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 /* file attributes which can be mapped to nfs attributes */
  78 #define NFS4_VALID_ATTRS (ATTR_MODE \
  79         | ATTR_UID \
  80         | ATTR_GID \
  81         | ATTR_SIZE \
  82         | ATTR_ATIME \
  83         | ATTR_MTIME \
  84         | ATTR_CTIME \
  85         | ATTR_ATIME_SET \
  86         | ATTR_MTIME_SET)
  87
  88 struct nfs4_opendata;
  89 static int _nfs4_proc_open(struct nfs4_opendata *data);
  90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs4_state *state, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 101                 struct rpc_cred *);
 102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
 103                 struct rpc_cred *);
 104 #endif
 105
 106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 107 static inline struct nfs4_label *
 108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 109         struct iattr *sattr, struct nfs4_label *label)
 110 {
 111         int err;
 112
 113         if (label == NULL)
 114                 return NULL;
 115
 116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 117                 return NULL;
 118
 119         err = security_dentry_init_security(dentry, sattr->ia_mode,
 120                                 &dentry->d_name, (void **)&label->label, &label->len);
 121         if (err == 0)
 122                 return label;
 123
 124         return NULL;
 125 }
 126 static inline void
 127 nfs4_label_release_security(struct nfs4_label *label)
 128 {
 129         if (label)
 130                 security_release_secctx(label->label, label->len);
 131 }
 132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 133 {
 134         if (label)
 135                 return server->attr_bitmask;
 136
 137         return server->attr_bitmask_nl;
 138 }
 139 #else
 140 static inline struct nfs4_label *
 141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 142         struct iattr *sattr, struct nfs4_label *l)
 143 { return NULL; }
 144 static inline void
 145 nfs4_label_release_security(struct nfs4_label *label)
 146 { return; }
 147 static inline u32 *
 148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 149 { return server->attr_bitmask; }
 150 #endif
 151
 152 /* Prevent leaks of NFSv4 errors into userland */
 153 static int nfs4_map_errors(int err)
 154 {
 155         if (err >= -1000)
 156                 return err;
 157         switch (err) {
 158         case -NFS4ERR_RESOURCE:
 159         case -NFS4ERR_LAYOUTTRYLATER:
 160         case -NFS4ERR_RECALLCONFLICT:
 161                 return -EREMOTEIO;
 162         case -NFS4ERR_WRONGSEC:
 163         case -NFS4ERR_WRONG_CRED:
 164                 return -EPERM;
 165         case -NFS4ERR_BADOWNER:
 166         case -NFS4ERR_BADNAME:
 167                 return -EINVAL;
 168         case -NFS4ERR_SHARE_DENIED:
 169                 return -EACCES;
 170         case -NFS4ERR_MINOR_VERS_MISMATCH:
 171                 return -EPROTONOSUPPORT;
 172         case -NFS4ERR_FILE_OPEN:
 173                 return -EBUSY;
 174         default:
 175                 dprintk("%s could not handle NFSv4 error %d\n",
 176                                 __func__, -err);
 177                 break;
 178         }
 179         return -EIO;
 180 }
 181
 182 /*
 183  * This is our standard bitmap for GETATTR requests.
 184  */
 185 const u32 nfs4_fattr_bitmap[3] = {
 186         FATTR4_WORD0_TYPE
 187         | FATTR4_WORD0_CHANGE
 188         | FATTR4_WORD0_SIZE
 189         | FATTR4_WORD0_FSID
 190         | FATTR4_WORD0_FILEID,
 191         FATTR4_WORD1_MODE
 192         | FATTR4_WORD1_NUMLINKS
 193         | FATTR4_WORD1_OWNER
 194         | FATTR4_WORD1_OWNER_GROUP
 195         | FATTR4_WORD1_RAWDEV
 196         | FATTR4_WORD1_SPACE_USED
 197         | FATTR4_WORD1_TIME_ACCESS
 198         | FATTR4_WORD1_TIME_METADATA
 199         | FATTR4_WORD1_TIME_MODIFY
 200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 202         FATTR4_WORD2_SECURITY_LABEL
 203 #endif
 204 };
 205
 206 static const u32 nfs4_pnfs_open_bitmap[3] = {
 207         FATTR4_WORD0_TYPE
 208         | FATTR4_WORD0_CHANGE
 209         | FATTR4_WORD0_SIZE
 210         | FATTR4_WORD0_FSID
 211         | FATTR4_WORD0_FILEID,
 212         FATTR4_WORD1_MODE
 213         | FATTR4_WORD1_NUMLINKS
 214         | FATTR4_WORD1_OWNER
 215         | FATTR4_WORD1_OWNER_GROUP
 216         | FATTR4_WORD1_RAWDEV
 217         | FATTR4_WORD1_SPACE_USED
 218         | FATTR4_WORD1_TIME_ACCESS
 219         | FATTR4_WORD1_TIME_METADATA
 220         | FATTR4_WORD1_TIME_MODIFY,
 221         FATTR4_WORD2_MDSTHRESHOLD
 222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 223         | FATTR4_WORD2_SECURITY_LABEL
 224 #endif
 225 };
 226
 227 static const u32 nfs4_open_noattr_bitmap[3] = {
 228         FATTR4_WORD0_TYPE
 229         | FATTR4_WORD0_CHANGE
 230         | FATTR4_WORD0_FILEID,
 231 };
 232
 233 const u32 nfs4_statfs_bitmap[3] = {
 234         FATTR4_WORD0_FILES_AVAIL
 235         | FATTR4_WORD0_FILES_FREE
 236         | FATTR4_WORD0_FILES_TOTAL,
 237         FATTR4_WORD1_SPACE_AVAIL
 238         | FATTR4_WORD1_SPACE_FREE
 239         | FATTR4_WORD1_SPACE_TOTAL
 240 };
 241
 242 const u32 nfs4_pathconf_bitmap[3] = {
 243         FATTR4_WORD0_MAXLINK
 244         | FATTR4_WORD0_MAXNAME,
 245         0
 246 };
 247
 248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 249                         | FATTR4_WORD0_MAXREAD
 250                         | FATTR4_WORD0_MAXWRITE
 251                         | FATTR4_WORD0_LEASE_TIME,
 252                         FATTR4_WORD1_TIME_DELTA
 253                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 254                         FATTR4_WORD2_LAYOUT_BLKSIZE
 255                         | FATTR4_WORD2_CLONE_BLKSIZE
 256 };
 257
 258 const u32 nfs4_fs_locations_bitmap[3] = {
 259         FATTR4_WORD0_TYPE
 260         | FATTR4_WORD0_CHANGE
 261         | FATTR4_WORD0_SIZE
 262         | FATTR4_WORD0_FSID
 263         | FATTR4_WORD0_FILEID
 264         | FATTR4_WORD0_FS_LOCATIONS,
 265         FATTR4_WORD1_MODE
 266         | FATTR4_WORD1_NUMLINKS
 267         | FATTR4_WORD1_OWNER
 268         | FATTR4_WORD1_OWNER_GROUP
 269         | FATTR4_WORD1_RAWDEV
 270         | FATTR4_WORD1_SPACE_USED
 271         | FATTR4_WORD1_TIME_ACCESS
 272         | FATTR4_WORD1_TIME_METADATA
 273         | FATTR4_WORD1_TIME_MODIFY
 274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 275 };
 276
 277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 278                 struct nfs4_readdir_arg *readdir)
 279 {
 280         __be32 *start, *p;
 281
 282         if (cookie > 2) {
 283                 readdir->cookie = cookie;
 284                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 285                 return;
 286         }
 287
 288         readdir->cookie = 0;
 289         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 290         if (cookie == 2)
 291                 return;
 292
 293         /*
 294          * NFSv4 servers do not return entries for '.' and '..'
 295          * Therefore, we fake these entries here.  We let '.'
 296          * have cookie 0 and '..' have cookie 1.  Note that
 297          * when talking to the server, we always send cookie 0
 298          * instead of 1 or 2.
 299          */
 300         start = p = kmap_atomic(*readdir->pages);
 301
 302         if (cookie == 0) {
 303                 *p++ = xdr_one;                                  /* next */
 304                 *p++ = xdr_zero;                   /* cookie, first word */
 305                 *p++ = xdr_one;                   /* cookie, second word */
 306                 *p++ = xdr_one;                             /* entry len */
 307                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 308                 p++;
 309                 *p++ = xdr_one;                         /* bitmap length */
 310                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 311                 *p++ = htonl(8);              /* attribute buffer length */
 312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 313         }
 314
 315         *p++ = xdr_one;                                  /* next */
 316         *p++ = xdr_zero;                   /* cookie, first word */
 317         *p++ = xdr_two;                   /* cookie, second word */
 318         *p++ = xdr_two;                             /* entry len */
 319         memcpy(p, "..\0\0", 4);                         /* entry */
 320         p++;
 321         *p++ = xdr_one;                         /* bitmap length */
 322         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 323         *p++ = htonl(8);              /* attribute buffer length */
 324         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 325
 326         readdir->pgbase = (char *)p - (char *)start;
 327         readdir->count -= readdir->pgbase;
 328         kunmap_atomic(start);
 329 }
 330
 331 static long nfs4_update_delay(long *timeout)
 332 {
 333         long ret;
 334         if (!timeout)
 335                 return NFS4_POLL_RETRY_MAX;
 336         if (*timeout <= 0)
 337                 *timeout = NFS4_POLL_RETRY_MIN;
 338         if (*timeout > NFS4_POLL_RETRY_MAX)
 339                 *timeout = NFS4_POLL_RETRY_MAX;
 340         ret = *timeout;
 341         *timeout <<= 1;
 342         return ret;
 343 }
 344
 345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 346 {
 347         int res = 0;
 348
 349         might_sleep();
 350
 351         freezable_schedule_timeout_killable_unsafe(
 352                 nfs4_update_delay(timeout));
 353         if (fatal_signal_pending(current))
 354                 res = -ERESTARTSYS;
 355         return res;
 356 }
 357
 358 /* This is the error handling routine for processes that are allowed
 359  * to sleep.
 360  */
 361 static int nfs4_do_handle_exception(struct nfs_server *server,
 362                 int errorcode, struct nfs4_exception *exception)
 363 {
 364         struct nfs_client *clp = server->nfs_client;
 365         struct nfs4_state *state = exception->state;
 366         const nfs4_stateid *stateid = exception->stateid;
 367         struct inode *inode = exception->inode;
 368         int ret = errorcode;
 369
 370         exception->delay = 0;
 371         exception->recovering = 0;
 372         exception->retry = 0;
 373         switch(errorcode) {
 374                 case 0:
 375                         return 0;
 376                 case -NFS4ERR_OPENMODE:
 377                 case -NFS4ERR_DELEG_REVOKED:
 378                 case -NFS4ERR_ADMIN_REVOKED:
 379                 case -NFS4ERR_BAD_STATEID:
 380                         if (inode) {
 381                                 int err;
 382
 383                                 err = nfs_async_inode_return_delegation(inode,
 384                                                 stateid);
 385                                 if (err == 0)
 386                                         goto wait_on_recovery;
 387                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 388                                         exception->retry = 1;
 389                                         break;
 390                                 }
 391                         }
 392                         if (state == NULL)
 393                                 break;
 394                         ret = nfs4_schedule_stateid_recovery(server, state);
 395                         if (ret < 0)
 396                                 break;
 397                         goto wait_on_recovery;
 398                 case -NFS4ERR_EXPIRED:
 399                         if (state != NULL) {
 400                                 ret = nfs4_schedule_stateid_recovery(server, state);
 401                                 if (ret < 0)
 402                                         break;
 403                         }
 404                 case -NFS4ERR_STALE_STATEID:
 405                 case -NFS4ERR_STALE_CLIENTID:
 406                         nfs4_schedule_lease_recovery(clp);
 407                         goto wait_on_recovery;
 408                 case -NFS4ERR_MOVED:
 409                         ret = nfs4_schedule_migration_recovery(server);
 410                         if (ret < 0)
 411                                 break;
 412                         goto wait_on_recovery;
 413                 case -NFS4ERR_LEASE_MOVED:
 414                         nfs4_schedule_lease_moved_recovery(clp);
 415                         goto wait_on_recovery;
 416 #if defined(CONFIG_NFS_V4_1)
 417                 case -NFS4ERR_BADSESSION:
 418                 case -NFS4ERR_BADSLOT:
 419                 case -NFS4ERR_BAD_HIGH_SLOT:
 420                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 421                 case -NFS4ERR_DEADSESSION:
 422                 case -NFS4ERR_SEQ_FALSE_RETRY:
 423                 case -NFS4ERR_SEQ_MISORDERED:
 424                         dprintk("%s ERROR: %d Reset session\n", __func__,
 425                                 errorcode);
 426                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 427                         goto wait_on_recovery;
 428 #endif /* defined(CONFIG_NFS_V4_1) */
 429                 case -NFS4ERR_FILE_OPEN:
 430                         if (exception->timeout > HZ) {
 431                                 /* We have retried a decent amount, time to
 432                                  * fail
 433                                  */
 434                                 ret = -EBUSY;
 435                                 break;
 436                         }
 437                 case -NFS4ERR_DELAY:
 438                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 439                 case -NFS4ERR_GRACE:
 440                 case -NFS4ERR_LAYOUTTRYLATER:
 441                 case -NFS4ERR_RECALLCONFLICT:
 442                         exception->delay = 1;
 443                         return 0;
 444
 445                 case -NFS4ERR_RETRY_UNCACHED_REP:
 446                 case -NFS4ERR_OLD_STATEID:
 447                         exception->retry = 1;
 448                         break;
 449                 case -NFS4ERR_BADOWNER:
 450                         /* The following works around a Linux server bug! */
 451                 case -NFS4ERR_BADNAME:
 452                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 453                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 454                                 exception->retry = 1;
 455                                 printk(KERN_WARNING "NFS: v4 server %s "
 456                                                 "does not accept raw "
 457                                                 "uid/gids. "
 458                                                 "Reenabling the idmapper.\n",
 459                                                 server->nfs_client->cl_hostname);
 460                         }
 461         }
 462         /* We failed to handle the error */
 463         return nfs4_map_errors(ret);
 464 wait_on_recovery:
 465         exception->recovering = 1;
 466         return 0;
 467 }
 468
 469 /* This is the error handling routine for processes that are allowed
 470  * to sleep.
 471  */
 472 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 473 {
 474         struct nfs_client *clp = server->nfs_client;
 475         int ret;
 476
 477         ret = nfs4_do_handle_exception(server, errorcode, exception);
 478         if (exception->delay) {
 479                 ret = nfs4_delay(server->client, &exception->timeout);
 480                 goto out_retry;
 481         }
 482         if (exception->recovering) {
 483                 ret = nfs4_wait_clnt_recover(clp);
 484                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 485                         return -EIO;
 486                 goto out_retry;
 487         }
 488         return ret;
 489 out_retry:
 490         if (ret == 0)
 491                 exception->retry = 1;
 492         return ret;
 493 }
 494
 495 static int
 496 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 497                 int errorcode, struct nfs4_exception *exception)
 498 {
 499         struct nfs_client *clp = server->nfs_client;
 500         int ret;
 501
 502         ret = nfs4_do_handle_exception(server, errorcode, exception);
 503         if (exception->delay) {
 504                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 505                 goto out_retry;
 506         }
 507         if (exception->recovering) {
 508                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 509                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 510                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 511                 goto out_retry;
 512         }
 513         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 514                 ret = -EIO;
 515         return ret;
 516 out_retry:
 517         if (ret == 0)
 518                 exception->retry = 1;
 519         return ret;
 520 }
 521
 522 static int
 523 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 524                         struct nfs4_state *state, long *timeout)
 525 {
 526         struct nfs4_exception exception = {
 527                 .state = state,
 528         };
 529
 530         if (task->tk_status >= 0)
 531                 return 0;
 532         if (timeout)
 533                 exception.timeout = *timeout;
 534         task->tk_status = nfs4_async_handle_exception(task, server,
 535                         task->tk_status,
 536                         &exception);
 537         if (exception.delay && timeout)
 538                 *timeout = exception.timeout;
 539         if (exception.retry)
 540                 return -EAGAIN;
 541         return 0;
 542 }
 543
 544 /*
 545  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 546  * or 'false' otherwise.
 547  */
 548 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 549 {
 550         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 551
 552         if (flavor == RPC_AUTH_GSS_KRB5I ||
 553             flavor == RPC_AUTH_GSS_KRB5P)
 554                 return true;
 555
 556         return false;
 557 }
 558
 559 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 560 {
 561         spin_lock(&clp->cl_lock);
 562         if (time_before(clp->cl_last_renewal,timestamp))
 563                 clp->cl_last_renewal = timestamp;
 564         spin_unlock(&clp->cl_lock);
 565 }
 566
 567 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 568 {
 569         struct nfs_client *clp = server->nfs_client;
 570
 571         if (!nfs4_has_session(clp))
 572                 do_renew_lease(clp, timestamp);
 573 }
 574
 575 struct nfs4_call_sync_data {
 576         const struct nfs_server *seq_server;
 577         struct nfs4_sequence_args *seq_args;
 578         struct nfs4_sequence_res *seq_res;
 579 };
 580
 581 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 582                         struct nfs4_sequence_res *res, int cache_reply)
 583 {
 584         args->sa_slot = NULL;
 585         args->sa_cache_this = cache_reply;
 586         args->sa_privileged = 0;
 587
 588         res->sr_slot = NULL;
 589 }
 590
 591 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 592 {
 593         args->sa_privileged = 1;
 594 }
 595
 596 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 597                          struct nfs4_sequence_args *args,
 598                          struct nfs4_sequence_res *res,
 599                          struct rpc_task *task)
 600 {
 601         struct nfs4_slot *slot;
 602
 603         /* slot already allocated? */
 604         if (res->sr_slot != NULL)
 605                 goto out_start;
 606
 607         spin_lock(&tbl->slot_tbl_lock);
 608         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 609                 goto out_sleep;
 610
 611         slot = nfs4_alloc_slot(tbl);
 612         if (IS_ERR(slot)) {
 613                 if (slot == ERR_PTR(-ENOMEM))
 614                         task->tk_timeout = HZ >> 2;
 615                 goto out_sleep;
 616         }
 617         spin_unlock(&tbl->slot_tbl_lock);
 618
 619         slot->privileged = args->sa_privileged ? 1 : 0;
 620         args->sa_slot = slot;
 621         res->sr_slot = slot;
 622
 623 out_start:
 624         rpc_call_start(task);
 625         return 0;
 626
 627 out_sleep:
 628         if (args->sa_privileged)
 629                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 630                                 NULL, RPC_PRIORITY_PRIVILEGED);
 631         else
 632                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 633         spin_unlock(&tbl->slot_tbl_lock);
 634         return -EAGAIN;
 635 }
 636 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 637
 638 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 639 {
 640         struct nfs4_slot *slot = res->sr_slot;
 641         struct nfs4_slot_table *tbl;
 642
 643         tbl = slot->table;
 644         spin_lock(&tbl->slot_tbl_lock);
 645         if (!nfs41_wake_and_assign_slot(tbl, slot))
 646                 nfs4_free_slot(tbl, slot);
 647         spin_unlock(&tbl->slot_tbl_lock);
 648
 649         res->sr_slot = NULL;
 650 }
 651
 652 static int nfs40_sequence_done(struct rpc_task *task,
 653                                struct nfs4_sequence_res *res)
 654 {
 655         if (res->sr_slot != NULL)
 656                 nfs40_sequence_free_slot(res);
 657         return 1;
 658 }
 659
 660 #if defined(CONFIG_NFS_V4_1)
 661
 662 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 663 {
 664         struct nfs4_session *session;
 665         struct nfs4_slot_table *tbl;
 666         struct nfs4_slot *slot = res->sr_slot;
 667         bool send_new_highest_used_slotid = false;
 668
 669         tbl = slot->table;
 670         session = tbl->session;
 671
 672         /* Bump the slot sequence number */
 673         if (slot->seq_done)
 674                 slot->seq_nr++;
 675         slot->seq_done = 0;
 676
 677         spin_lock(&tbl->slot_tbl_lock);
 678         /* Be nice to the server: try to ensure that the last transmitted
 679          * value for highest_user_slotid <= target_highest_slotid
 680          */
 681         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 682                 send_new_highest_used_slotid = true;
 683
 684         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 685                 send_new_highest_used_slotid = false;
 686                 goto out_unlock;
 687         }
 688         nfs4_free_slot(tbl, slot);
 689
 690         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 691                 send_new_highest_used_slotid = false;
 692 out_unlock:
 693         spin_unlock(&tbl->slot_tbl_lock);
 694         res->sr_slot = NULL;
 695         if (send_new_highest_used_slotid)
 696                 nfs41_notify_server(session->clp);
 697         if (waitqueue_active(&tbl->slot_waitq))
 698                 wake_up_all(&tbl->slot_waitq);
 699 }
 700
 701 static int nfs41_sequence_process(struct rpc_task *task,
 702                 struct nfs4_sequence_res *res)
 703 {
 704         struct nfs4_session *session;
 705         struct nfs4_slot *slot = res->sr_slot;
 706         struct nfs_client *clp;
 707         bool interrupted = false;
 708         int ret = 1;
 709
 710         if (slot == NULL)
 711                 goto out_noaction;
 712         /* don't increment the sequence number if the task wasn't sent */
 713         if (!RPC_WAS_SENT(task))
 714                 goto out;
 715
 716         session = slot->table->session;
 717
 718         if (slot->interrupted) {
 719                 slot->interrupted = 0;
 720                 interrupted = true;
 721         }
 722
 723         trace_nfs4_sequence_done(session, res);
 724         /* Check the SEQUENCE operation status */
 725         switch (res->sr_status) {
 726         case 0:
 727                 /* Update the slot's sequence and clientid lease timer */
 728                 slot->seq_done = 1;
 729                 clp = session->clp;
 730                 do_renew_lease(clp, res->sr_timestamp);
 731                 /* Check sequence flags */
 732                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 733                                 !!slot->privileged);
 734                 nfs41_update_target_slotid(slot->table, slot, res);
 735                 break;
 736         case 1:
 737                 /*
 738                  * sr_status remains 1 if an RPC level error occurred.
 739                  * The server may or may not have processed the sequence
 740                  * operation..
 741                  * Mark the slot as having hosted an interrupted RPC call.
 742                  */
 743                 slot->interrupted = 1;
 744                 goto out;
 745         case -NFS4ERR_DELAY:
 746                 /* The server detected a resend of the RPC call and
 747                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 748                  * of RFC5661.
 749                  */
 750                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 751                         __func__,
 752                         slot->slot_nr,
 753                         slot->seq_nr);
 754                 goto out_retry;
 755         case -NFS4ERR_BADSLOT:
 756                 /*
 757                  * The slot id we used was probably retired. Try again
 758                  * using a different slot id.
 759                  */
 760                 goto retry_nowait;
 761         case -NFS4ERR_SEQ_MISORDERED:
 762                 /*
 763                  * Was the last operation on this sequence interrupted?
 764                  * If so, retry after bumping the sequence number.
 765                  */
 766                 if (interrupted) {
 767                         ++slot->seq_nr;
 768                         goto retry_nowait;
 769                 }
 770                 /*
 771                  * Could this slot have been previously retired?
 772                  * If so, then the server may be expecting seq_nr = 1!
 773                  */
 774                 if (slot->seq_nr != 1) {
 775                         slot->seq_nr = 1;
 776                         goto retry_nowait;
 777                 }
 778                 break;
 779         case -NFS4ERR_SEQ_FALSE_RETRY:
 780                 ++slot->seq_nr;
 781                 goto retry_nowait;
 782         default:
 783                 /* Just update the slot sequence no. */
 784                 slot->seq_done = 1;
 785         }
 786 out:
 787         /* The session may be reset by one of the error handlers. */
 788         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 789 out_noaction:
 790         return ret;
 791 retry_nowait:
 792         if (rpc_restart_call_prepare(task)) {
 793                 nfs41_sequence_free_slot(res);
 794                 task->tk_status = 0;
 795                 ret = 0;
 796         }
 797         goto out;
 798 out_retry:
 799         if (!rpc_restart_call(task))
 800                 goto out;
 801         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 802         return 0;
 803 }
 804
 805 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 806 {
 807         if (!nfs41_sequence_process(task, res))
 808                 return 0;
 809         if (res->sr_slot != NULL)
 810                 nfs41_sequence_free_slot(res);
 811         return 1;
 812
 813 }
 814 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 815
 816 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 817 {
 818         if (res->sr_slot == NULL)
 819                 return 1;
 820         if (res->sr_slot->table->session != NULL)
 821                 return nfs41_sequence_process(task, res);
 822         return nfs40_sequence_done(task, res);
 823 }
 824
 825 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 826 {
 827         if (res->sr_slot != NULL) {
 828                 if (res->sr_slot->table->session != NULL)
 829                         nfs41_sequence_free_slot(res);
 830                 else
 831                         nfs40_sequence_free_slot(res);
 832         }
 833 }
 834
 835 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 836 {
 837         if (res->sr_slot == NULL)
 838                 return 1;
 839         if (!res->sr_slot->table->session)
 840                 return nfs40_sequence_done(task, res);
 841         return nfs41_sequence_done(task, res);
 842 }
 843 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 844
 845 int nfs41_setup_sequence(struct nfs4_session *session,
 846                                 struct nfs4_sequence_args *args,
 847                                 struct nfs4_sequence_res *res,
 848                                 struct rpc_task *task)
 849 {
 850         struct nfs4_slot *slot;
 851         struct nfs4_slot_table *tbl;
 852
 853         dprintk("--> %s\n", __func__);
 854         /* slot already allocated? */
 855         if (res->sr_slot != NULL)
 856                 goto out_success;
 857
 858         tbl = &session->fc_slot_table;
 859
 860         task->tk_timeout = 0;
 861
 862         spin_lock(&tbl->slot_tbl_lock);
 863         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 864             !args->sa_privileged) {
 865                 /* The state manager will wait until the slot table is empty */
 866                 dprintk("%s session is draining\n", __func__);
 867                 goto out_sleep;
 868         }
 869
 870         slot = nfs4_alloc_slot(tbl);
 871         if (IS_ERR(slot)) {
 872                 /* If out of memory, try again in 1/4 second */
 873                 if (slot == ERR_PTR(-ENOMEM))
 874                         task->tk_timeout = HZ >> 2;
 875                 dprintk("<-- %s: no free slots\n", __func__);
 876                 goto out_sleep;
 877         }
 878         spin_unlock(&tbl->slot_tbl_lock);
 879
 880         slot->privileged = args->sa_privileged ? 1 : 0;
 881         args->sa_slot = slot;
 882
 883         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 884                         slot->slot_nr, slot->seq_nr);
 885
 886         res->sr_slot = slot;
 887         res->sr_timestamp = jiffies;
 888         res->sr_status_flags = 0;
 889         /*
 890          * sr_status is only set in decode_sequence, and so will remain
 891          * set to 1 if an rpc level failure occurs.
 892          */
 893         res->sr_status = 1;
 894         trace_nfs4_setup_sequence(session, args);
 895 out_success:
 896         rpc_call_start(task);
 897         return 0;
 898 out_sleep:
 899         /* Privileged tasks are queued with top priority */
 900         if (args->sa_privileged)
 901                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 902                                 NULL, RPC_PRIORITY_PRIVILEGED);
 903         else
 904                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 905         spin_unlock(&tbl->slot_tbl_lock);
 906         return -EAGAIN;
 907 }
 908 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 909
 910 static int nfs4_setup_sequence(const struct nfs_server *server,
 911                                struct nfs4_sequence_args *args,
 912                                struct nfs4_sequence_res *res,
 913                                struct rpc_task *task)
 914 {
 915         struct nfs4_session *session = nfs4_get_session(server);
 916         int ret = 0;
 917
 918         if (!session)
 919                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 920                                             args, res, task);
 921
 922         dprintk("--> %s clp %p session %p sr_slot %u\n",
 923                 __func__, session->clp, session, res->sr_slot ?
