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