
2966
2967         return status;
2968 }
2969
2970 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2971 {
2972         struct nfs4_exception exception = { };
2973         int err;
2974         do {
2975                 err = nfs4_handle_exception(server,
2976                                 _nfs4_server_capabilities(server, fhandle),
2977                                 &exception);
2978         } while (exception.retry);
2979         return err;
2980 }
2981
2982 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2983                 struct nfs_fsinfo *info)
2984 {
2985         u32 bitmask[3];
2986         struct nfs4_lookup_root_arg args = {
2987                 .bitmask = bitmask,
2988         };
2989         struct nfs4_lookup_res res = {
2990                 .server = server,
2991                 .fattr = info->fattr,
2992                 .fh = fhandle,
2993         };
2994         struct rpc_message msg = {
2995                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2996                 .rpc_argp = &args,
2997                 .rpc_resp = &res,
2998         };
2999
3000         bitmask[0] = nfs4_fattr_bitmap[0];
3001         bitmask[1] = nfs4_fattr_bitmap[1];
3002         /*
3003          * Process the label in the upcoming getfattr
3004          */
3005         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3006
3007         nfs_fattr_init(info->fattr);
3008         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3009 }
3010
3011 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3012                 struct nfs_fsinfo *info)
3013 {
3014         struct nfs4_exception exception = { };
3015         int err;
3016         do {
3017                 err = _nfs4_lookup_root(server, fhandle, info);
3018                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3019                 switch (err) {
3020                 case 0:
3021                 case -NFS4ERR_WRONGSEC:
3022                         goto out;
3023                 default:
3024                         err = nfs4_handle_exception(server, err, &exception);
3025                 }
3026         } while (exception.retry);
3027 out:
3028         return err;
3029 }
3030
3031 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3032                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3033 {
3034         struct rpc_auth_create_args auth_args = {
3035                 .pseudoflavor = flavor,
3036         };
3037         struct rpc_auth *auth;
3038         int ret;
3039
3040         auth = rpcauth_create(&auth_args, server->client);
3041         if (IS_ERR(auth)) {
3042                 ret = -EACCES;
3043                 goto out;
3044         }
3045         ret = nfs4_lookup_root(server, fhandle, info);
3046 out:
3047         return ret;
3048 }
3049
3050 /*
3051  * Retry pseudoroot lookup with various security flavors.  We do this when:
3052  *
3053  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3054  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3055  *
3056  * Returns zero on success, or a negative NFS4ERR value, or a
3057  * negative errno value.
3058  */
3059 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3060                               struct nfs_fsinfo *info)
3061 {
3062         /* Per 3530bis 15.33.5 */
3063         static const rpc_authflavor_t flav_array[] = {
3064                 RPC_AUTH_GSS_KRB5P,
3065                 RPC_AUTH_GSS_KRB5I,
3066                 RPC_AUTH_GSS_KRB5,
3067                 RPC_AUTH_UNIX,                  /* courtesy */
3068                 RPC_AUTH_NULL,
3069         };
3070         int status = -EPERM;
3071         size_t i;
3072
3073         if (server->auth_info.flavor_len > 0) {
3074                 /* try each flavor specified by user */
3075                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3076                         status = nfs4_lookup_root_sec(server, fhandle, info,
3077                                                 server->auth_info.flavors[i]);
3078                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3079                                 continue;
3080                         break;
3081                 }
3082         } else {
3083                 /* no flavors specified by user, try default list */
3084                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3085                         status = nfs4_lookup_root_sec(server, fhandle, info,
3086                                                       flav_array[i]);
3087                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3088                                 continue;
3089                         break;
3090                 }
3091         }
3092
3093         /*
3094          * -EACCESS could mean that the user doesn't have correct permissions
3095          * to access the mount.  It could also mean that we tried to mount
3096          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3097          * existing mount programs don't handle -EACCES very well so it should
3098          * be mapped to -EPERM instead.
3099          */
3100         if (status == -EACCES)
3101                 status = -EPERM;
3102         return status;
3103 }
3104
3105 static int nfs4_do_find_root_sec(struct nfs_server *server,
3106                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3107 {
3108         int mv = server->nfs_client->cl_minorversion;
3109         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3110 }
3111
3112 /**
3113  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3114  * @server: initialized nfs_server handle
3115  * @fhandle: we fill in the pseudo-fs root file handle
3116  * @info: we fill in an FSINFO struct
3117  * @auth_probe: probe the auth flavours
3118  *
3119  * Returns zero on success, or a negative errno.
3120  */
3121 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3122                          struct nfs_fsinfo *info,
3123                          bool auth_probe)
3124 {
3125         int status = 0;
3126
3127         if (!auth_probe)
3128                 status = nfs4_lookup_root(server, fhandle, info);
3129
3130         if (auth_probe || status == NFS4ERR_WRONGSEC)
3131                 status = nfs4_do_find_root_sec(server, fhandle, info);
3132
3133         if (status == 0)
3134                 status = nfs4_server_capabilities(server, fhandle);
3135         if (status == 0)
3136                 status = nfs4_do_fsinfo(server, fhandle, info);
3137
3138         return nfs4_map_errors(status);
3139 }
3140
3141 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3142                               struct nfs_fsinfo *info)
3143 {
3144         int error;
3145         struct nfs_fattr *fattr = info->fattr;
3146         struct nfs4_label *label = NULL;
3147
3148         error = nfs4_server_capabilities(server, mntfh);
3149         if (error < 0) {
3150                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3151                 return error;
3152         }
3153
3154         label = nfs4_label_alloc(server, GFP_KERNEL);
3155         if (IS_ERR(label))
3156                 return PTR_ERR(label);
3157
3158         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3159         if (error < 0) {
3160                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3161                 goto err_free_label;
3162         }
3163
3164         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3165             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3166                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3167
3168 err_free_label:
3169         nfs4_label_free(label);
3170
3171         return error;
3172 }
3173
3174 /*
3175  * Get locations and (maybe) other attributes of a referral.
3176  * Note that we'll actually follow the referral later when
3177  * we detect fsid mismatch in inode revalidation
3178  */
3179 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3180                              const struct qstr *name, struct nfs_fattr *fattr,
3181                              struct nfs_fh *fhandle)
3182 {
3183         int status = -ENOMEM;
3184         struct page *page = NULL;
3185         struct nfs4_fs_locations *locations = NULL;
3186
3187         page = alloc_page(GFP_KERNEL);
3188         if (page == NULL)
3189                 goto out;
3190         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3191         if (locations == NULL)
3192                 goto out;
3193
3194         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3195         if (status != 0)
3196                 goto out;
3197
3198         /*
3199          * If the fsid didn't change, this is a migration event, not a
3200          * referral.  Cause us to drop into the exception handler, which
3201          * will kick off migration recovery.
3202          */
3203         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3204                 dprintk("%s: server did not return a different fsid for"
3205                         " a referral at %s\n", __func__, name->name);
3206                 status = -NFS4ERR_MOVED;
3207                 goto out;
3208         }
3209         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3210         nfs_fixup_referral_attributes(&locations->fattr);
3211
3212         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3213         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3214         memset(fhandle, 0, sizeof(struct nfs_fh));
3215 out:
3216         if (page)
3217                 __free_page(page);
3218         kfree(locations);
3219         return status;
3220 }
3221
3222 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3223                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3224 {
3225         struct nfs4_getattr_arg args = {
3226                 .fh = fhandle,
3227                 .bitmask = server->attr_bitmask,
3228         };
3229         struct nfs4_getattr_res res = {
3230                 .fattr = fattr,
3231                 .label = label,
3232                 .server = server,
3233         };
3234         struct rpc_message msg = {
3235                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3236                 .rpc_argp = &args,
3237                 .rpc_resp = &res,
3238         };
3239
3240         args.bitmask = nfs4_bitmask(server, label);
3241
3242         nfs_fattr_init(fattr);
3243         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3244 }
3245
3246 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3247                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3248 {
3249         struct nfs4_exception exception = { };
3250         int err;
3251         do {
3252                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3253                 trace_nfs4_getattr(server, fhandle, fattr, err);
3254                 err = nfs4_handle_exception(server, err,
3255                                 &exception);
3256         } while (exception.retry);
3257         return err;
3258 }
3259
3260 /*
3261  * The file is not closed if it is opened due to the a request to change
3262  * the size of the file. The open call will not be needed once the
3263  * VFS layer lookup-intents are implemented.
3264  *
3265  * Close is called when the inode is destroyed.
3266  * If we haven't opened the file for O_WRONLY, we
3267  * need to in the size_change case to obtain a stateid.
3268  *
3269  * Got race?
3270  * Because OPEN is always done by name in nfsv4, it is
3271  * possible that we opened a different file by the same
3272  * name.  We can recognize this race condition, but we
3273  * can't do anything about it besides returning an error.
3274  *
3275  * This will be fixed with VFS changes (lookup-intent).
3276  */
3277 static int
3278 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3279                   struct iattr *sattr)
3280 {
3281         struct inode *inode = d_inode(dentry);
3282         struct rpc_cred *cred = NULL;
3283         struct nfs4_state *state = NULL;
3284         struct nfs4_label *label = NULL;
3285         int status;
3286
3287         if (pnfs_ld_layoutret_on_setattr(inode) &&
3288             sattr->ia_valid & ATTR_SIZE &&
3289             sattr->ia_size < i_size_read(inode))
3290                 pnfs_commit_and_return_layout(inode);
3291
3292         nfs_fattr_init(fattr);
3293
3294         /* Deal with open(O_TRUNC) */
3295         if (sattr->ia_valid & ATTR_OPEN)
3296                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3297
3298         /* Optimization: if the end result is no change, don't RPC */
3299         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3300                 return 0;
3301
3302         /* Search for an existing open(O_WRITE) file */
3303         if (sattr->ia_valid & ATTR_FILE) {
3304                 struct nfs_open_context *ctx;
3305
3306                 ctx = nfs_file_open_context(sattr->ia_file);
3307                 if (ctx) {
3308                         cred = ctx->cred;
3309                         state = ctx->state;
3310                 }
3311         }
3312
3313         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3314         if (IS_ERR(label))
3315                 return PTR_ERR(label);
3316
3317         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3318         if (status == 0) {
3319                 nfs_setattr_update_inode(inode, sattr, fattr);
3320                 nfs_setsecurity(inode, fattr, label);
3321         }
3322         nfs4_label_free(label);
3323         return status;
3324 }
3325
3326 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3327                 const struct qstr *name, struct nfs_fh *fhandle,
3328                 struct nfs_fattr *fattr, struct nfs4_label *label)
3329 {
3330         struct nfs_server *server = NFS_SERVER(dir);
3331         int                    status;
3332         struct nfs4_lookup_arg args = {
3333                 .bitmask = server->attr_bitmask,
3334                 .dir_fh = NFS_FH(dir),
3335                 .name = name,
3336         };
3337         struct nfs4_lookup_res res = {
3338                 .server = server,
3339                 .fattr = fattr,
3340                 .label = label,
3341                 .fh = fhandle,
3342         };
3343         struct rpc_message msg = {
3344                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3345                 .rpc_argp = &args,
3346                 .rpc_resp = &res,
3347         };
3348
3349         args.bitmask = nfs4_bitmask(server, label);
3350
3351         nfs_fattr_init(fattr);
3352
3353         dprintk("NFS call  lookup %s\n", name->name);
3354         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3355         dprintk("NFS reply lookup: %d\n", status);
3356         return status;
3357 }
3358
3359 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3360 {
3361         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3362                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3363         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3364         fattr->nlink = 2;
3365 }
3366
3367 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3368                                    struct qstr *name, struct nfs_fh *fhandle,
3369                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3370 {
3371         struct nfs4_exception exception = { };
3372         struct rpc_clnt *client = *clnt;
3373         int err;
3374         do {
3375                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3376                 trace_nfs4_lookup(dir, name, err);
3377                 switch (err) {
3378                 case -NFS4ERR_BADNAME:
3379                         err = -ENOENT;
3380                         goto out;
3381                 case -NFS4ERR_MOVED:
3382                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3383                         if (err == -NFS4ERR_MOVED)
3384                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3385                         goto out;
3386                 case -NFS4ERR_WRONGSEC:
3387                         err = -EPERM;
3388                         if (client != *clnt)
3389                                 goto out;
3390                         client = nfs4_negotiate_security(client, dir, name);
3391                         if (IS_ERR(client))
3392                                 return PTR_ERR(client);
3393
3394                         exception.retry = 1;
3395                         break;
3396                 default:
3397                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3398                 }
3399         } while (exception.retry);
3400
3401 out:
3402         if (err == 0)
3403                 *clnt = client;
3404         else if (client != *clnt)
3405                 rpc_shutdown_client(client);
3406
3407         return err;
3408 }
3409
3410 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3411                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3412                             struct nfs4_label *label)
3413 {
3414         int status;
3415         struct rpc_clnt *client = NFS_CLIENT(dir);
3416
3417         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3418         if (client != NFS_CLIENT(dir)) {
3419                 rpc_shutdown_client(client);
3420                 nfs_fixup_secinfo_attributes(fattr);
3421         }
3422         return status;
3423 }
3424
3425 struct rpc_clnt *
3426 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3427                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3428 {
3429         struct rpc_clnt *client = NFS_CLIENT(dir);
3430         int status;
3431
3432         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3433         if (status < 0)
3434                 return ERR_PTR(status);
3435         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3436 }
3437
3438 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3439 {
3440         struct nfs_server *server = NFS_SERVER(inode);
3441         struct nfs4_accessargs args = {
3442                 .fh = NFS_FH(inode),
3443                 .bitmask = server->cache_consistency_bitmask,
3444         };
3445         struct nfs4_accessres res = {
3446                 .server = server,
3447         };
3448         struct rpc_message msg = {
3449                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3450                 .rpc_argp = &args,
3451                 .rpc_resp = &res,
3452                 .rpc_cred = entry->cred,
3453         };
3454         int mode = entry->mask;
3455         int status = 0;
3456
3457         /*
3458          * Determine which access bits we want to ask for...
3459          */
3460         if (mode & MAY_READ)
3461                 args.access |= NFS4_ACCESS_READ;
3462         if (S_ISDIR(inode->i_mode)) {
3463                 if (mode & MAY_WRITE)
3464                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3465                 if (mode & MAY_EXEC)
3466                         args.access |= NFS4_ACCESS_LOOKUP;
3467         } else {
3468                 if (mode & MAY_WRITE)
3469                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3470                 if (mode & MAY_EXEC)
3471                         args.access |= NFS4_ACCESS_EXECUTE;
3472         }
3473
3474         res.fattr = nfs_alloc_fattr();
3475         if (res.fattr == NULL)
3476                 return -ENOMEM;
3477
3478         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3479         if (!status) {
3480                 nfs_access_set_mask(entry, res.access);
3481                 nfs_refresh_inode(inode, res.fattr);
3482         }
3483         nfs_free_fattr(res.fattr);
3484         return status;
3485 }
3486
3487 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3488 {
3489         struct nfs4_exception exception = { };
3490         int err;
3491         do {
3492                 err = _nfs4_proc_access(inode, entry);
3493                 trace_nfs4_access(inode, err);
3494                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3495                                 &exception);
3496         } while (exception.retry);
3497         return err;
3498 }
3499
3500 /*
3501  * TODO: For the time being, we don't try to get any attributes
3502  * along with any of the zero-copy operations READ, READDIR,
3503  * READLINK, WRITE.
3504  *
3505  * In the case of the first three, we want to put the GETATTR
3506  * after the read-type operation -- this is because it is hard
3507  * to predict the length of a GETATTR response in v4, and thus
3508  * align the READ data correctly.  This means that the GETATTR
3509  * may end up partially falling into the page cache, and we should
3510  * shift it into the 'tail' of the xdr_buf before processing.
3511  * To do this efficiently, we need to know the total length
3512  * of data received, which doesn't seem to be available outside
3513  * of the RPC layer.
3514  *
3515  * In the case of WRITE, we also want to put the GETATTR after
3516  * the operation -- in this case because we want to make sure
3517  * we get the post-operation mtime and size.
3518  *
3519  * Both of these changes to the XDR layer would in fact be quite
3520  * minor, but I decided to leave them for a subsequent patch.