 924                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 925
 926         ret = nfs41_setup_sequence(session, args, res, task);
 927
 928         dprintk("<-- %s status=%d\n", __func__, ret);
 929         return ret;
 930 }
 931
 932 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 933 {
 934         struct nfs4_call_sync_data *data = calldata;
 935         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 936
 937         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 938
 939         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 940 }
 941
 942 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 943 {
 944         struct nfs4_call_sync_data *data = calldata;
 945
 946         nfs41_sequence_done(task, data->seq_res);
 947 }
 948
 949 static const struct rpc_call_ops nfs41_call_sync_ops = {
 950         .rpc_call_prepare = nfs41_call_sync_prepare,
 951         .rpc_call_done = nfs41_call_sync_done,
 952 };
 953
 954 #else   /* !CONFIG_NFS_V4_1 */
 955
 956 static int nfs4_setup_sequence(const struct nfs_server *server,
 957                                struct nfs4_sequence_args *args,
 958                                struct nfs4_sequence_res *res,
 959                                struct rpc_task *task)
 960 {
 961         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 962                                     args, res, task);
 963 }
 964
 965 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 966 {
 967         return nfs40_sequence_done(task, res);
 968 }
 969
 970 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 971 {
 972         if (res->sr_slot != NULL)
 973                 nfs40_sequence_free_slot(res);
 974 }
 975
 976 int nfs4_sequence_done(struct rpc_task *task,
 977                        struct nfs4_sequence_res *res)
 978 {
 979         return nfs40_sequence_done(task, res);
 980 }
 981 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 982
 983 #endif  /* !CONFIG_NFS_V4_1 */
 984
 985 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 986 {
 987         struct nfs4_call_sync_data *data = calldata;
 988         nfs4_setup_sequence(data->seq_server,
 989                                 data->seq_args, data->seq_res, task);
 990 }
 991
 992 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 993 {
 994         struct nfs4_call_sync_data *data = calldata;
 995         nfs4_sequence_done(task, data->seq_res);
 996 }
 997
 998 static const struct rpc_call_ops nfs40_call_sync_ops = {
 999         .rpc_call_prepare = nfs40_call_sync_prepare,
1000         .rpc_call_done = nfs40_call_sync_done,
1001 };
1002
1003 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1004                                    struct nfs_server *server,
1005                                    struct rpc_message *msg,
1006                                    struct nfs4_sequence_args *args,
1007                                    struct nfs4_sequence_res *res)
1008 {
1009         int ret;
1010         struct rpc_task *task;
1011         struct nfs_client *clp = server->nfs_client;
1012         struct nfs4_call_sync_data data = {
1013                 .seq_server = server,
1014                 .seq_args = args,
1015                 .seq_res = res,
1016         };
1017         struct rpc_task_setup task_setup = {
1018                 .rpc_client = clnt,
1019                 .rpc_message = msg,
1020                 .callback_ops = clp->cl_mvops->call_sync_ops,
1021                 .callback_data = &data
1022         };
1023
1024         task = rpc_run_task(&task_setup);
1025         if (IS_ERR(task))
1026                 ret = PTR_ERR(task);
1027         else {
1028                 ret = task->tk_status;
1029                 rpc_put_task(task);
1030         }
1031         return ret;
1032 }
1033
1034 int nfs4_call_sync(struct rpc_clnt *clnt,
1035                    struct nfs_server *server,
1036                    struct rpc_message *msg,
1037                    struct nfs4_sequence_args *args,
1038                    struct nfs4_sequence_res *res,
1039                    int cache_reply)
1040 {
1041         nfs4_init_sequence(args, res, cache_reply);
1042         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1043 }
1044
1045 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1046 {
1047         struct nfs_inode *nfsi = NFS_I(dir);
1048
1049         spin_lock(&dir->i_lock);
1050         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1051         if (!cinfo->atomic || cinfo->before != dir->i_version)
1052                 nfs_force_lookup_revalidate(dir);
1053         dir->i_version = cinfo->after;
1054         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1055         nfs_fscache_invalidate(dir);
1056         spin_unlock(&dir->i_lock);
1057 }
1058
1059 struct nfs4_opendata {
1060         struct kref kref;
1061         struct nfs_openargs o_arg;
1062         struct nfs_openres o_res;
1063         struct nfs_open_confirmargs c_arg;
1064         struct nfs_open_confirmres c_res;
1065         struct nfs4_string owner_name;
1066         struct nfs4_string group_name;
1067         struct nfs4_label *a_label;
1068         struct nfs_fattr f_attr;
1069         struct nfs4_label *f_label;
1070         struct dentry *dir;
1071         struct dentry *dentry;
1072         struct nfs4_state_owner *owner;
1073         struct nfs4_state *state;
1074         struct iattr attrs;
1075         unsigned long timestamp;
1076         unsigned int rpc_done : 1;
1077         unsigned int file_created : 1;
1078         unsigned int is_recover : 1;
1079         int rpc_status;
1080         int cancelled;
1081 };
1082
1083 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1084                 int err, struct nfs4_exception *exception)
1085 {
1086         if (err != -EINVAL)
1087                 return false;
1088         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1089                 return false;
1090         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1091         exception->retry = 1;
1092         return true;
1093 }
1094
1095 static u32
1096 nfs4_map_atomic_open_share(struct nfs_server *server,
1097                 fmode_t fmode, int openflags)
1098 {
1099         u32 res = 0;
1100
1101         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1102         case FMODE_READ:
1103                 res = NFS4_SHARE_ACCESS_READ;
1104                 break;
1105         case FMODE_WRITE:
1106                 res = NFS4_SHARE_ACCESS_WRITE;
1107                 break;
1108         case FMODE_READ|FMODE_WRITE:
1109                 res = NFS4_SHARE_ACCESS_BOTH;
1110         }
1111         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1112                 goto out;
1113         /* Want no delegation if we're using O_DIRECT */
1114         if (openflags & O_DIRECT)
1115                 res |= NFS4_SHARE_WANT_NO_DELEG;
1116 out:
1117         return res;
1118 }
1119
1120 static enum open_claim_type4
1121 nfs4_map_atomic_open_claim(struct nfs_server *server,
1122                 enum open_claim_type4 claim)
1123 {
1124         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1125                 return claim;
1126         switch (claim) {
1127         default:
1128                 return claim;
1129         case NFS4_OPEN_CLAIM_FH:
1130                 return NFS4_OPEN_CLAIM_NULL;
1131         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1132                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1133         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1134                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1135         }
1136 }
1137
1138 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1139 {
1140         p->o_res.f_attr = &p->f_attr;
1141         p->o_res.f_label = p->f_label;
1142         p->o_res.seqid = p->o_arg.seqid;
1143         p->c_res.seqid = p->c_arg.seqid;
1144         p->o_res.server = p->o_arg.server;
1145         p->o_res.access_request = p->o_arg.access;
1146         nfs_fattr_init(&p->f_attr);
1147         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1148 }
1149
1150 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1151                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1152                 const struct iattr *attrs,
1153                 struct nfs4_label *label,
1154                 enum open_claim_type4 claim,
1155                 gfp_t gfp_mask)
1156 {
1157         struct dentry *parent = dget_parent(dentry);
1158         struct inode *dir = d_inode(parent);
1159         struct nfs_server *server = NFS_SERVER(dir);
1160         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1161         struct nfs4_opendata *p;
1162
1163         p = kzalloc(sizeof(*p), gfp_mask);
1164         if (p == NULL)
1165                 goto err;
1166
1167         p->f_label = nfs4_label_alloc(server, gfp_mask);
1168         if (IS_ERR(p->f_label))
1169                 goto err_free_p;
1170
1171         p->a_label = nfs4_label_alloc(server, gfp_mask);
1172         if (IS_ERR(p->a_label))
1173                 goto err_free_f;
1174
1175         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1176         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1177         if (IS_ERR(p->o_arg.seqid))
1178                 goto err_free_label;
1179         nfs_sb_active(dentry->d_sb);
1180         p->dentry = dget(dentry);
1181         p->dir = parent;
1182         p->owner = sp;
1183         atomic_inc(&sp->so_count);
1184         p->o_arg.open_flags = flags;
1185         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1186         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1187                         fmode, flags);
1188         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1189          * will return permission denied for all bits until close */
1190         if (!(flags & O_EXCL)) {
1191                 /* ask server to check for all possible rights as results
1192                  * are cached */
1193                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1194                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1195         }
1196         p->o_arg.clientid = server->nfs_client->cl_clientid;
1197         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1198         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1199         p->o_arg.name = &dentry->d_name;
1200         p->o_arg.server = server;
1201         p->o_arg.bitmask = nfs4_bitmask(server, label);
1202         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1203         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1204         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1205         switch (p->o_arg.claim) {
1206         case NFS4_OPEN_CLAIM_NULL:
1207         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1208         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1209                 p->o_arg.fh = NFS_FH(dir);
1210                 break;
1211         case NFS4_OPEN_CLAIM_PREVIOUS:
1212         case NFS4_OPEN_CLAIM_FH:
1213         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1214         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1215                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1216         }
1217         if (attrs != NULL && attrs->ia_valid != 0) {
1218                 __u32 verf[2];
1219
1220                 p->o_arg.u.attrs = &p->attrs;
1221                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1222
1223                 verf[0] = jiffies;
1224                 verf[1] = current->pid;
1225                 memcpy(p->o_arg.u.verifier.data, verf,
1226                                 sizeof(p->o_arg.u.verifier.data));
1227         }
1228         p->c_arg.fh = &p->o_res.fh;
1229         p->c_arg.stateid = &p->o_res.stateid;
1230         p->c_arg.seqid = p->o_arg.seqid;
1231         nfs4_init_opendata_res(p);
1232         kref_init(&p->kref);
1233         return p;
1234
1235 err_free_label:
1236         nfs4_label_free(p->a_label);
1237 err_free_f:
1238         nfs4_label_free(p->f_label);
1239 err_free_p:
1240         kfree(p);
1241 err:
1242         dput(parent);
1243         return NULL;
1244 }
1245
1246 static void nfs4_opendata_free(struct kref *kref)
1247 {
1248         struct nfs4_opendata *p = container_of(kref,
1249                         struct nfs4_opendata, kref);
1250         struct super_block *sb = p->dentry->d_sb;
1251
1252         nfs_free_seqid(p->o_arg.seqid);
1253         nfs4_sequence_free_slot(&p->o_res.seq_res);
1254         if (p->state != NULL)
1255                 nfs4_put_open_state(p->state);
1256         nfs4_put_state_owner(p->owner);
1257
1258         nfs4_label_free(p->a_label);
1259         nfs4_label_free(p->f_label);
1260
1261         dput(p->dir);
1262         dput(p->dentry);
1263         nfs_sb_deactive(sb);
1264         nfs_fattr_free_names(&p->f_attr);
1265         kfree(p->f_attr.mdsthreshold);
1266         kfree(p);
1267 }
1268
1269 static void nfs4_opendata_put(struct nfs4_opendata *p)
1270 {
1271         if (p != NULL)
1272                 kref_put(&p->kref, nfs4_opendata_free);
1273 }
1274
1275 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1276 {
1277         int ret;
1278
1279         ret = rpc_wait_for_completion_task(task);
1280         return ret;
1281 }
1282
1283 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1284                 fmode_t fmode)
1285 {
1286         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1287         case FMODE_READ|FMODE_WRITE:
1288                 return state->n_rdwr != 0;
1289         case FMODE_WRITE:
1290                 return state->n_wronly != 0;
1291         case FMODE_READ:
1292                 return state->n_rdonly != 0;
1293         }
1294         WARN_ON_ONCE(1);
1295         return false;
1296 }
1297
1298 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1299 {
1300         int ret = 0;
1301
1302         if (open_mode & (O_EXCL|O_TRUNC))
1303                 goto out;
1304         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1305                 case FMODE_READ:
1306                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1307                                 && state->n_rdonly != 0;
1308                         break;
1309                 case FMODE_WRITE:
1310                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1311                                 && state->n_wronly != 0;
1312                         break;
1313                 case FMODE_READ|FMODE_WRITE:
1314                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1315                                 && state->n_rdwr != 0;
1316         }
1317 out:
1318         return ret;
1319 }
1320
1321 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1322                 enum open_claim_type4 claim)
1323 {
1324         if (delegation == NULL)
1325                 return 0;
1326         if ((delegation->type & fmode) != fmode)
1327                 return 0;
1328         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1329                 return 0;
1330         switch (claim) {
1331         case NFS4_OPEN_CLAIM_NULL:
1332         case NFS4_OPEN_CLAIM_FH:
1333                 break;
1334         case NFS4_OPEN_CLAIM_PREVIOUS:
1335                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1336                         break;
1337         default:
1338                 return 0;
1339         }
1340         nfs_mark_delegation_referenced(delegation);
1341         return 1;
1342 }
1343
1344 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1345 {
1346         switch (fmode) {
1347                 case FMODE_WRITE:
1348                         state->n_wronly++;
1349                         break;
1350                 case FMODE_READ:
1351                         state->n_rdonly++;
1352                         break;
1353                 case FMODE_READ|FMODE_WRITE:
1354                         state->n_rdwr++;
1355         }
1356         nfs4_state_set_mode_locked(state, state->state | fmode);
1357 }
1358
1359 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1360 {
1361         struct nfs_client *clp = state->owner->so_server->nfs_client;
1362         bool need_recover = false;
1363
1364         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1365                 need_recover = true;
1366         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1367                 need_recover = true;
1368         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1369                 need_recover = true;
1370         if (need_recover)
1371                 nfs4_state_mark_reclaim_nograce(clp, state);
1372 }
1373
1374 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1375                 nfs4_stateid *stateid)
1376 {
1377         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1378                 return true;
1379         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1380                 nfs_test_and_clear_all_open_stateid(state);
1381                 return true;
1382         }
1383         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1384                 return true;
1385         return false;
1386 }
1387
1388 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1389 {
1390         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1391                 return;
1392         if (state->n_wronly)
1393                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1394         if (state->n_rdonly)
1395                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1396         if (state->n_rdwr)
1397                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1398         set_bit(NFS_OPEN_STATE, &state->flags);
1399 }
1400
1401 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1402                 nfs4_stateid *arg_stateid,
1403                 nfs4_stateid *stateid, fmode_t fmode)
1404 {
1405         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1406         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1407         case FMODE_WRITE:
1408                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1409                 break;
1410         case FMODE_READ:
1411                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1412                 break;
1413         case 0:
1414                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1415                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1416                 clear_bit(NFS_OPEN_STATE, &state->flags);
1417         }
1418         if (stateid == NULL)
1419                 return;
1420         /* Handle races with OPEN */
1421         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1422             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1423             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1424                 nfs_resync_open_stateid_locked(state);
1425                 return;
1426         }
1427         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1428                 nfs4_stateid_copy(&state->stateid, stateid);
1429         nfs4_stateid_copy(&state->open_stateid, stateid);
1430 }
1431
1432 static void nfs_clear_open_stateid(struct nfs4_state *state,
1433         nfs4_stateid *arg_stateid,
1434         nfs4_stateid *stateid, fmode_t fmode)
1435 {
1436         write_seqlock(&state->seqlock);
1437         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1438         write_sequnlock(&state->seqlock);
1439         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1440                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1441 }
1442
1443 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1444 {
1445         switch (fmode) {
1446                 case FMODE_READ:
1447                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1448                         break;
1449                 case FMODE_WRITE:
1450                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1451                         break;
1452                 case FMODE_READ|FMODE_WRITE:
1453                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1454         }
1455         if (!nfs_need_update_open_stateid(state, stateid))
1456                 return;
1457         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1458                 nfs4_stateid_copy(&state->stateid, stateid);
1459         nfs4_stateid_copy(&state->open_stateid, stateid);
1460 }
1461
1462 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1463 {
1464         /*
1465          * Protect the call to nfs4_state_set_mode_locked and
1466          * serialise the stateid update
1467          */
1468         spin_lock(&state->owner->so_lock);
1469         write_seqlock(&state->seqlock);
1470         if (deleg_stateid != NULL) {
1471                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1472                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1473         }
1474         if (open_stateid != NULL)
1475                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1476         write_sequnlock(&state->seqlock);
1477         update_open_stateflags(state, fmode);
1478         spin_unlock(&state->owner->so_lock);
1479 }
1480
1481 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1482 {
1483         struct nfs_inode *nfsi = NFS_I(state->inode);
1484         struct nfs_delegation *deleg_cur;
1485         int ret = 0;
1486
1487         fmode &= (FMODE_READ|FMODE_WRITE);
1488
1489         rcu_read_lock();
1490         deleg_cur = rcu_dereference(nfsi->delegation);
1491         if (deleg_cur == NULL)
1492                 goto no_delegation;
1493
1494         spin_lock(&deleg_cur->lock);
1495         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1496            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1497             (deleg_cur->type & fmode) != fmode)
1498                 goto no_delegation_unlock;
1499
1500         if (delegation == NULL)
1501                 delegation = &deleg_cur->stateid;
1502         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1503                 goto no_delegation_unlock;
1504
1505         nfs_mark_delegation_referenced(deleg_cur);
1506         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1507         ret = 1;
1508 no_delegation_unlock:
1509         spin_unlock(&deleg_cur->lock);
1510 no_delegation:
1511         rcu_read_unlock();
1512
1513         if (!ret && open_stateid != NULL) {
1514                 __update_open_stateid(state, open_stateid, NULL, fmode);
1515                 ret = 1;
1516         }
1517         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1518                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1519
1520         return ret;
1521 }
1522
1523 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1524                 const nfs4_stateid *stateid)
1525 {
1526         struct nfs4_state *state = lsp->ls_state;
1527         bool ret = false;
1528
1529         spin_lock(&state->state_lock);
1530         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1531                 goto out_noupdate;
1532         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1533                 goto out_noupdate;
1534         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1535         ret = true;
1536 out_noupdate:
1537         spin_unlock(&state->state_lock);
1538         return ret;
1539 }
1540
1541 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1542 {
1543         struct nfs_delegation *delegation;
1544
1545         rcu_read_lock();
1546         delegation = rcu_dereference(NFS_I(inode)->delegation);
1547         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1548                 rcu_read_unlock();
1549                 return;
1550         }
1551         rcu_read_unlock();
1552         nfs4_inode_return_delegation(inode);
1553 }
1554
1555 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1556 {
1557         struct nfs4_state *state = opendata->state;
1558         struct nfs_inode *nfsi = NFS_I(state->inode);
1559         struct nfs_delegation *delegation;
1560         int open_mode = opendata->o_arg.open_flags;
1561         fmode_t fmode = opendata->o_arg.fmode;
1562         enum open_claim_type4 claim = opendata->o_arg.claim;
1563         nfs4_stateid stateid;
1564         int ret = -EAGAIN;
1565
1566         for (;;) {
1567                 spin_lock(&state->owner->so_lock);
1568                 if (can_open_cached(state, fmode, open_mode)) {
1569                         update_open_stateflags(state, fmode);
1570                         spin_unlock(&state->owner->so_lock);
1571                         goto out_return_state;
1572                 }
1573                 spin_unlock(&state->owner->so_lock);
1574                 rcu_read_lock();
1575                 delegation = rcu_dereference(nfsi->delegation);
1576                 if (!can_open_delegated(delegation, fmode, claim)) {
1577                         rcu_read_unlock();
1578                         break;
1579                 }
1580                 /* Save the delegation */
1581                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1582                 rcu_read_unlock();
1583                 nfs_release_seqid(opendata->o_arg.seqid);
1584                 if (!opendata->is_recover) {
1585                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1586                         if (ret != 0)
1587                                 goto out;
1588                 }
1589                 ret = -EAGAIN;
1590
1591                 /* Try to update the stateid using the delegation */
1592                 if (update_open_stateid(state, NULL, &stateid, fmode))
1593                         goto out_return_state;
1594         }
1595 out:
1596         return ERR_PTR(ret);
1597 out_return_state:
1598         atomic_inc(&state->count);
1599         return state;
1600 }
1601
1602 static void
1603 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1604 {
1605         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1606         struct nfs_delegation *delegation;
1607         int delegation_flags = 0;
1608
1609         rcu_read_lock();
1610         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1611         if (delegation)
1612                 delegation_flags = delegation->flags;
1613         rcu_read_unlock();
1614         switch (data->o_arg.claim) {
1615         default:
1616                 break;
1617         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1618         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1619                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1620                                    "returning a delegation for "
1621                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1622                                    clp->cl_hostname);
1623                 return;
1624         }
1625         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1626                 nfs_inode_set_delegation(state->inode,
1627                                          data->owner->so_cred,
1628                                          &data->o_res);
1629         else
1630                 nfs_inode_reclaim_delegation(state->inode,
1631                                              data->owner->so_cred,
1632                                              &data->o_res);
1633 }
1634
1635 /*
1636  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1637  * and update the nfs4_state.
1638  */
1639 static struct nfs4_state *
1640 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1641 {
1642         struct inode *inode = data->state->inode;
1643         struct nfs4_state *state = data->state;
1644         int ret;
1645
1646         if (!data->rpc_done) {
1647                 if (data->rpc_status) {
1648                         ret = data->rpc_status;
1649                         goto err;
1650                 }
1651                 /* cached opens have already been processed */
1652                 goto update;
1653         }
1654
1655         ret = nfs_refresh_inode(inode, &data->f_attr);
1656         if (ret)
1657                 goto err;
1658
1659         if (data->o_res.delegation_type != 0)
1660                 nfs4_opendata_check_deleg(data, state);
1661 update:
1662         update_open_stateid(state, &data->o_res.stateid, NULL,
1663                             data->o_arg.fmode);
1664         atomic_inc(&state->count);
1665
1666         return state;
1667 err:
1668         return ERR_PTR(ret);
1669
1670 }
1671
1672 static struct nfs4_state *
1673 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1674 {
1675         struct inode *inode;
1676         struct nfs4_state *state = NULL;
1677         int ret;
1678
1679         if (!data->rpc_done) {
1680                 state = nfs4_try_open_cached(data);
1681                 trace_nfs4_cached_open(data->state);
1682                 goto out;
1683         }
1684
1685         ret = -EAGAIN;
1686         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1687                 goto err;
1688         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1689         ret = PTR_ERR(inode);
1690         if (IS_ERR(inode))
1691                 goto err;
1692         ret = -ENOMEM;
1693         state = nfs4_get_open_state(inode, data->owner);
1694         if (state == NULL)
1695                 goto err_put_inode;
1696         if (data->o_res.delegation_type != 0)
1697                 nfs4_opendata_check_deleg(data, state);
1698         update_open_stateid(state, &data->o_res.stateid, NULL,
1699                         data->o_arg.fmode);
1700         iput(inode);
1701 out:
1702         nfs_release_seqid(data->o_arg.seqid);
1703         return state;
1704 err_put_inode:
1705         iput(inode);
1706 err:
1707         return ERR_PTR(ret);
1708 }
1709
1710 static struct nfs4_state *
1711 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1712 {
1713         struct nfs4_state *ret;
1714
1715         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1716                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1717         else
1718                 ret = _nfs4_opendata_to_nfs4_state(data);
1719         nfs4_sequence_free_slot(&data->o_res.seq_res);
1720         return ret;
1721 }
1722
1723 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1724 {
1725         struct nfs_inode *nfsi = NFS_I(state->inode);
1726         struct nfs_open_context *ctx;
1727
1728         spin_lock(&state->inode->i_lock);
1729         list_for_each_entry(ctx, &nfsi->open_files, list) {
1730                 if (ctx->state != state)
1731                         continue;
1732                 get_nfs_open_context(ctx);
1733                 spin_unlock(&state->inode->i_lock);
1734                 return ctx;
1735         }
1736         spin_unlock(&state->inode->i_lock);
1737         return ERR_PTR(-ENOENT);
1738 }
1739
1740 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1741                 struct nfs4_state *state, enum open_claim_type4 claim)
1742 {
1743         struct nfs4_opendata *opendata;
1744
1745         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1746                         NULL, NULL, claim, GFP_NOFS);
1747         if (opendata == NULL)
1748                 return ERR_PTR(-ENOMEM);
1749         opendata->state = state;
1750         atomic_inc(&state->count);
1751         return opendata;
1752 }
1753
1754 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1755                 fmode_t fmode)
1756 {
1757         struct nfs4_state *newstate;
1758         int ret;
1759
1760         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1761                 return 0;
1762         opendata->o_arg.open_flags = 0;
1763         opendata->o_arg.fmode = fmode;
1764         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1765                         NFS_SB(opendata->dentry->d_sb),
1766                         fmode, 0);
1767         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1768         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1769         nfs4_init_opendata_res(opendata);
1770         ret = _nfs4_recover_proc_open(opendata);
1771         if (ret != 0)
1772                 return ret;
1773         newstate = nfs4_opendata_to_nfs4_state(opendata);
1774         if (IS_ERR(newstate))
1775                 return PTR_ERR(newstate);
1776         if (newstate != opendata->state)
1777                 ret = -ESTALE;
1778         nfs4_close_state(newstate, fmode);
1779         return ret;
1780 }
1781
1782 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1783 {
1784         int ret;
1785
1786         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1787         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1788         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1789         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1790         /* memory barrier prior to reading state->n_* */
1791         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1792         clear_bit(NFS_OPEN_STATE, &state->flags);
1793         smp_rmb();
1794         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1795         if (ret != 0)
1796                 return ret;
1797         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1798         if (ret != 0)
1799                 return ret;
1800         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1801         if (ret != 0)
1802                 return ret;
1803         /*
1804          * We may have performed cached opens for all three recoveries.
1805          * Check if we need to update the current stateid.
1806          */
1807         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1808             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1809                 write_seqlock(&state->seqlock);
1810                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1811                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1812                 write_sequnlock(&state->seqlock);
1813         }
1814         return 0;
1815 }
1816
1817 /*
1818  * OPEN_RECLAIM:
1819  *      reclaim state on the server after a reboot.
1820  */
1821 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1822 {
1823         struct nfs_delegation *delegation;
1824         struct nfs4_opendata *opendata;
1825         fmode_t delegation_type = 0;
1826         int status;
1827
1828         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1829                         NFS4_OPEN_CLAIM_PREVIOUS);
1830         if (IS_ERR(opendata))
1831                 return PTR_ERR(opendata);
1832         rcu_read_lock();
1833         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1834         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1835                 delegation_type = delegation->type;
1836         rcu_read_unlock();
1837         opendata->o_arg.u.delegation_type = delegation_type;
1838         status = nfs4_open_recover(opendata, state);
1839         nfs4_opendata_put(opendata);
1840         return status;
1841 }
1842
1843 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1844 {
1845         struct nfs_server *server = NFS_SERVER(state->inode);
1846         struct nfs4_exception exception = { };
1847         int err;
1848         do {
1849                 err = _nfs4_do_open_reclaim(ctx, state);
1850                 trace_nfs4_open_reclaim(ctx, 0, err);
1851                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1852                         continue;
1853                 if (err != -NFS4ERR_DELAY)
1854                         break;
1855                 nfs4_handle_exception(server, err, &exception);
1856         } while (exception.retry);
1857         return err;
1858 }
1859
1860 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1861 {
1862         struct nfs_open_context *ctx;
1863         int ret;
1864
1865         ctx = nfs4_state_find_open_context(state);
1866         if (IS_ERR(ctx))
1867                 return -EAGAIN;
1868         ret = nfs4_do_open_reclaim(ctx, state);
1869         put_nfs_open_context(ctx);
1870         return ret;
1871 }
1872
1873 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1874 {
1875         switch (err) {
1876                 default:
1877                         printk(KERN_ERR "NFS: %s: unhandled error "
1878                                         "%d.\n", __func__, err);
1879                 case 0:
1880                 case -ENOENT:
1881                 case -EAGAIN:
1882                 case -ESTALE:
1883                         break;
1884                 case -NFS4ERR_BADSESSION:
1885                 case -NFS4ERR_BADSLOT:
1886                 case -NFS4ERR_BAD_HIGH_SLOT:
1887                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1888                 case -NFS4ERR_DEADSESSION:
1889                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1890                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1891                         return -EAGAIN;
1892                 case -NFS4ERR_STALE_CLIENTID:
1893                 case -NFS4ERR_STALE_STATEID:
1894                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1895                 case -NFS4ERR_EXPIRED:
1896                         /* Don't recall a delegation if it was lost */
1897                         nfs4_schedule_lease_recovery(server->nfs_client);
1898                         return -EAGAIN;
1899                 case -NFS4ERR_MOVED:
1900                         nfs4_schedule_migration_recovery(server);
1901                         return -EAGAIN;
1902                 case -NFS4ERR_LEASE_MOVED:
1903                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1904                         return -EAGAIN;
1905                 case -NFS4ERR_DELEG_REVOKED:
1906                 case -NFS4ERR_ADMIN_REVOKED:
1907                 case -NFS4ERR_BAD_STATEID:
1908                 case -NFS4ERR_OPENMODE:
1909                         nfs_inode_find_state_and_recover(state->inode,
1910                                         stateid);
1911                         nfs4_schedule_stateid_recovery(server, state);
1912                         return -EAGAIN;
1913                 case -NFS4ERR_DELAY:
1914                 case -NFS4ERR_GRACE:
1915                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1916                         ssleep(1);
1917                         return -EAGAIN;
1918                 case -ENOMEM:
1919                 case -NFS4ERR_DENIED:
1920                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1921                         return 0;
1922         }
1923         return err;
1924 }
1925
1926 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1927                 struct nfs4_state *state, const nfs4_stateid *stateid,
1928                 fmode_t type)
1929 {
1930         struct nfs_server *server = NFS_SERVER(state->inode);
1931         struct nfs4_opendata *opendata;
1932         int err = 0;
1933
1934         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1935                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1936         if (IS_ERR(opendata))
1937                 return PTR_ERR(opendata);
1938         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1939         write_seqlock(&state->seqlock);
1940         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1941         write_sequnlock(&state->seqlock);
1942         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1943         switch (type & (FMODE_READ|FMODE_WRITE)) {
1944         case FMODE_READ|FMODE_WRITE:
1945         case FMODE_WRITE:
1946                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1947                 if (err)
1948                         break;
1949                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1950                 if (err)
1951                         break;
1952         case FMODE_READ:
1953                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1954         }
1955         nfs4_opendata_put(opendata);
1956         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1957 }
1958
1959 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1960 {
1961         struct nfs4_opendata *data = calldata;
1962
1963         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1964                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1965 }
1966
1967 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1968 {
1969         struct nfs4_opendata *data = calldata;
1970
1971         nfs40_sequence_done(task, &data->c_res.seq_res);
1972
1973         data->rpc_status = task->tk_status;
1974         if (data->rpc_status == 0) {
1975                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1976                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1977                 renew_lease(data->o_res.server, data->timestamp);
1978                 data->rpc_done = 1;
1979         }
1980 }
1981
1982 static void nfs4_open_confirm_release(void *calldata)
1983 {
1984         struct nfs4_opendata *data = calldata;
1985         struct nfs4_state *state = NULL;
1986
1987         /* If this request hasn't been cancelled, do nothing */
1988         if (data->cancelled == 0)
1989                 goto out_free;
1990         /* In case of error, no cleanup! */
1991         if (!data->rpc_done)
1992                 goto out_free;
1993         state = nfs4_opendata_to_nfs4_state(data);
1994         if (!IS_ERR(state))
1995                 nfs4_close_state(state, data->o_arg.fmode);
1996 out_free:
1997         nfs4_opendata_put(data);
1998 }
1999
2000 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2001         .rpc_call_prepare = nfs4_open_confirm_prepare,
2002         .rpc_call_done = nfs4_open_confirm_done,
2003         .rpc_release = nfs4_open_confirm_release,
2004 };
2005
2006 /*
2007  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2008  */
2009 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2010 {
2011         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2012         struct rpc_task *task;
2013         struct  rpc_message msg = {
2014                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2015                 .rpc_argp = &data->c_arg,
2016                 .rpc_resp = &data->c_res,
2017                 .rpc_cred = data->owner->so_cred,
2018         };
2019         struct rpc_task_setup task_setup_data = {
2020                 .rpc_client = server->client,
2021                 .rpc_message = &msg,
2022                 .callback_ops = &nfs4_open_confirm_ops,
2023                 .callback_data = data,
2024                 .workqueue = nfsiod_workqueue,
2025                 .flags = RPC_TASK_ASYNC,
2026         };
2027         int status;
2028
2029         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2030         kref_get(&data->kref);
2031         data->rpc_done = 0;
2032         data->rpc_status = 0;
2033         data->timestamp = jiffies;
2034         if (data->is_recover)
2035                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2036         task = rpc_run_task(&task_setup_data);
2037         if (IS_ERR(task))
2038                 return PTR_ERR(task);
2039         status = nfs4_wait_for_completion_rpc_task(task);
2040         if (status != 0) {
2041                 data->cancelled = 1;
2042                 smp_wmb();
2043         } else
2044                 status = data->rpc_status;
2045         rpc_put_task(task);
2046         return status;
2047 }
2048
2049 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2050 {
2051         struct nfs4_opendata *data = calldata;
2052         struct nfs4_state_owner *sp = data->owner;
2053         struct nfs_client *clp = sp->so_server->nfs_client;
2054         enum open_claim_type4 claim = data->o_arg.claim;
2055
2056         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2057                 goto out_wait;
2058         /*
2059          * Check if we still need to send an OPEN call, or if we can use
2060          * a delegation instead.
2061          */
2062         if (data->state != NULL) {
2063                 struct nfs_delegation *delegation;
2064
2065                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2066                         goto out_no_action;
2067                 rcu_read_lock();
2068                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2069                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2070                         goto unlock_no_action;
2071                 rcu_read_unlock();
2072         }
2073         /* Update client id. */
2074         data->o_arg.clientid = clp->cl_clientid;
2075         switch (claim) {
2076         default:
2077                 break;
2078         case NFS4_OPEN_CLAIM_PREVIOUS:
2079         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2080         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2081                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2082         case NFS4_OPEN_CLAIM_FH:
2083                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2084                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2085         }
2086         data->timestamp = jiffies;
2087         if (nfs4_setup_sequence(data->o_arg.server,
2088                                 &data->o_arg.seq_args,
2089                                 &data->o_res.seq_res,
2090                                 task) != 0)
2091                 nfs_release_seqid(data->o_arg.seqid);
2092
2093         /* Set the create mode (note dependency on the session type) */
2094         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2095         if (data->o_arg.open_flags & O_EXCL) {
2096                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2097                 if (nfs4_has_persistent_session(clp))
2098                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2099                 else if (clp->cl_mvops->minor_version > 0)
2100                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2101         }
2102         return;
2103 unlock_no_action:
2104         trace_nfs4_cached_open(data->state);
2105         rcu_read_unlock();
2106 out_no_action:
2107         task->tk_action = NULL;
2108 out_wait:
2109         nfs4_sequence_done(task, &data->o_res.seq_res);
2110 }
2111
2112 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2113 {
2114         struct nfs4_opendata *data = calldata;
2115
2116         data->rpc_status = task->tk_status;
2117
2118         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2119                 return;
2120
2121         if (task->tk_status == 0) {
2122                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2123                         switch (data->o_res.f_attr->mode & S_IFMT) {
2124                         case S_IFREG:
2125                                 break;
2126                         case S_IFLNK:
2127                                 data->rpc_status = -ELOOP;
2128                                 break;
2129                         case S_IFDIR:
2130                                 data->rpc_status = -EISDIR;
2131                                 break;
2132                         default:
2133                                 data->rpc_status = -ENOTDIR;
2134                         }
2135                 }
2136                 renew_lease(data->o_res.server, data->timestamp);
2137                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2138                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2139         }
2140         data->rpc_done = 1;
2141 }
2142
2143 static void nfs4_open_release(void *calldata)
2144 {
2145         struct nfs4_opendata *data = calldata;
2146         struct nfs4_state *state = NULL;
2147
2148         /* If this request hasn't been cancelled, do nothing */
2149         if (data->cancelled == 0)
2150                 goto out_free;
2151         /* In case of error, no cleanup! */
2152         if (data->rpc_status != 0 || !data->rpc_done)
2153                 goto out_free;
2154         /* In case we need an open_confirm, no cleanup! */
2155         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2156                 goto out_free;
2157         state = nfs4_opendata_to_nfs4_state(data);
2158         if (!IS_ERR(state))
2159                 nfs4_close_state(state, data->o_arg.fmode);
2160 out_free:
2161         nfs4_opendata_put(data);
2162 }
2163
2164 static const struct rpc_call_ops nfs4_open_ops = {
2165         .rpc_call_prepare = nfs4_open_prepare,
2166         .rpc_call_done = nfs4_open_done,
2167         .rpc_release = nfs4_open_release,
2168 };
2169
2170 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2171 {
2172         struct inode *dir = d_inode(data->dir);
2173         struct nfs_server *server = NFS_SERVER(dir);
2174         struct nfs_openargs *o_arg = &data->o_arg;
2175         struct nfs_openres *o_res = &data->o_res;
2176         struct rpc_task *task;
2177         struct rpc_message msg = {
2178                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2179                 .rpc_argp = o_arg,
2180                 .rpc_resp = o_res,
2181                 .rpc_cred = data->owner->so_cred,
2182         };
2183         struct rpc_task_setup task_setup_data = {
2184                 .rpc_client = server->client,
2185                 .rpc_message = &msg,
2186                 .callback_ops = &nfs4_open_ops,
2187                 .callback_data = data,
2188                 .workqueue = nfsiod_workqueue,
2189                 .flags = RPC_TASK_ASYNC,
2190         };
2191         int status;
2192
2193         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2194         kref_get(&data->kref);
2195         data->rpc_done = 0;
2196         data->rpc_status = 0;
2197         data->cancelled = 0;
2198         data->is_recover = 0;
2199         if (isrecover) {
2200                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2201                 data->is_recover = 1;
2202         }
2203         task = rpc_run_task(&task_setup_data);
2204         if (IS_ERR(task))
2205                 return PTR_ERR(task);
2206         status = nfs4_wait_for_completion_rpc_task(task);
2207         if (status != 0) {
2208                 data->cancelled = 1;
2209                 smp_wmb();
2210         } else
2211                 status = data->rpc_status;
2212         rpc_put_task(task);
2213
2214         return status;
2215 }
2216
2217 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2218 {
2219         struct inode *dir = d_inode(data->dir);
2220         struct nfs_openres *o_res = &data->o_res;
2221         int status;
2222
2223         status = nfs4_run_open_task(data, 1);
2224         if (status != 0 || !data->rpc_done)
2225                 return status;
2226
2227         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2228
2229         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2230                 status = _nfs4_proc_open_confirm(data);
2231                 if (status != 0)
2232                         return status;
2233         }
2234
2235         return status;
2236 }
2237
2238 /*
2239  * Additional permission checks in order to distinguish between an
2240  * open for read, and an open for execute. This works around the
2241  * fact that NFSv4 OPEN treats read and execute permissions as being
2242  * the same.
2243  * Note that in the non-execute case, we want to turn off permission
2244  * checking if we just created a new file (POSIX open() semantics).
2245  */
2246 static int nfs4_opendata_access(struct rpc_cred *cred,
2247                                 struct nfs4_opendata *opendata,
2248                                 struct nfs4_state *state, fmode_t fmode,
2249                                 int openflags)
2250 {
2251         struct nfs_access_entry cache;
2252         u32 mask;
2253
2254         /* access call failed or for some reason the server doesn't
2255          * support any access modes -- defer access call until later */
2256         if (opendata->o_res.access_supported == 0)
2257                 return 0;
2258
2259         mask = 0;
2260         /*
2261          * Use openflags to check for exec, because fmode won't
2262          * always have FMODE_EXEC set when file open for exec.
2263          */
2264         if (openflags & __FMODE_EXEC) {
2265                 /* ONLY check for exec rights */
2266                 mask = MAY_EXEC;
2267         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2268                 mask = MAY_READ;
2269
2270         cache.cred = cred;
2271         cache.jiffies = jiffies;
2272         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2273         nfs_access_add_cache(state->inode, &cache);
2274
2275         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2276                 return 0;
2277
2278         /* even though OPEN succeeded, access is denied. Close the file */
2279         nfs4_close_state(state, fmode);
2280         return -EACCES;
2281 }
2282
2283 /*
2284  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2285  */
2286 static int _nfs4_proc_open(struct nfs4_opendata *data)
2287 {
2288         struct inode *dir = d_inode(data->dir);
2289         struct nfs_server *server = NFS_SERVER(dir);
2290         struct nfs_openargs *o_arg = &data->o_arg;
2291         struct nfs_openres *o_res = &data->o_res;
2292         int status;
2293
2294         status = nfs4_run_open_task(data, 0);
2295         if (!data->rpc_done)
2296                 return status;
2297         if (status != 0) {
2298                 if (status == -NFS4ERR_BADNAME &&
2299                                 !(o_arg->open_flags & O_CREAT))
2300                         return -ENOENT;
2301                 return status;
2302         }
2303
2304         nfs_fattr_map_and_free_names(server, &data->f_attr);
2305
2306         if (o_arg->open_flags & O_CREAT) {
2307                 update_changeattr(dir, &o_res->cinfo);
2308                 if (o_arg->open_flags & O_EXCL)
2309                         data->file_created = 1;
2310                 else if (o_res->cinfo.before != o_res->cinfo.after)
2311                         data->file_created = 1;
2312         }
2313         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2314                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2315         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2316                 status = _nfs4_proc_open_confirm(data);
2317                 if (status != 0)
2318                         return status;
2319         }
2320         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2321                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2322         return 0;
2323 }
2324
2325 static int nfs4_recover_expired_lease(struct nfs_server *server)
2326 {
2327         return nfs4_client_recover_expired_lease(server->nfs_client);
2328 }
2329
2330 /*
2331  * OPEN_EXPIRED:
2332  *      reclaim state on the server after a network partition.
2333  *      Assumes caller holds the appropriate lock
2334  */
2335 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2336 {
2337         struct nfs4_opendata *opendata;
2338         int ret;
2339
2340         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2341                         NFS4_OPEN_CLAIM_FH);
2342         if (IS_ERR(opendata))
2343                 return PTR_ERR(opendata);
2344         ret = nfs4_open_recover(opendata, state);
2345         if (ret == -ESTALE)
2346                 d_drop(ctx->dentry);
2347         nfs4_opendata_put(opendata);
2348         return ret;
2349 }
2350
2351 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2352 {
2353         struct nfs_server *server = NFS_SERVER(state->inode);
2354         struct nfs4_exception exception = { };
2355         int err;
2356
2357         do {
2358                 err = _nfs4_open_expired(ctx, state);
2359                 trace_nfs4_open_expired(ctx, 0, err);
2360                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2361                         continue;
2362                 switch (err) {
2363                 default:
2364                         goto out;
2365                 case -NFS4ERR_GRACE:
2366                 case -NFS4ERR_DELAY:
2367                         nfs4_handle_exception(server, err, &exception);
2368                         err = 0;
2369                 }
2370         } while (exception.retry);
2371 out:
2372         return err;
2373 }
2374
2375 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2376 {
2377         struct nfs_open_context *ctx;
2378         int ret;
2379
2380         ctx = nfs4_state_find_open_context(state);
2381         if (IS_ERR(ctx))
2382                 return -EAGAIN;
2383         ret = nfs4_do_open_expired(ctx, state);
2384         put_nfs_open_context(ctx);
2385         return ret;
2386 }
2387
2388 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2389 {
2390         nfs_remove_bad_delegation(state->inode);
2391         write_seqlock(&state->seqlock);
2392         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2393         write_sequnlock(&state->seqlock);
2394         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2395 }
2396
2397 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2398 {
2399         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2400                 nfs_finish_clear_delegation_stateid(state);
2401 }
2402
2403 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2404 {
2405         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2406         nfs40_clear_delegation_stateid(state);
2407         return nfs4_open_expired(sp, state);
2408 }
2409
2410 #if defined(CONFIG_NFS_V4_1)
2411 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2412                 nfs4_stateid *stateid,
2413                 struct rpc_cred *cred)
2414 {
2415         int status;
2416
2417         status = nfs41_test_stateid(server, stateid, cred);
2418
2419         switch (status) {
2420         case -NFS4ERR_EXPIRED:
2421         case -NFS4ERR_ADMIN_REVOKED:
2422         case -NFS4ERR_DELEG_REVOKED:
2423                 /* Ack the revoked state to the server */
2424                 nfs41_free_stateid(server, stateid, cred);
2425         case -NFS4ERR_BAD_STATEID:
2426                 return status;
2427         }
2428         return NFS_OK;
2429 }
2430
2431 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2432 {
2433         struct nfs_server *server = NFS_SERVER(state->inode);
2434         nfs4_stateid stateid;
2435         struct nfs_delegation *delegation;
2436         struct rpc_cred *cred;
2437         int status;
2438
2439         /* Get the delegation credential for use by test/free_stateid */
2440         rcu_read_lock();
2441         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2442         if (delegation == NULL) {
2443                 rcu_read_unlock();
2444                 return;
2445         }
2446         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2447                 rcu_read_unlock();
2448                 nfs_finish_clear_delegation_stateid(state);
2449                 return;
2450         }
2451
2452         nfs4_stateid_copy(&stateid, &delegation->stateid);
2453         cred = get_rpccred(delegation->cred);
2454         rcu_read_unlock();
2455         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2456         trace_nfs4_test_delegation_stateid(state, NULL, status);
2457         if (status != NFS_OK)
2458                 nfs_finish_clear_delegation_stateid(state);
2459
2460         put_rpccred(cred);
2461 }
2462
2463 /**
2464  * nfs41_check_open_stateid - possibly free an open stateid
2465  *
2466  * @state: NFSv4 state for an inode
2467  *
2468  * Returns NFS_OK if recovery for this stateid is now finished.
2469  * Otherwise a negative NFS4ERR value is returned.
2470  */
2471 static int nfs41_check_open_stateid(struct nfs4_state *state)
2472 {
2473         struct nfs_server *server = NFS_SERVER(state->inode);
2474         nfs4_stateid *stateid = &state->open_stateid;
2475         struct rpc_cred *cred = state->owner->so_cred;
2476         int status;
2477
2478         /* If a state reset has been done, test_stateid is unneeded */
2479         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2480             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2481             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2482                 return -NFS4ERR_BAD_STATEID;
2483
2484         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2485         trace_nfs4_test_open_stateid(state, NULL, status);
2486         if (status != NFS_OK) {
2487                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2488                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2489                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2490                 clear_bit(NFS_OPEN_STATE, &state->flags);
2491         }
2492         return status;
2493 }
2494
2495 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2496 {
2497         int status;
2498
2499         nfs41_check_delegation_stateid(state);
2500         status = nfs41_check_open_stateid(state);
2501         if (status != NFS_OK)
2502                 status = nfs4_open_expired(sp, state);
2503         return status;
2504 }
2505 #endif
2506
2507 /*
2508  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2509  * fields corresponding to attributes that were used to store the verifier.
2510  * Make sure we clobber those fields in the later setattr call
2511  */
2512 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2513                                 struct iattr *sattr, struct nfs4_label **label)
2514 {
2515         const u32 *attrset = opendata->o_res.attrset;
2516
2517         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2518             !(sattr->ia_valid & ATTR_ATIME_SET))
2519                 sattr->ia_valid |= ATTR_ATIME;
2520
2521         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2522             !(sattr->ia_valid & ATTR_MTIME_SET))
2523                 sattr->ia_valid |= ATTR_MTIME;
2524
2525         /* Except MODE, it seems harmless of setting twice. */
2526         if ((attrset[1] & FATTR4_WORD1_MODE))
2527                 sattr->ia_valid &= ~ATTR_MODE;
2528
2529         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2530                 *label = NULL;
2531 }
2532
2533 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2534                 fmode_t fmode,
2535                 int flags,
2536                 struct nfs_open_context *ctx)
2537 {
2538         struct nfs4_state_owner *sp = opendata->owner;
2539         struct nfs_server *server = sp->so_server;
2540         struct dentry *dentry;
2541         struct nfs4_state *state;
2542         unsigned int seq;
2543         int ret;
2544
2545         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2546
2547         ret = _nfs4_proc_open(opendata);
2548         if (ret != 0)
2549                 goto out;
2550
2551         state = nfs4_opendata_to_nfs4_state(opendata);
2552         ret = PTR_ERR(state);
2553         if (IS_ERR(state))
2554                 goto out;
2555         if (server->caps & NFS_CAP_POSIX_LOCK)
2556                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2557         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2558                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2559
2560         dentry = opendata->dentry;
2561         if (d_really_is_negative(dentry)) {
2562                 struct dentry *alias;
2563                 d_drop(dentry);
2564                 alias = d_exact_alias(dentry, state->inode);
2565                 if (!alias)
2566                         alias = d_splice_alias(igrab(state->inode), dentry);
2567                 /* d_splice_alias() can't fail here - it's a non-directory */
2568                 if (alias) {
2569                         dput(ctx->dentry);
2570                         ctx->dentry = dentry = alias;
2571                 }
2572                 nfs_set_verifier(dentry,
2573                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2574         }
2575
2576         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2577         if (ret != 0)
2578                 goto out;
2579
2580         ctx->state = state;
2581         if (d_inode(dentry) == state->inode) {
2582                 nfs_inode_attach_open_context(ctx);
2583                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2584                         nfs4_schedule_stateid_recovery(server, state);
2585         }
2586 out:
2587         return ret;
2588 }
2589
2590 /*
2591  * Returns a referenced nfs4_state
2592  */
2593 static int _nfs4_do_open(struct inode *dir,
2594                         struct nfs_open_context *ctx,
2595                         int flags,
2596                         struct iattr *sattr,
2597                         struct nfs4_label *label,
2598                         int *opened)
2599 {
2600         struct nfs4_state_owner  *sp;
2601         struct nfs4_state     *state = NULL;
2602         struct nfs_server       *server = NFS_SERVER(dir);
2603         struct nfs4_opendata *opendata;
2604         struct dentry *dentry = ctx->dentry;
2605         struct rpc_cred *cred = ctx->cred;
2606         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2607         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2608         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2609         struct nfs4_label *olabel = NULL;
2610         int status;
2611
2612         /* Protect against reboot recovery conflicts */
2613         status = -ENOMEM;
2614         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2615         if (sp == NULL) {
2616                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2617                 goto out_err;
2618         }
2619         status = nfs4_recover_expired_lease(server);
2620         if (status != 0)
2621                 goto err_put_state_owner;
2622         if (d_really_is_positive(dentry))
2623                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2624         status = -ENOMEM;
2625         if (d_really_is_positive(dentry))
2626                 claim = NFS4_OPEN_CLAIM_FH;
2627         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2628                         label, claim, GFP_KERNEL);
2629         if (opendata == NULL)
2630                 goto err_put_state_owner;
2631
2632         if (label) {
2633                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2634                 if (IS_ERR(olabel)) {
2635                         status = PTR_ERR(olabel);
2636                         goto err_opendata_put;
2637                 }
2638         }
2639
2640         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2641                 if (!opendata->f_attr.mdsthreshold) {
2642                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2643                         if (!opendata->f_attr.mdsthreshold)
2644                                 goto err_free_label;
2645                 }
2646                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2647         }
2648         if (d_really_is_positive(dentry))
2649                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2650
2651         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2652         if (status != 0)
2653                 goto err_free_label;
2654         state = ctx->state;
2655
2656         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2657             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2658                 nfs4_exclusive_attrset(opendata, sattr, &label);
2659                 /*
2660                  * send create attributes which was not set by open
2661                  * with an extra setattr.
2662                  */
2663                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2664                         nfs_fattr_init(opendata->o_res.f_attr);
2665                         status = nfs4_do_setattr(state->inode, cred,
2666                                         opendata->o_res.f_attr, sattr,
2667                                         state, label, olabel);
2668                         if (status == 0) {
2669                                 nfs_setattr_update_inode(state->inode, sattr,
2670                                                 opendata->o_res.f_attr);
2671                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2672                         }
2673                 }
2674         }
2675         if (opened && opendata->file_created)
2676                 *opened |= FILE_CREATED;
2677
2678         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2679                 *ctx_th = opendata->f_attr.mdsthreshold;
2680                 opendata->f_attr.mdsthreshold = NULL;
2681         }
2682
2683         nfs4_label_free(olabel);
2684
2685         nfs4_opendata_put(opendata);
2686         nfs4_put_state_owner(sp);
2687         return 0;
2688 err_free_label:
2689         nfs4_label_free(olabel);
2690 err_opendata_put:
2691         nfs4_opendata_put(opendata);
2692 err_put_state_owner:
2693         nfs4_put_state_owner(sp);
2694 out_err:
2695         return status;
2696 }
2697
2698
2699 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2700                                         struct nfs_open_context *ctx,
2701                                         int flags,
2702                                         struct iattr *sattr,
2703                                         struct nfs4_label *label,
2704                                         int *opened)
2705 {
2706         struct nfs_server *server = NFS_SERVER(dir);
2707         struct nfs4_exception exception = { };
2708         struct nfs4_state *res;
2709         int status;
2710
2711         do {
2712                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2713                 res = ctx->state;
2714                 trace_nfs4_open_file(ctx, flags, status);
2715                 if (status == 0)
2716                         break;
2717                 /* NOTE: BAD_SEQID means the server and client disagree about the
2718                  * book-keeping w.r.t. state-changing operations
2719                  * (OPEN/CLOSE/LOCK/LOCKU...)
2720                  * It is actually a sign of a bug on the client or on the server.
2721                  *
2722                  * If we receive a BAD_SEQID error in the particular case of
2723                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2724                  * have unhashed the old state_owner for us, and that we can
2725                  * therefore safely retry using a new one. We should still warn
2726                  * the user though...
2727                  */
2728                 if (status == -NFS4ERR_BAD_SEQID) {
2729                         pr_warn_ratelimited("NFS: v4 server %s "
2730                                         " returned a bad sequence-id error!\n",
2731                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2732                         exception.retry = 1;
2733                         continue;
2734                 }
2735                 /*
2736                  * BAD_STATEID on OPEN means that the server cancelled our
2737                  * state before it received the OPEN_CONFIRM.
2738                  * Recover by retrying the request as per the discussion
2739                  * on Page 181 of RFC3530.
2740                  */
2741                 if (status == -NFS4ERR_BAD_STATEID) {
2742                         exception.retry = 1;
2743                         continue;
2744                 }
2745                 if (status == -EAGAIN) {
2746                         /* We must have found a delegation */
2747                         exception.retry = 1;
2748                         continue;
2749                 }
2750                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2751                         continue;
2752                 res = ERR_PTR(nfs4_handle_exception(server,
2753                                         status, &exception));
2754         } while (exception.retry);
2755         return res;
2756 }
2757
2758 static int _nfs4_do_setattr(struct inode *inode,
2759                             struct nfs_setattrargs *arg,
2760                             struct nfs_setattrres *res,
2761                             struct rpc_cred *cred,
2762                             struct nfs4_state *state)
2763 {
2764         struct nfs_server *server = NFS_SERVER(inode);
2765         struct rpc_message msg = {
2766                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2767                 .rpc_argp       = arg,
2768                 .rpc_resp       = res,
2769                 .rpc_cred       = cred,
2770         };
2771         struct rpc_cred *delegation_cred = NULL;
2772         unsigned long timestamp = jiffies;
2773         fmode_t fmode;
2774         bool truncate;
2775         int status;
2776
2777         nfs_fattr_init(res->fattr);
2778
2779         /* Servers should only apply open mode checks for file size changes */
2780         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2781         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2782
2783         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2784                 /* Use that stateid */
2785         } else if (truncate && state != NULL) {
2786                 struct nfs_lockowner lockowner = {
2787                         .l_owner = current->files,
2788                         .l_pid = current->tgid,
2789                 };
2790                 if (!nfs4_valid_open_stateid(state))
2791                         return -EBADF;
2792                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2793                                 &arg->stateid, &delegation_cred) == -EIO)
2794                         return -EBADF;
2795         } else
2796                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2797         if (delegation_cred)
2798                 msg.rpc_cred = delegation_cred;
2799
2800         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2801
2802         put_rpccred(delegation_cred);
2803         if (status == 0 && state != NULL)
2804                 renew_lease(server, timestamp);
2805         trace_nfs4_setattr(inode, &arg->stateid, status);
2806         return status;
2807 }
2808
2809 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2810                            struct nfs_fattr *fattr, struct iattr *sattr,
2811                            struct nfs4_state *state, struct nfs4_label *ilabel,
2812                            struct nfs4_label *olabel)
2813 {
2814         struct nfs_server *server = NFS_SERVER(inode);
2815         struct nfs_setattrargs  arg = {
2816                 .fh             = NFS_FH(inode),
2817                 .iap            = sattr,
2818                 .server         = server,
2819                 .bitmask = server->attr_bitmask,
2820                 .label          = ilabel,
2821         };
2822         struct nfs_setattrres  res = {
2823                 .fattr          = fattr,
2824                 .label          = olabel,
2825                 .server         = server,
2826         };
2827         struct nfs4_exception exception = {
2828                 .state = state,
2829                 .inode = inode,
2830                 .stateid = &arg.stateid,
2831         };
2832         int err;
2833
2834         arg.bitmask = nfs4_bitmask(server, ilabel);
2835         if (ilabel)
2836                 arg.bitmask = nfs4_bitmask(server, olabel);
2837
2838         do {
2839                 err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
2840                 switch (err) {
2841                 case -NFS4ERR_OPENMODE:
2842                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2843                                 pr_warn_once("NFSv4: server %s is incorrectly "
2844                                                 "applying open mode checks to "
2845                                                 "a SETATTR that is not "
2846                                                 "changing file size.\n",
2847                                                 server->nfs_client->cl_hostname);
2848                         }
2849                         if (state && !(state->state & FMODE_WRITE)) {
2850                                 err = -EBADF;
2851                                 if (sattr->ia_valid & ATTR_OPEN)
2852                                         err = -EACCES;
2853                                 goto out;
2854                         }
2855                 }
2856                 err = nfs4_handle_exception(server, err, &exception);
2857         } while (exception.retry);
2858 out:
2859         return err;
2860 }
2861
2862 static bool
2863 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2864 {
2865         if (inode == NULL || !nfs_have_layout(inode))
2866                 return false;
2867
2868         return pnfs_wait_on_layoutreturn(inode, task);
2869 }
2870
2871 struct nfs4_closedata {
2872         struct inode *inode;
2873         struct nfs4_state *state;
2874         struct nfs_closeargs arg;
2875         struct nfs_closeres res;
2876         struct nfs_fattr fattr;
2877         unsigned long timestamp;
2878         bool roc;
2879         u32 roc_barrier;
2880 };
2881
2882 static void nfs4_free_closedata(void *data)
2883 {
2884         struct nfs4_closedata *calldata = data;
2885         struct nfs4_state_owner *sp = calldata->state->owner;
2886         struct super_block *sb = calldata->state->inode->i_sb;
2887
2888         if (calldata->roc)
2889                 pnfs_roc_release(calldata->state->inode);
2890         nfs4_put_open_state(calldata->state);
2891         nfs_free_seqid(calldata->arg.seqid);
2892         nfs4_put_state_owner(sp);
2893         nfs_sb_deactive(sb);
2894         kfree(calldata);
2895 }
2896
2897 static void nfs4_close_done(struct rpc_task *task, void *data)
2898 {
2899         struct nfs4_closedata *calldata = data;
2900         struct nfs4_state *state = calldata->state;
2901         struct nfs_server *server = NFS_SERVER(calldata->inode);
2902         nfs4_stateid *res_stateid = NULL;
2903
2904         dprintk("%s: begin!\n", __func__);
2905         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2906                 return;
2907         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2908         /* hmm. we are done with the inode, and in the process of freeing
2909          * the state_owner. we keep this around to process errors
2910          */
2911         switch (task->tk_status) {
2912                 case 0:
2913                         res_stateid = &calldata->res.stateid;
2914                         if (calldata->roc)
2915                                 pnfs_roc_set_barrier(state->inode,
2916                                                      calldata->roc_barrier);
2917                         renew_lease(server, calldata->timestamp);
2918                         break;
2919                 case -NFS4ERR_ADMIN_REVOKED:
2920                 case -NFS4ERR_STALE_STATEID:
2921                 case -NFS4ERR_OLD_STATEID:
2922                 case -NFS4ERR_BAD_STATEID:
2923                 case -NFS4ERR_EXPIRED:
2924                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2925                                                 &state->open_stateid)) {
2926                                 rpc_restart_call_prepare(task);
2927                                 goto out_release;
2928                         }
2929                         if (calldata->arg.fmode == 0)
2930                                 break;
2931                 default:
2932                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2933                                 rpc_restart_call_prepare(task);
2934                                 goto out_release;
2935                         }
2936         }
2937         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2938                         res_stateid, calldata->arg.fmode);
2939 out_release:
2940         nfs_release_seqid(calldata->arg.seqid);
2941         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2942         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2943 }
2944
2945 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2946 {
2947         struct nfs4_closedata *calldata = data;
2948         struct nfs4_state *state = calldata->state;
2949         struct inode *inode = calldata->inode;
2950         bool is_rdonly, is_wronly, is_rdwr;
2951         int call_close = 0;
2952
2953         dprintk("%s: begin!\n", __func__);
2954         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2955                 goto out_wait;
2956
2957         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2958         spin_lock(&state->owner->so_lock);
2959         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2960         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2961         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2962         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2963         /* Calculate the change in open mode */
2964         calldata->arg.fmode = 0;
2965         if (state->n_rdwr == 0) {
2966                 if (state->n_rdonly == 0)
2967                         call_close |= is_rdonly;
2968                 else if (is_rdonly)
2969                         calldata->arg.fmode |= FMODE_READ;
2970                 if (state->n_wronly == 0)
2971                         call_close |= is_wronly;
2972                 else if (is_wronly)
2973                         calldata->arg.fmode |= FMODE_WRITE;
2974                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2975                         call_close |= is_rdwr;
2976         } else if (is_rdwr)
2977                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2978
2979         if (!nfs4_valid_open_stateid(state))
2980                 call_close = 0;
2981         spin_unlock(&state->owner->so_lock);
2982
2983         if (!call_close) {
2984                 /* Note: exit _without_ calling nfs4_close_done */
2985                 goto out_no_action;
2986         }
2987
2988         if (nfs4_wait_on_layoutreturn(inode, task)) {
2989                 nfs_release_seqid(calldata->arg.seqid);
2990                 goto out_wait;
2991         }
2992
2993         if (calldata->arg.fmode == 0)
2994                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2995         if (calldata->roc)
2996                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2997
2998         calldata->arg.share_access =
2999                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3000                                 calldata->arg.fmode, 0);
3001
3002         nfs_fattr_init(calldata->res.fattr);
3003         calldata->timestamp = jiffies;
3004         if (nfs4_setup_sequence(NFS_SERVER(inode),
3005                                 &calldata->arg.seq_args,
3006                                 &calldata->res.seq_res,
3007                                 task) != 0)
3008                 nfs_release_seqid(calldata->arg.seqid);
3009         dprintk("%s: done!\n", __func__);
3010         return;
3011 out_no_action:
3012         task->tk_action = NULL;
3013 out_wait:
3014         nfs4_sequence_done(task, &calldata->res.seq_res);
3015 }
3016
3017 static const struct rpc_call_ops nfs4_close_ops = {
3018         .rpc_call_prepare = nfs4_close_prepare,
3019         .rpc_call_done = nfs4_close_done,
3020         .rpc_release = nfs4_free_closedata,
3021 };
3022
3023 static bool nfs4_roc(struct inode *inode)
3024 {
3025         if (!nfs_have_layout(inode))
3026                 return false;
3027         return pnfs_roc(inode);
3028 }
3029
3030 /*
3031  * It is possible for data to be read/written from a mem-mapped file
3032  * after the sys_close call (which hits the vfs layer as a flush).
3033  * This means that we can't safely call nfsv4 close on a file until
3034  * the inode is cleared. This in turn means that we are not good
3035  * NFSv4 citizens - we do not indicate to the server to update the file's
3036  * share state even when we are done with one of the three share
3037  * stateid's in the inode.
3038  *
3039  * NOTE: Caller must be holding the sp->so_owner semaphore!
3040  */
3041 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3042 {
3043         struct nfs_server *server = NFS_SERVER(state->inode);
3044         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3045         struct nfs4_closedata *calldata;
3046         struct nfs4_state_owner *sp = state->owner;
3047         struct rpc_task *task;
3048         struct rpc_message msg = {
3049                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3050                 .rpc_cred = state->owner->so_cred,
3051         };
3052         struct rpc_task_setup task_setup_data = {
3053                 .rpc_client = server->client,
3054                 .rpc_message = &msg,
3055                 .callback_ops = &nfs4_close_ops,
3056                 .workqueue = nfsiod_workqueue,
3057                 .flags = RPC_TASK_ASYNC,
3058         };
3059         int status = -ENOMEM;
3060
3061         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3062                 &task_setup_data.rpc_client, &msg);
3063
3064         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3065         if (calldata == NULL)
3066                 goto out;
3067         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3068         calldata->inode = state->inode;
3069         calldata->state = state;
3070         calldata->arg.fh = NFS_FH(state->inode);
3071         /* Serialization for the sequence id */
3072         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3073         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3074         if (IS_ERR(calldata->arg.seqid))
3075                 goto out_free_calldata;
3076         calldata->arg.fmode = 0;
3077         calldata->arg.bitmask = server->cache_consistency_bitmask;
3078         calldata->res.fattr = &calldata->fattr;
3079         calldata->res.seqid = calldata->arg.seqid;
3080         calldata->res.server = server;
3081         calldata->roc = nfs4_roc(state->inode);
3082         nfs_sb_active(calldata->inode->i_sb);
3083
3084         msg.rpc_argp = &calldata->arg;
3085         msg.rpc_resp = &calldata->res;
3086         task_setup_data.callback_data = calldata;
3087         task = rpc_run_task(&task_setup_data);
3088         if (IS_ERR(task))
3089                 return PTR_ERR(task);
3090         status = 0;
3091         if (wait)
3092                 status = rpc_wait_for_completion_task(task);
3093         rpc_put_task(task);
3094         return status;
3095 out_free_calldata:
3096         kfree(calldata);
3097 out:
3098         nfs4_put_open_state(state);
3099         nfs4_put_state_owner(sp);
3100         return status;
3101 }
3102
3103 static struct inode *
3104 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3105                 int open_flags, struct iattr *attr, int *opened)
3106 {
3107         struct nfs4_state *state;
3108         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3109
3110         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3111
3112         /* Protect against concurrent sillydeletes */
3113         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3114
3115         nfs4_label_release_security(label);
3116
3117         if (IS_ERR(state))
3118                 return ERR_CAST(state);
3119         return state->inode;
3120 }
3121
3122 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3123 {
3124         if (ctx->state == NULL)
3125                 return;
3126         if (is_sync)
3127                 nfs4_close_sync(ctx->state, ctx->mode);
3128         else
3129                 nfs4_close_state(ctx->state, ctx->mode);
3130 }
3131
3132 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3133 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3134 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3135
3136 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3137 {
3138         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3139         struct nfs4_server_caps_arg args = {
3140                 .fhandle = fhandle,
3141                 .bitmask = bitmask,
3142         };
3143         struct nfs4_server_caps_res res = {};
3144         struct rpc_message msg = {
3145                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3146                 .rpc_argp = &args,
3147                 .rpc_resp = &res,
3148         };
3149         int status;
3150
3151         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3152                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3153                      FATTR4_WORD0_LINK_SUPPORT |
3154                      FATTR4_WORD0_SYMLINK_SUPPORT |
3155                      FATTR4_WORD0_ACLSUPPORT;
3156         if (minorversion)
3157                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3158
3159         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3160         if (status == 0) {
3161                 /* Sanity check the server answers */
3162                 switch (minorversion) {
3163                 case 0:
3164                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3165                         res.attr_bitmask[2] = 0;
3166                         break;
3167                 case 1:
3168                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3169                         break;
3170                 case 2:
3171                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3172                 }
3173                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3174                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3175                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3176                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3177                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3178                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3179                                 NFS_CAP_SECURITY_LABEL);
3180                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3181                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3182                         server->caps |= NFS_CAP_ACLS;
3183                 if (res.has_links != 0)
3184                         server->caps |= NFS_CAP_HARDLINKS;
3185                 if (res.has_symlinks != 0)
3186                         server->caps |= NFS_CAP_SYMLINKS;
3187                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3188                         server->caps |= NFS_CAP_FILEID;
3189                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3190                         server->caps |= NFS_CAP_MODE;
3191                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3192                         server->caps |= NFS_CAP_NLINK;
3193                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3194                         server->caps |= NFS_CAP_OWNER;
3195                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3196                         server->caps |= NFS_CAP_OWNER_GROUP;
3197                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3198                         server->caps |= NFS_CAP_ATIME;
3199                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3200                         server->caps |= NFS_CAP_CTIME;
3201                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3202                         server->caps |= NFS_CAP_MTIME;
3203 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3204                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3205                         server->caps |= NFS_CAP_SECURITY_LABEL;
3206 #endif
3207                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3208                                 sizeof(server->attr_bitmask));
3209                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3210
3211                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3212                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3213                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3214                 server->cache_consistency_bitmask[2] = 0;
3215                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3216                         sizeof(server->exclcreat_bitmask));
3217                 server->acl_bitmask = res.acl_bitmask;
3218                 server->fh_expire_type = res.fh_expire_type;
3219         }
3220
3221         return status;
3222 }
3223
3224 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3225 {
3226         struct nfs4_exception exception = { };
3227         int err;
3228         do {
3229                 err = nfs4_handle_exception(server,
3230                                 _nfs4_server_capabilities(server, fhandle),
3231                                 &exception);
3232         } while (exception.retry);
3233         return err;
3234 }
3235
3236 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3237                 struct nfs_fsinfo *info)
3238 {
3239         u32 bitmask[3];
3240         struct nfs4_lookup_root_arg args = {
3241                 .bitmask = bitmask,
3242         };
3243         struct nfs4_lookup_res res = {
3244                 .server = server,
3245                 .fattr = info->fattr,
3246                 .fh = fhandle,
3247         };
3248         struct rpc_message msg = {
3249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3250                 .rpc_argp = &args,
3251                 .rpc_resp = &res,
3252         };
3253
3254         bitmask[0] = nfs4_fattr_bitmap[0];
3255         bitmask[1] = nfs4_fattr_bitmap[1];
3256         /*
3257          * Process the label in the upcoming getfattr
3258          */
3259         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3260
3261         nfs_fattr_init(info->fattr);
3262         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3263 }
3264
3265 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3266                 struct nfs_fsinfo *info)
3267 {
3268         struct nfs4_exception exception = { };
3269         int err;
3270         do {
3271                 err = _nfs4_lookup_root(server, fhandle, info);
3272                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3273                 switch (err) {
3274                 case 0:
3275                 case -NFS4ERR_WRONGSEC:
3276                         goto out;
3277                 default:
3278                         err = nfs4_handle_exception(server, err, &exception);
3279                 }
3280         } while (exception.retry);
3281 out:
3282         return err;
3283 }
3284
3285 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3286                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3287 {
3288         struct rpc_auth_create_args auth_args = {
3289                 .pseudoflavor = flavor,
3290         };
3291         struct rpc_auth *auth;
3292         int ret;
3293
3294         auth = rpcauth_create(&auth_args, server->client);
3295         if (IS_ERR(auth)) {
3296                 ret = -EACCES;
3297                 goto out;
3298         }
3299         ret = nfs4_lookup_root(server, fhandle, info);
3300 out:
3301         return ret;
3302 }
3303
3304 /*
3305  * Retry pseudoroot lookup with various security flavors.  We do this when:
3306  *
3307  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3308  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3309  *
3310  * Returns zero on success, or a negative NFS4ERR value, or a
3311  * negative errno value.