3521  */
3522 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3523                 unsigned int pgbase, unsigned int pglen)
3524 {
3525         struct nfs4_readlink args = {
3526                 .fh       = NFS_FH(inode),
3527                 .pgbase   = pgbase,
3528                 .pglen    = pglen,
3529                 .pages    = &page,
3530         };
3531         struct nfs4_readlink_res res;
3532         struct rpc_message msg = {
3533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3534                 .rpc_argp = &args,
3535                 .rpc_resp = &res,
3536         };
3537
3538         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3539 }
3540
3541 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3542                 unsigned int pgbase, unsigned int pglen)
3543 {
3544         struct nfs4_exception exception = { };
3545         int err;
3546         do {
3547                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3548                 trace_nfs4_readlink(inode, err);
3549                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3550                                 &exception);
3551         } while (exception.retry);
3552         return err;
3553 }
3554
3555 /*
3556  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3557  */
3558 static int
3559 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3560                  int flags)
3561 {
3562         struct nfs4_label l, *ilabel = NULL;
3563         struct nfs_open_context *ctx;
3564         struct nfs4_state *state;
3565         int status = 0;
3566
3567         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3568         if (IS_ERR(ctx))
3569                 return PTR_ERR(ctx);
3570
3571         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3572
3573         sattr->ia_mode &= ~current_umask();
3574         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3575         if (IS_ERR(state)) {
3576                 status = PTR_ERR(state);
3577                 goto out;
3578         }
3579 out:
3580         nfs4_label_release_security(ilabel);
3581         put_nfs_open_context(ctx);
3582         return status;
3583 }
3584
3585 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3586 {
3587         struct nfs_server *server = NFS_SERVER(dir);
3588         struct nfs_removeargs args = {
3589                 .fh = NFS_FH(dir),
3590                 .name = *name,
3591         };
3592         struct nfs_removeres res = {
3593                 .server = server,
3594         };
3595         struct rpc_message msg = {
3596                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3597                 .rpc_argp = &args,
3598                 .rpc_resp = &res,
3599         };
3600         int status;
3601
3602         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3603         if (status == 0)
3604                 update_changeattr(dir, &res.cinfo);
3605         return status;
3606 }
3607
3608 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3609 {
3610         struct nfs4_exception exception = { };
3611         int err;
3612         do {
3613                 err = _nfs4_proc_remove(dir, name);
3614                 trace_nfs4_remove(dir, name, err);
3615                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3616                                 &exception);
3617         } while (exception.retry);
3618         return err;
3619 }
3620
3621 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3622 {
3623         struct nfs_server *server = NFS_SERVER(dir);
3624         struct nfs_removeargs *args = msg->rpc_argp;
3625         struct nfs_removeres *res = msg->rpc_resp;
3626
3627         res->server = server;
3628         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3629         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3630
3631         nfs_fattr_init(res->dir_attr);
3632 }
3633
3634 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3635 {
3636         nfs4_setup_sequence(NFS_SERVER(data->dir),
3637                         &data->args.seq_args,
3638                         &data->res.seq_res,
3639                         task);
3640 }
3641
3642 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3643 {
3644         struct nfs_unlinkdata *data = task->tk_calldata;
3645         struct nfs_removeres *res = &data->res;
3646
3647         if (!nfs4_sequence_done(task, &res->seq_res))
3648                 return 0;
3649         if (nfs4_async_handle_error(task, res->server, NULL,
3650                                     &data->timeout) == -EAGAIN)
3651                 return 0;
3652         update_changeattr(dir, &res->cinfo);
3653         return 1;
3654 }
3655
3656 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3657 {
3658         struct nfs_server *server = NFS_SERVER(dir);
3659         struct nfs_renameargs *arg = msg->rpc_argp;
3660         struct nfs_renameres *res = msg->rpc_resp;
3661
3662         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3663         res->server = server;
3664         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3665 }
3666
3667 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3668 {
3669         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3670                         &data->args.seq_args,
3671                         &data->res.seq_res,
3672                         task);
3673 }
3674
3675 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3676                                  struct inode *new_dir)
3677 {
3678         struct nfs_renamedata *data = task->tk_calldata;
3679         struct nfs_renameres *res = &data->res;
3680
3681         if (!nfs4_sequence_done(task, &res->seq_res))
3682                 return 0;
3683         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3684                 return 0;
3685
3686         update_changeattr(old_dir, &res->old_cinfo);
3687         update_changeattr(new_dir, &res->new_cinfo);
3688         return 1;
3689 }
3690
3691 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3692 {
3693         struct nfs_server *server = NFS_SERVER(inode);
3694         struct nfs4_link_arg arg = {
3695                 .fh     = NFS_FH(inode),
3696                 .dir_fh = NFS_FH(dir),
3697                 .name   = name,
3698                 .bitmask = server->attr_bitmask,
3699         };
3700         struct nfs4_link_res res = {
3701                 .server = server,
3702                 .label = NULL,
3703         };
3704         struct rpc_message msg = {
3705                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3706                 .rpc_argp = &arg,
3707                 .rpc_resp = &res,
3708         };
3709         int status = -ENOMEM;
3710
3711         res.fattr = nfs_alloc_fattr();
3712         if (res.fattr == NULL)
3713                 goto out;
3714
3715         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3716         if (IS_ERR(res.label)) {
3717                 status = PTR_ERR(res.label);
3718                 goto out;
3719         }
3720         arg.bitmask = nfs4_bitmask(server, res.label);
3721
3722         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3723         if (!status) {
3724                 update_changeattr(dir, &res.cinfo);
3725                 status = nfs_post_op_update_inode(inode, res.fattr);
3726                 if (!status)
3727                         nfs_setsecurity(inode, res.fattr, res.label);
3728         }
3729
3730
3731         nfs4_label_free(res.label);
3732
3733 out:
3734         nfs_free_fattr(res.fattr);
3735         return status;
3736 }
3737
3738 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3739 {
3740         struct nfs4_exception exception = { };
3741         int err;
3742         do {
3743                 err = nfs4_handle_exception(NFS_SERVER(inode),
3744                                 _nfs4_proc_link(inode, dir, name),
3745                                 &exception);
3746         } while (exception.retry);
3747         return err;
3748 }
3749
3750 struct nfs4_createdata {
3751         struct rpc_message msg;
3752         struct nfs4_create_arg arg;
3753         struct nfs4_create_res res;
3754         struct nfs_fh fh;
3755         struct nfs_fattr fattr;
3756         struct nfs4_label *label;
3757 };
3758
3759 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3760                 struct qstr *name, struct iattr *sattr, u32 ftype)
3761 {
3762         struct nfs4_createdata *data;
3763
3764         data = kzalloc(sizeof(*data), GFP_KERNEL);
3765         if (data != NULL) {
3766                 struct nfs_server *server = NFS_SERVER(dir);
3767
3768                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3769                 if (IS_ERR(data->label))
3770                         goto out_free;
3771
3772                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3773                 data->msg.rpc_argp = &data->arg;
3774                 data->msg.rpc_resp = &data->res;
3775                 data->arg.dir_fh = NFS_FH(dir);
3776                 data->arg.server = server;
3777                 data->arg.name = name;
3778                 data->arg.attrs = sattr;
3779                 data->arg.ftype = ftype;
3780                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3781                 data->res.server = server;
3782                 data->res.fh = &data->fh;
3783                 data->res.fattr = &data->fattr;
3784                 data->res.label = data->label;
3785                 nfs_fattr_init(data->res.fattr);
3786         }
3787         return data;
3788 out_free:
3789         kfree(data);
3790         return NULL;
3791 }
3792
3793 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3794 {
3795         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3796                                     &data->arg.seq_args, &data->res.seq_res, 1);
3797         if (status == 0) {
3798                 update_changeattr(dir, &data->res.dir_cinfo);
3799                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3800         }
3801         return status;
3802 }
3803
3804 static void nfs4_free_createdata(struct nfs4_createdata *data)
3805 {
3806         nfs4_label_free(data->label);
3807         kfree(data);
3808 }
3809
3810 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3811                 struct page *page, unsigned int len, struct iattr *sattr,
3812                 struct nfs4_label *label)
3813 {
3814         struct nfs4_createdata *data;
3815         int status = -ENAMETOOLONG;
3816
3817         if (len > NFS4_MAXPATHLEN)
3818                 goto out;
3819
3820         status = -ENOMEM;
3821         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3822         if (data == NULL)
3823                 goto out;
3824
3825         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3826         data->arg.u.symlink.pages = &page;
3827         data->arg.u.symlink.len = len;
3828         data->arg.label = label;
3829
3830         status = nfs4_do_create(dir, dentry, data);
3831
3832         nfs4_free_createdata(data);
3833 out:
3834         return status;
3835 }
3836
3837 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3838                 struct page *page, unsigned int len, struct iattr *sattr)
3839 {
3840         struct nfs4_exception exception = { };
3841         struct nfs4_label l, *label = NULL;
3842         int err;
3843
3844         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3845
3846         do {
3847                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3848                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3849                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3850                                 &exception);
3851         } while (exception.retry);
3852
3853         nfs4_label_release_security(label);
3854         return err;
3855 }
3856
3857 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3858                 struct iattr *sattr, struct nfs4_label *label)
3859 {
3860         struct nfs4_createdata *data;
3861         int status = -ENOMEM;
3862
3863         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3864         if (data == NULL)
3865                 goto out;
3866
3867         data->arg.label = label;
3868         status = nfs4_do_create(dir, dentry, data);
3869
3870         nfs4_free_createdata(data);
3871 out:
3872         return status;
3873 }
3874
3875 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3876                 struct iattr *sattr)
3877 {
3878         struct nfs4_exception exception = { };
3879         struct nfs4_label l, *label = NULL;
3880         int err;
3881
3882         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3883
3884         sattr->ia_mode &= ~current_umask();
3885         do {
3886                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3887                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3888                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3889                                 &exception);
3890         } while (exception.retry);
3891         nfs4_label_release_security(label);
3892
3893         return err;
3894 }
3895
3896 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3897                 u64 cookie, struct page **pages, unsigned int count, int plus)
3898 {
3899         struct inode            *dir = d_inode(dentry);
3900         struct nfs4_readdir_arg args = {
3901                 .fh = NFS_FH(dir),
3902                 .pages = pages,
3903                 .pgbase = 0,
3904                 .count = count,
3905                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3906                 .plus = plus,
3907         };
3908         struct nfs4_readdir_res res;
3909         struct rpc_message msg = {
3910                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3911                 .rpc_argp = &args,
3912                 .rpc_resp = &res,
3913                 .rpc_cred = cred,
3914         };
3915         int                     status;
3916
3917         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3918                         dentry,
3919                         (unsigned long long)cookie);
3920         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3921         res.pgbase = args.pgbase;
3922         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3923         if (status >= 0) {
3924                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3925                 status += args.pgbase;
3926         }
3927
3928         nfs_invalidate_atime(dir);
3929
3930         dprintk("%s: returns %d\n", __func__, status);
3931         return status;
3932 }
3933
3934 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3935                 u64 cookie, struct page **pages, unsigned int count, int plus)
3936 {
3937         struct nfs4_exception exception = { };
3938         int err;
3939         do {
3940                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3941                                 pages, count, plus);
3942                 trace_nfs4_readdir(d_inode(dentry), err);
3943                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3944                                 &exception);
3945         } while (exception.retry);
3946         return err;
3947 }
3948
3949 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3950                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3951 {
3952         struct nfs4_createdata *data;
3953         int mode = sattr->ia_mode;
3954         int status = -ENOMEM;
3955
3956         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3957         if (data == NULL)
3958                 goto out;
3959
3960         if (S_ISFIFO(mode))
3961                 data->arg.ftype = NF4FIFO;
3962         else if (S_ISBLK(mode)) {
3963                 data->arg.ftype = NF4BLK;
3964                 data->arg.u.device.specdata1 = MAJOR(rdev);
3965                 data->arg.u.device.specdata2 = MINOR(rdev);
3966         }
3967         else if (S_ISCHR(mode)) {
3968                 data->arg.ftype = NF4CHR;
3969                 data->arg.u.device.specdata1 = MAJOR(rdev);
3970                 data->arg.u.device.specdata2 = MINOR(rdev);
3971         } else if (!S_ISSOCK(mode)) {
3972                 status = -EINVAL;
3973                 goto out_free;
3974         }
3975
3976         data->arg.label = label;
3977         status = nfs4_do_create(dir, dentry, data);
3978 out_free:
3979         nfs4_free_createdata(data);
3980 out:
3981         return status;
3982 }
3983
3984 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3985                 struct iattr *sattr, dev_t rdev)
3986 {
3987         struct nfs4_exception exception = { };
3988         struct nfs4_label l, *label = NULL;
3989         int err;
3990
3991         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3992
3993         sattr->ia_mode &= ~current_umask();
3994         do {
3995                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3996                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3997                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3998                                 &exception);
3999         } while (exception.retry);
4000
4001         nfs4_label_release_security(label);
4002
4003         return err;
4004 }
4005
4006 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4007                  struct nfs_fsstat *fsstat)
4008 {
4009         struct nfs4_statfs_arg args = {
4010                 .fh = fhandle,
4011                 .bitmask = server->attr_bitmask,
4012         };
4013         struct nfs4_statfs_res res = {
4014                 .fsstat = fsstat,
4015         };
4016         struct rpc_message msg = {
4017                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4018                 .rpc_argp = &args,
4019                 .rpc_resp = &res,
4020         };
4021
4022         nfs_fattr_init(fsstat->fattr);
4023         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4024 }
4025
4026 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4027 {
4028         struct nfs4_exception exception = { };
4029         int err;
4030         do {
4031                 err = nfs4_handle_exception(server,
4032                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4033                                 &exception);
4034         } while (exception.retry);
4035         return err;
4036 }
4037
4038 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4039                 struct nfs_fsinfo *fsinfo)
4040 {
4041         struct nfs4_fsinfo_arg args = {
4042                 .fh = fhandle,
4043                 .bitmask = server->attr_bitmask,
4044         };
4045         struct nfs4_fsinfo_res res = {
4046                 .fsinfo = fsinfo,
4047         };
4048         struct rpc_message msg = {
4049                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4050                 .rpc_argp = &args,
4051                 .rpc_resp = &res,
4052         };
4053
4054         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4055 }
4056
4057 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4058 {
4059         struct nfs4_exception exception = { };
4060         unsigned long now = jiffies;
4061         int err;
4062
4063         do {
4064                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4065                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4066                 if (err == 0) {
4067                         struct nfs_client *clp = server->nfs_client;
4068
4069                         spin_lock(&clp->cl_lock);
4070                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4071                         clp->cl_last_renewal = now;
4072                         spin_unlock(&clp->cl_lock);
4073                         break;
4074                 }
4075                 err = nfs4_handle_exception(server, err, &exception);
4076         } while (exception.retry);
4077         return err;
4078 }
4079
4080 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4081 {
4082         int error;
4083
4084         nfs_fattr_init(fsinfo->fattr);
4085         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4086         if (error == 0) {
4087                 /* block layout checks this! */
4088                 server->pnfs_blksize = fsinfo->blksize;
4089                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4090         }
4091
4092         return error;
4093 }
4094
4095 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4096                 struct nfs_pathconf *pathconf)
4097 {
4098         struct nfs4_pathconf_arg args = {
4099                 .fh = fhandle,
4100                 .bitmask = server->attr_bitmask,
4101         };
4102         struct nfs4_pathconf_res res = {
4103                 .pathconf = pathconf,
4104         };
4105         struct rpc_message msg = {
4106                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4107                 .rpc_argp = &args,
4108                 .rpc_resp = &res,
4109         };
4110
4111         /* None of the pathconf attributes are mandatory to implement */
4112         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4113                 memset(pathconf, 0, sizeof(*pathconf));
4114                 return 0;
4115         }
4116
4117         nfs_fattr_init(pathconf->fattr);
4118         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4119 }
4120
4121 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4122                 struct nfs_pathconf *pathconf)
4123 {
4124         struct nfs4_exception exception = { };
4125         int err;
4126
4127         do {
4128                 err = nfs4_handle_exception(server,
4129                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4130                                 &exception);
4131         } while (exception.retry);
4132         return err;
4133 }
4134
4135 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4136                 const struct nfs_open_context *ctx,
4137                 const struct nfs_lock_context *l_ctx,
4138                 fmode_t fmode)
4139 {
4140         const struct nfs_lockowner *lockowner = NULL;
4141
4142         if (l_ctx != NULL)
4143                 lockowner = &l_ctx->lockowner;
4144         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4145 }
4146 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4147
4148 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4149                 const struct nfs_open_context *ctx,
4150                 const struct nfs_lock_context *l_ctx,
4151                 fmode_t fmode)
4152 {
4153         nfs4_stateid current_stateid;
4154
4155         /* If the current stateid represents a lost lock, then exit */
4156         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4157                 return true;
4158         return nfs4_stateid_match(stateid, &current_stateid);
4159 }
4160
4161 static bool nfs4_error_stateid_expired(int err)
4162 {
4163         switch (err) {
4164         case -NFS4ERR_DELEG_REVOKED:
4165         case -NFS4ERR_ADMIN_REVOKED:
4166         case -NFS4ERR_BAD_STATEID:
4167         case -NFS4ERR_STALE_STATEID:
4168         case -NFS4ERR_OLD_STATEID:
4169         case -NFS4ERR_OPENMODE:
4170         case -NFS4ERR_EXPIRED:
4171                 return true;
4172         }
4173         return false;
4174 }
4175
4176 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4177 {
4178         nfs_invalidate_atime(hdr->inode);
4179 }
4180
4181 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4182 {
4183         struct nfs_server *server = NFS_SERVER(hdr->inode);
4184
4185         trace_nfs4_read(hdr, task->tk_status);
4186         if (nfs4_async_handle_error(task, server,
4187                                     hdr->args.context->state,
4188                                     NULL) == -EAGAIN) {
4189                 rpc_restart_call_prepare(task);
4190                 return -EAGAIN;
4191         }
4192
4193         __nfs4_read_done_cb(hdr);
4194         if (task->tk_status > 0)
4195                 renew_lease(server, hdr->timestamp);
4196         return 0;
4197 }
4198
4199 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4200                 struct nfs_pgio_args *args)
4201 {
4202
4203         if (!nfs4_error_stateid_expired(task->tk_status) ||
4204                 nfs4_stateid_is_current(&args->stateid,
4205                                 args->context,
4206                                 args->lock_context,
4207                                 FMODE_READ))
4208                 return false;
4209         rpc_restart_call_prepare(task);
4210         return true;
4211 }
4212
4213 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4214 {
4215
4216         dprintk("--> %s\n", __func__);
4217
4218         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4219                 return -EAGAIN;
4220         if (nfs4_read_stateid_changed(task, &hdr->args))
4221                 return -EAGAIN;
4222         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4223                                     nfs4_read_done_cb(task, hdr);
4224 }
4225
4226 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4227                                  struct rpc_message *msg)
4228 {
4229         hdr->timestamp   = jiffies;
4230         hdr->pgio_done_cb = nfs4_read_done_cb;
4231         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4232         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4233 }
4234
4235 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4236                                       struct nfs_pgio_header *hdr)
4237 {
4238         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4239                         &hdr->args.seq_args,
4240                         &hdr->res.seq_res,
4241                         task))
4242                 return 0;
4243         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4244                                 hdr->args.lock_context,
4245                                 hdr->rw_ops->rw_mode) == -EIO)
4246                 return -EIO;
4247         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4248                 return -EIO;
4249         return 0;
4250 }
4251
4252 static int nfs4_write_done_cb(struct rpc_task *task,
4253                               struct nfs_pgio_header *hdr)
4254 {
4255         struct inode *inode = hdr->inode;
4256
4257         trace_nfs4_write(hdr, task->tk_status);
4258         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4259                                     hdr->args.context->state,
4260                                     NULL) == -EAGAIN) {
4261                 rpc_restart_call_prepare(task);
4262                 return -EAGAIN;
4263         }
4264         if (task->tk_status >= 0) {
4265                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4266                 nfs_writeback_update_inode(hdr);
4267         }
4268         return 0;
4269 }
4270
4271 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4272                 struct nfs_pgio_args *args)
4273 {
4274
4275         if (!nfs4_error_stateid_expired(task->tk_status) ||
4276                 nfs4_stateid_is_current(&args->stateid,
4277                                 args->context,
4278                                 args->lock_context,
4279                                 FMODE_WRITE))
4280                 return false;
4281         rpc_restart_call_prepare(task);
4282         return true;
4283 }
4284
4285 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4286 {
4287         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4288                 return -EAGAIN;
4289         if (nfs4_write_stateid_changed(task, &hdr->args))
4290                 return -EAGAIN;
4291         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4292                 nfs4_write_done_cb(task, hdr);
4293 }
4294
4295 static
4296 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4297 {
4298         /* Don't request attributes for pNFS or O_DIRECT writes */
4299         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4300                 return false;
4301         /* Otherwise, request attributes if and only if we don't hold
4302          * a delegation
4303          */
4304         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4305 }
4306
4307 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4308                                   struct rpc_message *msg)
4309 {
4310         struct nfs_server *server = NFS_SERVER(hdr->inode);
4311
4312         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4313                 hdr->args.bitmask = NULL;
4314                 hdr->res.fattr = NULL;
4315         } else
4316                 hdr->args.bitmask = server->cache_consistency_bitmask;
4317
4318         if (!hdr->pgio_done_cb)
4319                 hdr->pgio_done_cb = nfs4_write_done_cb;
4320         hdr->res.server = server;
4321         hdr->timestamp   = jiffies;
4322
4323         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4324         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4325 }
4326
4327 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4328 {
4329         nfs4_setup_sequence(NFS_SERVER(data->inode),
4330                         &data->args.seq_args,
4331                         &data->res.seq_res,
4332                         task);
4333 }
4334
4335 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4336 {
4337         struct inode *inode = data->inode;
4338
4339         trace_nfs4_commit(data, task->tk_status);
4340         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4341                                     NULL, NULL) == -EAGAIN) {
4342                 rpc_restart_call_prepare(task);
4343                 return -EAGAIN;
4344         }
4345         return 0;
4346 }
4347
4348 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4349 {
4350         if (!nfs4_sequence_done(task, &data->res.seq_res))
4351                 return -EAGAIN;
4352         return data->commit_done_cb(task, data);
4353 }
4354
4355 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4356 {
4357         struct nfs_server *server = NFS_SERVER(data->inode);
4358
4359         if (data->commit_done_cb == NULL)
4360                 data->commit_done_cb = nfs4_commit_done_cb;
4361         data->res.server = server;
4362         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4363         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4364 }
4365
4366 struct nfs4_renewdata {
4367         struct nfs_client       *client;
4368         unsigned long           timestamp;
4369 };
4370
4371 /*
4372  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4373  * standalone procedure for queueing an asynchronous RENEW.
4374  */
4375 static void nfs4_renew_release(void *calldata)
4376 {
4377         struct nfs4_renewdata *data = calldata;
4378         struct nfs_client *clp = data->client;
4379
4380         if (atomic_read(&clp->cl_count) > 1)
4381                 nfs4_schedule_state_renewal(clp);
4382         nfs_put_client(clp);
4383         kfree(data);
4384 }
4385
4386 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4387 {
4388         struct nfs4_renewdata *data = calldata;
4389         struct nfs_client *clp = data->client;
4390         unsigned long timestamp = data->timestamp;
4391
4392         trace_nfs4_renew_async(clp, task->tk_status);
4393         switch (task->tk_status) {
4394         case 0:
4395                 break;
4396         case -NFS4ERR_LEASE_MOVED:
4397                 nfs4_schedule_lease_moved_recovery(clp);
4398                 break;
4399         default:
4400                 /* Unless we're shutting down, schedule state recovery! */
4401                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4402                         return;
4403                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4404                         nfs4_schedule_lease_recovery(clp);
4405                         return;
4406                 }
4407                 nfs4_schedule_path_down_recovery(clp);
4408         }
4409         do_renew_lease(clp, timestamp);
4410 }
4411
4412 static const struct rpc_call_ops nfs4_renew_ops = {
4413         .rpc_call_done = nfs4_renew_done,
4414         .rpc_release = nfs4_renew_release,
4415 };
4416
4417 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4418 {
4419         struct rpc_message msg = {
4420                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4421                 .rpc_argp       = clp,
4422                 .rpc_cred       = cred,
4423         };
4424         struct nfs4_renewdata *data;
4425
4426         if (renew_flags == 0)
4427                 return 0;
4428         if (!atomic_inc_not_zero(&clp->cl_count))
4429                 return -EIO;
4430         data = kmalloc(sizeof(*data), GFP_NOFS);
4431         if (data == NULL)
4432                 return -ENOMEM;
4433         data->client = clp;
4434         data->timestamp = jiffies;
4435         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4436                         &nfs4_renew_ops, data);
4437 }
4438
4439 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4440 {
4441         struct rpc_message msg = {
4442                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4443                 .rpc_argp       = clp,
4444                 .rpc_cred       = cred,
4445         };
4446         unsigned long now = jiffies;
4447         int status;
4448
4449         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4450         if (status < 0)
4451                 return status;
4452         do_renew_lease(clp, now);
4453         return 0;
4454 }
4455
4456 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4457 {
4458         return server->caps & NFS_CAP_ACLS;
4459 }
4460
4461 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4462  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4463  * the stack.
4464  */
4465 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4466
4467 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4468                 struct page **pages, unsigned int *pgbase)
4469 {
4470         struct page *newpage, **spages;
4471         int rc = 0;
4472         size_t len;
4473         spages = pages;
4474
4475         do {
4476                 len = min_t(size_t, PAGE_SIZE, buflen);
4477                 newpage = alloc_page(GFP_KERNEL);
4478
4479                 if (newpage == NULL)
4480                         goto unwind;
4481                 memcpy(page_address(newpage), buf, len);
4482                 buf += len;
4483                 buflen -= len;
4484                 *pages++ = newpage;
4485                 rc++;
4486         } while (buflen != 0);
4487
4488         return rc;
4489
4490 unwind:
4491         for(; rc > 0; rc--)
4492                 __free_page(spages[rc-1]);
4493         return -ENOMEM;
4494 }
4495
4496 struct nfs4_cached_acl {
4497         int cached;
4498         size_t len;
4499         char data[0];
4500 };
4501
4502 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4503 {
4504         struct nfs_inode *nfsi = NFS_I(inode);
4505
4506         spin_lock(&inode->i_lock);
4507         kfree(nfsi->nfs4_acl);
4508         nfsi->nfs4_acl = acl;
4509         spin_unlock(&inode->i_lock);
4510 }
4511
4512 static void nfs4_zap_acl_attr(struct inode *inode)
4513 {
4514         nfs4_set_cached_acl(inode, NULL);
4515 }
4516
4517 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4518 {
4519         struct nfs_inode *nfsi = NFS_I(inode);
4520         struct nfs4_cached_acl *acl;
4521         int ret = -ENOENT;
4522
4523         spin_lock(&inode->i_lock);
4524         acl = nfsi->nfs4_acl;
4525         if (acl == NULL)
4526                 goto out;
4527         if (buf == NULL) /* user is just asking for length */
4528                 goto out_len;
4529         if (acl->cached == 0)
4530                 goto out;
4531         ret = -ERANGE; /* see getxattr(2) man page */
4532         if (acl->len > buflen)
4533                 goto out;
4534         memcpy(buf, acl->data, acl->len);
4535 out_len:
4536         ret = acl->len;
4537 out:
4538         spin_unlock(&inode->i_lock);
4539         return ret;
4540 }
4541
4542 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4543 {
4544         struct nfs4_cached_acl *acl;
4545         size_t buflen = sizeof(*acl) + acl_len;
4546
4547         if (buflen <= PAGE_SIZE) {
4548                 acl = kmalloc(buflen, GFP_KERNEL);
4549                 if (acl == NULL)
4550                         goto out;
4551                 acl->cached = 1;
4552                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4553         } else {
4554                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4555                 if (acl == NULL)
4556                         goto out;
4557                 acl->cached = 0;
4558         }
4559         acl->len = acl_len;
4560 out:
4561         nfs4_set_cached_acl(inode, acl);
4562 }
4563
4564 /*
4565  * The getxattr API returns the required buffer length when called with a
4566  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4567  * the required buf.  On a NULL buf, we send a page of data to the server
4568  * guessing that the ACL request can be serviced by a page. If so, we cache
4569  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4570  * the cache. If not so, we throw away the page, and cache the required
4571  * length. The next getxattr call will then produce another round trip to
4572  * the server, this time with the input buf of the required size.