3312  */
3313 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3314                               struct nfs_fsinfo *info)
3315 {
3316         /* Per 3530bis 15.33.5 */
3317         static const rpc_authflavor_t flav_array[] = {
3318                 RPC_AUTH_GSS_KRB5P,
3319                 RPC_AUTH_GSS_KRB5I,
3320                 RPC_AUTH_GSS_KRB5,
3321                 RPC_AUTH_UNIX,                  /* courtesy */
3322                 RPC_AUTH_NULL,
3323         };
3324         int status = -EPERM;
3325         size_t i;
3326
3327         if (server->auth_info.flavor_len > 0) {
3328                 /* try each flavor specified by user */
3329                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3330                         status = nfs4_lookup_root_sec(server, fhandle, info,
3331                                                 server->auth_info.flavors[i]);
3332                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3333                                 continue;
3334                         break;
3335                 }
3336         } else {
3337                 /* no flavors specified by user, try default list */
3338                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3339                         status = nfs4_lookup_root_sec(server, fhandle, info,
3340                                                       flav_array[i]);
3341                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3342                                 continue;
3343                         break;
3344                 }
3345         }
3346
3347         /*
3348          * -EACCESS could mean that the user doesn't have correct permissions
3349          * to access the mount.  It could also mean that we tried to mount
3350          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3351          * existing mount programs don't handle -EACCES very well so it should
3352          * be mapped to -EPERM instead.
3353          */
3354         if (status == -EACCES)
3355                 status = -EPERM;
3356         return status;
3357 }
3358
3359 /**
3360  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3361  * @server: initialized nfs_server handle
3362  * @fhandle: we fill in the pseudo-fs root file handle
3363  * @info: we fill in an FSINFO struct
3364  * @auth_probe: probe the auth flavours
3365  *
3366  * Returns zero on success, or a negative errno.
3367  */
3368 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3369                          struct nfs_fsinfo *info,
3370                          bool auth_probe)
3371 {
3372         int status = 0;
3373
3374         if (!auth_probe)
3375                 status = nfs4_lookup_root(server, fhandle, info);
3376
3377         if (auth_probe || status == NFS4ERR_WRONGSEC)
3378                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3379                                 fhandle, info);
3380
3381         if (status == 0)
3382                 status = nfs4_server_capabilities(server, fhandle);
3383         if (status == 0)
3384                 status = nfs4_do_fsinfo(server, fhandle, info);
3385
3386         return nfs4_map_errors(status);
3387 }
3388
3389 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3390                               struct nfs_fsinfo *info)
3391 {
3392         int error;
3393         struct nfs_fattr *fattr = info->fattr;
3394         struct nfs4_label *label = NULL;
3395
3396         error = nfs4_server_capabilities(server, mntfh);
3397         if (error < 0) {
3398                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3399                 return error;
3400         }
3401
3402         label = nfs4_label_alloc(server, GFP_KERNEL);
3403         if (IS_ERR(label))
3404                 return PTR_ERR(label);
3405
3406         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3407         if (error < 0) {
3408                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3409                 goto err_free_label;
3410         }
3411
3412         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3413             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3414                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3415
3416 err_free_label:
3417         nfs4_label_free(label);
3418
3419         return error;
3420 }
3421
3422 /*
3423  * Get locations and (maybe) other attributes of a referral.
3424  * Note that we'll actually follow the referral later when
3425  * we detect fsid mismatch in inode revalidation
3426  */
3427 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3428                              const struct qstr *name, struct nfs_fattr *fattr,
3429                              struct nfs_fh *fhandle)
3430 {
3431         int status = -ENOMEM;
3432         struct page *page = NULL;
3433         struct nfs4_fs_locations *locations = NULL;
3434
3435         page = alloc_page(GFP_KERNEL);
3436         if (page == NULL)
3437                 goto out;
3438         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3439         if (locations == NULL)
3440                 goto out;
3441
3442         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3443         if (status != 0)
3444                 goto out;
3445
3446         /*
3447          * If the fsid didn't change, this is a migration event, not a
3448          * referral.  Cause us to drop into the exception handler, which
3449          * will kick off migration recovery.
3450          */
3451         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3452                 dprintk("%s: server did not return a different fsid for"
3453                         " a referral at %s\n", __func__, name->name);
3454                 status = -NFS4ERR_MOVED;
3455                 goto out;
3456         }
3457         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3458         nfs_fixup_referral_attributes(&locations->fattr);
3459
3460         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3461         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3462         memset(fhandle, 0, sizeof(struct nfs_fh));
3463 out:
3464         if (page)
3465                 __free_page(page);
3466         kfree(locations);
3467         return status;
3468 }
3469
3470 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3471                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3472 {
3473         struct nfs4_getattr_arg args = {
3474                 .fh = fhandle,
3475                 .bitmask = server->attr_bitmask,
3476         };
3477         struct nfs4_getattr_res res = {
3478                 .fattr = fattr,
3479                 .label = label,
3480                 .server = server,
3481         };
3482         struct rpc_message msg = {
3483                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3484                 .rpc_argp = &args,
3485                 .rpc_resp = &res,
3486         };
3487
3488         args.bitmask = nfs4_bitmask(server, label);
3489
3490         nfs_fattr_init(fattr);
3491         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3492 }
3493
3494 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3495                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3496 {
3497         struct nfs4_exception exception = { };
3498         int err;
3499         do {
3500                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3501                 trace_nfs4_getattr(server, fhandle, fattr, err);
3502                 err = nfs4_handle_exception(server, err,
3503                                 &exception);
3504         } while (exception.retry);
3505         return err;
3506 }
3507
3508 /*
3509  * The file is not closed if it is opened due to the a request to change
3510  * the size of the file. The open call will not be needed once the
3511  * VFS layer lookup-intents are implemented.
3512  *
3513  * Close is called when the inode is destroyed.
3514  * If we haven't opened the file for O_WRONLY, we
3515  * need to in the size_change case to obtain a stateid.
3516  *
3517  * Got race?
3518  * Because OPEN is always done by name in nfsv4, it is
3519  * possible that we opened a different file by the same
3520  * name.  We can recognize this race condition, but we
3521  * can't do anything about it besides returning an error.
3522  *
3523  * This will be fixed with VFS changes (lookup-intent).
3524  */
3525 static int
3526 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3527                   struct iattr *sattr)
3528 {
3529         struct inode *inode = d_inode(dentry);
3530         struct rpc_cred *cred = NULL;
3531         struct nfs4_state *state = NULL;
3532         struct nfs4_label *label = NULL;
3533         int status;
3534
3535         if (pnfs_ld_layoutret_on_setattr(inode) &&
3536             sattr->ia_valid & ATTR_SIZE &&
3537             sattr->ia_size < i_size_read(inode))
3538                 pnfs_commit_and_return_layout(inode);
3539
3540         nfs_fattr_init(fattr);
3541
3542         /* Deal with open(O_TRUNC) */
3543         if (sattr->ia_valid & ATTR_OPEN)
3544                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3545
3546         /* Optimization: if the end result is no change, don't RPC */
3547         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3548                 return 0;
3549
3550         /* Search for an existing open(O_WRITE) file */
3551         if (sattr->ia_valid & ATTR_FILE) {
3552                 struct nfs_open_context *ctx;
3553
3554                 ctx = nfs_file_open_context(sattr->ia_file);
3555                 if (ctx) {
3556                         cred = ctx->cred;
3557                         state = ctx->state;
3558                 }
3559         }
3560
3561         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3562         if (IS_ERR(label))
3563                 return PTR_ERR(label);
3564
3565         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3566         if (status == 0) {
3567                 nfs_setattr_update_inode(inode, sattr, fattr);
3568                 nfs_setsecurity(inode, fattr, label);
3569         }
3570         nfs4_label_free(label);
3571         return status;
3572 }
3573
3574 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3575                 const struct qstr *name, struct nfs_fh *fhandle,
3576                 struct nfs_fattr *fattr, struct nfs4_label *label)
3577 {
3578         struct nfs_server *server = NFS_SERVER(dir);
3579         int                    status;
3580         struct nfs4_lookup_arg args = {
3581                 .bitmask = server->attr_bitmask,
3582                 .dir_fh = NFS_FH(dir),
3583                 .name = name,
3584         };
3585         struct nfs4_lookup_res res = {
3586                 .server = server,
3587                 .fattr = fattr,
3588                 .label = label,
3589                 .fh = fhandle,
3590         };
3591         struct rpc_message msg = {
3592                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3593                 .rpc_argp = &args,
3594                 .rpc_resp = &res,
3595         };
3596
3597         args.bitmask = nfs4_bitmask(server, label);
3598
3599         nfs_fattr_init(fattr);
3600
3601         dprintk("NFS call  lookup %s\n", name->name);
3602         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3603         dprintk("NFS reply lookup: %d\n", status);
3604         return status;
3605 }
3606
3607 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3608 {
3609         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3610                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3611         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3612         fattr->nlink = 2;
3613 }
3614
3615 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3616                                    const struct qstr *name, struct nfs_fh *fhandle,
3617                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3618 {
3619         struct nfs4_exception exception = { };
3620         struct rpc_clnt *client = *clnt;
3621         int err;
3622         do {
3623                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3624                 trace_nfs4_lookup(dir, name, err);
3625                 switch (err) {
3626                 case -NFS4ERR_BADNAME:
3627                         err = -ENOENT;
3628                         goto out;
3629                 case -NFS4ERR_MOVED:
3630                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3631                         if (err == -NFS4ERR_MOVED)
3632                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3633                         goto out;
3634                 case -NFS4ERR_WRONGSEC:
3635                         err = -EPERM;
3636                         if (client != *clnt)
3637                                 goto out;
3638                         client = nfs4_negotiate_security(client, dir, name);
3639                         if (IS_ERR(client))
3640                                 return PTR_ERR(client);
3641
3642                         exception.retry = 1;
3643                         break;
3644                 default:
3645                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3646                 }
3647         } while (exception.retry);
3648
3649 out:
3650         if (err == 0)
3651                 *clnt = client;
3652         else if (client != *clnt)
3653                 rpc_shutdown_client(client);
3654
3655         return err;
3656 }
3657
3658 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3659                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3660                             struct nfs4_label *label)
3661 {
3662         int status;
3663         struct rpc_clnt *client = NFS_CLIENT(dir);
3664
3665         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3666         if (client != NFS_CLIENT(dir)) {
3667                 rpc_shutdown_client(client);
3668                 nfs_fixup_secinfo_attributes(fattr);
3669         }
3670         return status;
3671 }
3672
3673 struct rpc_clnt *
3674 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3675                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3676 {
3677         struct rpc_clnt *client = NFS_CLIENT(dir);
3678         int status;
3679
3680         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3681         if (status < 0)
3682                 return ERR_PTR(status);
3683         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3684 }
3685
3686 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3687 {
3688         struct nfs_server *server = NFS_SERVER(inode);
3689         struct nfs4_accessargs args = {
3690                 .fh = NFS_FH(inode),
3691                 .bitmask = server->cache_consistency_bitmask,
3692         };
3693         struct nfs4_accessres res = {
3694                 .server = server,
3695         };
3696         struct rpc_message msg = {
3697                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3698                 .rpc_argp = &args,
3699                 .rpc_resp = &res,
3700                 .rpc_cred = entry->cred,
3701         };
3702         int mode = entry->mask;
3703         int status = 0;
3704
3705         /*
3706          * Determine which access bits we want to ask for...
3707          */
3708         if (mode & MAY_READ)
3709                 args.access |= NFS4_ACCESS_READ;
3710         if (S_ISDIR(inode->i_mode)) {
3711                 if (mode & MAY_WRITE)
3712                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3713                 if (mode & MAY_EXEC)
3714                         args.access |= NFS4_ACCESS_LOOKUP;
3715         } else {
3716                 if (mode & MAY_WRITE)
3717                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3718                 if (mode & MAY_EXEC)
3719                         args.access |= NFS4_ACCESS_EXECUTE;
3720         }
3721
3722         res.fattr = nfs_alloc_fattr();
3723         if (res.fattr == NULL)
3724                 return -ENOMEM;
3725
3726         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3727         if (!status) {
3728                 nfs_access_set_mask(entry, res.access);
3729                 nfs_refresh_inode(inode, res.fattr);
3730         }
3731         nfs_free_fattr(res.fattr);
3732         return status;
3733 }
3734
3735 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3736 {
3737         struct nfs4_exception exception = { };
3738         int err;
3739         do {
3740                 err = _nfs4_proc_access(inode, entry);
3741                 trace_nfs4_access(inode, err);
3742                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3743                                 &exception);
3744         } while (exception.retry);
3745         return err;
3746 }
3747
3748 /*
3749  * TODO: For the time being, we don't try to get any attributes
3750  * along with any of the zero-copy operations READ, READDIR,
3751  * READLINK, WRITE.
3752  *
3753  * In the case of the first three, we want to put the GETATTR
3754  * after the read-type operation -- this is because it is hard
3755  * to predict the length of a GETATTR response in v4, and thus
3756  * align the READ data correctly.  This means that the GETATTR
3757  * may end up partially falling into the page cache, and we should
3758  * shift it into the 'tail' of the xdr_buf before processing.
3759  * To do this efficiently, we need to know the total length
3760  * of data received, which doesn't seem to be available outside
3761  * of the RPC layer.
3762  *
3763  * In the case of WRITE, we also want to put the GETATTR after
3764  * the operation -- in this case because we want to make sure
3765  * we get the post-operation mtime and size.
3766  *
3767  * Both of these changes to the XDR layer would in fact be quite
3768  * minor, but I decided to leave them for a subsequent patch.
3769  */
3770 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3771                 unsigned int pgbase, unsigned int pglen)
3772 {
3773         struct nfs4_readlink args = {
3774                 .fh       = NFS_FH(inode),
3775                 .pgbase   = pgbase,
3776                 .pglen    = pglen,
3777                 .pages    = &page,
3778         };
3779         struct nfs4_readlink_res res;
3780         struct rpc_message msg = {
3781                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3782                 .rpc_argp = &args,
3783                 .rpc_resp = &res,
3784         };
3785
3786         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3787 }
3788
3789 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3790                 unsigned int pgbase, unsigned int pglen)
3791 {
3792         struct nfs4_exception exception = { };
3793         int err;
3794         do {
3795                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3796                 trace_nfs4_readlink(inode, err);
3797                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3798                                 &exception);
3799         } while (exception.retry);
3800         return err;
3801 }
3802
3803 /*
3804  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3805  */
3806 static int
3807 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3808                  int flags)
3809 {
3810         struct nfs4_label l, *ilabel = NULL;
3811         struct nfs_open_context *ctx;
3812         struct nfs4_state *state;
3813         int status = 0;
3814
3815         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3816         if (IS_ERR(ctx))
3817                 return PTR_ERR(ctx);
3818
3819         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3820
3821         sattr->ia_mode &= ~current_umask();
3822         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3823         if (IS_ERR(state)) {
3824                 status = PTR_ERR(state);
3825                 goto out;
3826         }
3827 out:
3828         nfs4_label_release_security(ilabel);
3829         put_nfs_open_context(ctx);
3830         return status;
3831 }
3832
3833 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3834 {
3835         struct nfs_server *server = NFS_SERVER(dir);
3836         struct nfs_removeargs args = {
3837                 .fh = NFS_FH(dir),
3838                 .name = *name,
3839         };
3840         struct nfs_removeres res = {
3841                 .server = server,
3842         };
3843         struct rpc_message msg = {
3844                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3845                 .rpc_argp = &args,
3846                 .rpc_resp = &res,
3847         };
3848         int status;
3849
3850         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3851         if (status == 0)
3852                 update_changeattr(dir, &res.cinfo);
3853         return status;
3854 }
3855
3856 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3857 {
3858         struct nfs4_exception exception = { };
3859         int err;
3860         do {
3861                 err = _nfs4_proc_remove(dir, name);
3862                 trace_nfs4_remove(dir, name, err);
3863                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3864                                 &exception);
3865         } while (exception.retry);
3866         return err;
3867 }
3868
3869 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3870 {
3871         struct nfs_server *server = NFS_SERVER(dir);
3872         struct nfs_removeargs *args = msg->rpc_argp;
3873         struct nfs_removeres *res = msg->rpc_resp;
3874
3875         res->server = server;
3876         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3877         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3878
3879         nfs_fattr_init(res->dir_attr);
3880 }
3881
3882 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3883 {
3884         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3885                         &data->args.seq_args,
3886                         &data->res.seq_res,
3887                         task);
3888 }
3889
3890 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3891 {
3892         struct nfs_unlinkdata *data = task->tk_calldata;
3893         struct nfs_removeres *res = &data->res;
3894
3895         if (!nfs4_sequence_done(task, &res->seq_res))
3896                 return 0;
3897         if (nfs4_async_handle_error(task, res->server, NULL,
3898                                     &data->timeout) == -EAGAIN)
3899                 return 0;
3900         update_changeattr(dir, &res->cinfo);
3901         return 1;
3902 }
3903
3904 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3905 {
3906         struct nfs_server *server = NFS_SERVER(dir);
3907         struct nfs_renameargs *arg = msg->rpc_argp;
3908         struct nfs_renameres *res = msg->rpc_resp;
3909
3910         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3911         res->server = server;
3912         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3913 }
3914
3915 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3916 {
3917         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3918                         &data->args.seq_args,
3919                         &data->res.seq_res,
3920                         task);
3921 }
3922
3923 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3924                                  struct inode *new_dir)
3925 {
3926         struct nfs_renamedata *data = task->tk_calldata;
3927         struct nfs_renameres *res = &data->res;
3928
3929         if (!nfs4_sequence_done(task, &res->seq_res))
3930                 return 0;
3931         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3932                 return 0;
3933
3934         update_changeattr(old_dir, &res->old_cinfo);
3935         update_changeattr(new_dir, &res->new_cinfo);
3936         return 1;
3937 }
3938
3939 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3940 {
3941         struct nfs_server *server = NFS_SERVER(inode);
3942         struct nfs4_link_arg arg = {
3943                 .fh     = NFS_FH(inode),
3944                 .dir_fh = NFS_FH(dir),
3945                 .name   = name,
3946                 .bitmask = server->attr_bitmask,
3947         };
3948         struct nfs4_link_res res = {
3949                 .server = server,
3950                 .label = NULL,
3951         };
3952         struct rpc_message msg = {
3953                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3954                 .rpc_argp = &arg,
3955                 .rpc_resp = &res,
3956         };
3957         int status = -ENOMEM;
3958
3959         res.fattr = nfs_alloc_fattr();
3960         if (res.fattr == NULL)
3961                 goto out;
3962
3963         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3964         if (IS_ERR(res.label)) {
3965                 status = PTR_ERR(res.label);
3966                 goto out;
3967         }
3968         arg.bitmask = nfs4_bitmask(server, res.label);
3969
3970         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3971         if (!status) {
3972                 update_changeattr(dir, &res.cinfo);
3973                 status = nfs_post_op_update_inode(inode, res.fattr);
3974                 if (!status)
3975                         nfs_setsecurity(inode, res.fattr, res.label);
3976         }
3977
3978
3979         nfs4_label_free(res.label);
3980
3981 out:
3982         nfs_free_fattr(res.fattr);
3983         return status;
3984 }
3985
3986 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3987 {
3988         struct nfs4_exception exception = { };
3989         int err;
3990         do {
3991                 err = nfs4_handle_exception(NFS_SERVER(inode),
3992                                 _nfs4_proc_link(inode, dir, name),
3993                                 &exception);
3994         } while (exception.retry);
3995         return err;
3996 }
3997
3998 struct nfs4_createdata {
3999         struct rpc_message msg;
4000         struct nfs4_create_arg arg;
4001         struct nfs4_create_res res;
4002         struct nfs_fh fh;
4003         struct nfs_fattr fattr;
4004         struct nfs4_label *label;
4005 };
4006
4007 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4008                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4009 {
4010         struct nfs4_createdata *data;
4011
4012         data = kzalloc(sizeof(*data), GFP_KERNEL);
4013         if (data != NULL) {
4014                 struct nfs_server *server = NFS_SERVER(dir);
4015
4016                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4017                 if (IS_ERR(data->label))
4018                         goto out_free;
4019
4020                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4021                 data->msg.rpc_argp = &data->arg;
4022                 data->msg.rpc_resp = &data->res;
4023                 data->arg.dir_fh = NFS_FH(dir);
4024                 data->arg.server = server;
4025                 data->arg.name = name;
4026                 data->arg.attrs = sattr;
4027                 data->arg.ftype = ftype;
4028                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4029                 data->res.server = server;
4030                 data->res.fh = &data->fh;
4031                 data->res.fattr = &data->fattr;
4032                 data->res.label = data->label;
4033                 nfs_fattr_init(data->res.fattr);
4034         }
4035         return data;
4036 out_free:
4037         kfree(data);
4038         return NULL;
4039 }
4040
4041 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4042 {
4043         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4044                                     &data->arg.seq_args, &data->res.seq_res, 1);
4045         if (status == 0) {
4046                 update_changeattr(dir, &data->res.dir_cinfo);
4047                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4048         }
4049         return status;
4050 }
4051
4052 static void nfs4_free_createdata(struct nfs4_createdata *data)
4053 {
4054         nfs4_label_free(data->label);
4055         kfree(data);
4056 }
4057
4058 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4059                 struct page *page, unsigned int len, struct iattr *sattr,
4060                 struct nfs4_label *label)
4061 {
4062         struct nfs4_createdata *data;
4063         int status = -ENAMETOOLONG;
4064
4065         if (len > NFS4_MAXPATHLEN)
4066                 goto out;
4067
4068         status = -ENOMEM;
4069         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4070         if (data == NULL)
4071                 goto out;
4072
4073         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4074         data->arg.u.symlink.pages = &page;
4075         data->arg.u.symlink.len = len;
4076         data->arg.label = label;
4077
4078         status = nfs4_do_create(dir, dentry, data);
4079
4080         nfs4_free_createdata(data);
4081 out:
4082         return status;
4083 }
4084
4085 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4086                 struct page *page, unsigned int len, struct iattr *sattr)
4087 {
4088         struct nfs4_exception exception = { };
4089         struct nfs4_label l, *label = NULL;
4090         int err;
4091
4092         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4093
4094         do {
4095                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4096                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4097                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4098                                 &exception);
4099         } while (exception.retry);
4100
4101         nfs4_label_release_security(label);
4102         return err;
4103 }
4104
4105 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4106                 struct iattr *sattr, struct nfs4_label *label)
4107 {
4108         struct nfs4_createdata *data;
4109         int status = -ENOMEM;
4110
4111         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4112         if (data == NULL)
4113                 goto out;
4114
4115         data->arg.label = label;
4116         status = nfs4_do_create(dir, dentry, data);
4117
4118         nfs4_free_createdata(data);
4119 out:
4120         return status;
4121 }
4122
4123 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4124                 struct iattr *sattr)
4125 {
4126         struct nfs4_exception exception = { };
4127         struct nfs4_label l, *label = NULL;
4128         int err;
4129
4130         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4131
4132         sattr->ia_mode &= ~current_umask();
4133         do {
4134                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4135                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4136                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4137                                 &exception);
4138         } while (exception.retry);
4139         nfs4_label_release_security(label);
4140
4141         return err;
4142 }
4143
4144 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4145                 u64 cookie, struct page **pages, unsigned int count, int plus)
4146 {
4147         struct inode            *dir = d_inode(dentry);
4148         struct nfs4_readdir_arg args = {
4149                 .fh = NFS_FH(dir),
4150                 .pages = pages,
4151                 .pgbase = 0,
4152                 .count = count,
4153                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4154                 .plus = plus,
4155         };
4156         struct nfs4_readdir_res res;
4157         struct rpc_message msg = {
4158                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4159                 .rpc_argp = &args,
4160                 .rpc_resp = &res,
4161                 .rpc_cred = cred,
4162         };
4163         int                     status;
4164
4165         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4166                         dentry,
4167                         (unsigned long long)cookie);
4168         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4169         res.pgbase = args.pgbase;
4170         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4171         if (status >= 0) {
4172                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4173                 status += args.pgbase;
4174         }
4175
4176         nfs_invalidate_atime(dir);
4177
4178         dprintk("%s: returns %d\n", __func__, status);
4179         return status;
4180 }
4181
4182 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4183                 u64 cookie, struct page **pages, unsigned int count, int plus)
4184 {
4185         struct nfs4_exception exception = { };
4186         int err;
4187         do {
4188                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4189                                 pages, count, plus);
4190                 trace_nfs4_readdir(d_inode(dentry), err);
4191                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4192                                 &exception);
4193         } while (exception.retry);
4194         return err;
4195 }
4196
4197 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4198                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4199 {
4200         struct nfs4_createdata *data;
4201         int mode = sattr->ia_mode;
4202         int status = -ENOMEM;
4203
4204         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4205         if (data == NULL)
4206                 goto out;
4207
4208         if (S_ISFIFO(mode))
4209                 data->arg.ftype = NF4FIFO;
4210         else if (S_ISBLK(mode)) {
4211                 data->arg.ftype = NF4BLK;
4212                 data->arg.u.device.specdata1 = MAJOR(rdev);
4213                 data->arg.u.device.specdata2 = MINOR(rdev);
4214         }
4215         else if (S_ISCHR(mode)) {
4216                 data->arg.ftype = NF4CHR;
4217                 data->arg.u.device.specdata1 = MAJOR(rdev);
4218                 data->arg.u.device.specdata2 = MINOR(rdev);
4219         } else if (!S_ISSOCK(mode)) {
4220                 status = -EINVAL;
4221                 goto out_free;
4222         }
4223
4224         data->arg.label = label;
4225         status = nfs4_do_create(dir, dentry, data);
4226 out_free:
4227         nfs4_free_createdata(data);
4228 out:
4229         return status;
4230 }
4231
4232 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4233                 struct iattr *sattr, dev_t rdev)
4234 {
4235         struct nfs4_exception exception = { };
4236         struct nfs4_label l, *label = NULL;
4237         int err;
4238
4239         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4240
4241         sattr->ia_mode &= ~current_umask();
4242         do {
4243                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4244                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4245                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4246                                 &exception);
4247         } while (exception.retry);
4248
4249         nfs4_label_release_security(label);
4250
4251         return err;
4252 }
4253
4254 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4255                  struct nfs_fsstat *fsstat)
4256 {
4257         struct nfs4_statfs_arg args = {
4258                 .fh = fhandle,
4259                 .bitmask = server->attr_bitmask,
4260         };
4261         struct nfs4_statfs_res res = {
4262                 .fsstat = fsstat,
4263         };
4264         struct rpc_message msg = {
4265                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4266                 .rpc_argp = &args,
4267                 .rpc_resp = &res,
4268         };
4269
4270         nfs_fattr_init(fsstat->fattr);
4271         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4272 }
4273
4274 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4275 {
4276         struct nfs4_exception exception = { };
4277         int err;
4278         do {
4279                 err = nfs4_handle_exception(server,
4280                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4281                                 &exception);
4282         } while (exception.retry);
4283         return err;
4284 }
4285
4286 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4287                 struct nfs_fsinfo *fsinfo)
4288 {
4289         struct nfs4_fsinfo_arg args = {
4290                 .fh = fhandle,
4291                 .bitmask = server->attr_bitmask,
4292         };
4293         struct nfs4_fsinfo_res res = {
4294                 .fsinfo = fsinfo,
4295         };
4296         struct rpc_message msg = {
4297                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4298                 .rpc_argp = &args,
4299                 .rpc_resp = &res,
4300         };
4301
4302         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4303 }
4304
4305 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4306 {
4307         struct nfs4_exception exception = { };
4308         unsigned long now = jiffies;
4309         int err;
4310
4311         do {
4312                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4313                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4314                 if (err == 0) {
4315                         nfs4_set_lease_period(server->nfs_client,
4316                                         fsinfo->lease_time * HZ,
4317                                         now);
4318                         break;
4319                 }
4320                 err = nfs4_handle_exception(server, err, &exception);
4321         } while (exception.retry);
4322         return err;
4323 }
4324
4325 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4326 {
4327         int error;
4328
4329         nfs_fattr_init(fsinfo->fattr);
4330         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4331         if (error == 0) {
4332                 /* block layout checks this! */
4333                 server->pnfs_blksize = fsinfo->blksize;
4334                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4335         }
4336
4337         return error;
4338 }
4339
4340 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4341                 struct nfs_pathconf *pathconf)
4342 {
4343         struct nfs4_pathconf_arg args = {
4344                 .fh = fhandle,
4345                 .bitmask = server->attr_bitmask,
4346         };
4347         struct nfs4_pathconf_res res = {
4348                 .pathconf = pathconf,
4349         };
4350         struct rpc_message msg = {
4351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4352                 .rpc_argp = &args,
4353                 .rpc_resp = &res,
4354         };
4355
4356         /* None of the pathconf attributes are mandatory to implement */
4357         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4358                 memset(pathconf, 0, sizeof(*pathconf));
4359                 return 0;
4360         }
4361
4362         nfs_fattr_init(pathconf->fattr);
4363         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4364 }
4365
4366 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4367                 struct nfs_pathconf *pathconf)
4368 {
4369         struct nfs4_exception exception = { };
4370         int err;
4371
4372         do {
4373                 err = nfs4_handle_exception(server,
4374                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4375                                 &exception);
4376         } while (exception.retry);
4377         return err;
4378 }
4379
4380 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4381                 const struct nfs_open_context *ctx,
4382                 const struct nfs_lock_context *l_ctx,
4383                 fmode_t fmode)
4384 {
4385         const struct nfs_lockowner *lockowner = NULL;
4386
4387         if (l_ctx != NULL)
4388                 lockowner = &l_ctx->lockowner;
4389         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4390 }
4391 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4392
4393 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4394                 const struct nfs_open_context *ctx,
4395                 const struct nfs_lock_context *l_ctx,
4396                 fmode_t fmode)
4397 {
4398         nfs4_stateid current_stateid;
4399
4400         /* If the current stateid represents a lost lock, then exit */
4401         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4402                 return true;
4403         return nfs4_stateid_match(stateid, &current_stateid);
4404 }
4405
4406 static bool nfs4_error_stateid_expired(int err)
4407 {
4408         switch (err) {
4409         case -NFS4ERR_DELEG_REVOKED:
4410         case -NFS4ERR_ADMIN_REVOKED:
4411         case -NFS4ERR_BAD_STATEID:
4412         case -NFS4ERR_STALE_STATEID:
4413         case -NFS4ERR_OLD_STATEID:
4414         case -NFS4ERR_OPENMODE:
4415         case -NFS4ERR_EXPIRED:
4416                 return true;
4417         }
4418         return false;
4419 }
4420
4421 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4422 {
4423         nfs_invalidate_atime(hdr->inode);
4424 }
4425
4426 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4427 {
4428         struct nfs_server *server = NFS_SERVER(hdr->inode);
4429
4430         trace_nfs4_read(hdr, task->tk_status);
4431         if (nfs4_async_handle_error(task, server,
4432                                     hdr->args.context->state,
4433                                     NULL) == -EAGAIN) {
4434                 rpc_restart_call_prepare(task);
4435                 return -EAGAIN;
4436         }
4437
4438         __nfs4_read_done_cb(hdr);
4439         if (task->tk_status > 0)
4440                 renew_lease(server, hdr->timestamp);
4441         return 0;
4442 }
4443
4444 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4445                 struct nfs_pgio_args *args)
4446 {
4447
4448         if (!nfs4_error_stateid_expired(task->tk_status) ||
4449                 nfs4_stateid_is_current(&args->stateid,
4450                                 args->context,
4451                                 args->lock_context,
4452                                 FMODE_READ))
4453                 return false;
4454         rpc_restart_call_prepare(task);
4455         return true;
4456 }
4457
4458 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4459 {
4460
4461         dprintk("--> %s\n", __func__);
4462
4463         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4464                 return -EAGAIN;
4465         if (nfs4_read_stateid_changed(task, &hdr->args))
4466                 return -EAGAIN;
4467         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4468                                     nfs4_read_done_cb(task, hdr);
4469 }
4470
4471 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4472                                  struct rpc_message *msg)
4473 {
4474         hdr->timestamp   = jiffies;
4475         if (!hdr->pgio_done_cb)
4476                 hdr->pgio_done_cb = nfs4_read_done_cb;
4477         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4478         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4479 }
4480
4481 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4482                                       struct nfs_pgio_header *hdr)
4483 {
4484         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4485                         &hdr->args.seq_args,
4486                         &hdr->res.seq_res,
4487                         task))
4488                 return 0;
4489         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4490                                 hdr->args.lock_context,
4491                                 hdr->rw_ops->rw_mode) == -EIO)
4492                 return -EIO;
4493         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4494                 return -EIO;
4495         return 0;
4496 }
4497
4498 static int nfs4_write_done_cb(struct rpc_task *task,
4499                               struct nfs_pgio_header *hdr)
4500 {
4501         struct inode *inode = hdr->inode;
4502
4503         trace_nfs4_write(hdr, task->tk_status);
4504         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4505                                     hdr->args.context->state,
4506                                     NULL) == -EAGAIN) {
4507                 rpc_restart_call_prepare(task);
4508                 return -EAGAIN;
4509         }
4510         if (task->tk_status >= 0) {
4511                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4512                 nfs_writeback_update_inode(hdr);
4513         }
4514         return 0;
4515 }
4516
4517 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4518                 struct nfs_pgio_args *args)
4519 {
4520
4521         if (!nfs4_error_stateid_expired(task->tk_status) ||
4522                 nfs4_stateid_is_current(&args->stateid,
4523                                 args->context,
4524                                 args->lock_context,
4525                                 FMODE_WRITE))
4526                 return false;
4527         rpc_restart_call_prepare(task);
4528         return true;
4529 }
4530
4531 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4532 {
4533         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4534                 return -EAGAIN;
4535         if (nfs4_write_stateid_changed(task, &hdr->args))
4536                 return -EAGAIN;
4537         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4538                 nfs4_write_done_cb(task, hdr);
4539 }
4540
4541 static
4542 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4543 {
4544         /* Don't request attributes for pNFS or O_DIRECT writes */
4545         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4546                 return false;
4547         /* Otherwise, request attributes if and only if we don't hold
4548          * a delegation
4549          */
4550         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4551 }
4552
4553 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4554                                   struct rpc_message *msg)
4555 {
4556         struct nfs_server *server = NFS_SERVER(hdr->inode);
4557
4558         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4559                 hdr->args.bitmask = NULL;
4560                 hdr->res.fattr = NULL;
4561         } else
4562                 hdr->args.bitmask = server->cache_consistency_bitmask;
4563
4564         if (!hdr->pgio_done_cb)
4565                 hdr->pgio_done_cb = nfs4_write_done_cb;
4566         hdr->res.server = server;
4567         hdr->timestamp   = jiffies;
4568
4569         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4570         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4571 }
4572
4573 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4574 {
4575         nfs4_setup_sequence(NFS_SERVER(data->inode),
4576                         &data->args.seq_args,
4577                         &data->res.seq_res,
4578                         task);
4579 }
4580
4581 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4582 {
4583         struct inode *inode = data->inode;
4584
4585         trace_nfs4_commit(data, task->tk_status);
4586         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4587                                     NULL, NULL) == -EAGAIN) {
4588                 rpc_restart_call_prepare(task);
4589                 return -EAGAIN;
4590         }
4591         return 0;
4592 }
4593
4594 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4595 {
4596         if (!nfs4_sequence_done(task, &data->res.seq_res))
4597                 return -EAGAIN;
4598         return data->commit_done_cb(task, data);
4599 }
4600
4601 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4602 {
4603         struct nfs_server *server = NFS_SERVER(data->inode);
4604
4605         if (data->commit_done_cb == NULL)
4606                 data->commit_done_cb = nfs4_commit_done_cb;
4607         data->res.server = server;
4608         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4609         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4610 }
4611
4612 struct nfs4_renewdata {
4613         struct nfs_client       *client;
4614         unsigned long           timestamp;
4615 };
4616
4617 /*
4618  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4619  * standalone procedure for queueing an asynchronous RENEW.