4573  */
4574 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4575 {
4576         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4577         struct nfs_getaclargs args = {
4578                 .fh = NFS_FH(inode),
4579                 .acl_pages = pages,
4580                 .acl_len = buflen,
4581         };
4582         struct nfs_getaclres res = {
4583                 .acl_len = buflen,
4584         };
4585         struct rpc_message msg = {
4586                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4587                 .rpc_argp = &args,
4588                 .rpc_resp = &res,
4589         };
4590         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4591         int ret = -ENOMEM, i;
4592
4593         /* As long as we're doing a round trip to the server anyway,
4594          * let's be prepared for a page of acl data. */
4595         if (npages == 0)
4596                 npages = 1;
4597         if (npages > ARRAY_SIZE(pages))
4598                 return -ERANGE;
4599
4600         for (i = 0; i < npages; i++) {
4601                 pages[i] = alloc_page(GFP_KERNEL);
4602                 if (!pages[i])
4603                         goto out_free;
4604         }
4605
4606         /* for decoding across pages */
4607         res.acl_scratch = alloc_page(GFP_KERNEL);
4608         if (!res.acl_scratch)
4609                 goto out_free;
4610
4611         args.acl_len = npages * PAGE_SIZE;
4612         args.acl_pgbase = 0;
4613
4614         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4615                 __func__, buf, buflen, npages, args.acl_len);
4616         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4617                              &msg, &args.seq_args, &res.seq_res, 0);
4618         if (ret)
4619                 goto out_free;
4620
4621         /* Handle the case where the passed-in buffer is too short */
4622         if (res.acl_flags & NFS4_ACL_TRUNC) {
4623                 /* Did the user only issue a request for the acl length? */
4624                 if (buf == NULL)
4625                         goto out_ok;
4626                 ret = -ERANGE;
4627                 goto out_free;
4628         }
4629         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4630         if (buf) {
4631                 if (res.acl_len > buflen) {
4632                         ret = -ERANGE;
4633                         goto out_free;
4634                 }
4635                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4636         }
4637 out_ok:
4638         ret = res.acl_len;
4639 out_free:
4640         for (i = 0; i < npages; i++)
4641                 if (pages[i])
4642                         __free_page(pages[i]);
4643         if (res.acl_scratch)
4644                 __free_page(res.acl_scratch);
4645         return ret;
4646 }
4647
4648 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4649 {
4650         struct nfs4_exception exception = { };
4651         ssize_t ret;
4652         do {
4653                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4654                 trace_nfs4_get_acl(inode, ret);
4655                 if (ret >= 0)
4656                         break;
4657                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4658         } while (exception.retry);
4659         return ret;
4660 }
4661
4662 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4663 {
4664         struct nfs_server *server = NFS_SERVER(inode);
4665         int ret;
4666
4667         if (!nfs4_server_supports_acls(server))
4668                 return -EOPNOTSUPP;
4669         ret = nfs_revalidate_inode(server, inode);
4670         if (ret < 0)
4671                 return ret;
4672         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4673                 nfs_zap_acl_cache(inode);
4674         ret = nfs4_read_cached_acl(inode, buf, buflen);
4675         if (ret != -ENOENT)
4676                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4677                  * but no cached acl data, just the acl length */
4678                 return ret;
4679         return nfs4_get_acl_uncached(inode, buf, buflen);
4680 }
4681
4682 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4683 {
4684         struct nfs_server *server = NFS_SERVER(inode);
4685         struct page *pages[NFS4ACL_MAXPAGES];
4686         struct nfs_setaclargs arg = {
4687                 .fh             = NFS_FH(inode),
4688                 .acl_pages      = pages,
4689                 .acl_len        = buflen,
4690         };
4691         struct nfs_setaclres res;
4692         struct rpc_message msg = {
4693                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4694                 .rpc_argp       = &arg,
4695                 .rpc_resp       = &res,
4696         };
4697         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4698         int ret, i;
4699
4700         if (!nfs4_server_supports_acls(server))
4701                 return -EOPNOTSUPP;
4702         if (npages > ARRAY_SIZE(pages))
4703                 return -ERANGE;
4704         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4705         if (i < 0)
4706                 return i;
4707         nfs4_inode_return_delegation(inode);
4708         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4709
4710         /*
4711          * Free each page after tx, so the only ref left is
4712          * held by the network stack
4713          */
4714         for (; i > 0; i--)
4715                 put_page(pages[i-1]);
4716
4717         /*
4718          * Acl update can result in inode attribute update.
4719          * so mark the attribute cache invalid.
4720          */
4721         spin_lock(&inode->i_lock);
4722         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4723         spin_unlock(&inode->i_lock);
4724         nfs_access_zap_cache(inode);
4725         nfs_zap_acl_cache(inode);
4726         return ret;
4727 }
4728
4729 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4730 {
4731         struct nfs4_exception exception = { };
4732         int err;
4733         do {
4734                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4735                 trace_nfs4_set_acl(inode, err);
4736                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4737                                 &exception);
4738         } while (exception.retry);
4739         return err;
4740 }
4741
4742 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4743 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4744                                         size_t buflen)
4745 {
4746         struct nfs_server *server = NFS_SERVER(inode);
4747         struct nfs_fattr fattr;
4748         struct nfs4_label label = {0, 0, buflen, buf};
4749
4750         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4751         struct nfs4_getattr_arg arg = {
4752                 .fh             = NFS_FH(inode),
4753                 .bitmask        = bitmask,
4754         };
4755         struct nfs4_getattr_res res = {
4756                 .fattr          = &fattr,
4757                 .label          = &label,
4758                 .server         = server,
4759         };
4760         struct rpc_message msg = {
4761                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4762                 .rpc_argp       = &arg,
4763                 .rpc_resp       = &res,
4764         };
4765         int ret;
4766
4767         nfs_fattr_init(&fattr);
4768
4769         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4770         if (ret)
4771                 return ret;
4772         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4773                 return -ENOENT;
4774         if (buflen < label.len)
4775                 return -ERANGE;
4776         return 0;
4777 }
4778
4779 static int nfs4_get_security_label(struct inode *inode, void *buf,
4780                                         size_t buflen)
4781 {
4782         struct nfs4_exception exception = { };
4783         int err;
4784
4785         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4786                 return -EOPNOTSUPP;
4787
4788         do {
4789                 err = _nfs4_get_security_label(inode, buf, buflen);
4790                 trace_nfs4_get_security_label(inode, err);
4791                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4792                                 &exception);
4793         } while (exception.retry);
4794         return err;
4795 }
4796
4797 static int _nfs4_do_set_security_label(struct inode *inode,
4798                 struct nfs4_label *ilabel,
4799                 struct nfs_fattr *fattr,
4800                 struct nfs4_label *olabel)
4801 {
4802
4803         struct iattr sattr = {0};
4804         struct nfs_server *server = NFS_SERVER(inode);
4805         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4806         struct nfs_setattrargs arg = {
4807                 .fh             = NFS_FH(inode),
4808                 .iap            = &sattr,
4809                 .server         = server,
4810                 .bitmask        = bitmask,
4811                 .label          = ilabel,
4812         };
4813         struct nfs_setattrres res = {
4814                 .fattr          = fattr,
4815                 .label          = olabel,
4816                 .server         = server,
4817         };
4818         struct rpc_message msg = {
4819                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4820                 .rpc_argp       = &arg,
4821                 .rpc_resp       = &res,
4822         };
4823         int status;
4824
4825         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4826
4827         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4828         if (status)
4829                 dprintk("%s failed: %d\n", __func__, status);
4830
4831         return status;
4832 }
4833
4834 static int nfs4_do_set_security_label(struct inode *inode,
4835                 struct nfs4_label *ilabel,
4836                 struct nfs_fattr *fattr,
4837                 struct nfs4_label *olabel)
4838 {
4839         struct nfs4_exception exception = { };
4840         int err;
4841
4842         do {
4843                 err = _nfs4_do_set_security_label(inode, ilabel,
4844                                 fattr, olabel);
4845                 trace_nfs4_set_security_label(inode, err);
4846                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4847                                 &exception);
4848         } while (exception.retry);
4849         return err;
4850 }
4851
4852 static int
4853 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4854 {
4855         struct nfs4_label ilabel, *olabel = NULL;
4856         struct nfs_fattr fattr;
4857         struct rpc_cred *cred;
4858         struct inode *inode = d_inode(dentry);
4859         int status;
4860
4861         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4862                 return -EOPNOTSUPP;
4863
4864         nfs_fattr_init(&fattr);
4865
4866         ilabel.pi = 0;
4867         ilabel.lfs = 0;
4868         ilabel.label = (char *)buf;
4869         ilabel.len = buflen;
4870
4871         cred = rpc_lookup_cred();
4872         if (IS_ERR(cred))
4873                 return PTR_ERR(cred);
4874
4875         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4876         if (IS_ERR(olabel)) {
4877                 status = -PTR_ERR(olabel);
4878                 goto out;
4879         }
4880
4881         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4882         if (status == 0)
4883                 nfs_setsecurity(inode, &fattr, olabel);
4884
4885         nfs4_label_free(olabel);
4886 out:
4887         put_rpccred(cred);
4888         return status;
4889 }
4890 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4891
4892
4893 static int
4894 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4895                         struct nfs4_state *state, long *timeout)
4896 {
4897         struct nfs_client *clp = server->nfs_client;
4898
4899         if (task->tk_status >= 0)
4900                 return 0;
4901         switch(task->tk_status) {
4902                 case -NFS4ERR_DELEG_REVOKED:
4903                 case -NFS4ERR_ADMIN_REVOKED:
4904                 case -NFS4ERR_BAD_STATEID:
4905                 case -NFS4ERR_OPENMODE:
4906                         if (state == NULL)
4907                                 break;
4908                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4909                                 goto recovery_failed;
4910                         goto wait_on_recovery;
4911                 case -NFS4ERR_EXPIRED:
4912                         if (state != NULL) {
4913                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4914                                         goto recovery_failed;
4915                         }
4916                 case -NFS4ERR_STALE_STATEID:
4917                 case -NFS4ERR_STALE_CLIENTID:
4918                         nfs4_schedule_lease_recovery(clp);
4919                         goto wait_on_recovery;
4920                 case -NFS4ERR_MOVED:
4921                         if (nfs4_schedule_migration_recovery(server) < 0)
4922                                 goto recovery_failed;
4923                         goto wait_on_recovery;
4924                 case -NFS4ERR_LEASE_MOVED:
4925                         nfs4_schedule_lease_moved_recovery(clp);
4926                         goto wait_on_recovery;
4927 #if defined(CONFIG_NFS_V4_1)
4928                 case -NFS4ERR_BADSESSION:
4929                 case -NFS4ERR_BADSLOT:
4930                 case -NFS4ERR_BAD_HIGH_SLOT:
4931                 case -NFS4ERR_DEADSESSION:
4932                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4933                 case -NFS4ERR_SEQ_FALSE_RETRY:
4934                 case -NFS4ERR_SEQ_MISORDERED:
4935                         dprintk("%s ERROR %d, Reset session\n", __func__,
4936                                 task->tk_status);
4937                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4938                         goto wait_on_recovery;
4939 #endif /* CONFIG_NFS_V4_1 */
4940                 case -NFS4ERR_DELAY:
4941                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4942                         rpc_delay(task, nfs4_update_delay(timeout));
4943                         goto restart_call;
4944                 case -NFS4ERR_GRACE:
4945                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4946                 case -NFS4ERR_RETRY_UNCACHED_REP:
4947                 case -NFS4ERR_OLD_STATEID:
4948                         goto restart_call;
4949         }
4950         task->tk_status = nfs4_map_errors(task->tk_status);
4951         return 0;
4952 recovery_failed:
4953         task->tk_status = -EIO;
4954         return 0;
4955 wait_on_recovery:
4956         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4957         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4958                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4959         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4960                 goto recovery_failed;
4961 restart_call:
4962         task->tk_status = 0;
4963         return -EAGAIN;
4964 }
4965
4966 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4967                                     nfs4_verifier *bootverf)
4968 {
4969         __be32 verf[2];
4970
4971         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4972                 /* An impossible timestamp guarantees this value
4973                  * will never match a generated boot time. */
4974                 verf[0] = 0;
4975                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4976         } else {
4977                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4978                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4979                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4980         }
4981         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4982 }
4983
4984 static int
4985 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
4986 {
4987         int result;
4988         size_t len;
4989         char *str;
4990         bool retried = false;
4991
4992         if (clp->cl_owner_id != NULL)
4993                 return 0;
4994 retry:
4995         rcu_read_lock();
4996         len = 10 + strlen(clp->cl_ipaddr) + 1 +
4997                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
4998                 1 +
4999                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5000                 1;
5001         rcu_read_unlock();
5002
5003         if (len > NFS4_OPAQUE_LIMIT + 1)
5004                 return -EINVAL;
5005
5006         /*
5007          * Since this string is allocated at mount time, and held until the
5008          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5009          * about a memory-reclaim deadlock.
5010          */
5011         str = kmalloc(len, GFP_KERNEL);
5012         if (!str)
5013                 return -ENOMEM;
5014
5015         rcu_read_lock();
5016         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5017                         clp->cl_ipaddr,
5018                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5019                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5020         rcu_read_unlock();
5021
5022         /* Did something change? */
5023         if (result >= len) {
5024                 kfree(str);
5025                 if (retried)
5026                         return -EINVAL;
5027                 retried = true;
5028                 goto retry;
5029         }
5030         clp->cl_owner_id = str;
5031         return 0;
5032 }
5033
5034 static int
5035 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5036 {
5037         int result;
5038         size_t len;
5039         char *str;
5040
5041         len = 10 + 10 + 1 + 10 + 1 +
5042                 strlen(nfs4_client_id_uniquifier) + 1 +
5043                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5044
5045         if (len > NFS4_OPAQUE_LIMIT + 1)
5046                 return -EINVAL;
5047
5048         /*
5049          * Since this string is allocated at mount time, and held until the
5050          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5051          * about a memory-reclaim deadlock.
5052          */
5053         str = kmalloc(len, GFP_KERNEL);
5054         if (!str)
5055                 return -ENOMEM;
5056
5057         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5058                         clp->rpc_ops->version, clp->cl_minorversion,
5059                         nfs4_client_id_uniquifier,
5060                         clp->cl_rpcclient->cl_nodename);
5061         if (result >= len) {
5062                 kfree(str);
5063                 return -EINVAL;
5064         }
5065         clp->cl_owner_id = str;
5066         return 0;
5067 }
5068
5069 static int
5070 nfs4_init_uniform_client_string(struct nfs_client *clp)
5071 {
5072         int result;
5073         size_t len;
5074         char *str;
5075
5076         if (clp->cl_owner_id != NULL)
5077                 return 0;
5078
5079         if (nfs4_client_id_uniquifier[0] != '\0')
5080                 return nfs4_init_uniquifier_client_string(clp);
5081
5082         len = 10 + 10 + 1 + 10 + 1 +
5083                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5084
5085         if (len > NFS4_OPAQUE_LIMIT + 1)
5086                 return -EINVAL;
5087
5088         /*
5089          * Since this string is allocated at mount time, and held until the
5090          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5091          * about a memory-reclaim deadlock.
5092          */
5093         str = kmalloc(len, GFP_KERNEL);
5094         if (!str)
5095                 return -ENOMEM;
5096
5097         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5098                         clp->rpc_ops->version, clp->cl_minorversion,
5099                         clp->cl_rpcclient->cl_nodename);
5100         if (result >= len) {
5101                 kfree(str);
5102                 return -EINVAL;
5103         }
5104         clp->cl_owner_id = str;
5105         return 0;
5106 }
5107
5108 /*
5109  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5110  * services.  Advertise one based on the address family of the
5111  * clientaddr.
5112  */
5113 static unsigned int
5114 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5115 {
5116         if (strchr(clp->cl_ipaddr, ':') != NULL)
5117                 return scnprintf(buf, len, "tcp6");
5118         else
5119                 return scnprintf(buf, len, "tcp");
5120 }
5121
5122 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5123 {
5124         struct nfs4_setclientid *sc = calldata;
5125
5126         if (task->tk_status == 0)
5127                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5128 }
5129
5130 static const struct rpc_call_ops nfs4_setclientid_ops = {
5131         .rpc_call_done = nfs4_setclientid_done,
5132 };
5133
5134 /**
5135  * nfs4_proc_setclientid - Negotiate client ID
5136  * @clp: state data structure
5137  * @program: RPC program for NFSv4 callback service
5138  * @port: IP port number for NFS4 callback service
5139  * @cred: RPC credential to use for this call
5140  * @res: where to place the result
5141  *
5142  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5143  */
5144 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5145                 unsigned short port, struct rpc_cred *cred,
5146                 struct nfs4_setclientid_res *res)
5147 {
5148         nfs4_verifier sc_verifier;
5149         struct nfs4_setclientid setclientid = {
5150                 .sc_verifier = &sc_verifier,
5151                 .sc_prog = program,
5152                 .sc_clnt = clp,
5153         };
5154         struct rpc_message msg = {
5155                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5156                 .rpc_argp = &setclientid,
5157                 .rpc_resp = res,
5158                 .rpc_cred = cred,
5159         };
5160         struct rpc_task *task;
5161         struct rpc_task_setup task_setup_data = {
5162                 .rpc_client = clp->cl_rpcclient,
5163                 .rpc_message = &msg,
5164                 .callback_ops = &nfs4_setclientid_ops,
5165                 .callback_data = &setclientid,
5166                 .flags = RPC_TASK_TIMEOUT,
5167         };
5168         int status;
5169
5170         /* nfs_client_id4 */
5171         nfs4_init_boot_verifier(clp, &sc_verifier);
5172
5173         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5174                 status = nfs4_init_uniform_client_string(clp);
5175         else
5176                 status = nfs4_init_nonuniform_client_string(clp);
5177
5178         if (status)
5179                 goto out;
5180
5181         /* cb_client4 */
5182         setclientid.sc_netid_len =
5183                                 nfs4_init_callback_netid(clp,
5184                                                 setclientid.sc_netid,
5185                                                 sizeof(setclientid.sc_netid));
5186         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5187                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5188                                 clp->cl_ipaddr, port >> 8, port & 255);
5189
5190         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5191                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5192                 clp->cl_owner_id);
5193         task = rpc_run_task(&task_setup_data);
5194         if (IS_ERR(task)) {
5195                 status = PTR_ERR(task);
5196                 goto out;
5197         }
5198         status = task->tk_status;
5199         if (setclientid.sc_cred) {
5200                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5201                 put_rpccred(setclientid.sc_cred);
5202         }
5203         rpc_put_task(task);
5204 out:
5205         trace_nfs4_setclientid(clp, status);
5206         dprintk("NFS reply setclientid: %d\n", status);
5207         return status;
5208 }
5209
5210 /**
5211  * nfs4_proc_setclientid_confirm - Confirm client ID
5212  * @clp: state data structure
5213  * @res: result of a previous SETCLIENTID
5214  * @cred: RPC credential to use for this call
5215  *
5216  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5217  */
5218 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5219                 struct nfs4_setclientid_res *arg,
5220                 struct rpc_cred *cred)
5221 {
5222         struct rpc_message msg = {
5223                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5224                 .rpc_argp = arg,
5225                 .rpc_cred = cred,
5226         };
5227         int status;
5228
5229         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5230                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5231                 clp->cl_clientid);
5232         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5233         trace_nfs4_setclientid_confirm(clp, status);
5234         dprintk("NFS reply setclientid_confirm: %d\n", status);
5235         return status;
5236 }
5237
5238 struct nfs4_delegreturndata {
5239         struct nfs4_delegreturnargs args;
5240         struct nfs4_delegreturnres res;
5241         struct nfs_fh fh;
5242         nfs4_stateid stateid;
5243         unsigned long timestamp;
5244         struct nfs_fattr fattr;
5245         int rpc_status;
5246         struct inode *inode;
5247         bool roc;
5248         u32 roc_barrier;
5249 };
5250
5251 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5252 {
5253         struct nfs4_delegreturndata *data = calldata;
5254
5255         if (!nfs4_sequence_done(task, &data->res.seq_res))
5256                 return;
5257
5258         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5259         switch (task->tk_status) {
5260         case 0:
5261                 renew_lease(data->res.server, data->timestamp);
5262         case -NFS4ERR_ADMIN_REVOKED:
5263         case -NFS4ERR_DELEG_REVOKED:
5264         case -NFS4ERR_BAD_STATEID:
5265         case -NFS4ERR_OLD_STATEID:
5266         case -NFS4ERR_STALE_STATEID:
5267         case -NFS4ERR_EXPIRED:
5268                 task->tk_status = 0;
5269                 if (data->roc)
5270                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5271                 break;
5272         default:
5273                 if (nfs4_async_handle_error(task, data->res.server,
5274                                             NULL, NULL) == -EAGAIN) {
5275                         rpc_restart_call_prepare(task);
5276                         return;
5277                 }
5278         }
5279         data->rpc_status = task->tk_status;
5280 }
5281
5282 static void nfs4_delegreturn_release(void *calldata)
5283 {
5284         struct nfs4_delegreturndata *data = calldata;
5285         struct inode *inode = data->inode;
5286
5287         if (inode) {
5288                 if (data->roc)
5289                         pnfs_roc_release(inode);
5290                 nfs_iput_and_deactive(inode);
5291         }
5292         kfree(calldata);
5293 }
5294
5295 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5296 {
5297         struct nfs4_delegreturndata *d_data;
5298
5299         d_data = (struct nfs4_delegreturndata *)data;
5300
5301         if (d_data->roc)
5302                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5303
5304         nfs4_setup_sequence(d_data->res.server,
5305                         &d_data->args.seq_args,
5306                         &d_data->res.seq_res,
5307                         task);
5308 }
5309
5310 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5311         .rpc_call_prepare = nfs4_delegreturn_prepare,
5312         .rpc_call_done = nfs4_delegreturn_done,
5313         .rpc_release = nfs4_delegreturn_release,
5314 };
5315
5316 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5317 {
5318         struct nfs4_delegreturndata *data;
5319         struct nfs_server *server = NFS_SERVER(inode);
5320         struct rpc_task *task;
5321         struct rpc_message msg = {
5322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5323                 .rpc_cred = cred,
5324         };
5325         struct rpc_task_setup task_setup_data = {
5326                 .rpc_client = server->client,
5327                 .rpc_message = &msg,
5328                 .callback_ops = &nfs4_delegreturn_ops,
5329                 .flags = RPC_TASK_ASYNC,
5330         };
5331         int status = 0;
5332
5333         data = kzalloc(sizeof(*data), GFP_NOFS);
5334         if (data == NULL)
5335                 return -ENOMEM;
5336         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5337         data->args.fhandle = &data->fh;
5338         data->args.stateid = &data->stateid;
5339         data->args.bitmask = server->cache_consistency_bitmask;
5340         nfs_copy_fh(&data->fh, NFS_FH(inode));
5341         nfs4_stateid_copy(&data->stateid, stateid);
5342         data->res.fattr = &data->fattr;
5343         data->res.server = server;
5344         nfs_fattr_init(data->res.fattr);
5345         data->timestamp = jiffies;
5346         data->rpc_status = 0;
5347         data->inode = nfs_igrab_and_active(inode);
5348         if (data->inode)
5349                 data->roc = nfs4_roc(inode);
5350
5351         task_setup_data.callback_data = data;
5352         msg.rpc_argp = &data->args;
5353         msg.rpc_resp = &data->res;
5354         task = rpc_run_task(&task_setup_data);
5355         if (IS_ERR(task))
5356                 return PTR_ERR(task);
5357         if (!issync)
5358                 goto out;
5359         status = nfs4_wait_for_completion_rpc_task(task);
5360         if (status != 0)
5361                 goto out;
5362         status = data->rpc_status;
5363         if (status == 0)
5364                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5365         else
5366                 nfs_refresh_inode(inode, &data->fattr);
5367 out:
5368         rpc_put_task(task);
5369         return status;
5370 }
5371
5372 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5373 {
5374         struct nfs_server *server = NFS_SERVER(inode);
5375         struct nfs4_exception exception = { };
5376         int err;
5377         do {
5378                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5379                 trace_nfs4_delegreturn(inode, err);
5380                 switch (err) {
5381                         case -NFS4ERR_STALE_STATEID:
5382                         case -NFS4ERR_EXPIRED:
5383                         case 0:
5384                                 return 0;
5385                 }
5386                 err = nfs4_handle_exception(server, err, &exception);
5387         } while (exception.retry);
5388         return err;
5389 }
5390
5391 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5392 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5393
5394 /*
5395  * sleep, with exponential backoff, and retry the LOCK operation.