4620  */
4621 static void nfs4_renew_release(void *calldata)
4622 {
4623         struct nfs4_renewdata *data = calldata;
4624         struct nfs_client *clp = data->client;
4625
4626         if (atomic_read(&clp->cl_count) > 1)
4627                 nfs4_schedule_state_renewal(clp);
4628         nfs_put_client(clp);
4629         kfree(data);
4630 }
4631
4632 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4633 {
4634         struct nfs4_renewdata *data = calldata;
4635         struct nfs_client *clp = data->client;
4636         unsigned long timestamp = data->timestamp;
4637
4638         trace_nfs4_renew_async(clp, task->tk_status);
4639         switch (task->tk_status) {
4640         case 0:
4641                 break;
4642         case -NFS4ERR_LEASE_MOVED:
4643                 nfs4_schedule_lease_moved_recovery(clp);
4644                 break;
4645         default:
4646                 /* Unless we're shutting down, schedule state recovery! */
4647                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4648                         return;
4649                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4650                         nfs4_schedule_lease_recovery(clp);
4651                         return;
4652                 }
4653                 nfs4_schedule_path_down_recovery(clp);
4654         }
4655         do_renew_lease(clp, timestamp);
4656 }
4657
4658 static const struct rpc_call_ops nfs4_renew_ops = {
4659         .rpc_call_done = nfs4_renew_done,
4660         .rpc_release = nfs4_renew_release,
4661 };
4662
4663 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4664 {
4665         struct rpc_message msg = {
4666                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4667                 .rpc_argp       = clp,
4668                 .rpc_cred       = cred,
4669         };
4670         struct nfs4_renewdata *data;
4671
4672         if (renew_flags == 0)
4673                 return 0;
4674         if (!atomic_inc_not_zero(&clp->cl_count))
4675                 return -EIO;
4676         data = kmalloc(sizeof(*data), GFP_NOFS);
4677         if (data == NULL)
4678                 return -ENOMEM;
4679         data->client = clp;
4680         data->timestamp = jiffies;
4681         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4682                         &nfs4_renew_ops, data);
4683 }
4684
4685 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4686 {
4687         struct rpc_message msg = {
4688                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4689                 .rpc_argp       = clp,
4690                 .rpc_cred       = cred,
4691         };
4692         unsigned long now = jiffies;
4693         int status;
4694
4695         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4696         if (status < 0)
4697                 return status;
4698         do_renew_lease(clp, now);
4699         return 0;
4700 }
4701
4702 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4703 {
4704         return server->caps & NFS_CAP_ACLS;
4705 }
4706
4707 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4708  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4709  * the stack.
4710  */
4711 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4712
4713 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4714                 struct page **pages)
4715 {
4716         struct page *newpage, **spages;
4717         int rc = 0;
4718         size_t len;
4719         spages = pages;
4720
4721         do {
4722                 len = min_t(size_t, PAGE_SIZE, buflen);
4723                 newpage = alloc_page(GFP_KERNEL);
4724
4725                 if (newpage == NULL)
4726                         goto unwind;
4727                 memcpy(page_address(newpage), buf, len);
4728                 buf += len;
4729                 buflen -= len;
4730                 *pages++ = newpage;
4731                 rc++;
4732         } while (buflen != 0);
4733
4734         return rc;
4735
4736 unwind:
4737         for(; rc > 0; rc--)
4738                 __free_page(spages[rc-1]);
4739         return -ENOMEM;
4740 }
4741
4742 struct nfs4_cached_acl {
4743         int cached;
4744         size_t len;
4745         char data[0];
4746 };
4747
4748 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4749 {
4750         struct nfs_inode *nfsi = NFS_I(inode);
4751
4752         spin_lock(&inode->i_lock);
4753         kfree(nfsi->nfs4_acl);
4754         nfsi->nfs4_acl = acl;
4755         spin_unlock(&inode->i_lock);
4756 }
4757
4758 static void nfs4_zap_acl_attr(struct inode *inode)
4759 {
4760         nfs4_set_cached_acl(inode, NULL);
4761 }
4762
4763 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4764 {
4765         struct nfs_inode *nfsi = NFS_I(inode);
4766         struct nfs4_cached_acl *acl;
4767         int ret = -ENOENT;
4768
4769         spin_lock(&inode->i_lock);
4770         acl = nfsi->nfs4_acl;
4771         if (acl == NULL)
4772                 goto out;
4773         if (buf == NULL) /* user is just asking for length */
4774                 goto out_len;
4775         if (acl->cached == 0)
4776                 goto out;
4777         ret = -ERANGE; /* see getxattr(2) man page */
4778         if (acl->len > buflen)
4779                 goto out;
4780         memcpy(buf, acl->data, acl->len);
4781 out_len:
4782         ret = acl->len;
4783 out:
4784         spin_unlock(&inode->i_lock);
4785         return ret;
4786 }
4787
4788 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4789 {
4790         struct nfs4_cached_acl *acl;
4791         size_t buflen = sizeof(*acl) + acl_len;
4792
4793         if (buflen <= PAGE_SIZE) {
4794                 acl = kmalloc(buflen, GFP_KERNEL);
4795                 if (acl == NULL)
4796                         goto out;
4797                 acl->cached = 1;
4798                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4799         } else {
4800                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4801                 if (acl == NULL)
4802                         goto out;
4803                 acl->cached = 0;
4804         }
4805         acl->len = acl_len;
4806 out:
4807         nfs4_set_cached_acl(inode, acl);
4808 }
4809
4810 /*
4811  * The getxattr API returns the required buffer length when called with a
4812  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4813  * the required buf.  On a NULL buf, we send a page of data to the server
4814  * guessing that the ACL request can be serviced by a page. If so, we cache
4815  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4816  * the cache. If not so, we throw away the page, and cache the required
4817  * length. The next getxattr call will then produce another round trip to
4818  * the server, this time with the input buf of the required size.
4819  */
4820 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4821 {
4822         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4823         struct nfs_getaclargs args = {
4824                 .fh = NFS_FH(inode),
4825                 .acl_pages = pages,
4826                 .acl_len = buflen,
4827         };
4828         struct nfs_getaclres res = {
4829                 .acl_len = buflen,
4830         };
4831         struct rpc_message msg = {
4832                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4833                 .rpc_argp = &args,
4834                 .rpc_resp = &res,
4835         };
4836         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4837         int ret = -ENOMEM, i;
4838
4839         /* As long as we're doing a round trip to the server anyway,
4840          * let's be prepared for a page of acl data. */
4841         if (npages == 0)
4842                 npages = 1;
4843         if (npages > ARRAY_SIZE(pages))
4844                 return -ERANGE;
4845
4846         for (i = 0; i < npages; i++) {
4847                 pages[i] = alloc_page(GFP_KERNEL);
4848                 if (!pages[i])
4849                         goto out_free;
4850         }
4851
4852         /* for decoding across pages */
4853         res.acl_scratch = alloc_page(GFP_KERNEL);
4854         if (!res.acl_scratch)
4855                 goto out_free;
4856
4857         args.acl_len = npages * PAGE_SIZE;
4858
4859         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4860                 __func__, buf, buflen, npages, args.acl_len);
4861         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4862                              &msg, &args.seq_args, &res.seq_res, 0);
4863         if (ret)
4864                 goto out_free;
4865
4866         /* Handle the case where the passed-in buffer is too short */
4867         if (res.acl_flags & NFS4_ACL_TRUNC) {
4868                 /* Did the user only issue a request for the acl length? */
4869                 if (buf == NULL)
4870                         goto out_ok;
4871                 ret = -ERANGE;
4872                 goto out_free;
4873         }
4874         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4875         if (buf) {
4876                 if (res.acl_len > buflen) {
4877                         ret = -ERANGE;
4878                         goto out_free;
4879                 }
4880                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4881         }
4882 out_ok:
4883         ret = res.acl_len;
4884 out_free:
4885         for (i = 0; i < npages; i++)
4886                 if (pages[i])
4887                         __free_page(pages[i]);
4888         if (res.acl_scratch)
4889                 __free_page(res.acl_scratch);
4890         return ret;
4891 }
4892
4893 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4894 {
4895         struct nfs4_exception exception = { };
4896         ssize_t ret;
4897         do {
4898                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4899                 trace_nfs4_get_acl(inode, ret);
4900                 if (ret >= 0)
4901                         break;
4902                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4903         } while (exception.retry);
4904         return ret;
4905 }
4906
4907 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4908 {
4909         struct nfs_server *server = NFS_SERVER(inode);
4910         int ret;
4911
4912         if (!nfs4_server_supports_acls(server))
4913                 return -EOPNOTSUPP;
4914         ret = nfs_revalidate_inode(server, inode);
4915         if (ret < 0)
4916                 return ret;
4917         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4918                 nfs_zap_acl_cache(inode);
4919         ret = nfs4_read_cached_acl(inode, buf, buflen);
4920         if (ret != -ENOENT)
4921                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4922                  * but no cached acl data, just the acl length */
4923                 return ret;
4924         return nfs4_get_acl_uncached(inode, buf, buflen);
4925 }
4926
4927 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4928 {
4929         struct nfs_server *server = NFS_SERVER(inode);
4930         struct page *pages[NFS4ACL_MAXPAGES];
4931         struct nfs_setaclargs arg = {
4932                 .fh             = NFS_FH(inode),
4933                 .acl_pages      = pages,
4934                 .acl_len        = buflen,
4935         };
4936         struct nfs_setaclres res;
4937         struct rpc_message msg = {
4938                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4939                 .rpc_argp       = &arg,
4940                 .rpc_resp       = &res,
4941         };
4942         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4943         int ret, i;
4944
4945         if (!nfs4_server_supports_acls(server))
4946                 return -EOPNOTSUPP;
4947         if (npages > ARRAY_SIZE(pages))
4948                 return -ERANGE;
4949         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4950         if (i < 0)
4951                 return i;
4952         nfs4_inode_return_delegation(inode);
4953         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4954
4955         /*
4956          * Free each page after tx, so the only ref left is
4957          * held by the network stack
4958          */
4959         for (; i > 0; i--)
4960                 put_page(pages[i-1]);
4961
4962         /*
4963          * Acl update can result in inode attribute update.
4964          * so mark the attribute cache invalid.
4965          */
4966         spin_lock(&inode->i_lock);
4967         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4968         spin_unlock(&inode->i_lock);
4969         nfs_access_zap_cache(inode);
4970         nfs_zap_acl_cache(inode);
4971         return ret;
4972 }
4973
4974 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4975 {
4976         struct nfs4_exception exception = { };
4977         int err;
4978         do {
4979                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4980                 trace_nfs4_set_acl(inode, err);
4981                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4982                                 &exception);
4983         } while (exception.retry);
4984         return err;
4985 }
4986
4987 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4988 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4989                                         size_t buflen)
4990 {
4991         struct nfs_server *server = NFS_SERVER(inode);
4992         struct nfs_fattr fattr;
4993         struct nfs4_label label = {0, 0, buflen, buf};
4994
4995         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4996         struct nfs4_getattr_arg arg = {
4997                 .fh             = NFS_FH(inode),
4998                 .bitmask        = bitmask,
4999         };
5000         struct nfs4_getattr_res res = {
5001                 .fattr          = &fattr,
5002                 .label          = &label,
5003                 .server         = server,
5004         };
5005         struct rpc_message msg = {
5006                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5007                 .rpc_argp       = &arg,
5008                 .rpc_resp       = &res,
5009         };
5010         int ret;
5011
5012         nfs_fattr_init(&fattr);
5013
5014         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5015         if (ret)
5016                 return ret;
5017         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5018                 return -ENOENT;
5019         if (buflen < label.len)
5020                 return -ERANGE;
5021         return 0;
5022 }
5023
5024 static int nfs4_get_security_label(struct inode *inode, void *buf,
5025                                         size_t buflen)
5026 {
5027         struct nfs4_exception exception = { };
5028         int err;
5029
5030         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5031                 return -EOPNOTSUPP;
5032
5033         do {
5034                 err = _nfs4_get_security_label(inode, buf, buflen);
5035                 trace_nfs4_get_security_label(inode, err);
5036                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5037                                 &exception);
5038         } while (exception.retry);
5039         return err;
5040 }
5041
5042 static int _nfs4_do_set_security_label(struct inode *inode,
5043                 struct nfs4_label *ilabel,
5044                 struct nfs_fattr *fattr,
5045                 struct nfs4_label *olabel)
5046 {
5047
5048         struct iattr sattr = {0};
5049         struct nfs_server *server = NFS_SERVER(inode);
5050         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5051         struct nfs_setattrargs arg = {
5052                 .fh             = NFS_FH(inode),
5053                 .iap            = &sattr,
5054                 .server         = server,
5055                 .bitmask        = bitmask,
5056                 .label          = ilabel,
5057         };
5058         struct nfs_setattrres res = {
5059                 .fattr          = fattr,
5060                 .label          = olabel,
5061                 .server         = server,
5062         };
5063         struct rpc_message msg = {
5064                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5065                 .rpc_argp       = &arg,
5066                 .rpc_resp       = &res,
5067         };
5068         int status;
5069
5070         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5071
5072         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5073         if (status)
5074                 dprintk("%s failed: %d\n", __func__, status);
5075
5076         return status;
5077 }
5078
5079 static int nfs4_do_set_security_label(struct inode *inode,
5080                 struct nfs4_label *ilabel,
5081                 struct nfs_fattr *fattr,
5082                 struct nfs4_label *olabel)
5083 {
5084         struct nfs4_exception exception = { };
5085         int err;
5086
5087         do {
5088                 err = _nfs4_do_set_security_label(inode, ilabel,
5089                                 fattr, olabel);
5090                 trace_nfs4_set_security_label(inode, err);
5091                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5092                                 &exception);
5093         } while (exception.retry);
5094         return err;
5095 }
5096
5097 static int
5098 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5099 {
5100         struct nfs4_label ilabel, *olabel = NULL;
5101         struct nfs_fattr fattr;
5102         struct rpc_cred *cred;
5103         int status;
5104
5105         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5106                 return -EOPNOTSUPP;
5107
5108         nfs_fattr_init(&fattr);
5109
5110         ilabel.pi = 0;
5111         ilabel.lfs = 0;
5112         ilabel.label = (char *)buf;
5113         ilabel.len = buflen;
5114
5115         cred = rpc_lookup_cred();
5116         if (IS_ERR(cred))
5117                 return PTR_ERR(cred);
5118
5119         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5120         if (IS_ERR(olabel)) {
5121                 status = -PTR_ERR(olabel);
5122                 goto out;
5123         }
5124
5125         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5126         if (status == 0)
5127                 nfs_setsecurity(inode, &fattr, olabel);
5128
5129         nfs4_label_free(olabel);
5130 out:
5131         put_rpccred(cred);
5132         return status;
5133 }
5134 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5135
5136
5137 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5138                                     nfs4_verifier *bootverf)
5139 {
5140         __be32 verf[2];
5141
5142         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5143                 /* An impossible timestamp guarantees this value
5144                  * will never match a generated boot time. */
5145                 verf[0] = 0;
5146                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5147         } else {
5148                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5149                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5150                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5151         }
5152         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5153 }
5154
5155 static int
5156 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5157 {
5158         size_t len;
5159         char *str;
5160
5161         if (clp->cl_owner_id != NULL)
5162                 return 0;
5163
5164         rcu_read_lock();
5165         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5166                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5167                 1 +
5168                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5169                 1;
5170         rcu_read_unlock();
5171
5172         if (len > NFS4_OPAQUE_LIMIT + 1)
5173                 return -EINVAL;
5174
5175         /*
5176          * Since this string is allocated at mount time, and held until the
5177          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5178          * about a memory-reclaim deadlock.
5179          */
5180         str = kmalloc(len, GFP_KERNEL);
5181         if (!str)
5182                 return -ENOMEM;
5183
5184         rcu_read_lock();
5185         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5186                         clp->cl_ipaddr,
5187                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5188                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5189         rcu_read_unlock();
5190
5191         clp->cl_owner_id = str;
5192         return 0;
5193 }
5194
5195 static int
5196 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5197 {
5198         size_t len;
5199         char *str;
5200
5201         len = 10 + 10 + 1 + 10 + 1 +
5202                 strlen(nfs4_client_id_uniquifier) + 1 +
5203                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5204
5205         if (len > NFS4_OPAQUE_LIMIT + 1)
5206                 return -EINVAL;
5207
5208         /*
5209          * Since this string is allocated at mount time, and held until the
5210          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5211          * about a memory-reclaim deadlock.
5212          */
5213         str = kmalloc(len, GFP_KERNEL);
5214         if (!str)
5215                 return -ENOMEM;
5216
5217         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5218                         clp->rpc_ops->version, clp->cl_minorversion,
5219                         nfs4_client_id_uniquifier,
5220                         clp->cl_rpcclient->cl_nodename);
5221         clp->cl_owner_id = str;
5222         return 0;
5223 }
5224
5225 static int
5226 nfs4_init_uniform_client_string(struct nfs_client *clp)
5227 {
5228         size_t len;
5229         char *str;
5230
5231         if (clp->cl_owner_id != NULL)
5232                 return 0;
5233
5234         if (nfs4_client_id_uniquifier[0] != '\0')
5235                 return nfs4_init_uniquifier_client_string(clp);
5236
5237         len = 10 + 10 + 1 + 10 + 1 +
5238                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5239
5240         if (len > NFS4_OPAQUE_LIMIT + 1)
5241                 return -EINVAL;
5242
5243         /*
5244          * Since this string is allocated at mount time, and held until the
5245          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5246          * about a memory-reclaim deadlock.
5247          */
5248         str = kmalloc(len, GFP_KERNEL);
5249         if (!str)
5250                 return -ENOMEM;
5251
5252         scnprintf(str, len, "Linux NFSv%u.%u %s",
5253                         clp->rpc_ops->version, clp->cl_minorversion,
5254                         clp->cl_rpcclient->cl_nodename);
5255         clp->cl_owner_id = str;
5256         return 0;
5257 }
5258
5259 /*
5260  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5261  * services.  Advertise one based on the address family of the
5262  * clientaddr.
5263  */
5264 static unsigned int
5265 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5266 {
5267         if (strchr(clp->cl_ipaddr, ':') != NULL)
5268                 return scnprintf(buf, len, "tcp6");
5269         else
5270                 return scnprintf(buf, len, "tcp");
5271 }
5272
5273 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5274 {
5275         struct nfs4_setclientid *sc = calldata;
5276
5277         if (task->tk_status == 0)
5278                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5279 }
5280
5281 static const struct rpc_call_ops nfs4_setclientid_ops = {
5282         .rpc_call_done = nfs4_setclientid_done,
5283 };
5284
5285 /**
5286  * nfs4_proc_setclientid - Negotiate client ID
5287  * @clp: state data structure
5288  * @program: RPC program for NFSv4 callback service
5289  * @port: IP port number for NFS4 callback service
5290  * @cred: RPC credential to use for this call
5291  * @res: where to place the result
5292  *
5293  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5294  */
5295 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5296                 unsigned short port, struct rpc_cred *cred,
5297                 struct nfs4_setclientid_res *res)
5298 {
5299         nfs4_verifier sc_verifier;
5300         struct nfs4_setclientid setclientid = {
5301                 .sc_verifier = &sc_verifier,
5302                 .sc_prog = program,
5303                 .sc_clnt = clp,
5304         };
5305         struct rpc_message msg = {
5306                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5307                 .rpc_argp = &setclientid,
5308                 .rpc_resp = res,
5309                 .rpc_cred = cred,
5310         };
5311         struct rpc_task *task;
5312         struct rpc_task_setup task_setup_data = {
5313                 .rpc_client = clp->cl_rpcclient,
5314                 .rpc_message = &msg,
5315                 .callback_ops = &nfs4_setclientid_ops,
5316                 .callback_data = &setclientid,
5317                 .flags = RPC_TASK_TIMEOUT,
5318         };
5319         int status;
5320
5321         /* nfs_client_id4 */
5322         nfs4_init_boot_verifier(clp, &sc_verifier);
5323
5324         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5325                 status = nfs4_init_uniform_client_string(clp);
5326         else
5327                 status = nfs4_init_nonuniform_client_string(clp);
5328
5329         if (status)
5330                 goto out;
5331
5332         /* cb_client4 */
5333         setclientid.sc_netid_len =
5334                                 nfs4_init_callback_netid(clp,
5335                                                 setclientid.sc_netid,
5336                                                 sizeof(setclientid.sc_netid));
5337         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5338                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5339                                 clp->cl_ipaddr, port >> 8, port & 255);
5340
5341         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5342                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5343                 clp->cl_owner_id);
5344         task = rpc_run_task(&task_setup_data);
5345         if (IS_ERR(task)) {
5346                 status = PTR_ERR(task);
5347                 goto out;
5348         }
5349         status = task->tk_status;
5350         if (setclientid.sc_cred) {
5351                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5352                 put_rpccred(setclientid.sc_cred);
5353         }
5354         rpc_put_task(task);
5355 out:
5356         trace_nfs4_setclientid(clp, status);
5357         dprintk("NFS reply setclientid: %d\n", status);
5358         return status;
5359 }
5360
5361 /**
5362  * nfs4_proc_setclientid_confirm - Confirm client ID
5363  * @clp: state data structure
5364  * @res: result of a previous SETCLIENTID
5365  * @cred: RPC credential to use for this call
5366  *
5367  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5368  */
5369 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5370                 struct nfs4_setclientid_res *arg,
5371                 struct rpc_cred *cred)
5372 {
5373         struct rpc_message msg = {
5374                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5375                 .rpc_argp = arg,
5376                 .rpc_cred = cred,
5377         };
5378         int status;
5379
5380         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5381                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5382                 clp->cl_clientid);
5383         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5384         trace_nfs4_setclientid_confirm(clp, status);
5385         dprintk("NFS reply setclientid_confirm: %d\n", status);
5386         return status;
5387 }
5388
5389 struct nfs4_delegreturndata {
5390         struct nfs4_delegreturnargs args;
5391         struct nfs4_delegreturnres res;
5392         struct nfs_fh fh;
5393         nfs4_stateid stateid;
5394         unsigned long timestamp;
5395         struct nfs_fattr fattr;
5396         int rpc_status;
5397         struct inode *inode;
5398         bool roc;
5399         u32 roc_barrier;
5400 };
5401
5402 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5403 {
5404         struct nfs4_delegreturndata *data = calldata;
5405
5406         if (!nfs4_sequence_done(task, &data->res.seq_res))
5407                 return;
5408
5409         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5410         switch (task->tk_status) {
5411         case 0:
5412                 renew_lease(data->res.server, data->timestamp);
5413         case -NFS4ERR_ADMIN_REVOKED:
5414         case -NFS4ERR_DELEG_REVOKED:
5415         case -NFS4ERR_BAD_STATEID:
5416         case -NFS4ERR_OLD_STATEID:
5417         case -NFS4ERR_STALE_STATEID:
5418         case -NFS4ERR_EXPIRED:
5419                 task->tk_status = 0;
5420                 if (data->roc)
5421                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5422                 break;
5423         default:
5424                 if (nfs4_async_handle_error(task, data->res.server,
5425                                             NULL, NULL) == -EAGAIN) {
5426                         rpc_restart_call_prepare(task);
5427                         return;
5428                 }
5429         }
5430         data->rpc_status = task->tk_status;
5431 }
5432
5433 static void nfs4_delegreturn_release(void *calldata)
5434 {
5435         struct nfs4_delegreturndata *data = calldata;
5436         struct inode *inode = data->inode;
5437
5438         if (inode) {
5439                 if (data->roc)
5440                         pnfs_roc_release(inode);
5441                 nfs_iput_and_deactive(inode);
5442         }
5443         kfree(calldata);
5444 }
5445
5446 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5447 {
5448         struct nfs4_delegreturndata *d_data;
5449
5450         d_data = (struct nfs4_delegreturndata *)data;
5451
5452         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5453                 return;
5454
5455         if (d_data->roc)
5456                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5457
5458         nfs4_setup_sequence(d_data->res.server,
5459                         &d_data->args.seq_args,
5460                         &d_data->res.seq_res,
5461                         task);
5462 }
5463
5464 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5465         .rpc_call_prepare = nfs4_delegreturn_prepare,
5466         .rpc_call_done = nfs4_delegreturn_done,
5467         .rpc_release = nfs4_delegreturn_release,
5468 };
5469
5470 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5471 {
5472         struct nfs4_delegreturndata *data;
5473         struct nfs_server *server = NFS_SERVER(inode);
5474         struct rpc_task *task;
5475         struct rpc_message msg = {
5476                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5477                 .rpc_cred = cred,
5478         };
5479         struct rpc_task_setup task_setup_data = {
5480                 .rpc_client = server->client,
5481                 .rpc_message = &msg,
5482                 .callback_ops = &nfs4_delegreturn_ops,
5483                 .flags = RPC_TASK_ASYNC,
5484         };
5485         int status = 0;
5486
5487         data = kzalloc(sizeof(*data), GFP_NOFS);
5488         if (data == NULL)
5489                 return -ENOMEM;
5490         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5491
5492         nfs4_state_protect(server->nfs_client,
5493                         NFS_SP4_MACH_CRED_CLEANUP,
5494                         &task_setup_data.rpc_client, &msg);
5495
5496         data->args.fhandle = &data->fh;
5497         data->args.stateid = &data->stateid;
5498         data->args.bitmask = server->cache_consistency_bitmask;
5499         nfs_copy_fh(&data->fh, NFS_FH(inode));
5500         nfs4_stateid_copy(&data->stateid, stateid);
5501         data->res.fattr = &data->fattr;
5502         data->res.server = server;
5503         nfs_fattr_init(data->res.fattr);
5504         data->timestamp = jiffies;
5505         data->rpc_status = 0;
5506         data->inode = nfs_igrab_and_active(inode);
5507         if (data->inode)
5508                 data->roc = nfs4_roc(inode);
5509
5510         task_setup_data.callback_data = data;
5511         msg.rpc_argp = &data->args;
5512         msg.rpc_resp = &data->res;
5513         task = rpc_run_task(&task_setup_data);
5514         if (IS_ERR(task))
5515                 return PTR_ERR(task);
5516         if (!issync)
5517                 goto out;
5518         status = nfs4_wait_for_completion_rpc_task(task);
5519         if (status != 0)
5520                 goto out;
5521         status = data->rpc_status;
5522         if (status == 0)
5523                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5524         else
5525                 nfs_refresh_inode(inode, &data->fattr);
5526 out:
5527         rpc_put_task(task);
5528         return status;
5529 }
5530
5531 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5532 {
5533         struct nfs_server *server = NFS_SERVER(inode);
5534         struct nfs4_exception exception = { };
5535         int err;
5536         do {
5537                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5538                 trace_nfs4_delegreturn(inode, stateid, err);
5539                 switch (err) {
5540                         case -NFS4ERR_STALE_STATEID:
5541                         case -NFS4ERR_EXPIRED:
5542                         case 0:
5543                                 return 0;
5544                 }
5545                 err = nfs4_handle_exception(server, err, &exception);
5546         } while (exception.retry);
5547         return err;
5548 }
5549
5550 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5551 {
5552         struct inode *inode = state->inode;
5553         struct nfs_server *server = NFS_SERVER(inode);
5554         struct nfs_client *clp = server->nfs_client;
5555         struct nfs_lockt_args arg = {
5556                 .fh = NFS_FH(inode),
5557                 .fl = request,
5558         };
5559         struct nfs_lockt_res res = {
5560                 .denied = request,
5561         };
5562         struct rpc_message msg = {
5563                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5564                 .rpc_argp       = &arg,
5565                 .rpc_resp       = &res,
5566                 .rpc_cred       = state->owner->so_cred,
5567         };
5568         struct nfs4_lock_state *lsp;
5569         int status;
5570
5571         arg.lock_owner.clientid = clp->cl_clientid;
5572         status = nfs4_set_lock_state(state, request);
5573         if (status != 0)
5574                 goto out;
5575         lsp = request->fl_u.nfs4_fl.owner;
5576         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5577         arg.lock_owner.s_dev = server->s_dev;
5578         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5579         switch (status) {
5580                 case 0:
5581                         request->fl_type = F_UNLCK;
5582                         break;
5583                 case -NFS4ERR_DENIED:
5584                         status = 0;
5585         }
5586         request->fl_ops->fl_release_private(request);
5587         request->fl_ops = NULL;
5588 out:
5589         return status;
5590 }
5591
5592 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5593 {
5594         struct nfs4_exception exception = { };
5595         int err;
5596
5597         do {
5598                 err = _nfs4_proc_getlk(state, cmd, request);
5599                 trace_nfs4_get_lock(request, state, cmd, err);
5600                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5601                                 &exception);
5602         } while (exception.retry);
5603         return err;
5604 }
5605
5606 struct nfs4_unlockdata {
5607         struct nfs_locku_args arg;
5608         struct nfs_locku_res res;
5609         struct nfs4_lock_state *lsp;
5610         struct nfs_open_context *ctx;
5611         struct file_lock fl;
5612         struct nfs_server *server;
5613         unsigned long timestamp;
5614 };
5615
5616 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5617                 struct nfs_open_context *ctx,
5618                 struct nfs4_lock_state *lsp,
5619                 struct nfs_seqid *seqid)
5620 {
5621         struct nfs4_unlockdata *p;
5622         struct inode *inode = lsp->ls_state->inode;
5623
5624         p = kzalloc(sizeof(*p), GFP_NOFS);
5625         if (p == NULL)
5626                 return NULL;
5627         p->arg.fh = NFS_FH(inode);
5628         p->arg.fl = &p->fl;
5629         p->arg.seqid = seqid;
5630         p->res.seqid = seqid;
5631         p->lsp = lsp;
5632         atomic_inc(&lsp->ls_count);
5633         /* Ensure we don't close file until we're done freeing locks! */
5634         p->ctx = get_nfs_open_context(ctx);
5635         memcpy(&p->fl, fl, sizeof(p->fl));
5636         p->server = NFS_SERVER(inode);
5637         return p;
5638 }
5639
5640 static void nfs4_locku_release_calldata(void *data)
5641 {
5642         struct nfs4_unlockdata *calldata = data;
5643         nfs_free_seqid(calldata->arg.seqid);
5644         nfs4_put_lock_state(calldata->lsp);
5645         put_nfs_open_context(calldata->ctx);
5646         kfree(calldata);
5647 }
5648
5649 static void nfs4_locku_done(struct rpc_task *task, void *data)
5650 {
5651         struct nfs4_unlockdata *calldata = data;
5652
5653         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5654                 return;
5655         switch (task->tk_status) {
5656                 case 0:
5657                         renew_lease(calldata->server, calldata->timestamp);
5658                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5659                         if (nfs4_update_lock_stateid(calldata->lsp,
5660                                         &calldata->res.stateid))
5661                                 break;
5662                 case -NFS4ERR_BAD_STATEID:
5663                 case -NFS4ERR_OLD_STATEID:
5664                 case -NFS4ERR_STALE_STATEID:
5665                 case -NFS4ERR_EXPIRED:
5666                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5667                                                 &calldata->lsp->ls_stateid))
5668                                 rpc_restart_call_prepare(task);
5669                         break;
5670                 default:
5671                         if (nfs4_async_handle_error(task, calldata->server,
5672                                                     NULL, NULL) == -EAGAIN)
5673                                 rpc_restart_call_prepare(task);
5674         }
5675         nfs_release_seqid(calldata->arg.seqid);
5676 }
5677
5678 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5679 {
5680         struct nfs4_unlockdata *calldata = data;
5681
5682         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5683                 goto out_wait;
5684         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5685         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5686                 /* Note: exit _without_ running nfs4_locku_done */
5687                 goto out_no_action;
5688         }
5689         calldata->timestamp = jiffies;
5690         if (nfs4_setup_sequence(calldata->server,
5691                                 &calldata->arg.seq_args,
5692                                 &calldata->res.seq_res,
5693                                 task) != 0)
5694                 nfs_release_seqid(calldata->arg.seqid);
5695         return;
5696 out_no_action:
5697         task->tk_action = NULL;
5698 out_wait:
5699         nfs4_sequence_done(task, &calldata->res.seq_res);
5700 }
5701
5702 static const struct rpc_call_ops nfs4_locku_ops = {
5703         .rpc_call_prepare = nfs4_locku_prepare,
5704         .rpc_call_done = nfs4_locku_done,
5705         .rpc_release = nfs4_locku_release_calldata,
5706 };
5707
5708 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5709                 struct nfs_open_context *ctx,
5710                 struct nfs4_lock_state *lsp,
5711                 struct nfs_seqid *seqid)
5712 {
5713         struct nfs4_unlockdata *data;
5714         struct rpc_message msg = {
5715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5716                 .rpc_cred = ctx->cred,
5717         };
5718         struct rpc_task_setup task_setup_data = {
5719                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5720                 .rpc_message = &msg,
5721                 .callback_ops = &nfs4_locku_ops,
5722                 .workqueue = nfsiod_workqueue,
5723                 .flags = RPC_TASK_ASYNC,
5724         };
5725
5726         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5727                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5728
5729         /* Ensure this is an unlock - when canceling a lock, the
5730          * canceled lock is passed in, and it won't be an unlock.
5731          */
5732         fl->fl_type = F_UNLCK;
5733
5734         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5735         if (data == NULL) {
5736                 nfs_free_seqid(seqid);
5737                 return ERR_PTR(-ENOMEM);
5738         }
5739
5740         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5741         msg.rpc_argp = &data->arg;
5742         msg.rpc_resp = &data->res;
5743         task_setup_data.callback_data = data;
5744         return rpc_run_task(&task_setup_data);
5745 }
5746
5747 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5748 {
5749         struct inode *inode = state->inode;
5750         struct nfs4_state_owner *sp = state->owner;
5751         struct nfs_inode *nfsi = NFS_I(inode);
5752         struct nfs_seqid *seqid;
5753         struct nfs4_lock_state *lsp;
5754         struct rpc_task *task;
5755         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5756         int status = 0;
5757         unsigned char fl_flags = request->fl_flags;
5758
5759         status = nfs4_set_lock_state(state, request);
5760         /* Unlock _before_ we do the RPC call */
5761         request->fl_flags |= FL_EXISTS;
5762         /* Exclude nfs_delegation_claim_locks() */
5763         mutex_lock(&sp->so_delegreturn_mutex);
5764         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5765         down_read(&nfsi->rwsem);
5766         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
5767                 up_read(&nfsi->rwsem);
5768                 mutex_unlock(&sp->so_delegreturn_mutex);
5769                 goto out;
5770         }
5771         up_read(&nfsi->rwsem);
5772         mutex_unlock(&sp->so_delegreturn_mutex);
5773         if (status != 0)
5774                 goto out;
5775         /* Is this a delegated lock? */
5776         lsp = request->fl_u.nfs4_fl.owner;
5777         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5778                 goto out;
5779         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5780         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5781         status = -ENOMEM;
5782         if (IS_ERR(seqid))
5783                 goto out;
5784         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5785         status = PTR_ERR(task);
5786         if (IS_ERR(task))
5787                 goto out;
5788         status = nfs4_wait_for_completion_rpc_task(task);
5789         rpc_put_task(task);
5790 out:
5791         request->fl_flags = fl_flags;
5792         trace_nfs4_unlock(request, state, F_SETLK, status);
5793         return status;
5794 }
5795
5796 struct nfs4_lockdata {
5797         struct nfs_lock_args arg;
5798         struct nfs_lock_res res;
5799         struct nfs4_lock_state *lsp;
5800         struct nfs_open_context *ctx;
5801         struct file_lock fl;
5802         unsigned long timestamp;
5803         int rpc_status;
5804         int cancelled;
5805         struct nfs_server *server;
5806 };
5807
5808 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5809                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5810                 gfp_t gfp_mask)
5811 {
5812         struct nfs4_lockdata *p;
5813         struct inode *inode = lsp->ls_state->inode;
5814         struct nfs_server *server = NFS_SERVER(inode);
5815         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5816
5817         p = kzalloc(sizeof(*p), gfp_mask);
5818         if (p == NULL)
5819                 return NULL;
5820
5821         p->arg.fh = NFS_FH(inode);
5822         p->arg.fl = &p->fl;
5823         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5824         if (IS_ERR(p->arg.open_seqid))
5825                 goto out_free;
5826         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5827         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5828         if (IS_ERR(p->arg.lock_seqid))
5829                 goto out_free_seqid;
5830         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5831         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5832         p->arg.lock_owner.s_dev = server->s_dev;
5833         p->res.lock_seqid = p->arg.lock_seqid;
5834         p->lsp = lsp;
5835         p->server = server;
5836         atomic_inc(&lsp->ls_count);
5837         p->ctx = get_nfs_open_context(ctx);
5838         get_file(fl->fl_file);
5839         memcpy(&p->fl, fl, sizeof(p->fl));
5840         return p;
5841 out_free_seqid:
5842         nfs_free_seqid(p->arg.open_seqid);
5843 out_free:
5844         kfree(p);
5845         return NULL;
5846 }
5847
5848 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5849 {
5850         struct nfs4_lockdata *data = calldata;
5851         struct nfs4_state *state = data->lsp->ls_state;
5852
5853         dprintk("%s: begin!\n", __func__);
5854         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5855                 goto out_wait;
5856         /* Do we need to do an open_to_lock_owner? */
5857         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5858                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5859                         goto out_release_lock_seqid;
5860                 }
5861                 nfs4_stateid_copy(&data->arg.open_stateid,
5862                                 &state->open_stateid);
5863                 data->arg.new_lock_owner = 1;
5864                 data->res.open_seqid = data->arg.open_seqid;
5865         } else {
5866                 data->arg.new_lock_owner = 0;
5867                 nfs4_stateid_copy(&data->arg.lock_stateid,
5868                                 &data->lsp->ls_stateid);
5869         }
5870         if (!nfs4_valid_open_stateid(state)) {
5871                 data->rpc_status = -EBADF;
5872                 task->tk_action = NULL;
5873                 goto out_release_open_seqid;
5874         }
5875         data->timestamp = jiffies;
5876         if (nfs4_setup_sequence(data->server,
5877                                 &data->arg.seq_args,
5878                                 &data->res.seq_res,
5879                                 task) == 0)
5880                 return;
5881 out_release_open_seqid:
5882         nfs_release_seqid(data->arg.open_seqid);
5883 out_release_lock_seqid:
5884         nfs_release_seqid(data->arg.lock_seqid);
5885 out_wait:
5886         nfs4_sequence_done(task, &data->res.seq_res);
5887         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5888 }
5889
5890 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5891 {
5892         struct nfs4_lockdata *data = calldata;
5893         struct nfs4_lock_state *lsp = data->lsp;
5894
5895         dprintk("%s: begin!\n", __func__);
5896
5897         if (!nfs4_sequence_done(task, &data->res.seq_res))
5898                 return;
5899
5900         data->rpc_status = task->tk_status;
5901         switch (task->tk_status) {
5902         case 0:
5903                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5904                                 data->timestamp);
5905                 if (data->arg.new_lock) {
5906                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5907                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
5908                                 rpc_restart_call_prepare(task);
5909                                 break;
5910                         }
5911                 }
5912                 if (data->arg.new_lock_owner != 0) {
5913                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5914                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5915                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5916                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5917                         rpc_restart_call_prepare(task);
5918                 break;
5919         case -NFS4ERR_BAD_STATEID:
5920         case -NFS4ERR_OLD_STATEID:
5921         case -NFS4ERR_STALE_STATEID:
5922         case -NFS4ERR_EXPIRED:
5923                 if (data->arg.new_lock_owner != 0) {
5924                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5925                                                 &lsp->ls_state->open_stateid))
5926                                 rpc_restart_call_prepare(task);
5927                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5928                                                 &lsp->ls_stateid))
5929                                 rpc_restart_call_prepare(task);
5930         }
5931         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5932 }
5933
5934 static void nfs4_lock_release(void *calldata)
5935 {
5936         struct nfs4_lockdata *data = calldata;
5937
5938         dprintk("%s: begin!\n", __func__);
5939         nfs_free_seqid(data->arg.open_seqid);
5940         if (data->cancelled != 0) {
5941                 struct rpc_task *task;
5942                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5943                                 data->arg.lock_seqid);
5944                 if (!IS_ERR(task))
5945                         rpc_put_task_async(task);
5946                 dprintk("%s: cancelling lock!\n", __func__);
5947         } else
5948                 nfs_free_seqid(data->arg.lock_seqid);
5949         nfs4_put_lock_state(data->lsp);
5950         put_nfs_open_context(data->ctx);
5951         fput(data->fl.fl_file);
5952         kfree(data);
5953         dprintk("%s: done!\n", __func__);
5954 }
5955
5956 static const struct rpc_call_ops nfs4_lock_ops = {
5957         .rpc_call_prepare = nfs4_lock_prepare,
5958         .rpc_call_done = nfs4_lock_done,
5959         .rpc_release = nfs4_lock_release,
5960 };
5961
5962 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5963 {
5964         switch (error) {
5965         case -NFS4ERR_ADMIN_REVOKED:
5966         case -NFS4ERR_BAD_STATEID:
5967                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5968                 if (new_lock_owner != 0 ||
5969                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5970                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5971                 break;
5972         case -NFS4ERR_STALE_STATEID:
5973                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5974         case -NFS4ERR_EXPIRED:
5975                 nfs4_schedule_lease_recovery(server->nfs_client);
5976         };
5977 }
5978
5979 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5980 {
5981         struct nfs4_lockdata *data;
5982         struct rpc_task *task;
5983         struct rpc_message msg = {
5984                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5985                 .rpc_cred = state->owner->so_cred,
5986         };
5987         struct rpc_task_setup task_setup_data = {
5988                 .rpc_client = NFS_CLIENT(state->inode),
5989                 .rpc_message = &msg,
5990                 .callback_ops = &nfs4_lock_ops,
5991                 .workqueue = nfsiod_workqueue,
5992                 .flags = RPC_TASK_ASYNC,
5993         };
5994         int ret;
5995
5996         dprintk("%s: begin!\n", __func__);
5997         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5998                         fl->fl_u.nfs4_fl.owner,
5999                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6000         if (data == NULL)
6001                 return -ENOMEM;
6002         if (IS_SETLKW(cmd))
6003                 data->arg.block = 1;
6004         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6005         msg.rpc_argp = &data->arg;
6006         msg.rpc_resp = &data->res;
6007         task_setup_data.callback_data = data;
6008         if (recovery_type > NFS_LOCK_NEW) {
6009                 if (recovery_type == NFS_LOCK_RECLAIM)
6010                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6011                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6012         } else
6013                 data->arg.new_lock = 1;
6014         task = rpc_run_task(&task_setup_data);
6015         if (IS_ERR(task))
6016                 return PTR_ERR(task);
6017         ret = nfs4_wait_for_completion_rpc_task(task);
6018         if (ret == 0) {
6019                 ret = data->rpc_status;
6020                 if (ret)
6021                         nfs4_handle_setlk_error(data->server, data->lsp,
6022                                         data->arg.new_lock_owner, ret);
6023         } else
6024                 data->cancelled = 1;
6025         rpc_put_task(task);
6026         dprintk("%s: done, ret = %d!\n", __func__, ret);
6027         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6028         return ret;
6029 }
6030
6031 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6032 {
6033         struct nfs_server *server = NFS_SERVER(state->inode);
6034         struct nfs4_exception exception = {
6035                 .inode = state->inode,
6036         };
6037         int err;
6038
6039         do {
6040                 /* Cache the lock if possible... */
6041                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6042                         return 0;
6043                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6044                 if (err != -NFS4ERR_DELAY)
6045                         break;
6046                 nfs4_handle_exception(server, err, &exception);
6047         } while (exception.retry);
6048         return err;
6049 }
6050
6051 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6052 {
6053         struct nfs_server *server = NFS_SERVER(state->inode);
6054         struct nfs4_exception exception = {
6055                 .inode = state->inode,
6056         };
6057         int err;
6058
6059         err = nfs4_set_lock_state(state, request);
6060         if (err != 0)
6061                 return err;
6062         if (!recover_lost_locks) {
6063                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6064                 return 0;
6065         }
6066         do {
6067                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6068                         return 0;
6069                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6070                 switch (err) {
6071                 default:
6072                         goto out;
6073                 case -NFS4ERR_GRACE:
6074                 case -NFS4ERR_DELAY:
6075                         nfs4_handle_exception(server, err, &exception);
6076                         err = 0;
6077                 }
6078         } while (exception.retry);
6079 out:
6080         return err;
6081 }
6082
6083 #if defined(CONFIG_NFS_V4_1)
6084 /**
6085  * nfs41_check_expired_locks - possibly free a lock stateid
6086  *
6087  * @state: NFSv4 state for an inode
6088  *
6089  * Returns NFS_OK if recovery for this stateid is now finished.
6090  * Otherwise a negative NFS4ERR value is returned.
6091  */
6092 static int nfs41_check_expired_locks(struct nfs4_state *state)
6093 {
6094         int status, ret = -NFS4ERR_BAD_STATEID;
6095         struct nfs4_lock_state *lsp;
6096         struct nfs_server *server = NFS_SERVER(state->inode);
6097
6098         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6099                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6100                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6101
6102                         status = nfs41_test_and_free_expired_stateid(server,
6103                                         &lsp->ls_stateid,
6104                                         cred);
6105                         trace_nfs4_test_lock_stateid(state, lsp, status);
6106                         if (status != NFS_OK) {
6107                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6108                                 ret = status;
6109                         }
6110                 }
6111         };
6112
6113         return ret;
6114 }
6115
6116 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6117 {
6118         int status = NFS_OK;
6119
6120         if (test_bit(LK_STATE_IN_USE, &state->flags))
6121                 status = nfs41_check_expired_locks(state);
6122         if (status != NFS_OK)
6123                 status = nfs4_lock_expired(state, request);
6124         return status;
6125 }
6126 #endif
6127
6128 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6129 {
6130         struct nfs_inode *nfsi = NFS_I(state->inode);
6131         struct nfs4_state_owner *sp = state->owner;
6132         unsigned char fl_flags = request->fl_flags;
6133         int status;
6134
6135         request->fl_flags |= FL_ACCESS;
6136         status = locks_lock_inode_wait(state->inode, request);
6137         if (status < 0)
6138                 goto out;
6139         mutex_lock(&sp->so_delegreturn_mutex);
6140         down_read(&nfsi->rwsem);
6141         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6142                 /* Yes: cache locks! */
6143                 /* ...but avoid races with delegation recall... */
6144                 request->fl_flags = fl_flags & ~FL_SLEEP;
6145                 status = locks_lock_inode_wait(state->inode, request);
6146                 up_read(&nfsi->rwsem);
6147                 mutex_unlock(&sp->so_delegreturn_mutex);
6148                 goto out;
6149         }
6150         up_read(&nfsi->rwsem);
6151         mutex_unlock(&sp->so_delegreturn_mutex);
6152         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6153 out:
6154         request->fl_flags = fl_flags;
6155         return status;
6156 }
6157
6158 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6159 {
6160         struct nfs4_exception exception = {
6161                 .state = state,
6162                 .inode = state->inode,
6163         };
6164         int err;
6165
6166         do {
6167                 err = _nfs4_proc_setlk(state, cmd, request);
6168                 if (err == -NFS4ERR_DENIED)
6169                         err = -EAGAIN;
6170                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6171                                 err, &exception);
6172         } while (exception.retry);
6173         return err;
6174 }
6175
6176 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6177 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6178
6179 static int
6180 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6181                         struct file_lock *request)
6182 {
6183         int             status = -ERESTARTSYS;
6184         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6185
6186         while(!signalled()) {
6187                 status = nfs4_proc_setlk(state, cmd, request);
6188                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6189                         break;
6190                 freezable_schedule_timeout_interruptible(timeout);
6191                 timeout *= 2;
6192                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6193                 status = -ERESTARTSYS;
6194         }
6195         return status;
6196 }
6197
6198 #ifdef CONFIG_NFS_V4_1
6199 struct nfs4_lock_waiter {
6200         struct task_struct      *task;
6201         struct inode            *inode;
6202         struct nfs_lowner       *owner;
6203         bool                    notified;
6204 };
6205
6206 static int
6207 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6208 {
6209         int ret;
6210         struct cb_notify_lock_args *cbnl = key;
6211         struct nfs4_lock_waiter *waiter = wait->private;
6212         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6213                                 *wowner = waiter->owner;
6214
6215         /* Only wake if the callback was for the same owner */
6216         if (lowner->clientid != wowner->clientid ||
6217             lowner->id != wowner->id             ||
6218             lowner->s_dev != wowner->s_dev)
6219                 return 0;
6220
6221         /* Make sure it's for the right inode */
6222         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6223                 return 0;
6224
6225         waiter->notified = true;
6226
6227         /* override "private" so we can use default_wake_function */
6228         wait->private = waiter->task;
6229         ret = autoremove_wake_function(wait, mode, flags, key);
6230         wait->private = waiter;
6231         return ret;
6232 }
6233
6234 static int
6235 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6236 {
6237         int status = -ERESTARTSYS;
6238         unsigned long flags;
6239         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6240         struct nfs_server *server = NFS_SERVER(state->inode);
6241         struct nfs_client *clp = server->nfs_client;
6242         wait_queue_head_t *q = &clp->cl_lock_waitq;
6243         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6244                                     .id = lsp->ls_seqid.owner_id,
6245                                     .s_dev = server->s_dev };
6246         struct nfs4_lock_waiter waiter = { .task  = current,
6247                                            .inode = state->inode,
6248                                            .owner = &owner,
6249                                            .notified = false };
6250         wait_queue_t wait;
6251
6252         /* Don't bother with waitqueue if we don't expect a callback */
6253         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6254                 return nfs4_retry_setlk_simple(state, cmd, request);
6255
6256         init_wait(&wait);
6257         wait.private = &waiter;
6258         wait.func = nfs4_wake_lock_waiter;
6259         add_wait_queue(q, &wait);
6260
6261         while(!signalled()) {
6262                 status = nfs4_proc_setlk(state, cmd, request);
6263                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6264                         break;
6265
6266                 status = -ERESTARTSYS;
6267                 spin_lock_irqsave(&q->lock, flags);
6268                 if (waiter.notified) {
6269                         spin_unlock_irqrestore(&q->lock, flags);
6270                         continue;
6271                 }
6272                 set_current_state(TASK_INTERRUPTIBLE);
6273                 spin_unlock_irqrestore(&q->lock, flags);
6274
6275                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6276         }
6277
6278         finish_wait(q, &wait);
6279         return status;
6280 }
6281 #else /* !CONFIG_NFS_V4_1 */
6282 static inline int
6283 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6284 {
6285         return nfs4_retry_setlk_simple(state, cmd, request);
6286 }
6287 #endif
6288
6289 static int
6290 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6291 {
6292         struct nfs_open_context *ctx;
6293         struct nfs4_state *state;
6294         int status;
6295
6296         /* verify open state */
6297         ctx = nfs_file_open_context(filp);
6298         state = ctx->state;
6299
6300         if (request->fl_start < 0 || request->fl_end < 0)
6301                 return -EINVAL;
6302
6303         if (IS_GETLK(cmd)) {
6304                 if (state != NULL)
6305                         return nfs4_proc_getlk(state, F_GETLK, request);
6306                 return 0;
6307         }
6308
6309         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6310                 return -EINVAL;
6311
6312         if (request->fl_type == F_UNLCK) {
6313                 if (state != NULL)
6314                         return nfs4_proc_unlck(state, cmd, request);
6315                 return 0;
6316         }
6317
6318         if (state == NULL)
6319                 return -ENOLCK;
6320
6321         if ((request->fl_flags & FL_POSIX) &&
6322             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6323                 return -ENOLCK;
6324
6325         /*
6326          * Don't rely on the VFS having checked the file open mode,
6327          * since it won't do this for flock() locks.
6328          */
6329         switch (request->fl_type) {
6330         case F_RDLCK:
6331                 if (!(filp->f_mode & FMODE_READ))
6332                         return -EBADF;
6333                 break;
6334         case F_WRLCK:
6335                 if (!(filp->f_mode & FMODE_WRITE))
6336                         return -EBADF;
6337         }
6338
6339         status = nfs4_set_lock_state(state, request);
6340         if (status != 0)
6341                 return status;
6342
6343         return nfs4_retry_setlk(state, cmd, request);
6344 }
6345
6346 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6347 {
6348         struct nfs_server *server = NFS_SERVER(state->inode);
6349         int err;
6350
6351         err = nfs4_set_lock_state(state, fl);
6352         if (err != 0)
6353                 return err;
6354         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6355         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6356 }
6357
6358 struct nfs_release_lockowner_data {
6359         struct nfs4_lock_state *lsp;
6360         struct nfs_server *server;
6361         struct nfs_release_lockowner_args args;
6362         struct nfs_release_lockowner_res res;
6363         unsigned long timestamp;
6364 };
6365
6366 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6367 {
6368         struct nfs_release_lockowner_data *data = calldata;
6369         struct nfs_server *server = data->server;
6370         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6371                              &data->args.seq_args, &data->res.seq_res, task);
6372         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6373         data->timestamp = jiffies;
6374 }
6375
6376 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6377 {
6378         struct nfs_release_lockowner_data *data = calldata;
6379         struct nfs_server *server = data->server;
6380
6381         nfs40_sequence_done(task, &data->res.seq_res);
6382
6383         switch (task->tk_status) {
6384         case 0:
6385                 renew_lease(server, data->timestamp);
6386                 break;
6387         case -NFS4ERR_STALE_CLIENTID:
6388         case -NFS4ERR_EXPIRED:
6389                 nfs4_schedule_lease_recovery(server->nfs_client);
6390                 break;
6391         case -NFS4ERR_LEASE_MOVED:
6392         case -NFS4ERR_DELAY:
6393                 if (nfs4_async_handle_error(task, server,
6394                                             NULL, NULL) == -EAGAIN)
6395                         rpc_restart_call_prepare(task);
6396         }
6397 }
6398
6399 static void nfs4_release_lockowner_release(void *calldata)
6400 {
6401         struct nfs_release_lockowner_data *data = calldata;
6402         nfs4_free_lock_state(data->server, data->lsp);
6403         kfree(calldata);
6404 }
6405
6406 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6407         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6408         .rpc_call_done = nfs4_release_lockowner_done,
6409         .rpc_release = nfs4_release_lockowner_release,
6410 };
6411
6412 static void
6413 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6414 {
6415         struct nfs_release_lockowner_data *data;
6416         struct rpc_message msg = {
6417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6418         };
6419
6420         if (server->nfs_client->cl_mvops->minor_version != 0)
6421                 return;
6422
6423         data = kmalloc(sizeof(*data), GFP_NOFS);
6424         if (!data)
6425                 return;
6426         data->lsp = lsp;
6427         data->server = server;
6428         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6429         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6430         data->args.lock_owner.s_dev = server->s_dev;
6431
6432         msg.rpc_argp = &data->args;
6433         msg.rpc_resp = &data->res;
6434         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6435         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6436 }
6437
6438 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6439
6440 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6441                                    struct dentry *unused, struct inode *inode,
6442                                    const char *key, const void *buf,
6443                                    size_t buflen, int flags)
6444 {
6445         return nfs4_proc_set_acl(inode, buf, buflen);
6446 }
6447
6448 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6449                                    struct dentry *unused, struct inode *inode,
6450                                    const char *key, void *buf, size_t buflen)
6451 {
6452         return nfs4_proc_get_acl(inode, buf, buflen);
6453 }
6454
6455 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6456 {
6457         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6458 }
6459
6460 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6461
6462 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6463                                      struct dentry *unused, struct inode *inode,
6464                                      const char *key, const void *buf,
6465                                      size_t buflen, int flags)
6466 {
6467         if (security_ismaclabel(key))
6468                 return nfs4_set_security_label(inode, buf, buflen);
6469
6470         return -EOPNOTSUPP;
6471 }
6472
6473 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6474                                      struct dentry *unused, struct inode *inode,
6475                                      const char *key, void *buf, size_t buflen)
6476 {
6477         if (security_ismaclabel(key))
6478                 return nfs4_get_security_label(inode, buf, buflen);
6479         return -EOPNOTSUPP;
6480 }
6481
6482 static ssize_t
6483 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6484 {
6485         int len = 0;
6486
6487         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6488                 len = security_inode_listsecurity(inode, list, list_len);
6489                 if (list_len && len > list_len)
6490                         return -ERANGE;
6491         }
6492         return len;
6493 }
6494
6495 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6496         .prefix = XATTR_SECURITY_PREFIX,
6497         .get    = nfs4_xattr_get_nfs4_label,
6498         .set    = nfs4_xattr_set_nfs4_label,
6499 };
6500
6501 #else
6502
6503 static ssize_t
6504 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6505 {
6506         return 0;
6507 }
6508
6509 #endif
6510
6511 /*
6512  * nfs_fhget will use either the mounted_on_fileid or the fileid
6513  */
6514 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6515 {
6516         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6517                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6518               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6519               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6520                 return;
6521
6522         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6523                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6524         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6525         fattr->nlink = 2;
6526 }
6527
6528 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6529                                    const struct qstr *name,
6530                                    struct nfs4_fs_locations *fs_locations,
6531                                    struct page *page)
6532 {
6533         struct nfs_server *server = NFS_SERVER(dir);
6534         u32 bitmask[3] = {
6535                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6536         };
6537         struct nfs4_fs_locations_arg args = {
6538                 .dir_fh = NFS_FH(dir),
6539                 .name = name,
6540                 .page = page,
6541                 .bitmask = bitmask,
6542         };
6543         struct nfs4_fs_locations_res res = {
6544                 .fs_locations = fs_locations,
6545         };
6546         struct rpc_message msg = {
6547                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6548                 .rpc_argp = &args,
6549                 .rpc_resp = &res,
6550         };
6551         int status;
6552
6553         dprintk("%s: start\n", __func__);
6554
6555         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6556          * is not supported */
6557         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6558                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6559         else
6560                 bitmask[0] |= FATTR4_WORD0_FILEID;
6561
6562         nfs_fattr_init(&fs_locations->fattr);
6563         fs_locations->server = server;
6564         fs_locations->nlocations = 0;
6565         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6566         dprintk("%s: returned status = %d\n", __func__, status);
6567         return status;
6568 }
6569
6570 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6571                            const struct qstr *name,
6572                            struct nfs4_fs_locations *fs_locations,
6573                            struct page *page)
6574 {
6575         struct nfs4_exception exception = { };
6576         int err;
6577         do {
6578                 err = _nfs4_proc_fs_locations(client, dir, name,
6579                                 fs_locations, page);
6580                 trace_nfs4_get_fs_locations(dir, name, err);
6581                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6582                                 &exception);
6583         } while (exception.retry);
6584         return err;
6585 }
6586
6587 /*
6588  * This operation also signals the server that this client is
6589  * performing migration recovery.  The server can stop returning
6590  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6591  * appended to this compound to identify the client ID which is
6592  * performing recovery.
6593  */
6594 static int _nfs40_proc_get_locations(struct inode *inode,
6595                                      struct nfs4_fs_locations *locations,
6596                                      struct page *page, struct rpc_cred *cred)
6597 {
6598         struct nfs_server *server = NFS_SERVER(inode);
6599         struct rpc_clnt *clnt = server->client;
6600         u32 bitmask[2] = {
6601                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6602         };
6603         struct nfs4_fs_locations_arg args = {
6604                 .clientid       = server->nfs_client->cl_clientid,
6605                 .fh             = NFS_FH(inode),
6606                 .page           = page,
6607                 .bitmask        = bitmask,
6608                 .migration      = 1,            /* skip LOOKUP */
6609                 .renew          = 1,            /* append RENEW */
6610         };
6611         struct nfs4_fs_locations_res res = {
6612                 .fs_locations   = locations,
6613                 .migration      = 1,
6614                 .renew          = 1,
6615         };
6616         struct rpc_message msg = {
6617                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6618                 .rpc_argp       = &args,
6619                 .rpc_resp       = &res,
6620                 .rpc_cred       = cred,
6621         };
6622         unsigned long now = jiffies;
6623         int status;
6624
6625         nfs_fattr_init(&locations->fattr);
6626         locations->server = server;
6627         locations->nlocations = 0;
6628
6629         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6630         nfs4_set_sequence_privileged(&args.seq_args);
6631         status = nfs4_call_sync_sequence(clnt, server, &msg,
6632                                         &args.seq_args, &res.seq_res);
6633         if (status)
6634                 return status;
6635
6636         renew_lease(server, now);
6637         return 0;
6638 }
6639
6640 #ifdef CONFIG_NFS_V4_1
6641
6642 /*
6643  * This operation also signals the server that this client is
6644  * performing migration recovery.  The server can stop asserting
6645  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6646  * performing this operation is identified in the SEQUENCE
6647  * operation in this compound.