5396  */
5397 static unsigned long
5398 nfs4_set_lock_task_retry(unsigned long timeout)
5399 {
5400         freezable_schedule_timeout_killable_unsafe(timeout);
5401         timeout <<= 1;
5402         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5403                 return NFS4_LOCK_MAXTIMEOUT;
5404         return timeout;
5405 }
5406
5407 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5408 {
5409         struct inode *inode = state->inode;
5410         struct nfs_server *server = NFS_SERVER(inode);
5411         struct nfs_client *clp = server->nfs_client;
5412         struct nfs_lockt_args arg = {
5413                 .fh = NFS_FH(inode),
5414                 .fl = request,
5415         };
5416         struct nfs_lockt_res res = {
5417                 .denied = request,
5418         };
5419         struct rpc_message msg = {
5420                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5421                 .rpc_argp       = &arg,
5422                 .rpc_resp       = &res,
5423                 .rpc_cred       = state->owner->so_cred,
5424         };
5425         struct nfs4_lock_state *lsp;
5426         int status;
5427
5428         arg.lock_owner.clientid = clp->cl_clientid;
5429         status = nfs4_set_lock_state(state, request);
5430         if (status != 0)
5431                 goto out;
5432         lsp = request->fl_u.nfs4_fl.owner;
5433         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5434         arg.lock_owner.s_dev = server->s_dev;
5435         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5436         switch (status) {
5437                 case 0:
5438                         request->fl_type = F_UNLCK;
5439                         break;
5440                 case -NFS4ERR_DENIED:
5441                         status = 0;
5442         }
5443         request->fl_ops->fl_release_private(request);
5444         request->fl_ops = NULL;
5445 out:
5446         return status;
5447 }
5448
5449 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5450 {
5451         struct nfs4_exception exception = { };
5452         int err;
5453
5454         do {
5455                 err = _nfs4_proc_getlk(state, cmd, request);
5456                 trace_nfs4_get_lock(request, state, cmd, err);
5457                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5458                                 &exception);
5459         } while (exception.retry);
5460         return err;
5461 }
5462
5463 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5464 {
5465         int res = 0;
5466         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5467                 case FL_POSIX:
5468                         res = posix_lock_inode_wait(inode, fl);
5469                         break;
5470                 case FL_FLOCK:
5471                         res = flock_lock_inode_wait(inode, fl);
5472                         break;
5473                 default:
5474                         BUG();
5475         }
5476         return res;
5477 }
5478
5479 struct nfs4_unlockdata {
5480         struct nfs_locku_args arg;
5481         struct nfs_locku_res res;
5482         struct nfs4_lock_state *lsp;
5483         struct nfs_open_context *ctx;
5484         struct file_lock fl;
5485         const struct nfs_server *server;
5486         unsigned long timestamp;
5487 };
5488
5489 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5490                 struct nfs_open_context *ctx,
5491                 struct nfs4_lock_state *lsp,
5492                 struct nfs_seqid *seqid)
5493 {
5494         struct nfs4_unlockdata *p;
5495         struct inode *inode = lsp->ls_state->inode;
5496
5497         p = kzalloc(sizeof(*p), GFP_NOFS);
5498         if (p == NULL)
5499                 return NULL;
5500         p->arg.fh = NFS_FH(inode);
5501         p->arg.fl = &p->fl;
5502         p->arg.seqid = seqid;
5503         p->res.seqid = seqid;
5504         p->lsp = lsp;
5505         atomic_inc(&lsp->ls_count);
5506         /* Ensure we don't close file until we're done freeing locks! */
5507         p->ctx = get_nfs_open_context(ctx);
5508         memcpy(&p->fl, fl, sizeof(p->fl));
5509         p->server = NFS_SERVER(inode);
5510         return p;
5511 }
5512
5513 static void nfs4_locku_release_calldata(void *data)
5514 {
5515         struct nfs4_unlockdata *calldata = data;
5516         nfs_free_seqid(calldata->arg.seqid);
5517         nfs4_put_lock_state(calldata->lsp);
5518         put_nfs_open_context(calldata->ctx);
5519         kfree(calldata);
5520 }
5521
5522 static void nfs4_locku_done(struct rpc_task *task, void *data)
5523 {
5524         struct nfs4_unlockdata *calldata = data;
5525
5526         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5527                 return;
5528         switch (task->tk_status) {
5529                 case 0:
5530                         renew_lease(calldata->server, calldata->timestamp);
5531                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5532                         if (nfs4_update_lock_stateid(calldata->lsp,
5533                                         &calldata->res.stateid))
5534                                 break;
5535                 case -NFS4ERR_BAD_STATEID:
5536                 case -NFS4ERR_OLD_STATEID:
5537                 case -NFS4ERR_STALE_STATEID:
5538                 case -NFS4ERR_EXPIRED:
5539                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5540                                                 &calldata->lsp->ls_stateid))
5541                                 rpc_restart_call_prepare(task);
5542                         break;
5543                 default:
5544                         if (nfs4_async_handle_error(task, calldata->server,
5545                                                     NULL, NULL) == -EAGAIN)
5546                                 rpc_restart_call_prepare(task);
5547         }
5548         nfs_release_seqid(calldata->arg.seqid);
5549 }
5550
5551 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5552 {
5553         struct nfs4_unlockdata *calldata = data;
5554
5555         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5556                 goto out_wait;
5557         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5558         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5559                 /* Note: exit _without_ running nfs4_locku_done */
5560                 goto out_no_action;
5561         }
5562         calldata->timestamp = jiffies;
5563         if (nfs4_setup_sequence(calldata->server,
5564                                 &calldata->arg.seq_args,
5565                                 &calldata->res.seq_res,
5566                                 task) != 0)
5567                 nfs_release_seqid(calldata->arg.seqid);
5568         return;
5569 out_no_action:
5570         task->tk_action = NULL;
5571 out_wait:
5572         nfs4_sequence_done(task, &calldata->res.seq_res);
5573 }
5574
5575 static const struct rpc_call_ops nfs4_locku_ops = {
5576         .rpc_call_prepare = nfs4_locku_prepare,
5577         .rpc_call_done = nfs4_locku_done,
5578         .rpc_release = nfs4_locku_release_calldata,
5579 };
5580
5581 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5582                 struct nfs_open_context *ctx,
5583                 struct nfs4_lock_state *lsp,
5584                 struct nfs_seqid *seqid)
5585 {
5586         struct nfs4_unlockdata *data;
5587         struct rpc_message msg = {
5588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5589                 .rpc_cred = ctx->cred,
5590         };
5591         struct rpc_task_setup task_setup_data = {
5592                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5593                 .rpc_message = &msg,
5594                 .callback_ops = &nfs4_locku_ops,
5595                 .workqueue = nfsiod_workqueue,
5596                 .flags = RPC_TASK_ASYNC,
5597         };
5598
5599         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5600                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5601
5602         /* Ensure this is an unlock - when canceling a lock, the
5603          * canceled lock is passed in, and it won't be an unlock.
5604          */
5605         fl->fl_type = F_UNLCK;
5606
5607         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5608         if (data == NULL) {
5609                 nfs_free_seqid(seqid);
5610                 return ERR_PTR(-ENOMEM);
5611         }
5612
5613         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5614         msg.rpc_argp = &data->arg;
5615         msg.rpc_resp = &data->res;
5616         task_setup_data.callback_data = data;
5617         return rpc_run_task(&task_setup_data);
5618 }
5619
5620 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5621 {
5622         struct inode *inode = state->inode;
5623         struct nfs4_state_owner *sp = state->owner;
5624         struct nfs_inode *nfsi = NFS_I(inode);
5625         struct nfs_seqid *seqid;
5626         struct nfs4_lock_state *lsp;
5627         struct rpc_task *task;
5628         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5629         int status = 0;
5630         unsigned char fl_flags = request->fl_flags;
5631
5632         status = nfs4_set_lock_state(state, request);
5633         /* Unlock _before_ we do the RPC call */
5634         request->fl_flags |= FL_EXISTS;
5635         /* Exclude nfs_delegation_claim_locks() */
5636         mutex_lock(&sp->so_delegreturn_mutex);
5637         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5638         down_read(&nfsi->rwsem);
5639         if (do_vfs_lock(inode, request) == -ENOENT) {
5640                 up_read(&nfsi->rwsem);
5641                 mutex_unlock(&sp->so_delegreturn_mutex);
5642                 goto out;
5643         }
5644         up_read(&nfsi->rwsem);
5645         mutex_unlock(&sp->so_delegreturn_mutex);
5646         if (status != 0)
5647                 goto out;
5648         /* Is this a delegated lock? */
5649         lsp = request->fl_u.nfs4_fl.owner;
5650         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5651                 goto out;
5652         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5653         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5654         status = -ENOMEM;
5655         if (IS_ERR(seqid))
5656                 goto out;
5657         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5658         status = PTR_ERR(task);
5659         if (IS_ERR(task))
5660                 goto out;
5661         status = nfs4_wait_for_completion_rpc_task(task);
5662         rpc_put_task(task);
5663 out:
5664         request->fl_flags = fl_flags;
5665         trace_nfs4_unlock(request, state, F_SETLK, status);
5666         return status;
5667 }
5668
5669 struct nfs4_lockdata {
5670         struct nfs_lock_args arg;
5671         struct nfs_lock_res res;
5672         struct nfs4_lock_state *lsp;
5673         struct nfs_open_context *ctx;
5674         struct file_lock fl;
5675         unsigned long timestamp;
5676         int rpc_status;
5677         int cancelled;
5678         struct nfs_server *server;
5679 };
5680
5681 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5682                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5683                 gfp_t gfp_mask)
5684 {
5685         struct nfs4_lockdata *p;
5686         struct inode *inode = lsp->ls_state->inode;
5687         struct nfs_server *server = NFS_SERVER(inode);
5688         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5689
5690         p = kzalloc(sizeof(*p), gfp_mask);
5691         if (p == NULL)
5692                 return NULL;
5693
5694         p->arg.fh = NFS_FH(inode);
5695         p->arg.fl = &p->fl;
5696         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5697         if (IS_ERR(p->arg.open_seqid))
5698                 goto out_free;
5699         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5700         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5701         if (IS_ERR(p->arg.lock_seqid))
5702                 goto out_free_seqid;
5703         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5704         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5705         p->arg.lock_owner.s_dev = server->s_dev;
5706         p->res.lock_seqid = p->arg.lock_seqid;
5707         p->lsp = lsp;
5708         p->server = server;
5709         atomic_inc(&lsp->ls_count);
5710         p->ctx = get_nfs_open_context(ctx);
5711         get_file(fl->fl_file);
5712         memcpy(&p->fl, fl, sizeof(p->fl));
5713         return p;
5714 out_free_seqid:
5715         nfs_free_seqid(p->arg.open_seqid);
5716 out_free:
5717         kfree(p);
5718         return NULL;
5719 }
5720
5721 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5722 {
5723         struct nfs4_lockdata *data = calldata;
5724         struct nfs4_state *state = data->lsp->ls_state;
5725
5726         dprintk("%s: begin!\n", __func__);
5727         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5728                 goto out_wait;
5729         /* Do we need to do an open_to_lock_owner? */
5730         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5731                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5732                         goto out_release_lock_seqid;
5733                 }
5734                 nfs4_stateid_copy(&data->arg.open_stateid,
5735                                 &state->open_stateid);
5736                 data->arg.new_lock_owner = 1;
5737                 data->res.open_seqid = data->arg.open_seqid;
5738         } else {
5739                 data->arg.new_lock_owner = 0;
5740                 nfs4_stateid_copy(&data->arg.lock_stateid,
5741                                 &data->lsp->ls_stateid);
5742         }
5743         if (!nfs4_valid_open_stateid(state)) {
5744                 data->rpc_status = -EBADF;
5745                 task->tk_action = NULL;
5746                 goto out_release_open_seqid;
5747         }
5748         data->timestamp = jiffies;
5749         if (nfs4_setup_sequence(data->server,
5750                                 &data->arg.seq_args,
5751                                 &data->res.seq_res,
5752                                 task) == 0)
5753                 return;
5754 out_release_open_seqid:
5755         nfs_release_seqid(data->arg.open_seqid);
5756 out_release_lock_seqid:
5757         nfs_release_seqid(data->arg.lock_seqid);
5758 out_wait:
5759         nfs4_sequence_done(task, &data->res.seq_res);
5760         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5761 }
5762
5763 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5764 {
5765         struct nfs4_lockdata *data = calldata;
5766         struct nfs4_lock_state *lsp = data->lsp;
5767
5768         dprintk("%s: begin!\n", __func__);
5769
5770         if (!nfs4_sequence_done(task, &data->res.seq_res))
5771                 return;
5772
5773         data->rpc_status = task->tk_status;
5774         switch (task->tk_status) {
5775         case 0:
5776                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5777                                 data->timestamp);
5778                 if (data->arg.new_lock) {
5779                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5780                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5781                                 rpc_restart_call_prepare(task);
5782                                 break;
5783                         }
5784                 }
5785                 if (data->arg.new_lock_owner != 0) {
5786                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5787                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5788                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5789                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5790                         rpc_restart_call_prepare(task);
5791                 break;
5792         case -NFS4ERR_BAD_STATEID:
5793         case -NFS4ERR_OLD_STATEID:
5794         case -NFS4ERR_STALE_STATEID:
5795         case -NFS4ERR_EXPIRED:
5796                 if (data->arg.new_lock_owner != 0) {
5797                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5798                                                 &lsp->ls_state->open_stateid))
5799                                 rpc_restart_call_prepare(task);
5800                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5801                                                 &lsp->ls_stateid))
5802                                 rpc_restart_call_prepare(task);
5803         }
5804         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5805 }
5806
5807 static void nfs4_lock_release(void *calldata)
5808 {
5809         struct nfs4_lockdata *data = calldata;
5810
5811         dprintk("%s: begin!\n", __func__);
5812         nfs_free_seqid(data->arg.open_seqid);
5813         if (data->cancelled != 0) {
5814                 struct rpc_task *task;
5815                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5816                                 data->arg.lock_seqid);
5817                 if (!IS_ERR(task))
5818                         rpc_put_task_async(task);
5819                 dprintk("%s: cancelling lock!\n", __func__);
5820         } else
5821                 nfs_free_seqid(data->arg.lock_seqid);
5822         nfs4_put_lock_state(data->lsp);
5823         put_nfs_open_context(data->ctx);
5824         fput(data->fl.fl_file);
5825         kfree(data);
5826         dprintk("%s: done!\n", __func__);
5827 }
5828
5829 static const struct rpc_call_ops nfs4_lock_ops = {
5830         .rpc_call_prepare = nfs4_lock_prepare,
5831         .rpc_call_done = nfs4_lock_done,
5832         .rpc_release = nfs4_lock_release,
5833 };
5834
5835 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5836 {
5837         switch (error) {
5838         case -NFS4ERR_ADMIN_REVOKED:
5839         case -NFS4ERR_BAD_STATEID:
5840                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5841                 if (new_lock_owner != 0 ||
5842                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5843                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5844                 break;
5845         case -NFS4ERR_STALE_STATEID:
5846                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5847         case -NFS4ERR_EXPIRED:
5848                 nfs4_schedule_lease_recovery(server->nfs_client);
5849         };
5850 }
5851
5852 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5853 {
5854         struct nfs4_lockdata *data;
5855         struct rpc_task *task;
5856         struct rpc_message msg = {
5857                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5858                 .rpc_cred = state->owner->so_cred,
5859         };
5860         struct rpc_task_setup task_setup_data = {
5861                 .rpc_client = NFS_CLIENT(state->inode),
5862                 .rpc_message = &msg,
5863                 .callback_ops = &nfs4_lock_ops,
5864                 .workqueue = nfsiod_workqueue,
5865                 .flags = RPC_TASK_ASYNC,
5866         };
5867         int ret;
5868
5869         dprintk("%s: begin!\n", __func__);
5870         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5871                         fl->fl_u.nfs4_fl.owner,
5872                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5873         if (data == NULL)
5874                 return -ENOMEM;
5875         if (IS_SETLKW(cmd))
5876                 data->arg.block = 1;
5877         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5878         msg.rpc_argp = &data->arg;
5879         msg.rpc_resp = &data->res;
5880         task_setup_data.callback_data = data;
5881         if (recovery_type > NFS_LOCK_NEW) {
5882                 if (recovery_type == NFS_LOCK_RECLAIM)
5883                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5884                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5885         } else
5886                 data->arg.new_lock = 1;
5887         task = rpc_run_task(&task_setup_data);
5888         if (IS_ERR(task))
5889                 return PTR_ERR(task);
5890         ret = nfs4_wait_for_completion_rpc_task(task);
5891         if (ret == 0) {
5892                 ret = data->rpc_status;
5893                 if (ret)
5894                         nfs4_handle_setlk_error(data->server, data->lsp,
5895                                         data->arg.new_lock_owner, ret);
5896         } else
5897                 data->cancelled = 1;
5898         rpc_put_task(task);
5899         dprintk("%s: done, ret = %d!\n", __func__, ret);
5900         return ret;
5901 }
5902
5903 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5904 {
5905         struct nfs_server *server = NFS_SERVER(state->inode);
5906         struct nfs4_exception exception = {
5907                 .inode = state->inode,
5908         };
5909         int err;
5910
5911         do {
5912                 /* Cache the lock if possible... */
5913                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5914                         return 0;
5915                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5916                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5917                 if (err != -NFS4ERR_DELAY)
5918                         break;
5919                 nfs4_handle_exception(server, err, &exception);
5920         } while (exception.retry);
5921         return err;
5922 }
5923
5924 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5925 {
5926         struct nfs_server *server = NFS_SERVER(state->inode);
5927         struct nfs4_exception exception = {
5928                 .inode = state->inode,
5929         };
5930         int err;
5931
5932         err = nfs4_set_lock_state(state, request);
5933         if (err != 0)
5934                 return err;
5935         if (!recover_lost_locks) {
5936                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5937                 return 0;
5938         }
5939         do {
5940                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5941                         return 0;
5942                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5943                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5944                 switch (err) {
5945                 default:
5946                         goto out;
5947                 case -NFS4ERR_GRACE:
5948                 case -NFS4ERR_DELAY:
5949                         nfs4_handle_exception(server, err, &exception);
5950                         err = 0;
5951                 }
5952         } while (exception.retry);
5953 out:
5954         return err;
5955 }
5956
5957 #if defined(CONFIG_NFS_V4_1)
5958 /**
5959  * nfs41_check_expired_locks - possibly free a lock stateid
5960  *
5961  * @state: NFSv4 state for an inode
5962  *
5963  * Returns NFS_OK if recovery for this stateid is now finished.