6648  *
6649  * When the client supports GETATTR(fs_locations_info), it can
6650  * be plumbed in here.
6651  */
6652 static int _nfs41_proc_get_locations(struct inode *inode,
6653                                      struct nfs4_fs_locations *locations,
6654                                      struct page *page, struct rpc_cred *cred)
6655 {
6656         struct nfs_server *server = NFS_SERVER(inode);
6657         struct rpc_clnt *clnt = server->client;
6658         u32 bitmask[2] = {
6659                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6660         };
6661         struct nfs4_fs_locations_arg args = {
6662                 .fh             = NFS_FH(inode),
6663                 .page           = page,
6664                 .bitmask        = bitmask,
6665                 .migration      = 1,            /* skip LOOKUP */
6666         };
6667         struct nfs4_fs_locations_res res = {
6668                 .fs_locations   = locations,
6669                 .migration      = 1,
6670         };
6671         struct rpc_message msg = {
6672                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6673                 .rpc_argp       = &args,
6674                 .rpc_resp       = &res,
6675                 .rpc_cred       = cred,
6676         };
6677         int status;
6678
6679         nfs_fattr_init(&locations->fattr);
6680         locations->server = server;
6681         locations->nlocations = 0;
6682
6683         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6684         nfs4_set_sequence_privileged(&args.seq_args);
6685         status = nfs4_call_sync_sequence(clnt, server, &msg,
6686                                         &args.seq_args, &res.seq_res);
6687         if (status == NFS4_OK &&
6688             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6689                 status = -NFS4ERR_LEASE_MOVED;
6690         return status;
6691 }
6692
6693 #endif  /* CONFIG_NFS_V4_1 */
6694
6695 /**
6696  * nfs4_proc_get_locations - discover locations for a migrated FSID
6697  * @inode: inode on FSID that is migrating
6698  * @locations: result of query
6699  * @page: buffer
6700  * @cred: credential to use for this operation
6701  *
6702  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6703  * operation failed, or a negative errno if a local error occurred.
6704  *
6705  * On success, "locations" is filled in, but if the server has
6706  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6707  * asserted.
6708  *
6709  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6710  * from this client that require migration recovery.
6711  */
6712 int nfs4_proc_get_locations(struct inode *inode,
6713                             struct nfs4_fs_locations *locations,
6714                             struct page *page, struct rpc_cred *cred)
6715 {
6716         struct nfs_server *server = NFS_SERVER(inode);
6717         struct nfs_client *clp = server->nfs_client;
6718         const struct nfs4_mig_recovery_ops *ops =
6719                                         clp->cl_mvops->mig_recovery_ops;
6720         struct nfs4_exception exception = { };
6721         int status;
6722
6723         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6724                 (unsigned long long)server->fsid.major,
6725                 (unsigned long long)server->fsid.minor,
6726                 clp->cl_hostname);
6727         nfs_display_fhandle(NFS_FH(inode), __func__);
6728
6729         do {
6730                 status = ops->get_locations(inode, locations, page, cred);
6731                 if (status != -NFS4ERR_DELAY)
6732                         break;
6733                 nfs4_handle_exception(server, status, &exception);
6734         } while (exception.retry);
6735         return status;
6736 }
6737
6738 /*
6739  * This operation also signals the server that this client is
6740  * performing "lease moved" recovery.  The server can stop
6741  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6742  * is appended to this compound to identify the client ID which is
6743  * performing recovery.
6744  */
6745 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6746 {
6747         struct nfs_server *server = NFS_SERVER(inode);
6748         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6749         struct rpc_clnt *clnt = server->client;
6750         struct nfs4_fsid_present_arg args = {
6751                 .fh             = NFS_FH(inode),
6752                 .clientid       = clp->cl_clientid,
6753                 .renew          = 1,            /* append RENEW */
6754         };
6755         struct nfs4_fsid_present_res res = {
6756                 .renew          = 1,
6757         };
6758         struct rpc_message msg = {
6759                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6760                 .rpc_argp       = &args,
6761                 .rpc_resp       = &res,
6762                 .rpc_cred       = cred,
6763         };
6764         unsigned long now = jiffies;
6765         int status;
6766
6767         res.fh = nfs_alloc_fhandle();
6768         if (res.fh == NULL)
6769                 return -ENOMEM;
6770
6771         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6772         nfs4_set_sequence_privileged(&args.seq_args);
6773         status = nfs4_call_sync_sequence(clnt, server, &msg,
6774                                                 &args.seq_args, &res.seq_res);
6775         nfs_free_fhandle(res.fh);
6776         if (status)
6777                 return status;
6778
6779         do_renew_lease(clp, now);
6780         return 0;
6781 }
6782
6783 #ifdef CONFIG_NFS_V4_1
6784
6785 /*
6786  * This operation also signals the server that this client is
6787  * performing "lease moved" recovery.  The server can stop asserting
6788  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6789  * this operation is identified in the SEQUENCE operation in this
6790  * compound.
6791  */
6792 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6793 {
6794         struct nfs_server *server = NFS_SERVER(inode);
6795         struct rpc_clnt *clnt = server->client;
6796         struct nfs4_fsid_present_arg args = {
6797                 .fh             = NFS_FH(inode),
6798         };
6799         struct nfs4_fsid_present_res res = {
6800         };
6801         struct rpc_message msg = {
6802                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6803                 .rpc_argp       = &args,
6804                 .rpc_resp       = &res,
6805                 .rpc_cred       = cred,
6806         };
6807         int status;
6808
6809         res.fh = nfs_alloc_fhandle();
6810         if (res.fh == NULL)
6811                 return -ENOMEM;
6812
6813         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6814         nfs4_set_sequence_privileged(&args.seq_args);
6815         status = nfs4_call_sync_sequence(clnt, server, &msg,
6816                                                 &args.seq_args, &res.seq_res);
6817         nfs_free_fhandle(res.fh);
6818         if (status == NFS4_OK &&
6819             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6820                 status = -NFS4ERR_LEASE_MOVED;
6821         return status;
6822 }
6823
6824 #endif  /* CONFIG_NFS_V4_1 */
6825
6826 /**
6827  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6828  * @inode: inode on FSID to check
6829  * @cred: credential to use for this operation
6830  *
6831  * Server indicates whether the FSID is present, moved, or not
6832  * recognized.  This operation is necessary to clear a LEASE_MOVED
6833  * condition for this client ID.
6834  *
6835  * Returns NFS4_OK if the FSID is present on this server,
6836  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6837  *  NFS4ERR code if some error occurred on the server, or a
6838  *  negative errno if a local failure occurred.
6839  */
6840 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6841 {
6842         struct nfs_server *server = NFS_SERVER(inode);
6843         struct nfs_client *clp = server->nfs_client;
6844         const struct nfs4_mig_recovery_ops *ops =
6845                                         clp->cl_mvops->mig_recovery_ops;
6846         struct nfs4_exception exception = { };
6847         int status;
6848
6849         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6850                 (unsigned long long)server->fsid.major,
6851                 (unsigned long long)server->fsid.minor,
6852                 clp->cl_hostname);
6853         nfs_display_fhandle(NFS_FH(inode), __func__);
6854
6855         do {
6856                 status = ops->fsid_present(inode, cred);
6857                 if (status != -NFS4ERR_DELAY)
6858                         break;
6859                 nfs4_handle_exception(server, status, &exception);
6860         } while (exception.retry);
6861         return status;
6862 }
6863
6864 /**
6865  * If 'use_integrity' is true and the state managment nfs_client
6866  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6867  * and the machine credential as per RFC3530bis and RFC5661 Security
6868  * Considerations sections. Otherwise, just use the user cred with the
6869  * filesystem's rpc_client.
6870  */
6871 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6872 {
6873         int status;
6874         struct nfs4_secinfo_arg args = {
6875                 .dir_fh = NFS_FH(dir),
6876                 .name   = name,
6877         };
6878         struct nfs4_secinfo_res res = {
6879                 .flavors     = flavors,
6880         };
6881         struct rpc_message msg = {
6882                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6883                 .rpc_argp = &args,
6884                 .rpc_resp = &res,
6885         };
6886         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6887         struct rpc_cred *cred = NULL;
6888
6889         if (use_integrity) {
6890                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6891                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6892                 msg.rpc_cred = cred;
6893         }
6894
6895         dprintk("NFS call  secinfo %s\n", name->name);
6896
6897         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6898                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6899
6900         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6901                                 &res.seq_res, 0);
6902         dprintk("NFS reply  secinfo: %d\n", status);
6903
6904         if (cred)
6905                 put_rpccred(cred);
6906
6907         return status;
6908 }
6909
6910 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6911                       struct nfs4_secinfo_flavors *flavors)
6912 {
6913         struct nfs4_exception exception = { };
6914         int err;
6915         do {
6916                 err = -NFS4ERR_WRONGSEC;
6917
6918                 /* try to use integrity protection with machine cred */
6919                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6920                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6921
6922                 /*
6923                  * if unable to use integrity protection, or SECINFO with
6924                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6925                  * disallowed by spec, but exists in deployed servers) use
6926                  * the current filesystem's rpc_client and the user cred.
6927                  */
6928                 if (err == -NFS4ERR_WRONGSEC)
6929                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6930
6931                 trace_nfs4_secinfo(dir, name, err);
6932                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6933                                 &exception);
6934         } while (exception.retry);
6935         return err;
6936 }
6937
6938 #ifdef CONFIG_NFS_V4_1
6939 /*
6940  * Check the exchange flags returned by the server for invalid flags, having
6941  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6942  * DS flags set.
6943  */
6944 static int nfs4_check_cl_exchange_flags(u32 flags)
6945 {
6946         if (flags & ~EXCHGID4_FLAG_MASK_R)
6947                 goto out_inval;
6948         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6949             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6950                 goto out_inval;
6951         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6952                 goto out_inval;
6953         return NFS_OK;
6954 out_inval:
6955         return -NFS4ERR_INVAL;
6956 }
6957
6958 static bool
6959 nfs41_same_server_scope(struct nfs41_server_scope *a,
6960                         struct nfs41_server_scope *b)
6961 {
6962         if (a->server_scope_sz == b->server_scope_sz &&
6963             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6964                 return true;
6965
6966         return false;
6967 }
6968
6969 static void
6970 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6971 {
6972 }
6973
6974 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6975         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6976 };
6977
6978 /*
6979  * nfs4_proc_bind_one_conn_to_session()
6980  *
6981  * The 4.1 client currently uses the same TCP connection for the
6982  * fore and backchannel.
6983  */
6984 static
6985 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6986                 struct rpc_xprt *xprt,
6987                 struct nfs_client *clp,
6988                 struct rpc_cred *cred)
6989 {
6990         int status;
6991         struct nfs41_bind_conn_to_session_args args = {
6992                 .client = clp,
6993                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6994         };
6995         struct nfs41_bind_conn_to_session_res res;
6996         struct rpc_message msg = {
6997                 .rpc_proc =
6998                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6999                 .rpc_argp = &args,
7000                 .rpc_resp = &res,
7001                 .rpc_cred = cred,
7002         };
7003         struct rpc_task_setup task_setup_data = {
7004                 .rpc_client = clnt,
7005                 .rpc_xprt = xprt,
7006                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7007                 .rpc_message = &msg,
7008                 .flags = RPC_TASK_TIMEOUT,
7009         };
7010         struct rpc_task *task;
7011
7012         dprintk("--> %s\n", __func__);
7013
7014         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7015         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7016                 args.dir = NFS4_CDFC4_FORE;
7017
7018         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7019         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7020                 args.dir = NFS4_CDFC4_FORE;
7021
7022         task = rpc_run_task(&task_setup_data);
7023         if (!IS_ERR(task)) {
7024                 status = task->tk_status;
7025                 rpc_put_task(task);
7026         } else
7027                 status = PTR_ERR(task);
7028         trace_nfs4_bind_conn_to_session(clp, status);
7029         if (status == 0) {
7030                 if (memcmp(res.sessionid.data,
7031                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7032                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7033                         status = -EIO;
7034                         goto out;
7035                 }
7036                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7037                         dprintk("NFS: %s: Unexpected direction from server\n",
7038                                 __func__);
7039                         status = -EIO;
7040                         goto out;
7041                 }
7042                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7043                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7044                                 __func__);
7045                         status = -EIO;
7046                         goto out;
7047                 }
7048         }
7049 out:
7050         dprintk("<-- %s status= %d\n", __func__, status);
7051         return status;
7052 }
7053
7054 struct rpc_bind_conn_calldata {
7055         struct nfs_client *clp;
7056         struct rpc_cred *cred;
7057 };
7058
7059 static int
7060 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7061                 struct rpc_xprt *xprt,
7062                 void *calldata)
7063 {
7064         struct rpc_bind_conn_calldata *p = calldata;
7065
7066         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7067 }
7068
7069 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7070 {
7071         struct rpc_bind_conn_calldata data = {
7072                 .clp = clp,
7073                 .cred = cred,
7074         };
7075         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7076                         nfs4_proc_bind_conn_to_session_callback, &data);
7077 }
7078
7079 /*
7080  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7081  * and operations we'd like to see to enable certain features in the allow map
7082  */
7083 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7084         .how = SP4_MACH_CRED,
7085         .enforce.u.words = {
7086                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7087                       1 << (OP_EXCHANGE_ID - 32) |
7088                       1 << (OP_CREATE_SESSION - 32) |
7089                       1 << (OP_DESTROY_SESSION - 32) |
7090                       1 << (OP_DESTROY_CLIENTID - 32)
7091         },
7092         .allow.u.words = {
7093                 [0] = 1 << (OP_CLOSE) |
7094                       1 << (OP_OPEN_DOWNGRADE) |
7095                       1 << (OP_LOCKU) |
7096                       1 << (OP_DELEGRETURN) |
7097                       1 << (OP_COMMIT),
7098                 [1] = 1 << (OP_SECINFO - 32) |
7099                       1 << (OP_SECINFO_NO_NAME - 32) |
7100                       1 << (OP_LAYOUTRETURN - 32) |
7101                       1 << (OP_TEST_STATEID - 32) |
7102                       1 << (OP_FREE_STATEID - 32) |
7103                       1 << (OP_WRITE - 32)
7104         }
7105 };
7106
7107 /*
7108  * Select the state protection mode for client `clp' given the server results
7109  * from exchange_id in `sp'.
7110  *
7111  * Returns 0 on success, negative errno otherwise.
7112  */
7113 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7114                                  struct nfs41_state_protection *sp)
7115 {
7116         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7117                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7118                       1 << (OP_EXCHANGE_ID - 32) |
7119                       1 << (OP_CREATE_SESSION - 32) |
7120                       1 << (OP_DESTROY_SESSION - 32) |
7121                       1 << (OP_DESTROY_CLIENTID - 32)
7122         };
7123         unsigned int i;
7124
7125         if (sp->how == SP4_MACH_CRED) {
7126                 /* Print state protect result */
7127                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7128                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7129                         if (test_bit(i, sp->enforce.u.longs))
7130                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7131                         if (test_bit(i, sp->allow.u.longs))
7132                                 dfprintk(MOUNT, "  allow op %d\n", i);
7133                 }
7134
7135                 /* make sure nothing is on enforce list that isn't supported */
7136                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7137                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7138                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7139                                 return -EINVAL;
7140                         }
7141                 }
7142
7143                 /*
7144                  * Minimal mode - state operations are allowed to use machine
7145                  * credential.  Note this already happens by default, so the
7146                  * client doesn't have to do anything more than the negotiation.
7147                  *
7148                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7149                  *       we're already using the machine cred for exchange_id
7150                  *       and will never use a different cred.
7151                  */
7152                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7153                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7154                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7155                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7156                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7157                         dfprintk(MOUNT, "  minimal mode enabled\n");
7158                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7159                 } else {
7160                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7161                         return -EINVAL;
7162                 }
7163
7164                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7165                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7166                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7167                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7168                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7169                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7170                 }
7171
7172                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7173                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7174                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7175                                 &clp->cl_sp4_flags);
7176                 }
7177
7178                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7179                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7180                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7181                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7182                 }
7183
7184                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7185                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7186                         dfprintk(MOUNT, "  stateid mode enabled\n");
7187                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7188                 }
7189
7190                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7191                         dfprintk(MOUNT, "  write mode enabled\n");
7192                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7193                 }
7194
7195                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7196                         dfprintk(MOUNT, "  commit mode enabled\n");
7197                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7198                 }
7199         }
7200
7201         return 0;
7202 }
7203
7204 struct nfs41_exchange_id_data {
7205         struct nfs41_exchange_id_res res;
7206         struct nfs41_exchange_id_args args;
7207         struct rpc_xprt *xprt;
7208         int rpc_status;
7209 };
7210
7211 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7212 {
7213         struct nfs41_exchange_id_data *cdata =
7214                                         (struct nfs41_exchange_id_data *)data;
7215         struct nfs_client *clp = cdata->args.client;
7216         int status = task->tk_status;
7217
7218         trace_nfs4_exchange_id(clp, status);
7219
7220         if (status == 0)
7221                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7222
7223         if (cdata->xprt && status == 0) {
7224                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7225                                                       cdata->xprt);
7226                 goto out;
7227         }
7228
7229         if (status  == 0)
7230                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7231
7232         if (status == 0) {
7233                 clp->cl_clientid = cdata->res.clientid;
7234                 clp->cl_exchange_flags = cdata->res.flags;
7235                 /* Client ID is not confirmed */
7236                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7237                         clear_bit(NFS4_SESSION_ESTABLISHED,
7238                         &clp->cl_session->session_state);
7239                         clp->cl_seqid = cdata->res.seqid;
7240                 }
7241
7242                 kfree(clp->cl_serverowner);
7243                 clp->cl_serverowner = cdata->res.server_owner;
7244                 cdata->res.server_owner = NULL;
7245
7246                 /* use the most recent implementation id */
7247                 kfree(clp->cl_implid);
7248                 clp->cl_implid = cdata->res.impl_id;
7249                 cdata->res.impl_id = NULL;
7250
7251                 if (clp->cl_serverscope != NULL &&
7252                     !nfs41_same_server_scope(clp->cl_serverscope,
7253                                         cdata->res.server_scope)) {
7254                         dprintk("%s: server_scope mismatch detected\n",
7255                                 __func__);
7256                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7257                         kfree(clp->cl_serverscope);
7258                         clp->cl_serverscope = NULL;
7259                 }
7260
7261                 if (clp->cl_serverscope == NULL) {
7262                         clp->cl_serverscope = cdata->res.server_scope;
7263                         cdata->res.server_scope = NULL;
7264                 }
7265                 /* Save the EXCHANGE_ID verifier session trunk tests */
7266                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7267                        sizeof(clp->cl_confirm.data));
7268         }
7269 out:
7270         cdata->rpc_status = status;
7271         return;
7272 }
7273
7274 static void nfs4_exchange_id_release(void *data)
7275 {
7276         struct nfs41_exchange_id_data *cdata =
7277                                         (struct nfs41_exchange_id_data *)data;
7278
7279         nfs_put_client(cdata->args.client);
7280         if (cdata->xprt) {
7281                 xprt_put(cdata->xprt);
7282                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7283         }
7284         kfree(cdata->res.impl_id);
7285         kfree(cdata->res.server_scope);
7286         kfree(cdata->res.server_owner);
7287         kfree(cdata);
7288 }
7289
7290 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7291         .rpc_call_done = nfs4_exchange_id_done,
7292         .rpc_release = nfs4_exchange_id_release,
7293 };
7294
7295 /*
7296  * _nfs4_proc_exchange_id()
7297  *
7298  * Wrapper for EXCHANGE_ID operation.
7299  */
7300 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7301                         u32 sp4_how, struct rpc_xprt *xprt)
7302 {
7303         nfs4_verifier verifier;
7304         struct rpc_message msg = {
7305                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7306                 .rpc_cred = cred,
7307         };
7308         struct rpc_task_setup task_setup_data = {
7309                 .rpc_client = clp->cl_rpcclient,
7310                 .callback_ops = &nfs4_exchange_id_call_ops,
7311                 .rpc_message = &msg,
7312                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7313         };
7314         struct nfs41_exchange_id_data *calldata;
7315         struct rpc_task *task;
7316         int status = -EIO;
7317
7318         if (!atomic_inc_not_zero(&clp->cl_count))
7319                 goto out;
7320
7321         status = -ENOMEM;
7322         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7323         if (!calldata)
7324                 goto out;
7325
7326         if (!xprt)
7327                 nfs4_init_boot_verifier(clp, &verifier);
7328
7329         status = nfs4_init_uniform_client_string(clp);
7330         if (status)
7331                 goto out_calldata;
7332
7333         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7334                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7335                 clp->cl_owner_id);
7336
7337         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7338                                                 GFP_NOFS);
7339         status = -ENOMEM;
7340         if (unlikely(calldata->res.server_owner == NULL))
7341                 goto out_calldata;
7342
7343         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7344                                         GFP_NOFS);
7345         if (unlikely(calldata->res.server_scope == NULL))
7346                 goto out_server_owner;
7347
7348         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7349         if (unlikely(calldata->res.impl_id == NULL))
7350                 goto out_server_scope;
7351
7352         switch (sp4_how) {
7353         case SP4_NONE:
7354                 calldata->args.state_protect.how = SP4_NONE;
7355                 break;
7356
7357         case SP4_MACH_CRED:
7358                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7359                 break;
7360
7361         default:
7362                 /* unsupported! */
7363                 WARN_ON_ONCE(1);
7364                 status = -EINVAL;
7365                 goto out_impl_id;
7366         }
7367         if (xprt) {
7368                 calldata->xprt = xprt;
7369                 task_setup_data.rpc_xprt = xprt;
7370                 task_setup_data.flags =
7371                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7372                 calldata->args.verifier = &clp->cl_confirm;
7373         } else {
7374                 calldata->args.verifier = &verifier;
7375         }
7376         calldata->args.client = clp;
7377 #ifdef CONFIG_NFS_V4_1_MIGRATION
7378         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7379         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7380         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7381 #else
7382         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7383         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7384 #endif
7385         msg.rpc_argp = &calldata->args;
7386         msg.rpc_resp = &calldata->res;
7387         task_setup_data.callback_data = calldata;
7388
7389         task = rpc_run_task(&task_setup_data);
7390         if (IS_ERR(task)) {
7391         status = PTR_ERR(task);
7392                 goto out_impl_id;
7393         }
7394
7395         if (!xprt) {
7396                 status = rpc_wait_for_completion_task(task);
7397                 if (!status)
7398                         status = calldata->rpc_status;
7399         } else  /* session trunking test */
7400                 status = calldata->rpc_status;
7401
7402         rpc_put_task(task);
7403 out:
7404         if (clp->cl_implid != NULL)
7405                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7406                         "domain: %s, name: %s, date: %llu,%u\n",
7407                         clp->cl_implid->domain, clp->cl_implid->name,
7408                         clp->cl_implid->date.seconds,
7409                         clp->cl_implid->date.nseconds);
7410         dprintk("NFS reply exchange_id: %d\n", status);
7411         return status;
7412
7413 out_impl_id:
7414         kfree(calldata->res.impl_id);
7415 out_server_scope:
7416         kfree(calldata->res.server_scope);
7417 out_server_owner:
7418         kfree(calldata->res.server_owner);
7419 out_calldata:
7420         kfree(calldata);
7421         goto out;
7422 }
7423
7424 /*
7425  * nfs4_proc_exchange_id()
7426  *
7427  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7428  *
7429  * Since the clientid has expired, all compounds using sessions
7430  * associated with the stale clientid will be returning
7431  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7432  * be in some phase of session reset.
7433  *
7434  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7435  */
7436 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7437 {
7438         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7439         int status;
7440
7441         /* try SP4_MACH_CRED if krb5i/p */
7442         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7443             authflavor == RPC_AUTH_GSS_KRB5P) {
7444                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7445                 if (!status)
7446                         return 0;
7447         }
7448
7449         /* try SP4_NONE */
7450         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7451 }
7452
7453 /**
7454  * nfs4_test_session_trunk
7455  *
7456  * This is an add_xprt_test() test function called from
7457  * rpc_clnt_setup_test_and_add_xprt.
7458  *
7459  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7460  * and is dereferrenced in nfs4_exchange_id_release
7461  *
7462  * Upon success, add the new transport to the rpc_clnt
7463  *
7464  * @clnt: struct rpc_clnt to get new transport
7465  * @xprt: the rpc_xprt to test
7466  * @data: call data for _nfs4_proc_exchange_id.
7467  */
7468 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7469                             void *data)
7470 {
7471         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7472         u32 sp4_how;
7473
7474         dprintk("--> %s try %s\n", __func__,
7475                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7476
7477         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7478
7479         /* Test connection for session trunking. Async exchange_id call */
7480         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7481 }
7482 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7483
7484 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7485                 struct rpc_cred *cred)
7486 {
7487         struct rpc_message msg = {
7488                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7489                 .rpc_argp = clp,
7490                 .rpc_cred = cred,
7491         };
7492         int status;
7493
7494         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7495         trace_nfs4_destroy_clientid(clp, status);
7496         if (status)
7497                 dprintk("NFS: Got error %d from the server %s on "
7498                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7499         return status;
7500 }
7501
7502 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7503                 struct rpc_cred *cred)
7504 {
7505         unsigned int loop;
7506         int ret;
7507
7508         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7509                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7510                 switch (ret) {
7511                 case -NFS4ERR_DELAY:
7512                 case -NFS4ERR_CLIENTID_BUSY:
7513                         ssleep(1);
7514                         break;
7515                 default:
7516                         return ret;
7517                 }
7518         }
7519         return 0;
7520 }
7521
7522 int nfs4_destroy_clientid(struct nfs_client *clp)
7523 {
7524         struct rpc_cred *cred;
7525         int ret = 0;
7526
7527         if (clp->cl_mvops->minor_version < 1)
7528                 goto out;
7529         if (clp->cl_exchange_flags == 0)
7530                 goto out;
7531         if (clp->cl_preserve_clid)
7532                 goto out;
7533         cred = nfs4_get_clid_cred(clp);
7534         ret = nfs4_proc_destroy_clientid(clp, cred);
7535         if (cred)
7536                 put_rpccred(cred);
7537         switch (ret) {
7538         case 0:
7539         case -NFS4ERR_STALE_CLIENTID:
7540                 clp->cl_exchange_flags = 0;
7541         }
7542 out:
7543         return ret;
7544 }
7545
7546 struct nfs4_get_lease_time_data {
7547         struct nfs4_get_lease_time_args *args;
7548         struct nfs4_get_lease_time_res *res;
7549         struct nfs_client *clp;
7550 };
7551
7552 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7553                                         void *calldata)
7554 {
7555         struct nfs4_get_lease_time_data *data =
7556                         (struct nfs4_get_lease_time_data *)calldata;
7557
7558         dprintk("--> %s\n", __func__);
7559         /* just setup sequence, do not trigger session recovery
7560            since we're invoked within one */
7561         nfs41_setup_sequence(data->clp->cl_session,
7562                         &data->args->la_seq_args,
7563                         &data->res->lr_seq_res,
7564                         task);
7565         dprintk("<-- %s\n", __func__);
7566 }
7567
7568 /*
7569  * Called from nfs4_state_manager thread for session setup, so don't recover
7570  * from sequence operation or clientid errors.
7571  */
7572 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7573 {
7574         struct nfs4_get_lease_time_data *data =
7575                         (struct nfs4_get_lease_time_data *)calldata;
7576
7577         dprintk("--> %s\n", __func__);
7578         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7579                 return;
7580         switch (task->tk_status) {
7581         case -NFS4ERR_DELAY:
7582         case -NFS4ERR_GRACE:
7583                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7584                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7585                 task->tk_status = 0;
7586                 /* fall through */
7587         case -NFS4ERR_RETRY_UNCACHED_REP:
7588                 rpc_restart_call_prepare(task);
7589                 return;
7590         }
7591         dprintk("<-- %s\n", __func__);
7592 }
7593
7594 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7595         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7596         .rpc_call_done = nfs4_get_lease_time_done,
7597 };
7598
7599 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7600 {
7601         struct rpc_task *task;
7602         struct nfs4_get_lease_time_args args;
7603         struct nfs4_get_lease_time_res res = {
7604                 .lr_fsinfo = fsinfo,
7605         };
7606         struct nfs4_get_lease_time_data data = {
7607                 .args = &args,
7608                 .res = &res,
7609                 .clp = clp,
7610         };
7611         struct rpc_message msg = {
7612                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7613                 .rpc_argp = &args,
7614                 .rpc_resp = &res,
7615         };
7616         struct rpc_task_setup task_setup = {
7617                 .rpc_client = clp->cl_rpcclient,
7618                 .rpc_message = &msg,
7619                 .callback_ops = &nfs4_get_lease_time_ops,
7620                 .callback_data = &data,
7621                 .flags = RPC_TASK_TIMEOUT,
7622         };
7623         int status;
7624
7625         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7626         nfs4_set_sequence_privileged(&args.la_seq_args);
7627         dprintk("--> %s\n", __func__);
7628         task = rpc_run_task(&task_setup);
7629
7630         if (IS_ERR(task))
7631                 status = PTR_ERR(task);
7632         else {
7633                 status = task->tk_status;
7634                 rpc_put_task(task);
7635         }
7636         dprintk("<-- %s return %d\n", __func__, status);
7637
7638         return status;
7639 }
7640
7641 /*
7642  * Initialize the values to be used by the client in CREATE_SESSION
7643  * If nfs4_init_session set the fore channel request and response sizes,
7644  * use them.
7645  *
7646  * Set the back channel max_resp_sz_cached to zero to force the client to
7647  * always set csa_cachethis to FALSE because the current implementation
7648  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7649  */
7650 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7651                                     struct rpc_clnt *clnt)
7652 {
7653         unsigned int max_rqst_sz, max_resp_sz;
7654         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7655
7656         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7657         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7658
7659         /* Fore channel attributes */
7660         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7661         args->fc_attrs.max_resp_sz = max_resp_sz;
7662         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7663         args->fc_attrs.max_reqs = max_session_slots;
7664
7665         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7666                 "max_ops=%u max_reqs=%u\n",
7667                 __func__,
7668                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7669                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7670
7671         /* Back channel attributes */
7672         args->bc_attrs.max_rqst_sz = max_bc_payload;
7673         args->bc_attrs.max_resp_sz = max_bc_payload;
7674         args->bc_attrs.max_resp_sz_cached = 0;
7675         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7676         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7677
7678         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7679                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7680                 __func__,
7681                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7682                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7683                 args->bc_attrs.max_reqs);
7684 }
7685
7686 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7687                 struct nfs41_create_session_res *res)
7688 {
7689         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7690         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7691
7692         if (rcvd->max_resp_sz > sent->max_resp_sz)
7693                 return -EINVAL;
7694         /*
7695          * Our requested max_ops is the minimum we need; we're not
7696          * prepared to break up compounds into smaller pieces than that.
7697          * So, no point even trying to continue if the server won't
7698          * cooperate:
7699          */
7700         if (rcvd->max_ops < sent->max_ops)
7701                 return -EINVAL;
7702         if (rcvd->max_reqs == 0)
7703                 return -EINVAL;
7704         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7705                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7706         return 0;
7707 }
7708
7709 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7710                 struct nfs41_create_session_res *res)
7711 {
7712         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7713         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7714
7715         if (!(res->flags & SESSION4_BACK_CHAN))
7716                 goto out;
7717         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7718                 return -EINVAL;
7719         if (rcvd->max_resp_sz < sent->max_resp_sz)
7720                 return -EINVAL;
7721         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7722                 return -EINVAL;
7723         if (rcvd->max_ops > sent->max_ops)
7724                 return -EINVAL;
7725         if (rcvd->max_reqs > sent->max_reqs)
7726                 return -EINVAL;
7727 out:
7728         return 0;
7729 }
7730
7731 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7732                                      struct nfs41_create_session_res *res)
7733 {
7734         int ret;
7735
7736         ret = nfs4_verify_fore_channel_attrs(args, res);
7737         if (ret)
7738                 return ret;
7739         return nfs4_verify_back_channel_attrs(args, res);
7740 }
7741
7742 static void nfs4_update_session(struct nfs4_session *session,
7743                 struct nfs41_create_session_res *res)
7744 {
7745         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7746         /* Mark client id and session as being confirmed */
7747         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7748         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7749         session->flags = res->flags;
7750         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7751         if (res->flags & SESSION4_BACK_CHAN)
7752                 memcpy(&session->bc_attrs, &res->bc_attrs,
7753                                 sizeof(session->bc_attrs));
7754 }
7755
7756 static int _nfs4_proc_create_session(struct nfs_client *clp,
7757                 struct rpc_cred *cred)
7758 {
7759         struct nfs4_session *session = clp->cl_session;
7760         struct nfs41_create_session_args args = {
7761                 .client = clp,
7762                 .clientid = clp->cl_clientid,
7763                 .seqid = clp->cl_seqid,
7764                 .cb_program = NFS4_CALLBACK,
7765         };
7766         struct nfs41_create_session_res res;
7767
7768         struct rpc_message msg = {
7769                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7770                 .rpc_argp = &args,
7771                 .rpc_resp = &res,
7772                 .rpc_cred = cred,
7773         };
7774         int status;
7775
7776         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7777         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7778
7779         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7780         trace_nfs4_create_session(clp, status);
7781
7782         switch (status) {
7783         case -NFS4ERR_STALE_CLIENTID:
7784         case -NFS4ERR_DELAY:
7785         case -ETIMEDOUT:
7786         case -EACCES:
7787         case -EAGAIN:
7788                 goto out;
7789         };
7790
7791         clp->cl_seqid++;
7792         if (!status) {
7793                 /* Verify the session's negotiated channel_attrs values */
7794                 status = nfs4_verify_channel_attrs(&args, &res);
7795                 /* Increment the clientid slot sequence id */
7796                 if (status)
7797                         goto out;
7798                 nfs4_update_session(session, &res);
7799         }
7800 out:
7801         return status;
7802 }
7803
7804 /*
7805  * Issues a CREATE_SESSION operation to the server.