5964  * Otherwise a negative NFS4ERR value is returned.
5965  */
5966 static int nfs41_check_expired_locks(struct nfs4_state *state)
5967 {
5968         int status, ret = -NFS4ERR_BAD_STATEID;
5969         struct nfs4_lock_state *lsp;
5970         struct nfs_server *server = NFS_SERVER(state->inode);
5971
5972         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5973                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5974                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5975
5976                         status = nfs41_test_stateid(server,
5977                                         &lsp->ls_stateid,
5978                                         cred);
5979                         trace_nfs4_test_lock_stateid(state, lsp, status);
5980                         if (status != NFS_OK) {
5981                                 /* Free the stateid unless the server
5982                                  * informs us the stateid is unrecognized. */
5983                                 if (status != -NFS4ERR_BAD_STATEID)
5984                                         nfs41_free_stateid(server,
5985                                                         &lsp->ls_stateid,
5986                                                         cred);
5987                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5988                                 ret = status;
5989                         }
5990                 }
5991         };
5992
5993         return ret;
5994 }
5995
5996 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5997 {
5998         int status = NFS_OK;
5999
6000         if (test_bit(LK_STATE_IN_USE, &state->flags))
6001                 status = nfs41_check_expired_locks(state);
6002         if (status != NFS_OK)
6003                 status = nfs4_lock_expired(state, request);
6004         return status;
6005 }
6006 #endif
6007
6008 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6009 {
6010         struct nfs_inode *nfsi = NFS_I(state->inode);
6011         unsigned char fl_flags = request->fl_flags;
6012         int status = -ENOLCK;
6013
6014         if ((fl_flags & FL_POSIX) &&
6015                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6016                 goto out;
6017         /* Is this a delegated open? */
6018         status = nfs4_set_lock_state(state, request);
6019         if (status != 0)
6020                 goto out;
6021         request->fl_flags |= FL_ACCESS;
6022         status = do_vfs_lock(state->inode, request);
6023         if (status < 0)
6024                 goto out;
6025         down_read(&nfsi->rwsem);
6026         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6027                 /* Yes: cache locks! */
6028                 /* ...but avoid races with delegation recall... */
6029                 request->fl_flags = fl_flags & ~FL_SLEEP;
6030                 status = do_vfs_lock(state->inode, request);
6031                 up_read(&nfsi->rwsem);
6032                 goto out;
6033         }
6034         up_read(&nfsi->rwsem);
6035         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6036 out:
6037         request->fl_flags = fl_flags;
6038         return status;
6039 }
6040
6041 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6042 {
6043         struct nfs4_exception exception = {
6044                 .state = state,
6045                 .inode = state->inode,
6046         };
6047         int err;
6048
6049         do {
6050                 err = _nfs4_proc_setlk(state, cmd, request);
6051                 trace_nfs4_set_lock(request, state, cmd, err);
6052                 if (err == -NFS4ERR_DENIED)
6053                         err = -EAGAIN;
6054                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6055                                 err, &exception);
6056         } while (exception.retry);
6057         return err;
6058 }
6059
6060 static int
6061 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6062 {
6063         struct nfs_open_context *ctx;
6064         struct nfs4_state *state;
6065         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6066         int status;
6067
6068         /* verify open state */
6069         ctx = nfs_file_open_context(filp);
6070         state = ctx->state;
6071
6072         if (request->fl_start < 0 || request->fl_end < 0)
6073                 return -EINVAL;
6074
6075         if (IS_GETLK(cmd)) {
6076                 if (state != NULL)
6077                         return nfs4_proc_getlk(state, F_GETLK, request);
6078                 return 0;
6079         }
6080
6081         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6082                 return -EINVAL;
6083
6084         if (request->fl_type == F_UNLCK) {
6085                 if (state != NULL)
6086                         return nfs4_proc_unlck(state, cmd, request);
6087                 return 0;
6088         }
6089
6090         if (state == NULL)
6091                 return -ENOLCK;
6092         /*
6093          * Don't rely on the VFS having checked the file open mode,
6094          * since it won't do this for flock() locks.
6095          */
6096         switch (request->fl_type) {
6097         case F_RDLCK:
6098                 if (!(filp->f_mode & FMODE_READ))
6099                         return -EBADF;
6100                 break;
6101         case F_WRLCK:
6102                 if (!(filp->f_mode & FMODE_WRITE))
6103                         return -EBADF;
6104         }
6105
6106         do {
6107                 status = nfs4_proc_setlk(state, cmd, request);
6108                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6109                         break;
6110                 timeout = nfs4_set_lock_task_retry(timeout);
6111                 status = -ERESTARTSYS;
6112                 if (signalled())
6113                         break;
6114         } while(status < 0);
6115         return status;
6116 }
6117
6118 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6119 {
6120         struct nfs_server *server = NFS_SERVER(state->inode);
6121         int err;
6122
6123         err = nfs4_set_lock_state(state, fl);
6124         if (err != 0)
6125                 return err;
6126         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6127         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6128 }
6129
6130 struct nfs_release_lockowner_data {
6131         struct nfs4_lock_state *lsp;
6132         struct nfs_server *server;
6133         struct nfs_release_lockowner_args args;
6134         struct nfs_release_lockowner_res res;
6135         unsigned long timestamp;
6136 };
6137
6138 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6139 {
6140         struct nfs_release_lockowner_data *data = calldata;
6141         struct nfs_server *server = data->server;
6142         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6143                              &data->args.seq_args, &data->res.seq_res, task);
6144         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6145         data->timestamp = jiffies;
6146 }
6147
6148 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6149 {
6150         struct nfs_release_lockowner_data *data = calldata;
6151         struct nfs_server *server = data->server;
6152
6153         nfs40_sequence_done(task, &data->res.seq_res);
6154
6155         switch (task->tk_status) {
6156         case 0:
6157                 renew_lease(server, data->timestamp);
6158                 break;
6159         case -NFS4ERR_STALE_CLIENTID:
6160         case -NFS4ERR_EXPIRED:
6161                 nfs4_schedule_lease_recovery(server->nfs_client);
6162                 break;
6163         case -NFS4ERR_LEASE_MOVED:
6164         case -NFS4ERR_DELAY:
6165                 if (nfs4_async_handle_error(task, server,
6166                                             NULL, NULL) == -EAGAIN)
6167                         rpc_restart_call_prepare(task);
6168         }
6169 }
6170
6171 static void nfs4_release_lockowner_release(void *calldata)
6172 {
6173         struct nfs_release_lockowner_data *data = calldata;
6174         nfs4_free_lock_state(data->server, data->lsp);
6175         kfree(calldata);
6176 }
6177
6178 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6179         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6180         .rpc_call_done = nfs4_release_lockowner_done,
6181         .rpc_release = nfs4_release_lockowner_release,
6182 };
6183
6184 static void
6185 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6186 {
6187         struct nfs_release_lockowner_data *data;
6188         struct rpc_message msg = {
6189                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6190         };
6191
6192         if (server->nfs_client->cl_mvops->minor_version != 0)
6193                 return;
6194
6195         data = kmalloc(sizeof(*data), GFP_NOFS);
6196         if (!data)
6197                 return;
6198         data->lsp = lsp;
6199         data->server = server;
6200         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6201         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6202         data->args.lock_owner.s_dev = server->s_dev;
6203
6204         msg.rpc_argp = &data->args;
6205         msg.rpc_resp = &data->res;
6206         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6207         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6208 }
6209
6210 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6211
6212 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6213                                    const void *buf, size_t buflen,
6214                                    int flags, int type)
6215 {
6216         if (strcmp(key, "") != 0)
6217                 return -EINVAL;
6218
6219         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6220 }
6221
6222 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6223                                    void *buf, size_t buflen, int type)
6224 {
6225         if (strcmp(key, "") != 0)
6226                 return -EINVAL;
6227
6228         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6229 }
6230
6231 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6232                                        size_t list_len, const char *name,
6233                                        size_t name_len, int type)
6234 {
6235         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6236
6237         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6238                 return 0;
6239
6240         if (list && len <= list_len)
6241                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6242         return len;
6243 }
6244
6245 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6246 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6247 {
6248         return server->caps & NFS_CAP_SECURITY_LABEL;
6249 }
6250
6251 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6252                                    const void *buf, size_t buflen,
6253                                    int flags, int type)
6254 {
6255         if (security_ismaclabel(key))
6256                 return nfs4_set_security_label(dentry, buf, buflen);
6257
6258         return -EOPNOTSUPP;
6259 }
6260
6261 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6262                                    void *buf, size_t buflen, int type)
6263 {
6264         if (security_ismaclabel(key))
6265                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6266         return -EOPNOTSUPP;
6267 }
6268
6269 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6270                                        size_t list_len, const char *name,
6271                                        size_t name_len, int type)
6272 {
6273         size_t len = 0;
6274
6275         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6276                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6277                 if (list && len <= list_len)
6278                         security_inode_listsecurity(d_inode(dentry), list, len);
6279         }
6280         return len;
6281 }
6282
6283 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6284         .prefix = XATTR_SECURITY_PREFIX,
6285         .list   = nfs4_xattr_list_nfs4_label,
6286         .get    = nfs4_xattr_get_nfs4_label,
6287         .set    = nfs4_xattr_set_nfs4_label,
6288 };
6289 #endif
6290
6291
6292 /*
6293  * nfs_fhget will use either the mounted_on_fileid or the fileid
6294  */
6295 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6296 {
6297         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6298                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6299               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6300               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6301                 return;
6302
6303         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6304                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6305         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6306         fattr->nlink = 2;
6307 }
6308
6309 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6310                                    const struct qstr *name,
6311                                    struct nfs4_fs_locations *fs_locations,
6312                                    struct page *page)
6313 {
6314         struct nfs_server *server = NFS_SERVER(dir);
6315         u32 bitmask[3] = {
6316                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6317         };
6318         struct nfs4_fs_locations_arg args = {
6319                 .dir_fh = NFS_FH(dir),
6320                 .name = name,
6321                 .page = page,
6322                 .bitmask = bitmask,
6323         };
6324         struct nfs4_fs_locations_res res = {
6325                 .fs_locations = fs_locations,
6326         };
6327         struct rpc_message msg = {
6328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6329                 .rpc_argp = &args,
6330                 .rpc_resp = &res,
6331         };
6332         int status;
6333
6334         dprintk("%s: start\n", __func__);
6335
6336         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6337          * is not supported */
6338         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6339                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6340         else
6341                 bitmask[0] |= FATTR4_WORD0_FILEID;
6342
6343         nfs_fattr_init(&fs_locations->fattr);
6344         fs_locations->server = server;
6345         fs_locations->nlocations = 0;
6346         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6347         dprintk("%s: returned status = %d\n", __func__, status);
6348         return status;
6349 }
6350
6351 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6352                            const struct qstr *name,
6353                            struct nfs4_fs_locations *fs_locations,
6354                            struct page *page)
6355 {
6356         struct nfs4_exception exception = { };
6357         int err;
6358         do {
6359                 err = _nfs4_proc_fs_locations(client, dir, name,
6360                                 fs_locations, page);
6361                 trace_nfs4_get_fs_locations(dir, name, err);
6362                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6363                                 &exception);
6364         } while (exception.retry);
6365         return err;
6366 }
6367
6368 /*
6369  * This operation also signals the server that this client is
6370  * performing migration recovery.  The server can stop returning
6371  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6372  * appended to this compound to identify the client ID which is
6373  * performing recovery.
6374  */
6375 static int _nfs40_proc_get_locations(struct inode *inode,
6376                                      struct nfs4_fs_locations *locations,
6377                                      struct page *page, struct rpc_cred *cred)
6378 {
6379         struct nfs_server *server = NFS_SERVER(inode);
6380         struct rpc_clnt *clnt = server->client;
6381         u32 bitmask[2] = {
6382                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6383         };
6384         struct nfs4_fs_locations_arg args = {
6385                 .clientid       = server->nfs_client->cl_clientid,
6386                 .fh             = NFS_FH(inode),
6387                 .page           = page,
6388                 .bitmask        = bitmask,
6389                 .migration      = 1,            /* skip LOOKUP */
6390                 .renew          = 1,            /* append RENEW */
6391         };
6392         struct nfs4_fs_locations_res res = {
6393                 .fs_locations   = locations,
6394                 .migration      = 1,
6395                 .renew          = 1,
6396         };
6397         struct rpc_message msg = {
6398                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6399                 .rpc_argp       = &args,
6400                 .rpc_resp       = &res,
6401                 .rpc_cred       = cred,
6402         };
6403         unsigned long now = jiffies;
6404         int status;
6405
6406         nfs_fattr_init(&locations->fattr);
6407         locations->server = server;
6408         locations->nlocations = 0;
6409
6410         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6411         nfs4_set_sequence_privileged(&args.seq_args);
6412         status = nfs4_call_sync_sequence(clnt, server, &msg,
6413                                         &args.seq_args, &res.seq_res);
6414         if (status)
6415                 return status;
6416
6417         renew_lease(server, now);
6418         return 0;
6419 }
6420
6421 #ifdef CONFIG_NFS_V4_1
6422
6423 /*
6424  * This operation also signals the server that this client is
6425  * performing migration recovery.  The server can stop asserting
6426  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6427  * performing this operation is identified in the SEQUENCE
6428  * operation in this compound.
6429  *
6430  * When the client supports GETATTR(fs_locations_info), it can
6431  * be plumbed in here.
6432  */
6433 static int _nfs41_proc_get_locations(struct inode *inode,
6434                                      struct nfs4_fs_locations *locations,
6435                                      struct page *page, struct rpc_cred *cred)
6436 {
6437         struct nfs_server *server = NFS_SERVER(inode);
6438         struct rpc_clnt *clnt = server->client;
6439         u32 bitmask[2] = {
6440                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6441         };
6442         struct nfs4_fs_locations_arg args = {
6443                 .fh             = NFS_FH(inode),
6444                 .page           = page,
6445                 .bitmask        = bitmask,
6446                 .migration      = 1,            /* skip LOOKUP */
6447         };
6448         struct nfs4_fs_locations_res res = {
6449                 .fs_locations   = locations,
6450                 .migration      = 1,
6451         };
6452         struct rpc_message msg = {
6453                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6454                 .rpc_argp       = &args,
6455                 .rpc_resp       = &res,
6456                 .rpc_cred       = cred,
6457         };
6458         int status;
6459
6460         nfs_fattr_init(&locations->fattr);
6461         locations->server = server;
6462         locations->nlocations = 0;
6463
6464         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6465         nfs4_set_sequence_privileged(&args.seq_args);
6466         status = nfs4_call_sync_sequence(clnt, server, &msg,
6467                                         &args.seq_args, &res.seq_res);
6468         if (status == NFS4_OK &&
6469             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6470                 status = -NFS4ERR_LEASE_MOVED;
6471         return status;
6472 }
6473
6474 #endif  /* CONFIG_NFS_V4_1 */
6475
6476 /**
6477  * nfs4_proc_get_locations - discover locations for a migrated FSID
6478  * @inode: inode on FSID that is migrating
6479  * @locations: result of query
6480  * @page: buffer
6481  * @cred: credential to use for this operation
6482  *
6483  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6484  * operation failed, or a negative errno if a local error occurred.
6485  *
6486  * On success, "locations" is filled in, but if the server has
6487  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6488  * asserted.
6489  *
6490  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6491  * from this client that require migration recovery.
6492  */
6493 int nfs4_proc_get_locations(struct inode *inode,
6494                             struct nfs4_fs_locations *locations,
6495                             struct page *page, struct rpc_cred *cred)
6496 {
6497         struct nfs_server *server = NFS_SERVER(inode);
6498         struct nfs_client *clp = server->nfs_client;
6499         const struct nfs4_mig_recovery_ops *ops =
6500                                         clp->cl_mvops->mig_recovery_ops;
6501         struct nfs4_exception exception = { };
6502         int status;
6503
6504         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6505                 (unsigned long long)server->fsid.major,
6506                 (unsigned long long)server->fsid.minor,
6507                 clp->cl_hostname);
6508         nfs_display_fhandle(NFS_FH(inode), __func__);
6509
6510         do {
6511                 status = ops->get_locations(inode, locations, page, cred);
6512                 if (status != -NFS4ERR_DELAY)
6513                         break;
6514                 nfs4_handle_exception(server, status, &exception);
6515         } while (exception.retry);
6516         return status;
6517 }
6518
6519 /*
6520  * This operation also signals the server that this client is
6521  * performing "lease moved" recovery.  The server can stop
6522  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6523  * is appended to this compound to identify the client ID which is
6524  * performing recovery.
6525  */
6526 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6527 {
6528         struct nfs_server *server = NFS_SERVER(inode);
6529         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6530         struct rpc_clnt *clnt = server->client;
6531         struct nfs4_fsid_present_arg args = {
6532                 .fh             = NFS_FH(inode),
6533                 .clientid       = clp->cl_clientid,
6534                 .renew          = 1,            /* append RENEW */
6535         };
6536         struct nfs4_fsid_present_res res = {
6537                 .renew          = 1,
6538         };
6539         struct rpc_message msg = {
6540                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6541                 .rpc_argp       = &args,
6542                 .rpc_resp       = &res,
6543                 .rpc_cred       = cred,
6544         };
6545         unsigned long now = jiffies;
6546         int status;
6547
6548         res.fh = nfs_alloc_fhandle();
6549         if (res.fh == NULL)
6550                 return -ENOMEM;
6551
6552         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6553         nfs4_set_sequence_privileged(&args.seq_args);
6554         status = nfs4_call_sync_sequence(clnt, server, &msg,
6555                                                 &args.seq_args, &res.seq_res);
6556         nfs_free_fhandle(res.fh);
6557         if (status)
6558                 return status;
6559
6560         do_renew_lease(clp, now);
6561         return 0;
6562 }
6563
6564 #ifdef CONFIG_NFS_V4_1
6565
6566 /*
6567  * This operation also signals the server that this client is
6568  * performing "lease moved" recovery.  The server can stop asserting
6569  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6570  * this operation is identified in the SEQUENCE operation in this
6571  * compound.
6572  */
6573 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6574 {
6575         struct nfs_server *server = NFS_SERVER(inode);
6576         struct rpc_clnt *clnt = server->client;
6577         struct nfs4_fsid_present_arg args = {
6578                 .fh             = NFS_FH(inode),
6579         };
6580         struct nfs4_fsid_present_res res = {
6581         };
6582         struct rpc_message msg = {
6583                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6584                 .rpc_argp       = &args,
6585                 .rpc_resp       = &res,
6586                 .rpc_cred       = cred,
6587         };
6588         int status;
6589
6590         res.fh = nfs_alloc_fhandle();
6591         if (res.fh == NULL)
6592                 return -ENOMEM;
6593
6594         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6595         nfs4_set_sequence_privileged(&args.seq_args);
6596         status = nfs4_call_sync_sequence(clnt, server, &msg,
6597                                                 &args.seq_args, &res.seq_res);
6598         nfs_free_fhandle(res.fh);
6599         if (status == NFS4_OK &&
6600             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6601                 status = -NFS4ERR_LEASE_MOVED;
6602         return status;
6603 }
6604
6605 #endif  /* CONFIG_NFS_V4_1 */
6606
6607 /**
6608  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6609  * @inode: inode on FSID to check
6610  * @cred: credential to use for this operation
6611  *
6612  * Server indicates whether the FSID is present, moved, or not
6613  * recognized.  This operation is necessary to clear a LEASE_MOVED
6614  * condition for this client ID.
6615  *
6616  * Returns NFS4_OK if the FSID is present on this server,
6617  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6618  *  NFS4ERR code if some error occurred on the server, or a
6619  *  negative errno if a local failure occurred.
6620  */
6621 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6622 {
6623         struct nfs_server *server = NFS_SERVER(inode);
6624         struct nfs_client *clp = server->nfs_client;
6625         const struct nfs4_mig_recovery_ops *ops =
6626                                         clp->cl_mvops->mig_recovery_ops;
6627         struct nfs4_exception exception = { };
6628         int status;
6629
6630         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6631                 (unsigned long long)server->fsid.major,
6632                 (unsigned long long)server->fsid.minor,
6633                 clp->cl_hostname);
6634         nfs_display_fhandle(NFS_FH(inode), __func__);
6635
6636         do {
6637                 status = ops->fsid_present(inode, cred);
6638                 if (status != -NFS4ERR_DELAY)
6639                         break;
6640                 nfs4_handle_exception(server, status, &exception);
6641         } while (exception.retry);
6642         return status;
6643 }
6644
6645 /**
6646  * If 'use_integrity' is true and the state managment nfs_client
6647  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6648  * and the machine credential as per RFC3530bis and RFC5661 Security
6649  * Considerations sections. Otherwise, just use the user cred with the
6650  * filesystem's rpc_client.
6651  */
6652 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6653 {
6654         int status;
6655         struct nfs4_secinfo_arg args = {
6656                 .dir_fh = NFS_FH(dir),
6657                 .name   = name,
6658         };
6659         struct nfs4_secinfo_res res = {
6660                 .flavors     = flavors,
6661         };
6662         struct rpc_message msg = {
6663                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6664                 .rpc_argp = &args,
6665                 .rpc_resp = &res,
6666         };
6667         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6668         struct rpc_cred *cred = NULL;
6669
6670         if (use_integrity) {
6671                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6672                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6673                 msg.rpc_cred = cred;
6674         }
6675
6676         dprintk("NFS call  secinfo %s\n", name->name);
6677
6678         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6679                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6680
6681         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6682                                 &res.seq_res, 0);
6683         dprintk("NFS reply  secinfo: %d\n", status);
6684
6685         if (cred)
6686                 put_rpccred(cred);
6687
6688         return status;
6689 }
6690
6691 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6692                       struct nfs4_secinfo_flavors *flavors)
6693 {
6694         struct nfs4_exception exception = { };
6695         int err;
6696         do {
6697                 err = -NFS4ERR_WRONGSEC;
6698
6699                 /* try to use integrity protection with machine cred */
6700                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6701                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6702
6703                 /*
6704                  * if unable to use integrity protection, or SECINFO with
6705                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6706                  * disallowed by spec, but exists in deployed servers) use
6707                  * the current filesystem's rpc_client and the user cred.
6708                  */
6709                 if (err == -NFS4ERR_WRONGSEC)
6710                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6711
6712                 trace_nfs4_secinfo(dir, name, err);
6713                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6714                                 &exception);
6715         } while (exception.retry);
6716         return err;
6717 }
6718
6719 #ifdef CONFIG_NFS_V4_1
6720 /*
6721  * Check the exchange flags returned by the server for invalid flags, having
6722  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6723  * DS flags set.
6724  */
6725 static int nfs4_check_cl_exchange_flags(u32 flags)
6726 {
6727         if (flags & ~EXCHGID4_FLAG_MASK_R)
6728                 goto out_inval;
6729         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6730             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6731                 goto out_inval;
6732         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6733                 goto out_inval;
6734         return NFS_OK;
6735 out_inval:
6736         return -NFS4ERR_INVAL;
6737 }
6738
6739 static bool
6740 nfs41_same_server_scope(struct nfs41_server_scope *a,
6741                         struct nfs41_server_scope *b)
6742 {
6743         if (a->server_scope_sz == b->server_scope_sz &&
6744             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6745                 return true;
6746
6747         return false;
6748 }
6749
6750 /*
6751  * nfs4_proc_bind_conn_to_session()
6752  *
6753  * The 4.1 client currently uses the same TCP connection for the
6754  * fore and backchannel.
6755  */
6756 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6757 {
6758         int status;
6759         struct nfs41_bind_conn_to_session_args args = {
6760                 .client = clp,
6761                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6762         };
6763         struct nfs41_bind_conn_to_session_res res;
6764         struct rpc_message msg = {
6765                 .rpc_proc =
6766                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6767                 .rpc_argp = &args,
6768                 .rpc_resp = &res,
6769                 .rpc_cred = cred,
6770         };
6771
6772         dprintk("--> %s\n", __func__);
6773
6774         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6775         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6776                 args.dir = NFS4_CDFC4_FORE;
6777
6778         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6779         trace_nfs4_bind_conn_to_session(clp, status);
6780         if (status == 0) {
6781                 if (memcmp(res.sessionid.data,
6782                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6783                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6784                         status = -EIO;
6785                         goto out;
6786                 }
6787                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6788                         dprintk("NFS: %s: Unexpected direction from server\n",
6789                                 __func__);
6790                         status = -EIO;
6791                         goto out;
6792                 }
6793                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6794                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6795                                 __func__);
6796                         status = -EIO;
6797                         goto out;
6798                 }
6799         }
6800 out:
6801         dprintk("<-- %s status= %d\n", __func__, status);
6802         return status;
6803 }
6804
6805 /*
6806  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6807  * and operations we'd like to see to enable certain features in the allow map
6808  */
6809 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6810         .how = SP4_MACH_CRED,
6811         .enforce.u.words = {
6812                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6813                       1 << (OP_EXCHANGE_ID - 32) |
6814                       1 << (OP_CREATE_SESSION - 32) |
6815                       1 << (OP_DESTROY_SESSION - 32) |
6816                       1 << (OP_DESTROY_CLIENTID - 32)
6817         },
6818         .allow.u.words = {
6819                 [0] = 1 << (OP_CLOSE) |
6820                       1 << (OP_LOCKU) |
6821                       1 << (OP_COMMIT),
6822                 [1] = 1 << (OP_SECINFO - 32) |
6823                       1 << (OP_SECINFO_NO_NAME - 32) |
6824                       1 << (OP_TEST_STATEID - 32) |
6825                       1 << (OP_FREE_STATEID - 32) |
6826                       1 << (OP_WRITE - 32)
6827         }
6828 };
6829
6830 /*
6831  * Select the state protection mode for client `clp' given the server results
6832  * from exchange_id in `sp'.
6833  *
6834  * Returns 0 on success, negative errno otherwise.
6835  */
6836 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6837                                  struct nfs41_state_protection *sp)
6838 {
6839         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6840                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6841                       1 << (OP_EXCHANGE_ID - 32) |
6842                       1 << (OP_CREATE_SESSION - 32) |
6843                       1 << (OP_DESTROY_SESSION - 32) |
6844                       1 << (OP_DESTROY_CLIENTID - 32)
6845         };
6846         unsigned int i;
6847
6848         if (sp->how == SP4_MACH_CRED) {
6849                 /* Print state protect result */
6850                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6851                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6852                         if (test_bit(i, sp->enforce.u.longs))
6853                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6854                         if (test_bit(i, sp->allow.u.longs))
6855                                 dfprintk(MOUNT, "  allow op %d\n", i);
6856                 }
6857
6858                 /* make sure nothing is on enforce list that isn't supported */
6859                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6860                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6861                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6862                                 return -EINVAL;
6863                         }
6864                 }
6865
6866                 /*
6867                  * Minimal mode - state operations are allowed to use machine
6868                  * credential.  Note this already happens by default, so the
6869                  * client doesn't have to do anything more than the negotiation.
6870                  *
6871                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6872                  *       we're already using the machine cred for exchange_id
6873                  *       and will never use a different cred.
6874                  */
6875                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6876                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6877                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6878                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6879                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6880                         dfprintk(MOUNT, "  minimal mode enabled\n");
6881                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6882                 } else {
6883                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6884                         return -EINVAL;
6885                 }
6886
6887                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6888                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6889                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6890                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6891                 }
6892
6893                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6894                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6895                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6896                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6897                 }
6898
6899                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6900                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6901                         dfprintk(MOUNT, "  stateid mode enabled\n");
6902                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6903                 }
6904
6905                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6906                         dfprintk(MOUNT, "  write mode enabled\n");
6907                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6908                 }
6909
6910                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6911                         dfprintk(MOUNT, "  commit mode enabled\n");
6912                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6913                 }
6914         }
6915
6916         return 0;
6917 }
6918
6919 /*
6920  * _nfs4_proc_exchange_id()
6921  *
6922  * Wrapper for EXCHANGE_ID operation.
6923  */
6924 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6925         u32 sp4_how)
6926 {
6927         nfs4_verifier verifier;
6928         struct nfs41_exchange_id_args args = {
6929                 .verifier = &verifier,
6930                 .client = clp,
6931 #ifdef CONFIG_NFS_V4_1_MIGRATION
6932                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6933                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6934                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6935 #else
6936                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6937                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6938 #endif
6939         };
6940         struct nfs41_exchange_id_res res = {
6941                 0
6942         };
6943         int status;
6944         struct rpc_message msg = {
6945                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6946                 .rpc_argp = &args,
6947                 .rpc_resp = &res,
6948                 .rpc_cred = cred,
6949         };
6950
6951         nfs4_init_boot_verifier(clp, &verifier);
6952
6953         status = nfs4_init_uniform_client_string(clp);
6954         if (status)
6955                 goto out;
6956
6957         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6958                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6959                 clp->cl_owner_id);
6960
6961         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6962                                         GFP_NOFS);
6963         if (unlikely(res.server_owner == NULL)) {
6964                 status = -ENOMEM;
6965                 goto out;
6966         }
6967
6968         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6969                                         GFP_NOFS);
6970         if (unlikely(res.server_scope == NULL)) {
6971                 status = -ENOMEM;
6972                 goto out_server_owner;
6973         }
6974
6975         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6976         if (unlikely(res.impl_id == NULL)) {
6977                 status = -ENOMEM;
6978                 goto out_server_scope;
6979         }
6980
6981         switch (sp4_how) {
6982         case SP4_NONE:
6983                 args.state_protect.how = SP4_NONE;
6984                 break;
6985
6986         case SP4_MACH_CRED:
6987                 args.state_protect = nfs4_sp4_mach_cred_request;
6988                 break;
6989
6990         default:
6991                 /* unsupported! */
6992                 WARN_ON_ONCE(1);
6993                 status = -EINVAL;
6994                 goto out_impl_id;
6995         }
6996
6997         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6998         trace_nfs4_exchange_id(clp, status);
6999         if (status == 0)
7000                 status = nfs4_check_cl_exchange_flags(res.flags);
7001
7002         if (status == 0)
7003                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7004
7005         if (status == 0) {
7006                 clp->cl_clientid = res.clientid;
7007                 clp->cl_exchange_flags = res.flags;
7008                 /* Client ID is not confirmed */
7009                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7010                         clear_bit(NFS4_SESSION_ESTABLISHED,
7011                                         &clp->cl_session->session_state);
7012                         clp->cl_seqid = res.seqid;
7013                 }
7014
7015                 kfree(clp->cl_serverowner);
7016                 clp->cl_serverowner = res.server_owner;
7017                 res.server_owner = NULL;
7018
7019                 /* use the most recent implementation id */
7020                 kfree(clp->cl_implid);
7021                 clp->cl_implid = res.impl_id;
7022                 res.impl_id = NULL;
7023
7024                 if (clp->cl_serverscope != NULL &&
7025                     !nfs41_same_server_scope(clp->cl_serverscope,
7026                                              res.server_scope)) {
7027                         dprintk("%s: server_scope mismatch detected\n",
7028                                 __func__);
7029                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7030                         kfree(clp->cl_serverscope);
7031                         clp->cl_serverscope = NULL;
7032                 }
7033
7034                 if (clp->cl_serverscope == NULL) {
7035                         clp->cl_serverscope = res.server_scope;
7036                         res.server_scope = NULL;
7037                 }
7038         }
7039
7040 out_impl_id:
7041         kfree(res.impl_id);
7042 out_server_scope:
7043         kfree(res.server_scope);
7044 out_server_owner:
7045         kfree(res.server_owner);
7046 out:
7047         if (clp->cl_implid != NULL)
7048                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7049                         "domain: %s, name: %s, date: %llu,%u\n",
7050                         clp->cl_implid->domain, clp->cl_implid->name,
7051                         clp->cl_implid->date.seconds,
7052                         clp->cl_implid->date.nseconds);
7053         dprintk("NFS reply exchange_id: %d\n", status);
7054         return status;
7055 }
7056
7057 /*
7058  * nfs4_proc_exchange_id()
7059  *
7060  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7061  *
7062  * Since the clientid has expired, all compounds using sessions
7063  * associated with the stale clientid will be returning
7064  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7065  * be in some phase of session reset.
7066  *
7067  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7068  */
7069 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7070 {
7071         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7072         int status;
7073
7074         /* try SP4_MACH_CRED if krb5i/p */
7075         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7076             authflavor == RPC_AUTH_GSS_KRB5P) {
7077                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7078                 if (!status)
7079                         return 0;
7080         }
7081
7082         /* try SP4_NONE */
7083         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7084 }
7085
7086 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7087                 struct rpc_cred *cred)
7088 {
7089         struct rpc_message msg = {
7090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7091                 .rpc_argp = clp,
7092                 .rpc_cred = cred,
7093         };
7094         int status;
7095
7096         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7097         trace_nfs4_destroy_clientid(clp, status);
7098         if (status)
7099                 dprintk("NFS: Got error %d from the server %s on "
7100                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7101         return status;
7102 }
7103
7104 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7105                 struct rpc_cred *cred)
7106 {
7107         unsigned int loop;
7108         int ret;
7109
7110         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7111                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7112                 switch (ret) {
7113                 case -NFS4ERR_DELAY:
7114                 case -NFS4ERR_CLIENTID_BUSY:
7115                         ssleep(1);
7116                         break;
7117                 default:
7118                         return ret;
7119                 }
7120         }
7121         return 0;
7122 }
7123
7124 int nfs4_destroy_clientid(struct nfs_client *clp)
7125 {
7126         struct rpc_cred *cred;
7127         int ret = 0;
7128
7129         if (clp->cl_mvops->minor_version < 1)
7130                 goto out;
7131         if (clp->cl_exchange_flags == 0)
7132                 goto out;
7133         if (clp->cl_preserve_clid)
7134                 goto out;
7135         cred = nfs4_get_clid_cred(clp);
7136         ret = nfs4_proc_destroy_clientid(clp, cred);
7137         if (cred)
7138                 put_rpccred(cred);
7139         switch (ret) {
7140         case 0:
7141         case -NFS4ERR_STALE_CLIENTID:
7142                 clp->cl_exchange_flags = 0;
7143         }
7144 out:
7145         return ret;
7146 }
7147
7148 struct nfs4_get_lease_time_data {
7149         struct nfs4_get_lease_time_args *args;
7150         struct nfs4_get_lease_time_res *res;
7151         struct nfs_client *clp;
7152 };
7153
7154 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7155                                         void *calldata)
7156 {
7157         struct nfs4_get_lease_time_data *data =
7158                         (struct nfs4_get_lease_time_data *)calldata;
7159
7160         dprintk("--> %s\n", __func__);
7161         /* just setup sequence, do not trigger session recovery
7162            since we're invoked within one */
7163         nfs41_setup_sequence(data->clp->cl_session,
7164                         &data->args->la_seq_args,
7165                         &data->res->lr_seq_res,
7166                         task);
7167         dprintk("<-- %s\n", __func__);
7168 }
7169
7170 /*
7171  * Called from nfs4_state_manager thread for session setup, so don't recover
7172  * from sequence operation or clientid errors.
7173  */
7174 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7175 {
7176         struct nfs4_get_lease_time_data *data =
7177                         (struct nfs4_get_lease_time_data *)calldata;
7178
7179         dprintk("--> %s\n", __func__);
7180         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7181                 return;
7182         switch (task->tk_status) {
7183         case -NFS4ERR_DELAY:
7184         case -NFS4ERR_GRACE:
7185                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7186                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7187                 task->tk_status = 0;
7188                 /* fall through */
7189         case -NFS4ERR_RETRY_UNCACHED_REP:
7190                 rpc_restart_call_prepare(task);
7191                 return;
7192         }
7193         dprintk("<-- %s\n", __func__);
7194 }
7195
7196 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7197         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7198         .rpc_call_done = nfs4_get_lease_time_done,
7199 };
7200
7201 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7202 {
7203         struct rpc_task *task;
7204         struct nfs4_get_lease_time_args args;
7205         struct nfs4_get_lease_time_res res = {
7206                 .lr_fsinfo = fsinfo,
7207         };
7208         struct nfs4_get_lease_time_data data = {
7209                 .args = &args,
7210                 .res = &res,
7211                 .clp = clp,
7212         };
7213         struct rpc_message msg = {
7214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7215                 .rpc_argp = &args,
7216                 .rpc_resp = &res,
7217         };
7218         struct rpc_task_setup task_setup = {
7219                 .rpc_client = clp->cl_rpcclient,
7220                 .rpc_message = &msg,
7221                 .callback_ops = &nfs4_get_lease_time_ops,
7222                 .callback_data = &data,
7223                 .flags = RPC_TASK_TIMEOUT,
7224         };
7225         int status;
7226
7227         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7228         nfs4_set_sequence_privileged(&args.la_seq_args);
7229         dprintk("--> %s\n", __func__);
7230         task = rpc_run_task(&task_setup);
7231
7232         if (IS_ERR(task))
7233                 status = PTR_ERR(task);
7234         else {
7235                 status = task->tk_status;
7236                 rpc_put_task(task);
7237         }
7238         dprintk("<-- %s return %d\n", __func__, status);
7239
7240         return status;
7241 }
7242
7243 /*
7244  * Initialize the values to be used by the client in CREATE_SESSION
7245  * If nfs4_init_session set the fore channel request and response sizes,
7246  * use them.
7247  *
7248  * Set the back channel max_resp_sz_cached to zero to force the client to
7249  * always set csa_cachethis to FALSE because the current implementation
7250  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7251  */
7252 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7253 {
7254         unsigned int max_rqst_sz, max_resp_sz;
7255
7256         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7257         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7258
7259         /* Fore channel attributes */
7260         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7261         args->fc_attrs.max_resp_sz = max_resp_sz;
7262         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7263         args->fc_attrs.max_reqs = max_session_slots;
7264
7265         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7266                 "max_ops=%u max_reqs=%u\n",
7267                 __func__,
7268                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7269                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7270
7271         /* Back channel attributes */
7272         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7273         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7274         args->bc_attrs.max_resp_sz_cached = 0;
7275         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7276         args->bc_attrs.max_reqs = 1;
7277
7278         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7279                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7280                 __func__,
7281                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7282                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7283                 args->bc_attrs.max_reqs);
7284 }
7285
7286 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7287                 struct nfs41_create_session_res *res)
7288 {
7289         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7290         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7291
7292         if (rcvd->max_resp_sz > sent->max_resp_sz)
7293                 return -EINVAL;
7294         /*
7295          * Our requested max_ops is the minimum we need; we're not
7296          * prepared to break up compounds into smaller pieces than that.
7297          * So, no point even trying to continue if the server won't
7298          * cooperate:
7299          */
7300         if (rcvd->max_ops < sent->max_ops)
7301                 return -EINVAL;
7302         if (rcvd->max_reqs == 0)
7303                 return -EINVAL;
7304         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7305                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7306         return 0;
7307 }
7308
7309 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7310                 struct nfs41_create_session_res *res)
7311 {
7312         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7313         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7314
7315         if (!(res->flags & SESSION4_BACK_CHAN))
7316                 goto out;
7317         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7318                 return -EINVAL;
7319         if (rcvd->max_resp_sz < sent->max_resp_sz)
7320                 return -EINVAL;
7321         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7322                 return -EINVAL;
7323         /* These would render the backchannel useless: */
7324         if (rcvd->max_ops != sent->max_ops)
7325                 return -EINVAL;
7326         if (rcvd->max_reqs != sent->max_reqs)
7327                 return -EINVAL;
7328 out:
7329         return 0;
7330 }
7331
7332 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7333                                      struct nfs41_create_session_res *res)
7334 {
7335         int ret;
7336
7337         ret = nfs4_verify_fore_channel_attrs(args, res);
7338         if (ret)
7339                 return ret;
7340         return nfs4_verify_back_channel_attrs(args, res);
7341 }
7342
7343 static void nfs4_update_session(struct nfs4_session *session,
7344                 struct nfs41_create_session_res *res)
7345 {
7346         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7347         /* Mark client id and session as being confirmed */
7348         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7349         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7350         session->flags = res->flags;
7351         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7352         if (res->flags & SESSION4_BACK_CHAN)
7353                 memcpy(&session->bc_attrs, &res->bc_attrs,
7354                                 sizeof(session->bc_attrs));
7355 }
7356
7357 static int _nfs4_proc_create_session(struct nfs_client *clp,
7358                 struct rpc_cred *cred)
7359 {
7360         struct nfs4_session *session = clp->cl_session;
7361         struct nfs41_create_session_args args = {
7362                 .client = clp,
7363                 .clientid = clp->cl_clientid,
7364                 .seqid = clp->cl_seqid,
7365                 .cb_program = NFS4_CALLBACK,
7366         };
7367         struct nfs41_create_session_res res;
7368
7369         struct rpc_message msg = {
7370                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7371                 .rpc_argp = &args,
7372                 .rpc_resp = &res,
7373                 .rpc_cred = cred,
7374         };
7375         int status;
7376
7377         nfs4_init_channel_attrs(&args);
7378         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7379
7380         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7381         trace_nfs4_create_session(clp, status);
7382
7383         if (!status) {
7384                 /* Verify the session's negotiated channel_attrs values */
7385                 status = nfs4_verify_channel_attrs(&args, &res);
7386                 /* Increment the clientid slot sequence id */
7387                 if (clp->cl_seqid == res.seqid)
7388                         clp->cl_seqid++;
7389                 if (status)
7390                         goto out;
7391                 nfs4_update_session(session, &res);
7392         }
7393 out:
7394         return status;
7395 }
7396
7397 /*
7398  * Issues a CREATE_SESSION operation to the server.
7399  * It is the responsibility of the caller to verify the session is
7400  * expired before calling this routine.
7401  */
7402 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7403 {
7404         int status;
7405         unsigned *ptr;
7406         struct nfs4_session *session = clp->cl_session;
7407
7408         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7409
7410         status = _nfs4_proc_create_session(clp, cred);
7411         if (status)
7412                 goto out;
7413
7414         /* Init or reset the session slot tables */
7415         status = nfs4_setup_session_slot_tables(session);
7416         dprintk("slot table setup returned %d\n", status);
7417         if (status)
7418                 goto out;
7419
7420         ptr = (unsigned *)&session->sess_id.data[0];
7421         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7422                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7423 out:
7424         dprintk("<-- %s\n", __func__);
7425         return status;
7426 }
7427
7428 /*
7429  * Issue the over-the-wire RPC DESTROY_SESSION.
7430  * The caller must serialize access to this routine.
7431  */
7432 int nfs4_proc_destroy_session(struct nfs4_session *session,
7433                 struct rpc_cred *cred)
7434 {
7435         struct rpc_message msg = {
7436                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7437                 .rpc_argp = session,
7438                 .rpc_cred = cred,
7439         };
7440         int status = 0;
7441
7442         dprintk("--> nfs4_proc_destroy_session\n");
7443
7444         /* session is still being setup */
7445         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7446                 return 0;
7447
7448         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7449         trace_nfs4_destroy_session(session->clp, status);
7450
7451         if (status)
7452                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7453                         "Session has been destroyed regardless...\n", status);
7454
7455         dprintk("<-- nfs4_proc_destroy_session\n");
7456         return status;
7457 }
7458
7459 /*
7460  * Renew the cl_session lease.
7461  */
7462 struct nfs4_sequence_data {
7463         struct nfs_client *clp;
7464         struct nfs4_sequence_args args;
7465         struct nfs4_sequence_res res;
7466 };
7467
7468 static void nfs41_sequence_release(void *data)
7469 {
7470         struct nfs4_sequence_data *calldata = data;
7471         struct nfs_client *clp = calldata->clp;
7472
7473         if (atomic_read(&clp->cl_count) > 1)
7474                 nfs4_schedule_state_renewal(clp);
7475         nfs_put_client(clp);
7476         kfree(calldata);
7477 }
7478
7479 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7480 {
7481         switch(task->tk_status) {
7482         case -NFS4ERR_DELAY:
7483                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7484                 return -EAGAIN;
7485         default:
7486                 nfs4_schedule_lease_recovery(clp);
7487         }
7488         return 0;
7489 }
7490
7491 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7492 {
7493         struct nfs4_sequence_data *calldata = data;
7494         struct nfs_client *clp = calldata->clp;
7495
7496         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7497                 return;
7498
7499         trace_nfs4_sequence(clp, task->tk_status);
7500         if (task->tk_status < 0) {
7501                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7502                 if (atomic_read(&clp->cl_count) == 1)
7503                         goto out;
7504
7505                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7506                         rpc_restart_call_prepare(task);
7507                         return;
7508                 }
7509         }
7510         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7511 out:
7512         dprintk("<-- %s\n", __func__);
7513 }
7514
7515 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7516 {
7517         struct nfs4_sequence_data *calldata = data;
7518         struct nfs_client *clp = calldata->clp;
7519         struct nfs4_sequence_args *args;
7520         struct nfs4_sequence_res *res;
7521
7522         args = task->tk_msg.rpc_argp;
7523         res = task->tk_msg.rpc_resp;
7524
7525         nfs41_setup_sequence(clp->cl_session, args, res, task);
7526 }
7527
7528 static const struct rpc_call_ops nfs41_sequence_ops = {
7529         .rpc_call_done = nfs41_sequence_call_done,
7530         .rpc_call_prepare = nfs41_sequence_prepare,
7531         .rpc_release = nfs41_sequence_release,
7532 };
7533
7534 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7535                 struct rpc_cred *cred,
7536                 bool is_privileged)
7537 {
7538         struct nfs4_sequence_data *calldata;
7539         struct rpc_message msg = {
7540                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7541                 .rpc_cred = cred,
7542         };
7543         struct rpc_task_setup task_setup_data = {
7544                 .rpc_client = clp->cl_rpcclient,
7545                 .rpc_message = &msg,
7546                 .callback_ops = &nfs41_sequence_ops,
7547                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7548         };
7549
7550         if (!atomic_inc_not_zero(&clp->cl_count))
7551                 return ERR_PTR(-EIO);
7552         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7553         if (calldata == NULL) {
7554                 nfs_put_client(clp);
7555                 return ERR_PTR(-ENOMEM);
7556         }
7557         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7558         if (is_privileged)
7559                 nfs4_set_sequence_privileged(&calldata->args);
7560         msg.rpc_argp = &calldata->args;
7561         msg.rpc_resp = &calldata->res;
7562         calldata->clp = clp;
7563         task_setup_data.callback_data = calldata;
7564
7565         return rpc_run_task(&task_setup_data);
7566 }
7567
7568 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7569 {
7570         struct rpc_task *task;
7571         int ret = 0;
7572
7573         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7574                 return -EAGAIN;
7575         task = _nfs41_proc_sequence(clp, cred, false);
7576         if (IS_ERR(task))
7577                 ret = PTR_ERR(task);
7578         else
7579                 rpc_put_task_async(task);
7580         dprintk("<-- %s status=%d\n", __func__, ret);
7581         return ret;
7582 }
7583
7584 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7585 {
7586         struct rpc_task *task;
7587         int ret;
7588
7589         task = _nfs41_proc_sequence(clp, cred, true);
7590         if (IS_ERR(task)) {
7591                 ret = PTR_ERR(task);
7592                 goto out;
7593         }
7594         ret = rpc_wait_for_completion_task(task);
7595         if (!ret)
7596                 ret = task->tk_status;
7597         rpc_put_task(task);
7598 out:
7599         dprintk("<-- %s status=%d\n", __func__, ret);
7600         return ret;
7601 }
7602
7603 struct nfs4_reclaim_complete_data {
7604         struct nfs_client *clp;
7605         struct nfs41_reclaim_complete_args arg;
7606         struct nfs41_reclaim_complete_res res;
7607 };
7608
7609 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7610 {
7611         struct nfs4_reclaim_complete_data *calldata = data;
7612
7613         nfs41_setup_sequence(calldata->clp->cl_session,
7614                         &calldata->arg.seq_args,
7615                         &calldata->res.seq_res,
7616                         task);
7617 }
7618
7619 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7620 {
7621         switch(task->tk_status) {
7622         case 0:
7623         case -NFS4ERR_COMPLETE_ALREADY:
7624         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7625                 break;
7626         case -NFS4ERR_DELAY:
7627                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7628                 /* fall through */
7629         case -NFS4ERR_RETRY_UNCACHED_REP:
7630                 return -EAGAIN;
7631         default:
7632                 nfs4_schedule_lease_recovery(clp);
7633         }
7634         return 0;
7635 }
7636
7637 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7638 {
7639         struct nfs4_reclaim_complete_data *calldata = data;
7640         struct nfs_client *clp = calldata->clp;
7641         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7642
7643         dprintk("--> %s\n", __func__);
7644         if (!nfs41_sequence_done(task, res))
7645                 return;
7646
7647         trace_nfs4_reclaim_complete(clp, task->tk_status);
7648         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7649                 rpc_restart_call_prepare(task);
7650                 return;
7651         }
7652         dprintk("<-- %s\n", __func__);
7653 }
7654
7655 static void nfs4_free_reclaim_complete_data(void *data)
7656 {
7657         struct nfs4_reclaim_complete_data *calldata = data;
7658
7659         kfree(calldata);
7660 }
7661
7662 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7663         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7664         .rpc_call_done = nfs4_reclaim_complete_done,
7665         .rpc_release = nfs4_free_reclaim_complete_data,
7666 };
7667
7668 /*
7669  * Issue a global reclaim complete.
7670  */
7671 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7672                 struct rpc_cred *cred)
7673 {
7674         struct nfs4_reclaim_complete_data *calldata;
7675         struct rpc_task *task;
7676         struct rpc_message msg = {
7677                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7678                 .rpc_cred = cred,
7679         };
7680         struct rpc_task_setup task_setup_data = {
7681                 .rpc_client = clp->cl_rpcclient,
7682                 .rpc_message = &msg,
7683                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7684                 .flags = RPC_TASK_ASYNC,
7685         };
7686         int status = -ENOMEM;
7687
7688         dprintk("--> %s\n", __func__);
7689         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7690         if (calldata == NULL)
7691                 goto out;
7692         calldata->clp = clp;
7693         calldata->arg.one_fs = 0;
7694
7695         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7696         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7697         msg.rpc_argp = &calldata->arg;
7698         msg.rpc_resp = &calldata->res;
7699         task_setup_data.callback_data = calldata;
7700         task = rpc_run_task(&task_setup_data);
7701         if (IS_ERR(task)) {
7702                 status = PTR_ERR(task);
7703                 goto out;
7704         }
7705         status = nfs4_wait_for_completion_rpc_task(task);
7706         if (status == 0)
7707                 status = task->tk_status;
7708         rpc_put_task(task);
7709         return 0;
7710 out:
7711         dprintk("<-- %s status=%d\n", __func__, status);
7712         return status;
7713 }
7714
7715 static void
7716 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7717 {
7718         struct nfs4_layoutget *lgp = calldata;
7719         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7720         struct nfs4_session *session = nfs4_get_session(server);
7721
7722         dprintk("--> %s\n", __func__);
7723         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7724          * right now covering the LAYOUTGET we are about to send.
7725          * However, that is not so catastrophic, and there seems
7726          * to be no way to prevent it completely.
7727          */
7728         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7729                                 &lgp->res.seq_res, task))
7730                 return;
7731         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7732                                           NFS_I(lgp->args.inode)->layout,
7733                                           &lgp->args.range,
7734                                           lgp->args.ctx->state)) {
7735                 rpc_exit(task, NFS4_OK);
7736         }
7737 }
7738
7739 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7740 {
7741         struct nfs4_layoutget *lgp = calldata;
7742         struct inode *inode = lgp->args.inode;
7743         struct nfs_server *server = NFS_SERVER(inode);
7744         struct pnfs_layout_hdr *lo;
7745         struct nfs4_state *state = NULL;
7746         unsigned long timeo, now, giveup;
7747
7748         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7749
7750         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7751                 goto out;
7752
7753         switch (task->tk_status) {
7754         case 0:
7755                 goto out;
7756         /*
7757          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7758          * (or clients) writing to the same RAID stripe
7759          */
7760         case -NFS4ERR_LAYOUTTRYLATER:
7761         /*
7762          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7763          * existing layout before getting a new one).
7764          */
7765         case -NFS4ERR_RECALLCONFLICT:
7766                 timeo = rpc_get_timeout(task->tk_client);
7767                 giveup = lgp->args.timestamp + timeo;
7768                 now = jiffies;
7769                 if (time_after(giveup, now)) {
7770                         unsigned long delay;
7771
7772                         /* Delay for:
7773                          * - Not less then NFS4_POLL_RETRY_MIN.
7774                          * - One last time a jiffie before we give up
7775                          * - exponential backoff (time_now minus start_attempt)
7776                          */
7777                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7778                                     min((giveup - now - 1),
7779                                         now - lgp->args.timestamp));
7780
7781                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7782                                 __func__, delay);
7783                         rpc_delay(task, delay);
7784                         task->tk_status = 0;
7785                         rpc_restart_call_prepare(task);
7786                         goto out; /* Do not call nfs4_async_handle_error() */
7787                 }
7788                 break;
7789         case -NFS4ERR_EXPIRED:
7790         case -NFS4ERR_BAD_STATEID:
7791                 spin_lock(&inode->i_lock);
7792                 lo = NFS_I(inode)->layout;
7793                 if (!lo || list_empty(&lo->plh_segs)) {
7794                         spin_unlock(&inode->i_lock);
7795                         /* If the open stateid was bad, then recover it. */
7796                         state = lgp->args.ctx->state;
7797                 } else {
7798                         LIST_HEAD(head);
7799
7800                         /*
7801                          * Mark the bad layout state as invalid, then retry
7802                          * with the current stateid.
7803                          */
7804                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7805                         spin_unlock(&inode->i_lock);
7806                         pnfs_free_lseg_list(&head);
7807
7808                         task->tk_status = 0;
7809                         rpc_restart_call_prepare(task);
7810                 }
7811         }
7812         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7813                 rpc_restart_call_prepare(task);
7814 out:
7815         dprintk("<-- %s\n", __func__);
7816 }
7817
7818 static size_t max_response_pages(struct nfs_server *server)
7819 {
7820         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7821         return nfs_page_array_len(0, max_resp_sz);
7822 }
7823
7824 static void nfs4_free_pages(struct page **pages, size_t size)
7825 {
7826         int i;
7827
7828         if (!pages)
7829                 return;
7830
7831         for (i = 0; i < size; i++) {
7832                 if (!pages[i])
7833                         break;
7834                 __free_page(pages[i]);
7835         }
7836         kfree(pages);
7837 }
7838
7839 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7840 {
7841         struct page **pages;
7842         int i;
7843
7844         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7845         if (!pages) {
7846                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7847                 return NULL;
7848         }
7849
7850         for (i = 0; i < size; i++) {
7851                 pages[i] = alloc_page(gfp_flags);
7852                 if (!pages[i]) {
7853                         dprintk("%s: failed to allocate page\n", __func__);
7854                         nfs4_free_pages(pages, size);
7855                         return NULL;
7856                 }
7857         }
7858
7859         return pages;
7860 }
7861
7862 static void nfs4_layoutget_release(void *calldata)
7863 {
7864         struct nfs4_layoutget *lgp = calldata;
7865         struct inode *inode = lgp->args.inode;
7866         struct nfs_server *server = NFS_SERVER(inode);
7867         size_t max_pages = max_response_pages(server);
7868
7869         dprintk("--> %s\n", __func__);
7870         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7871         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7872         put_nfs_open_context(lgp->args.ctx);
7873         kfree(calldata);
7874         dprintk("<-- %s\n", __func__);
7875 }
7876
7877 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7878         .rpc_call_prepare = nfs4_layoutget_prepare,
7879         .rpc_call_done = nfs4_layoutget_done,
7880         .rpc_release = nfs4_layoutget_release,
7881 };
7882
7883 struct pnfs_layout_segment *
7884 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7885 {
7886         struct inode *inode = lgp->args.inode;
7887         struct nfs_server *server = NFS_SERVER(inode);
7888         size_t max_pages = max_response_pages(server);
7889         struct rpc_task *task;
7890         struct rpc_message msg = {
7891                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7892                 .rpc_argp = &lgp->args,
7893                 .rpc_resp = &lgp->res,
7894                 .rpc_cred = lgp->cred,
7895         };
7896         struct rpc_task_setup task_setup_data = {
7897                 .rpc_client = server->client,
7898                 .rpc_message = &msg,
7899                 .callback_ops = &nfs4_layoutget_call_ops,
7900                 .callback_data = lgp,
7901                 .flags = RPC_TASK_ASYNC,
7902         };
7903         struct pnfs_layout_segment *lseg = NULL;
7904         int status = 0;
7905
7906         dprintk("--> %s\n", __func__);
7907
7908         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7909         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7910
7911         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7912         if (!lgp->args.layout.pages) {
7913                 nfs4_layoutget_release(lgp);
7914                 return ERR_PTR(-ENOMEM);
7915         }
7916         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7917         lgp->args.timestamp = jiffies;
7918
7919         lgp->res.layoutp = &lgp->args.layout;
7920         lgp->res.seq_res.sr_slot = NULL;
7921         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7922
7923         task = rpc_run_task(&task_setup_data);
7924         if (IS_ERR(task))
7925                 return ERR_CAST(task);
7926         status = nfs4_wait_for_completion_rpc_task(task);
7927         if (status == 0)
7928                 status = task->tk_status;
7929         trace_nfs4_layoutget(lgp->args.ctx,
7930                         &lgp->args.range,
7931                         &lgp->res.range,
7932                         status);
7933         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7934         if (status == 0 && lgp->res.layoutp->len)
7935                 lseg = pnfs_layout_process(lgp);
7936         rpc_put_task(task);
7937         dprintk("<-- %s status=%d\n", __func__, status);
7938         if (status)
7939                 return ERR_PTR(status);
7940         return lseg;
7941 }
7942
7943 static void
7944 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7945 {
7946         struct nfs4_layoutreturn *lrp = calldata;
7947
7948         dprintk("--> %s\n", __func__);
7949         nfs41_setup_sequence(lrp->clp->cl_session,
7950                         &lrp->args.seq_args,
7951                         &lrp->res.seq_res,
7952                         task);
7953 }
7954
7955 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7956 {
7957         struct nfs4_layoutreturn *lrp = calldata;
7958         struct nfs_server *server;
7959
7960         dprintk("--> %s\n", __func__);
7961
7962         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7963                 return;
7964
7965         server = NFS_SERVER(lrp->args.inode);
7966         switch (task->tk_status) {
7967         default:
7968                 task->tk_status = 0;
7969         case 0:
7970                 break;
7971         case -NFS4ERR_DELAY:
7972                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7973                         break;
7974                 rpc_restart_call_prepare(task);
7975                 return;
7976         }
7977         dprintk("<-- %s\n", __func__);
7978 }
7979
7980 static void nfs4_layoutreturn_release(void *calldata)
7981 {
7982         struct nfs4_layoutreturn *lrp = calldata;
7983         struct pnfs_layout_hdr *lo = lrp->args.layout;
7984         LIST_HEAD(freeme);
7985
7986         dprintk("--> %s\n", __func__);
7987         spin_lock(&lo->plh_inode->i_lock);
7988         if (lrp->res.lrs_present)
7989                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7990         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
7991         pnfs_clear_layoutreturn_waitbit(lo);
7992         lo->plh_block_lgets--;
7993         spin_unlock(&lo->plh_inode->i_lock);
7994         pnfs_free_lseg_list(&freeme);
7995         pnfs_put_layout_hdr(lrp->args.layout);
7996         nfs_iput_and_deactive(lrp->inode);
7997         kfree(calldata);
7998         dprintk("<-- %s\n", __func__);
7999 }
8000
8001 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8002         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8003         .rpc_call_done = nfs4_layoutreturn_done,
8004         .rpc_release = nfs4_layoutreturn_release,
8005 };
8006
8007 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8008 {
8009         struct rpc_task *task;
8010         struct rpc_message msg = {
8011                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8012                 .rpc_argp = &lrp->args,
8013                 .rpc_resp = &lrp->res,
8014                 .rpc_cred = lrp->cred,
8015         };
8016         struct rpc_task_setup task_setup_data = {
8017                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8018                 .rpc_message = &msg,
8019                 .callback_ops = &nfs4_layoutreturn_call_ops,
8020                 .callback_data = lrp,
8021         };
8022         int status = 0;
8023
8024         dprintk("--> %s\n", __func__);
8025         if (!sync) {
8026                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8027                 if (!lrp->inode) {
8028                         nfs4_layoutreturn_release(lrp);
8029                         return -EAGAIN;
8030                 }
8031                 task_setup_data.flags |= RPC_TASK_ASYNC;
8032         }
8033         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8034         task = rpc_run_task(&task_setup_data);
8035         if (IS_ERR(task))
8036                 return PTR_ERR(task);
8037         if (sync)
8038                 status = task->tk_status;
8039         trace_nfs4_layoutreturn(lrp->args.inode, status);
8040         dprintk("<-- %s status=%d\n", __func__, status);
8041         rpc_put_task(task);
8042         return status;
8043 }
8044
8045 static int
8046 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8047                 struct pnfs_device *pdev,
8048                 struct rpc_cred *cred)
8049 {
8050         struct nfs4_getdeviceinfo_args args = {
8051                 .pdev = pdev,
8052                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8053                         NOTIFY_DEVICEID4_DELETE,
8054         };
8055         struct nfs4_getdeviceinfo_res res = {
8056                 .pdev = pdev,
8057         };
8058         struct rpc_message msg = {
8059                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8060                 .rpc_argp = &args,
8061                 .rpc_resp = &res,
8062                 .rpc_cred = cred,
8063         };
8064         int status;
8065
8066         dprintk("--> %s\n", __func__);
8067         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8068         if (res.notification & ~args.notify_types)
8069                 dprintk("%s: unsupported notification\n", __func__);
8070         if (res.notification != args.notify_types)
8071                 pdev->nocache = 1;
8072
8073         dprintk("<-- %s status=%d\n", __func__, status);
8074
8075         return status;
8076 }
8077
8078 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8079                 struct pnfs_device *pdev,
8080                 struct rpc_cred *cred)
8081 {
8082         struct nfs4_exception exception = { };
8083         int err;
8084
8085         do {
8086                 err = nfs4_handle_exception(server,
8087                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8088                                         &exception);
8089         } while (exception.retry);
8090         return err;
8091 }
8092 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8093
8094 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8095 {
8096         struct nfs4_layoutcommit_data *data = calldata;
8097         struct nfs_server *server = NFS_SERVER(data->args.inode);
8098         struct nfs4_session *session = nfs4_get_session(server);
8099
8100         nfs41_setup_sequence(session,
8101                         &data->args.seq_args,
8102                         &data->res.seq_res,
8103                         task);
8104 }
8105
8106 static void
8107 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8108 {
8109         struct nfs4_layoutcommit_data *data = calldata;
8110         struct nfs_server *server = NFS_SERVER(data->args.inode);
8111
8112         if (!nfs41_sequence_done(task, &data->res.seq_res))
8113                 return;
8114
8115         switch (task->tk_status) { /* Just ignore these failures */
8116         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8117         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8118         case -NFS4ERR_BADLAYOUT:     /* no layout */
8119         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8120                 task->tk_status = 0;
8121         case 0:
8122                 break;
8123         default:
8124                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8125                         rpc_restart_call_prepare(task);
8126                         return;
8127                 }
8128         }
8129 }
8130
8131 static void nfs4_layoutcommit_release(void *calldata)
8132 {
8133         struct nfs4_layoutcommit_data *data = calldata;
8134
8135         pnfs_cleanup_layoutcommit(data);
8136         nfs_post_op_update_inode_force_wcc(data->args.inode,
8137                                            data->res.fattr);
8138         put_rpccred(data->cred);
8139         nfs_iput_and_deactive(data->inode);
8140         kfree(data);
8141 }
8142
8143 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8144         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8145         .rpc_call_done = nfs4_layoutcommit_done,
8146         .rpc_release = nfs4_layoutcommit_release,
8147 };
8148
8149 int
8150 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8151 {
8152         struct rpc_message msg = {
8153                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8154                 .rpc_argp = &data->args,
8155                 .rpc_resp = &data->res,
8156                 .rpc_cred = data->cred,
8157         };
8158         struct rpc_task_setup task_setup_data = {
8159                 .task = &data->task,
8160                 .rpc_client = NFS_CLIENT(data->args.inode),
8161                 .rpc_message = &msg,
8162                 .callback_ops = &nfs4_layoutcommit_ops,
8163                 .callback_data = data,
8164         };
8165         struct rpc_task *task;
8166         int status = 0;
8167
8168         dprintk("NFS: initiating layoutcommit call. sync %d "
8169                 "lbw: %llu inode %lu\n", sync,
8170                 data->args.lastbytewritten,
8171                 data->args.inode->i_ino);
8172
8173         if (!sync) {
8174                 data->inode = nfs_igrab_and_active(data->args.inode);
8175                 if (data->inode == NULL) {
8176                         nfs4_layoutcommit_release(data);
8177                         return -EAGAIN;
8178                 }
8179                 task_setup_data.flags = RPC_TASK_ASYNC;
8180         }
8181         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8182         task = rpc_run_task(&task_setup_data);
8183         if (IS_ERR(task))
8184                 return PTR_ERR(task);
8185         if (sync)
8186                 status = task->tk_status;
8187         trace_nfs4_layoutcommit(data->args.inode, status);
8188         dprintk("%s: status %d\n", __func__, status);
8189         rpc_put_task(task);
8190         return status;
8191 }
8192
8193 /**
8194  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8195  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8196  */
8197 static int
8198 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8199                     struct nfs_fsinfo *info,
8200                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8201 {
8202         struct nfs41_secinfo_no_name_args args = {
8203                 .style = SECINFO_STYLE_CURRENT_FH,
8204         };
8205         struct nfs4_secinfo_res res = {
8206                 .flavors = flavors,
8207         };
8208         struct rpc_message msg = {
8209                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8210                 .rpc_argp = &args,
8211                 .rpc_resp = &res,
8212         };
8213         struct rpc_clnt *clnt = server->client;
8214         struct rpc_cred *cred = NULL;
8215         int status;
8216
8217         if (use_integrity) {
8218                 clnt = server->nfs_client->cl_rpcclient;
8219                 cred = nfs4_get_clid_cred(server->nfs_client);
8220                 msg.rpc_cred = cred;
8221         }
8222
8223         dprintk("--> %s\n", __func__);
8224         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8225                                 &res.seq_res, 0);
8226         dprintk("<-- %s status=%d\n", __func__, status);
8227
8228         if (cred)
8229                 put_rpccred(cred);
8230
8231         return status;
8232 }
8233
8234 static int
8235 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8236                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8237 {
8238         struct nfs4_exception exception = { };
8239         int err;
8240         do {
8241                 /* first try using integrity protection */
8242                 err = -NFS4ERR_WRONGSEC;
8243
8244                 /* try to use integrity protection with machine cred */
8245                 if (_nfs4_is_integrity_protected(server->nfs_client))
8246                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8247                                                           flavors, true);
8248
8249                 /*
8250                  * if unable to use integrity protection, or SECINFO with
8251                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8252                  * disallowed by spec, but exists in deployed servers) use
8253                  * the current filesystem's rpc_client and the user cred.
8254                  */
8255                 if (err == -NFS4ERR_WRONGSEC)
8256                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8257                                                           flavors, false);
8258
8259                 switch (err) {
8260                 case 0:
8261                 case -NFS4ERR_WRONGSEC:
8262                 case -ENOTSUPP:
8263                         goto out;
8264                 default:
8265                         err = nfs4_handle_exception(server, err, &exception);
8266                 }
8267         } while (exception.retry);
8268 out:
8269         return err;
8270 }
8271
8272 static int
8273 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8274                     struct nfs_fsinfo *info)
8275 {
8276         int err;
8277         struct page *page;
8278         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8279         struct nfs4_secinfo_flavors *flavors;
8280         struct nfs4_secinfo4 *secinfo;
8281         int i;
8282
8283         page = alloc_page(GFP_KERNEL);
8284         if (!page) {
8285                 err = -ENOMEM;
8286                 goto out;
8287         }
8288
8289         flavors = page_address(page);
8290         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8291
8292         /*
8293          * Fall back on "guess and check" method if
8294          * the server doesn't support SECINFO_NO_NAME
8295          */
8296         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8297                 err = nfs4_find_root_sec(server, fhandle, info);
8298                 goto out_freepage;
8299         }
8300         if (err)
8301                 goto out_freepage;
8302
8303         for (i = 0; i < flavors->num_flavors; i++) {
8304                 secinfo = &flavors->flavors[i];
8305
8306                 switch (secinfo->flavor) {
8307                 case RPC_AUTH_NULL:
8308                 case RPC_AUTH_UNIX:
8309                 case RPC_AUTH_GSS:
8310                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8311                                         &secinfo->flavor_info);
8312                         break;
8313                 default:
8314                         flavor = RPC_AUTH_MAXFLAVOR;
8315                         break;
8316                 }
8317
8318                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8319                         flavor = RPC_AUTH_MAXFLAVOR;
8320
8321                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8322                         err = nfs4_lookup_root_sec(server, fhandle,
8323                                                    info, flavor);
8324                         if (!err)
8325                                 break;
8326                 }
8327         }
8328
8329         if (flavor == RPC_AUTH_MAXFLAVOR)
8330                 err = -EPERM;
8331
8332 out_freepage:
8333         put_page(page);
8334         if (err == -EACCES)
8335                 return -EPERM;
8336 out:
8337         return err;
8338 }
8339
8340 static int _nfs41_test_stateid(struct nfs_server *server,
8341                 nfs4_stateid *stateid,
8342                 struct rpc_cred *cred)
8343 {
8344         int status;
8345         struct nfs41_test_stateid_args args = {
8346                 .stateid = stateid,
8347         };
8348         struct nfs41_test_stateid_res res;
8349         struct rpc_message msg = {
8350                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8351                 .rpc_argp = &args,
8352                 .rpc_resp = &res,
8353                 .rpc_cred = cred,
8354         };
8355         struct rpc_clnt *rpc_client = server->client;
8356
8357         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8358                 &rpc_client, &msg);
8359
8360         dprintk("NFS call  test_stateid %p\n", stateid);
8361         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8362         nfs4_set_sequence_privileged(&args.seq_args);
8363         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8364                         &args.seq_args, &res.seq_res);
8365         if (status != NFS_OK) {
8366                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8367                 return status;
8368         }
8369         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8370         return -res.status;
8371 }
8372
8373 /**
8374  * nfs41_test_stateid - perform a TEST_STATEID operation
8375  *
8376  * @server: server / transport on which to perform the operation
8377  * @stateid: state ID to test
8378  * @cred: credential
8379  *
8380  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8381  * Otherwise a negative NFS4ERR value is returned if the operation
8382  * failed or the state ID is not currently valid.
8383  */
8384 static int nfs41_test_stateid(struct nfs_server *server,
8385                 nfs4_stateid *stateid,
8386                 struct rpc_cred *cred)
8387 {
8388         struct nfs4_exception exception = { };
8389         int err;
8390         do {
8391                 err = _nfs41_test_stateid(server, stateid, cred);
8392                 if (err != -NFS4ERR_DELAY)
8393                         break;
8394                 nfs4_handle_exception(server, err, &exception);
8395         } while (exception.retry);
8396         return err;
8397 }
8398
8399 struct nfs_free_stateid_data {
8400         struct nfs_server *server;
8401         struct nfs41_free_stateid_args args;
8402         struct nfs41_free_stateid_res res;
8403 };
8404
8405 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8406 {
8407         struct nfs_free_stateid_data *data = calldata;
8408         nfs41_setup_sequence(nfs4_get_session(data->server),
8409                         &data->args.seq_args,
8410                         &data->res.seq_res,
8411                         task);
8412 }
8413
8414 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8415 {
8416         struct nfs_free_stateid_data *data = calldata;
8417
8418         nfs41_sequence_done(task, &data->res.seq_res);
8419
8420         switch (task->tk_status) {
8421         case -NFS4ERR_DELAY:
8422                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8423                         rpc_restart_call_prepare(task);
8424         }
8425 }
8426
8427 static void nfs41_free_stateid_release(void *calldata)
8428 {
8429         kfree(calldata);
8430 }
8431
8432 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8433         .rpc_call_prepare = nfs41_free_stateid_prepare,
8434         .rpc_call_done = nfs41_free_stateid_done,
8435         .rpc_release = nfs41_free_stateid_release,
8436 };
8437
8438 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8439                 nfs4_stateid *stateid,
8440                 struct rpc_cred *cred,
8441                 bool privileged)
8442 {
8443         struct rpc_message msg = {
8444                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8445                 .rpc_cred = cred,
8446         };
8447         struct rpc_task_setup task_setup = {
8448                 .rpc_client = server->client,
8449                 .rpc_message = &msg,
8450                 .callback_ops = &nfs41_free_stateid_ops,
8451                 .flags = RPC_TASK_ASYNC,
8452         };
8453         struct nfs_free_stateid_data *data;
8454
8455         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8456                 &task_setup.rpc_client, &msg);
8457
8458         dprintk("NFS call  free_stateid %p\n", stateid);
8459         data = kmalloc(sizeof(*data), GFP_NOFS);
8460         if (!data)
8461                 return ERR_PTR(-ENOMEM);
8462         data->server = server;
8463         nfs4_stateid_copy(&data->args.stateid, stateid);
8464
8465         task_setup.callback_data = data;
8466
8467         msg.rpc_argp = &data->args;
8468         msg.rpc_resp = &data->res;
8469         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8470         if (privileged)
8471                 nfs4_set_sequence_privileged(&data->args.seq_args);
8472
8473         return rpc_run_task(&task_setup);
8474 }
8475
8476 /**
8477  * nfs41_free_stateid - perform a FREE_STATEID operation
8478  *
8479  * @server: server / transport on which to perform the operation
8480  * @stateid: state ID to release
8481  * @cred: credential
8482  *
8483  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8484  * negative NFS4ERR value is returned.
8485  */
8486 static int nfs41_free_stateid(struct nfs_server *server,
8487                 nfs4_stateid *stateid,
8488                 struct rpc_cred *cred)
8489 {
8490         struct rpc_task *task;
8491         int ret;
8492
8493         task = _nfs41_free_stateid(server, stateid, cred, true);
8494         if (IS_ERR(task))
8495                 return PTR_ERR(task);
8496         ret = rpc_wait_for_completion_task(task);
8497         if (!ret)
8498                 ret = task->tk_status;
8499         rpc_put_task(task);
8500         return ret;
8501 }
8502
8503 static void
8504 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8505 {
8506         struct rpc_task *task;
8507         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8508
8509         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8510         nfs4_free_lock_state(server, lsp);
8511         if (IS_ERR(task))
8512                 return;
8513         rpc_put_task(task);
8514 }
8515
8516 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8517                 const nfs4_stateid *s2)
8518 {
8519         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8520                 return false;
8521
8522         if (s1->seqid == s2->seqid)
8523                 return true;
8524         if (s1->seqid == 0 || s2->seqid == 0)
8525                 return true;
8526
8527         return false;
8528 }
8529
8530 #endif /* CONFIG_NFS_V4_1 */
8531
8532 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8533                 const nfs4_stateid *s2)
8534 {
8535         return nfs4_stateid_match(s1, s2);
8536 }
8537
8538
8539 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8540         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8541         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8542         .recover_open   = nfs4_open_reclaim,
8543         .recover_lock   = nfs4_lock_reclaim,
8544         .establish_clid = nfs4_init_clientid,
8545         .detect_trunking = nfs40_discover_server_trunking,
8546 };
8547
8548 #if defined(CONFIG_NFS_V4_1)
8549 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8550         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8551         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8552         .recover_open   = nfs4_open_reclaim,
8553         .recover_lock   = nfs4_lock_reclaim,
8554         .establish_clid = nfs41_init_clientid,
8555         .reclaim_complete = nfs41_proc_reclaim_complete,
8556         .detect_trunking = nfs41_discover_server_trunking,
8557 };
8558 #endif /* CONFIG_NFS_V4_1 */
8559
8560 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8561         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8562         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8563         .recover_open   = nfs40_open_expired,
8564         .recover_lock   = nfs4_lock_expired,
8565         .establish_clid = nfs4_init_clientid,
8566 };
8567
8568 #if defined(CONFIG_NFS_V4_1)
8569 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8570         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8571         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8572         .recover_open   = nfs41_open_expired,
8573         .recover_lock   = nfs41_lock_expired,
8574         .establish_clid = nfs41_init_clientid,
8575 };
8576 #endif /* CONFIG_NFS_V4_1 */
8577
8578 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8579         .sched_state_renewal = nfs4_proc_async_renew,
8580         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8581         .renew_lease = nfs4_proc_renew,
8582 };
8583
8584 #if defined(CONFIG_NFS_V4_1)
8585 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8586         .sched_state_renewal = nfs41_proc_async_sequence,
8587         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8588         .renew_lease = nfs4_proc_sequence,
8589 };
8590 #endif
8591
8592 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8593         .get_locations = _nfs40_proc_get_locations,
8594         .fsid_present = _nfs40_proc_fsid_present,
8595 };
8596
8597 #if defined(CONFIG_NFS_V4_1)
8598 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8599         .get_locations = _nfs41_proc_get_locations,
8600         .fsid_present = _nfs41_proc_fsid_present,
8601 };
8602 #endif  /* CONFIG_NFS_V4_1 */
8603
8604 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8605         .minor_version = 0,
8606         .init_caps = NFS_CAP_READDIRPLUS
8607                 | NFS_CAP_ATOMIC_OPEN
8608                 | NFS_CAP_POSIX_LOCK,
8609         .init_client = nfs40_init_client,
8610         .shutdown_client = nfs40_shutdown_client,
8611         .match_stateid = nfs4_match_stateid,
8612         .find_root_sec = nfs4_find_root_sec,
8613         .free_lock_state = nfs4_release_lockowner,
8614         .alloc_seqid = nfs_alloc_seqid,
8615         .call_sync_ops = &nfs40_call_sync_ops,
8616         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8617         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8618         .state_renewal_ops = &nfs40_state_renewal_ops,
8619         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8620 };
8621
8622 #if defined(CONFIG_NFS_V4_1)
8623 static struct nfs_seqid *
8624 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8625 {
8626         return NULL;
8627 }
8628
8629 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8630         .minor_version = 1,
8631         .init_caps = NFS_CAP_READDIRPLUS
8632                 | NFS_CAP_ATOMIC_OPEN
8633                 | NFS_CAP_POSIX_LOCK
8634                 | NFS_CAP_STATEID_NFSV41
8635                 | NFS_CAP_ATOMIC_OPEN_V1,
8636         .init_client = nfs41_init_client,
8637         .shutdown_client = nfs41_shutdown_client,
8638         .match_stateid = nfs41_match_stateid,
8639         .find_root_sec = nfs41_find_root_sec,
8640         .free_lock_state = nfs41_free_lock_state,
8641         .alloc_seqid = nfs_alloc_no_seqid,
8642         .call_sync_ops = &nfs41_call_sync_ops,
8643         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8644         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8645         .state_renewal_ops = &nfs41_state_renewal_ops,
8646         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8647 };
8648 #endif
8649
8650 #if defined(CONFIG_NFS_V4_2)
8651 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8652         .minor_version = 2,
8653         .init_caps = NFS_CAP_READDIRPLUS
8654                 | NFS_CAP_ATOMIC_OPEN
8655                 | NFS_CAP_POSIX_LOCK
8656                 | NFS_CAP_STATEID_NFSV41
8657                 | NFS_CAP_ATOMIC_OPEN_V1
8658                 | NFS_CAP_ALLOCATE
8659                 | NFS_CAP_DEALLOCATE
8660                 | NFS_CAP_SEEK
8661                 | NFS_CAP_LAYOUTSTATS,
8662         .init_client = nfs41_init_client,
8663         .shutdown_client = nfs41_shutdown_client,
8664         .match_stateid = nfs41_match_stateid,
8665         .find_root_sec = nfs41_find_root_sec,
8666         .free_lock_state = nfs41_free_lock_state,
8667         .call_sync_ops = &nfs41_call_sync_ops,
8668         .alloc_seqid = nfs_alloc_no_seqid,
8669         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8670         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8671         .state_renewal_ops = &nfs41_state_renewal_ops,
8672         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8673 };
8674 #endif
8675
8676 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8677         [0] = &nfs_v4_0_minor_ops,
8678 #if defined(CONFIG_NFS_V4_1)
8679         [1] = &nfs_v4_1_minor_ops,
8680 #endif
8681 #if defined(CONFIG_NFS_V4_2)
8682         [2] = &nfs_v4_2_minor_ops,
8683 #endif
8684 };
8685
8686 static const struct inode_operations nfs4_dir_inode_operations = {
8687         .create         = nfs_create,
8688         .lookup         = nfs_lookup,
8689         .atomic_open    = nfs_atomic_open,
8690         .link           = nfs_link,
8691         .unlink         = nfs_unlink,
8692         .symlink        = nfs_symlink,
8693         .mkdir          = nfs_mkdir,
8694         .rmdir          = nfs_rmdir,
8695         .mknod          = nfs_mknod,
8696         .rename         = nfs_rename,
8697         .permission     = nfs_permission,
8698         .getattr        = nfs_getattr,
8699         .setattr        = nfs_setattr,
8700         .getxattr       = generic_getxattr,
8701         .setxattr       = generic_setxattr,
8702         .listxattr      = generic_listxattr,
8703         .removexattr    = generic_removexattr,
8704 };
8705
8706 static const struct inode_operations nfs4_file_inode_operations = {
8707         .permission     = nfs_permission,
8708         .getattr        = nfs_getattr,
8709         .setattr        = nfs_setattr,
8710         .getxattr       = generic_getxattr,
8711         .setxattr       = generic_setxattr,
8712         .listxattr      = generic_listxattr,
8713         .removexattr    = generic_removexattr,
8714 };
8715
8716 const struct nfs_rpc_ops nfs_v4_clientops = {
8717         .version        = 4,                    /* protocol version */
8718         .dentry_ops     = &nfs4_dentry_operations,
8719         .dir_inode_ops  = &nfs4_dir_inode_operations,
8720         .file_inode_ops = &nfs4_file_inode_operations,
8721         .file_ops       = &nfs4_file_operations,
8722         .getroot        = nfs4_proc_get_root,
8723         .submount       = nfs4_submount,
8724         .try_mount      = nfs4_try_mount,
8725         .getattr        = nfs4_proc_getattr,
8726         .setattr        = nfs4_proc_setattr,
8727         .lookup         = nfs4_proc_lookup,
8728         .access         = nfs4_proc_access,
8729         .readlink       = nfs4_proc_readlink,
8730         .create         = nfs4_proc_create,
8731         .remove         = nfs4_proc_remove,
8732         .unlink_setup   = nfs4_proc_unlink_setup,
8733         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8734         .unlink_done    = nfs4_proc_unlink_done,
8735         .rename_setup   = nfs4_proc_rename_setup,
8736         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8737         .rename_done    = nfs4_proc_rename_done,
8738         .link           = nfs4_proc_link,
8739         .symlink        = nfs4_proc_symlink,
8740         .mkdir          = nfs4_proc_mkdir,
8741         .rmdir          = nfs4_proc_remove,
8742         .readdir        = nfs4_proc_readdir,
8743         .mknod          = nfs4_proc_mknod,
8744         .statfs         = nfs4_proc_statfs,
8745         .fsinfo         = nfs4_proc_fsinfo,
8746         .pathconf       = nfs4_proc_pathconf,
8747         .set_capabilities = nfs4_server_capabilities,
8748         .decode_dirent  = nfs4_decode_dirent,
8749         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8750         .read_setup     = nfs4_proc_read_setup,
8751         .read_done      = nfs4_read_done,
8752         .write_setup    = nfs4_proc_write_setup,
8753         .write_done     = nfs4_write_done,
8754         .commit_setup   = nfs4_proc_commit_setup,
8755         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8756         .commit_done    = nfs4_commit_done,
8757         .lock           = nfs4_proc_lock,
8758         .clear_acl_cache = nfs4_zap_acl_attr,
8759         .close_context  = nfs4_close_context,
8760         .open_context   = nfs4_atomic_open,
8761         .have_delegation = nfs4_have_delegation,
8762         .return_delegation = nfs4_inode_return_delegation,
8763         .alloc_client   = nfs4_alloc_client,
8764         .init_client    = nfs4_init_client,
8765         .free_client    = nfs4_free_client,
8766         .create_server  = nfs4_create_server,
8767         .clone_server   = nfs_clone_server,
8768 };
8769
8770 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8771         .prefix = XATTR_NAME_NFSV4_ACL,
8772         .list   = nfs4_xattr_list_nfs4_acl,
8773         .get    = nfs4_xattr_get_nfs4_acl,
8774         .set    = nfs4_xattr_set_nfs4_acl,
8775 };
8776
8777 const struct xattr_handler *nfs4_xattr_handlers[] = {
8778         &nfs4_xattr_nfs4_acl_handler,
8779 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8780         &nfs4_xattr_nfs4_label_handler,
8781 #endif
8782         NULL
8783 };
8784
8785 /*
8786  * Local variables:
8787  *  c-basic-offset: 8
8788  * End:
8789  */