7806  * It is the responsibility of the caller to verify the session is
7807  * expired before calling this routine.
7808  */
7809 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7810 {
7811         int status;
7812         unsigned *ptr;
7813         struct nfs4_session *session = clp->cl_session;
7814
7815         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7816
7817         status = _nfs4_proc_create_session(clp, cred);
7818         if (status)
7819                 goto out;
7820
7821         /* Init or reset the session slot tables */
7822         status = nfs4_setup_session_slot_tables(session);
7823         dprintk("slot table setup returned %d\n", status);
7824         if (status)
7825                 goto out;
7826
7827         ptr = (unsigned *)&session->sess_id.data[0];
7828         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7829                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7830 out:
7831         dprintk("<-- %s\n", __func__);
7832         return status;
7833 }
7834
7835 /*
7836  * Issue the over-the-wire RPC DESTROY_SESSION.
7837  * The caller must serialize access to this routine.
7838  */
7839 int nfs4_proc_destroy_session(struct nfs4_session *session,
7840                 struct rpc_cred *cred)
7841 {
7842         struct rpc_message msg = {
7843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7844                 .rpc_argp = session,
7845                 .rpc_cred = cred,
7846         };
7847         int status = 0;
7848
7849         dprintk("--> nfs4_proc_destroy_session\n");
7850
7851         /* session is still being setup */
7852         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7853                 return 0;
7854
7855         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7856         trace_nfs4_destroy_session(session->clp, status);
7857
7858         if (status)
7859                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7860                         "Session has been destroyed regardless...\n", status);
7861
7862         dprintk("<-- nfs4_proc_destroy_session\n");
7863         return status;
7864 }
7865
7866 /*
7867  * Renew the cl_session lease.
7868  */
7869 struct nfs4_sequence_data {
7870         struct nfs_client *clp;
7871         struct nfs4_sequence_args args;
7872         struct nfs4_sequence_res res;
7873 };
7874
7875 static void nfs41_sequence_release(void *data)
7876 {
7877         struct nfs4_sequence_data *calldata = data;
7878         struct nfs_client *clp = calldata->clp;
7879
7880         if (atomic_read(&clp->cl_count) > 1)
7881                 nfs4_schedule_state_renewal(clp);
7882         nfs_put_client(clp);
7883         kfree(calldata);
7884 }
7885
7886 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7887 {
7888         switch(task->tk_status) {
7889         case -NFS4ERR_DELAY:
7890                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7891                 return -EAGAIN;
7892         default:
7893                 nfs4_schedule_lease_recovery(clp);
7894         }
7895         return 0;
7896 }
7897
7898 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7899 {
7900         struct nfs4_sequence_data *calldata = data;
7901         struct nfs_client *clp = calldata->clp;
7902
7903         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7904                 return;
7905
7906         trace_nfs4_sequence(clp, task->tk_status);
7907         if (task->tk_status < 0) {
7908                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7909                 if (atomic_read(&clp->cl_count) == 1)
7910                         goto out;
7911
7912                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7913                         rpc_restart_call_prepare(task);
7914                         return;
7915                 }
7916         }
7917         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7918 out:
7919         dprintk("<-- %s\n", __func__);
7920 }
7921
7922 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7923 {
7924         struct nfs4_sequence_data *calldata = data;
7925         struct nfs_client *clp = calldata->clp;
7926         struct nfs4_sequence_args *args;
7927         struct nfs4_sequence_res *res;
7928
7929         args = task->tk_msg.rpc_argp;
7930         res = task->tk_msg.rpc_resp;
7931
7932         nfs41_setup_sequence(clp->cl_session, args, res, task);
7933 }
7934
7935 static const struct rpc_call_ops nfs41_sequence_ops = {
7936         .rpc_call_done = nfs41_sequence_call_done,
7937         .rpc_call_prepare = nfs41_sequence_prepare,
7938         .rpc_release = nfs41_sequence_release,
7939 };
7940
7941 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7942                 struct rpc_cred *cred,
7943                 bool is_privileged)
7944 {
7945         struct nfs4_sequence_data *calldata;
7946         struct rpc_message msg = {
7947                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7948                 .rpc_cred = cred,
7949         };
7950         struct rpc_task_setup task_setup_data = {
7951                 .rpc_client = clp->cl_rpcclient,
7952                 .rpc_message = &msg,
7953                 .callback_ops = &nfs41_sequence_ops,
7954                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7955         };
7956
7957         if (!atomic_inc_not_zero(&clp->cl_count))
7958                 return ERR_PTR(-EIO);
7959         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7960         if (calldata == NULL) {
7961                 nfs_put_client(clp);
7962                 return ERR_PTR(-ENOMEM);
7963         }
7964         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7965         if (is_privileged)
7966                 nfs4_set_sequence_privileged(&calldata->args);
7967         msg.rpc_argp = &calldata->args;
7968         msg.rpc_resp = &calldata->res;
7969         calldata->clp = clp;
7970         task_setup_data.callback_data = calldata;
7971
7972         return rpc_run_task(&task_setup_data);
7973 }
7974
7975 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7976 {
7977         struct rpc_task *task;
7978         int ret = 0;
7979
7980         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7981                 return -EAGAIN;
7982         task = _nfs41_proc_sequence(clp, cred, false);
7983         if (IS_ERR(task))
7984                 ret = PTR_ERR(task);
7985         else
7986                 rpc_put_task_async(task);
7987         dprintk("<-- %s status=%d\n", __func__, ret);
7988         return ret;
7989 }
7990
7991 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7992 {
7993         struct rpc_task *task;
7994         int ret;
7995
7996         task = _nfs41_proc_sequence(clp, cred, true);
7997         if (IS_ERR(task)) {
7998                 ret = PTR_ERR(task);
7999                 goto out;
8000         }
8001         ret = rpc_wait_for_completion_task(task);
8002         if (!ret)
8003                 ret = task->tk_status;
8004         rpc_put_task(task);
8005 out:
8006         dprintk("<-- %s status=%d\n", __func__, ret);
8007         return ret;
8008 }
8009
8010 struct nfs4_reclaim_complete_data {
8011         struct nfs_client *clp;
8012         struct nfs41_reclaim_complete_args arg;
8013         struct nfs41_reclaim_complete_res res;
8014 };
8015
8016 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8017 {
8018         struct nfs4_reclaim_complete_data *calldata = data;
8019
8020         nfs41_setup_sequence(calldata->clp->cl_session,
8021                         &calldata->arg.seq_args,
8022                         &calldata->res.seq_res,
8023                         task);
8024 }
8025
8026 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8027 {
8028         switch(task->tk_status) {
8029         case 0:
8030         case -NFS4ERR_COMPLETE_ALREADY:
8031         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8032                 break;
8033         case -NFS4ERR_DELAY:
8034                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8035                 /* fall through */
8036         case -NFS4ERR_RETRY_UNCACHED_REP:
8037                 return -EAGAIN;
8038         default:
8039                 nfs4_schedule_lease_recovery(clp);
8040         }
8041         return 0;
8042 }
8043
8044 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8045 {
8046         struct nfs4_reclaim_complete_data *calldata = data;
8047         struct nfs_client *clp = calldata->clp;
8048         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8049
8050         dprintk("--> %s\n", __func__);
8051         if (!nfs41_sequence_done(task, res))
8052                 return;
8053
8054         trace_nfs4_reclaim_complete(clp, task->tk_status);
8055         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8056                 rpc_restart_call_prepare(task);
8057                 return;
8058         }
8059         dprintk("<-- %s\n", __func__);
8060 }
8061
8062 static void nfs4_free_reclaim_complete_data(void *data)
8063 {
8064         struct nfs4_reclaim_complete_data *calldata = data;
8065
8066         kfree(calldata);
8067 }
8068
8069 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8070         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8071         .rpc_call_done = nfs4_reclaim_complete_done,
8072         .rpc_release = nfs4_free_reclaim_complete_data,
8073 };
8074
8075 /*
8076  * Issue a global reclaim complete.
8077  */
8078 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8079                 struct rpc_cred *cred)
8080 {
8081         struct nfs4_reclaim_complete_data *calldata;
8082         struct rpc_task *task;
8083         struct rpc_message msg = {
8084                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8085                 .rpc_cred = cred,
8086         };
8087         struct rpc_task_setup task_setup_data = {
8088                 .rpc_client = clp->cl_rpcclient,
8089                 .rpc_message = &msg,
8090                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8091                 .flags = RPC_TASK_ASYNC,
8092         };
8093         int status = -ENOMEM;
8094
8095         dprintk("--> %s\n", __func__);
8096         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8097         if (calldata == NULL)
8098                 goto out;
8099         calldata->clp = clp;
8100         calldata->arg.one_fs = 0;
8101
8102         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8103         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8104         msg.rpc_argp = &calldata->arg;
8105         msg.rpc_resp = &calldata->res;
8106         task_setup_data.callback_data = calldata;
8107         task = rpc_run_task(&task_setup_data);
8108         if (IS_ERR(task)) {
8109                 status = PTR_ERR(task);
8110                 goto out;
8111         }
8112         status = nfs4_wait_for_completion_rpc_task(task);
8113         if (status == 0)
8114                 status = task->tk_status;
8115         rpc_put_task(task);
8116         return 0;
8117 out:
8118         dprintk("<-- %s status=%d\n", __func__, status);
8119         return status;
8120 }
8121
8122 static void
8123 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8124 {
8125         struct nfs4_layoutget *lgp = calldata;
8126         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8127         struct nfs4_session *session = nfs4_get_session(server);
8128
8129         dprintk("--> %s\n", __func__);
8130         nfs41_setup_sequence(session, &lgp->args.seq_args,
8131                                 &lgp->res.seq_res, task);
8132         dprintk("<-- %s\n", __func__);
8133 }
8134
8135 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8136 {
8137         struct nfs4_layoutget *lgp = calldata;
8138
8139         dprintk("--> %s\n", __func__);
8140         nfs41_sequence_process(task, &lgp->res.seq_res);
8141         dprintk("<-- %s\n", __func__);
8142 }
8143
8144 static int
8145 nfs4_layoutget_handle_exception(struct rpc_task *task,
8146                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8147 {
8148         struct inode *inode = lgp->args.inode;
8149         struct nfs_server *server = NFS_SERVER(inode);
8150         struct pnfs_layout_hdr *lo;
8151         int nfs4err = task->tk_status;
8152         int err, status = 0;
8153         LIST_HEAD(head);
8154
8155         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8156
8157         switch (nfs4err) {
8158         case 0:
8159                 goto out;
8160
8161         /*
8162          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8163          * on the file. set tk_status to -ENODATA to tell upper layer to
8164          * retry go inband.
8165          */
8166         case -NFS4ERR_LAYOUTUNAVAILABLE:
8167                 status = -ENODATA;
8168                 goto out;
8169         /*
8170          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8171          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8172          */
8173         case -NFS4ERR_BADLAYOUT:
8174                 status = -EOVERFLOW;
8175                 goto out;
8176         /*
8177          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8178          * (or clients) writing to the same RAID stripe except when
8179          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8180          *
8181          * Treat it like we would RECALLCONFLICT -- we retry for a little
8182          * while, and then eventually give up.
8183          */
8184         case -NFS4ERR_LAYOUTTRYLATER:
8185                 if (lgp->args.minlength == 0) {
8186                         status = -EOVERFLOW;
8187                         goto out;
8188                 }
8189                 status = -EBUSY;
8190                 break;
8191         case -NFS4ERR_RECALLCONFLICT:
8192                 status = -ERECALLCONFLICT;
8193                 break;
8194         case -NFS4ERR_EXPIRED:
8195         case -NFS4ERR_BAD_STATEID:
8196                 exception->timeout = 0;
8197                 spin_lock(&inode->i_lock);
8198                 lo = NFS_I(inode)->layout;
8199                 /* If the open stateid was bad, then recover it. */
8200                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8201                     nfs4_stateid_match_other(&lgp->args.stateid,
8202                                         &lgp->args.ctx->state->stateid)) {
8203                         spin_unlock(&inode->i_lock);
8204                         exception->state = lgp->args.ctx->state;
8205                         break;
8206                 }
8207
8208                 /*
8209                  * Mark the bad layout state as invalid, then retry
8210                  */
8211                 pnfs_mark_layout_stateid_invalid(lo, &head);
8212                 spin_unlock(&inode->i_lock);
8213                 pnfs_free_lseg_list(&head);
8214                 status = -EAGAIN;
8215                 goto out;
8216         }
8217
8218         err = nfs4_handle_exception(server, nfs4err, exception);
8219         if (!status) {
8220                 if (exception->retry)
8221                         status = -EAGAIN;
8222                 else
8223                         status = err;
8224         }
8225 out:
8226         dprintk("<-- %s\n", __func__);
8227         return status;
8228 }
8229
8230 static size_t max_response_pages(struct nfs_server *server)
8231 {
8232         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8233         return nfs_page_array_len(0, max_resp_sz);
8234 }
8235
8236 static void nfs4_free_pages(struct page **pages, size_t size)
8237 {
8238         int i;
8239
8240         if (!pages)
8241                 return;
8242
8243         for (i = 0; i < size; i++) {
8244                 if (!pages[i])
8245                         break;
8246                 __free_page(pages[i]);
8247         }
8248         kfree(pages);
8249 }
8250
8251 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8252 {
8253         struct page **pages;
8254         int i;
8255
8256         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8257         if (!pages) {
8258                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8259                 return NULL;
8260         }
8261
8262         for (i = 0; i < size; i++) {
8263                 pages[i] = alloc_page(gfp_flags);
8264                 if (!pages[i]) {
8265                         dprintk("%s: failed to allocate page\n", __func__);
8266                         nfs4_free_pages(pages, size);
8267                         return NULL;
8268                 }
8269         }
8270
8271         return pages;
8272 }
8273
8274 static void nfs4_layoutget_release(void *calldata)
8275 {
8276         struct nfs4_layoutget *lgp = calldata;
8277         struct inode *inode = lgp->args.inode;
8278         struct nfs_server *server = NFS_SERVER(inode);
8279         size_t max_pages = max_response_pages(server);
8280
8281         dprintk("--> %s\n", __func__);
8282         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8283         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8284         put_nfs_open_context(lgp->args.ctx);
8285         kfree(calldata);
8286         dprintk("<-- %s\n", __func__);
8287 }
8288
8289 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8290         .rpc_call_prepare = nfs4_layoutget_prepare,
8291         .rpc_call_done = nfs4_layoutget_done,
8292         .rpc_release = nfs4_layoutget_release,
8293 };
8294
8295 struct pnfs_layout_segment *
8296 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8297 {
8298         struct inode *inode = lgp->args.inode;
8299         struct nfs_server *server = NFS_SERVER(inode);
8300         size_t max_pages = max_response_pages(server);
8301         struct rpc_task *task;
8302         struct rpc_message msg = {
8303                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8304                 .rpc_argp = &lgp->args,
8305                 .rpc_resp = &lgp->res,
8306                 .rpc_cred = lgp->cred,
8307         };
8308         struct rpc_task_setup task_setup_data = {
8309                 .rpc_client = server->client,
8310                 .rpc_message = &msg,
8311                 .callback_ops = &nfs4_layoutget_call_ops,
8312                 .callback_data = lgp,
8313                 .flags = RPC_TASK_ASYNC,
8314         };
8315         struct pnfs_layout_segment *lseg = NULL;
8316         struct nfs4_exception exception = {
8317                 .inode = inode,
8318                 .timeout = *timeout,
8319         };
8320         int status = 0;
8321
8322         dprintk("--> %s\n", __func__);
8323
8324         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8325         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8326
8327         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8328         if (!lgp->args.layout.pages) {
8329                 nfs4_layoutget_release(lgp);
8330                 return ERR_PTR(-ENOMEM);
8331         }
8332         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8333
8334         lgp->res.layoutp = &lgp->args.layout;
8335         lgp->res.seq_res.sr_slot = NULL;
8336         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8337
8338         task = rpc_run_task(&task_setup_data);
8339         if (IS_ERR(task))
8340                 return ERR_CAST(task);
8341         status = nfs4_wait_for_completion_rpc_task(task);
8342         if (status == 0) {
8343                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8344                 *timeout = exception.timeout;
8345         }
8346
8347         trace_nfs4_layoutget(lgp->args.ctx,
8348                         &lgp->args.range,
8349                         &lgp->res.range,
8350                         &lgp->res.stateid,
8351                         status);
8352
8353         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8354         if (status == 0 && lgp->res.layoutp->len)
8355                 lseg = pnfs_layout_process(lgp);
8356         nfs4_sequence_free_slot(&lgp->res.seq_res);
8357         rpc_put_task(task);
8358         dprintk("<-- %s status=%d\n", __func__, status);
8359         if (status)
8360                 return ERR_PTR(status);
8361         return lseg;
8362 }
8363
8364 static void
8365 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8366 {
8367         struct nfs4_layoutreturn *lrp = calldata;
8368
8369         dprintk("--> %s\n", __func__);
8370         nfs41_setup_sequence(lrp->clp->cl_session,
8371                         &lrp->args.seq_args,
8372                         &lrp->res.seq_res,
8373                         task);
8374 }
8375
8376 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8377 {
8378         struct nfs4_layoutreturn *lrp = calldata;
8379         struct nfs_server *server;
8380
8381         dprintk("--> %s\n", __func__);
8382
8383         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8384                 return;
8385
8386         server = NFS_SERVER(lrp->args.inode);
8387         switch (task->tk_status) {
8388         default:
8389                 task->tk_status = 0;
8390         case 0:
8391                 break;
8392         case -NFS4ERR_DELAY:
8393                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8394                         break;
8395                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8396                 rpc_restart_call_prepare(task);
8397                 return;
8398         }
8399         dprintk("<-- %s\n", __func__);
8400 }
8401
8402 static void nfs4_layoutreturn_release(void *calldata)
8403 {
8404         struct nfs4_layoutreturn *lrp = calldata;
8405         struct pnfs_layout_hdr *lo = lrp->args.layout;
8406         LIST_HEAD(freeme);
8407
8408         dprintk("--> %s\n", __func__);
8409         spin_lock(&lo->plh_inode->i_lock);
8410         if (lrp->res.lrs_present) {
8411                 pnfs_mark_matching_lsegs_invalid(lo, &freeme,
8412                                 &lrp->args.range,
8413                                 be32_to_cpu(lrp->args.stateid.seqid));
8414                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8415         } else
8416                 pnfs_mark_layout_stateid_invalid(lo, &freeme);
8417         pnfs_clear_layoutreturn_waitbit(lo);
8418         spin_unlock(&lo->plh_inode->i_lock);
8419         nfs4_sequence_free_slot(&lrp->res.seq_res);
8420         pnfs_free_lseg_list(&freeme);
8421         pnfs_put_layout_hdr(lrp->args.layout);
8422         nfs_iput_and_deactive(lrp->inode);
8423         kfree(calldata);
8424         dprintk("<-- %s\n", __func__);
8425 }
8426
8427 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8428         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8429         .rpc_call_done = nfs4_layoutreturn_done,
8430         .rpc_release = nfs4_layoutreturn_release,
8431 };
8432
8433 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8434 {
8435         struct rpc_task *task;
8436         struct rpc_message msg = {
8437                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8438                 .rpc_argp = &lrp->args,
8439                 .rpc_resp = &lrp->res,
8440                 .rpc_cred = lrp->cred,
8441         };
8442         struct rpc_task_setup task_setup_data = {
8443                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8444                 .rpc_message = &msg,
8445                 .callback_ops = &nfs4_layoutreturn_call_ops,
8446                 .callback_data = lrp,
8447         };
8448         int status = 0;
8449
8450         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8451                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8452                         &task_setup_data.rpc_client, &msg);
8453
8454         dprintk("--> %s\n", __func__);
8455         if (!sync) {
8456                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8457                 if (!lrp->inode) {
8458                         nfs4_layoutreturn_release(lrp);
8459                         return -EAGAIN;
8460                 }
8461                 task_setup_data.flags |= RPC_TASK_ASYNC;
8462         }
8463         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8464         task = rpc_run_task(&task_setup_data);
8465         if (IS_ERR(task))
8466                 return PTR_ERR(task);
8467         if (sync)
8468                 status = task->tk_status;
8469         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8470         dprintk("<-- %s status=%d\n", __func__, status);
8471         rpc_put_task(task);
8472         return status;
8473 }
8474
8475 static int
8476 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8477                 struct pnfs_device *pdev,
8478                 struct rpc_cred *cred)
8479 {
8480         struct nfs4_getdeviceinfo_args args = {
8481                 .pdev = pdev,
8482                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8483                         NOTIFY_DEVICEID4_DELETE,
8484         };
8485         struct nfs4_getdeviceinfo_res res = {
8486                 .pdev = pdev,
8487         };
8488         struct rpc_message msg = {
8489                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8490                 .rpc_argp = &args,
8491                 .rpc_resp = &res,
8492                 .rpc_cred = cred,
8493         };
8494         int status;
8495
8496         dprintk("--> %s\n", __func__);
8497         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8498         if (res.notification & ~args.notify_types)
8499                 dprintk("%s: unsupported notification\n", __func__);
8500         if (res.notification != args.notify_types)
8501                 pdev->nocache = 1;
8502
8503         dprintk("<-- %s status=%d\n", __func__, status);
8504
8505         return status;
8506 }
8507
8508 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8509                 struct pnfs_device *pdev,
8510                 struct rpc_cred *cred)
8511 {
8512         struct nfs4_exception exception = { };
8513         int err;
8514
8515         do {
8516                 err = nfs4_handle_exception(server,
8517                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8518                                         &exception);
8519         } while (exception.retry);
8520         return err;
8521 }
8522 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8523
8524 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8525 {
8526         struct nfs4_layoutcommit_data *data = calldata;
8527         struct nfs_server *server = NFS_SERVER(data->args.inode);
8528         struct nfs4_session *session = nfs4_get_session(server);
8529
8530         nfs41_setup_sequence(session,
8531                         &data->args.seq_args,
8532                         &data->res.seq_res,
8533                         task);
8534 }
8535
8536 static void
8537 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8538 {
8539         struct nfs4_layoutcommit_data *data = calldata;
8540         struct nfs_server *server = NFS_SERVER(data->args.inode);
8541
8542         if (!nfs41_sequence_done(task, &data->res.seq_res))
8543                 return;
8544
8545         switch (task->tk_status) { /* Just ignore these failures */
8546         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8547         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8548         case -NFS4ERR_BADLAYOUT:     /* no layout */
8549         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8550                 task->tk_status = 0;
8551         case 0:
8552                 break;
8553         default:
8554                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8555                         rpc_restart_call_prepare(task);
8556                         return;
8557                 }
8558         }
8559 }
8560
8561 static void nfs4_layoutcommit_release(void *calldata)
8562 {
8563         struct nfs4_layoutcommit_data *data = calldata;
8564
8565         pnfs_cleanup_layoutcommit(data);
8566         nfs_post_op_update_inode_force_wcc(data->args.inode,
8567                                            data->res.fattr);
8568         put_rpccred(data->cred);
8569         nfs_iput_and_deactive(data->inode);
8570         kfree(data);
8571 }
8572
8573 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8574         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8575         .rpc_call_done = nfs4_layoutcommit_done,
8576         .rpc_release = nfs4_layoutcommit_release,
8577 };
8578
8579 int
8580 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8581 {
8582         struct rpc_message msg = {
8583                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8584                 .rpc_argp = &data->args,
8585                 .rpc_resp = &data->res,
8586                 .rpc_cred = data->cred,
8587         };
8588         struct rpc_task_setup task_setup_data = {
8589                 .task = &data->task,
8590                 .rpc_client = NFS_CLIENT(data->args.inode),
8591                 .rpc_message = &msg,
8592                 .callback_ops = &nfs4_layoutcommit_ops,
8593                 .callback_data = data,
8594         };
8595         struct rpc_task *task;
8596         int status = 0;
8597
8598         dprintk("NFS: initiating layoutcommit call. sync %d "
8599                 "lbw: %llu inode %lu\n", sync,
8600                 data->args.lastbytewritten,
8601                 data->args.inode->i_ino);
8602
8603         if (!sync) {
8604                 data->inode = nfs_igrab_and_active(data->args.inode);
8605                 if (data->inode == NULL) {
8606                         nfs4_layoutcommit_release(data);
8607                         return -EAGAIN;
8608                 }
8609                 task_setup_data.flags = RPC_TASK_ASYNC;
8610         }
8611         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8612         task = rpc_run_task(&task_setup_data);
8613         if (IS_ERR(task))
8614                 return PTR_ERR(task);
8615         if (sync)
8616                 status = task->tk_status;
8617         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8618         dprintk("%s: status %d\n", __func__, status);
8619         rpc_put_task(task);
8620         return status;
8621 }
8622
8623 /**
8624  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8625  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8626  */
8627 static int
8628 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8629                     struct nfs_fsinfo *info,
8630                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8631 {
8632         struct nfs41_secinfo_no_name_args args = {
8633                 .style = SECINFO_STYLE_CURRENT_FH,
8634         };
8635         struct nfs4_secinfo_res res = {
8636                 .flavors = flavors,
8637         };
8638         struct rpc_message msg = {
8639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8640                 .rpc_argp = &args,
8641                 .rpc_resp = &res,
8642         };
8643         struct rpc_clnt *clnt = server->client;
8644         struct rpc_cred *cred = NULL;
8645         int status;
8646
8647         if (use_integrity) {
8648                 clnt = server->nfs_client->cl_rpcclient;
8649                 cred = nfs4_get_clid_cred(server->nfs_client);
8650                 msg.rpc_cred = cred;
8651         }
8652
8653         dprintk("--> %s\n", __func__);
8654         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8655                                 &res.seq_res, 0);
8656         dprintk("<-- %s status=%d\n", __func__, status);
8657
8658         if (cred)
8659                 put_rpccred(cred);
8660
8661         return status;
8662 }
8663
8664 static int
8665 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8666                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8667 {
8668         struct nfs4_exception exception = { };
8669         int err;
8670         do {
8671                 /* first try using integrity protection */
8672                 err = -NFS4ERR_WRONGSEC;
8673
8674                 /* try to use integrity protection with machine cred */
8675                 if (_nfs4_is_integrity_protected(server->nfs_client))
8676                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8677                                                           flavors, true);
8678
8679                 /*
8680                  * if unable to use integrity protection, or SECINFO with
8681                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8682                  * disallowed by spec, but exists in deployed servers) use
8683                  * the current filesystem's rpc_client and the user cred.
8684                  */
8685                 if (err == -NFS4ERR_WRONGSEC)
8686                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8687                                                           flavors, false);
8688
8689                 switch (err) {
8690                 case 0:
8691                 case -NFS4ERR_WRONGSEC:
8692                 case -ENOTSUPP:
8693                         goto out;
8694                 default:
8695                         err = nfs4_handle_exception(server, err, &exception);
8696                 }
8697         } while (exception.retry);
8698 out:
8699         return err;
8700 }
8701
8702 static int
8703 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8704                     struct nfs_fsinfo *info)
8705 {
8706         int err;
8707         struct page *page;
8708         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8709         struct nfs4_secinfo_flavors *flavors;
8710         struct nfs4_secinfo4 *secinfo;
8711         int i;
8712
8713         page = alloc_page(GFP_KERNEL);
8714         if (!page) {
8715                 err = -ENOMEM;
8716                 goto out;
8717         }
8718
8719         flavors = page_address(page);
8720         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8721
8722         /*
8723          * Fall back on "guess and check" method if
8724          * the server doesn't support SECINFO_NO_NAME
8725          */
8726         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8727                 err = nfs4_find_root_sec(server, fhandle, info);
8728                 goto out_freepage;
8729         }
8730         if (err)
8731                 goto out_freepage;
8732
8733         for (i = 0; i < flavors->num_flavors; i++) {
8734                 secinfo = &flavors->flavors[i];
8735
8736                 switch (secinfo->flavor) {
8737                 case RPC_AUTH_NULL:
8738                 case RPC_AUTH_UNIX:
8739                 case RPC_AUTH_GSS:
8740                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8741                                         &secinfo->flavor_info);
8742                         break;
8743                 default:
8744                         flavor = RPC_AUTH_MAXFLAVOR;
8745                         break;
8746                 }
8747
8748                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8749                         flavor = RPC_AUTH_MAXFLAVOR;
8750
8751                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8752                         err = nfs4_lookup_root_sec(server, fhandle,
8753                                                    info, flavor);
8754                         if (!err)
8755                                 break;
8756                 }
8757         }
8758
8759         if (flavor == RPC_AUTH_MAXFLAVOR)
8760                 err = -EPERM;
8761
8762 out_freepage:
8763         put_page(page);
8764         if (err == -EACCES)
8765                 return -EPERM;
8766 out:
8767         return err;
8768 }
8769
8770 static int _nfs41_test_stateid(struct nfs_server *server,
8771                 nfs4_stateid *stateid,
8772                 struct rpc_cred *cred)
8773 {
8774         int status;
8775         struct nfs41_test_stateid_args args = {
8776                 .stateid = stateid,
8777         };
8778         struct nfs41_test_stateid_res res;
8779         struct rpc_message msg = {
8780                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8781                 .rpc_argp = &args,
8782                 .rpc_resp = &res,
8783                 .rpc_cred = cred,
8784         };
8785         struct rpc_clnt *rpc_client = server->client;
8786
8787         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8788                 &rpc_client, &msg);
8789
8790         dprintk("NFS call  test_stateid %p\n", stateid);
8791         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8792         nfs4_set_sequence_privileged(&args.seq_args);
8793         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8794                         &args.seq_args, &res.seq_res);
8795         if (status != NFS_OK) {
8796                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8797                 return status;
8798         }
8799         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8800         return -res.status;
8801 }
8802
8803 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8804                 int err, struct nfs4_exception *exception)
8805 {
8806         exception->retry = 0;
8807         switch(err) {
8808         case -NFS4ERR_DELAY:
8809                 nfs4_handle_exception(server, err, exception);
8810                 break;
8811         case -NFS4ERR_BADSESSION:
8812         case -NFS4ERR_BADSLOT:
8813         case -NFS4ERR_BAD_HIGH_SLOT:
8814         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8815         case -NFS4ERR_DEADSESSION:
8816                 nfs4_do_handle_exception(server, err, exception);
8817         }
8818 }
8819
8820 /**
8821  * nfs41_test_stateid - perform a TEST_STATEID operation
8822  *
8823  * @server: server / transport on which to perform the operation
8824  * @stateid: state ID to test
8825  * @cred: credential
8826  *
8827  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8828  * Otherwise a negative NFS4ERR value is returned if the operation
8829  * failed or the state ID is not currently valid.
8830  */
8831 static int nfs41_test_stateid(struct nfs_server *server,
8832                 nfs4_stateid *stateid,
8833                 struct rpc_cred *cred)
8834 {
8835         struct nfs4_exception exception = { };
8836         int err;
8837         do {
8838                 err = _nfs41_test_stateid(server, stateid, cred);
8839                 nfs4_handle_delay_or_session_error(server, err, &exception);
8840         } while (exception.retry);
8841         return err;
8842 }
8843
8844 struct nfs_free_stateid_data {
8845         struct nfs_server *server;
8846         struct nfs41_free_stateid_args args;
8847         struct nfs41_free_stateid_res res;
8848 };
8849
8850 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8851 {
8852         struct nfs_free_stateid_data *data = calldata;
8853         nfs41_setup_sequence(nfs4_get_session(data->server),
8854                         &data->args.seq_args,
8855                         &data->res.seq_res,
8856                         task);
8857 }
8858
8859 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8860 {
8861         struct nfs_free_stateid_data *data = calldata;
8862
8863         nfs41_sequence_done(task, &data->res.seq_res);
8864
8865         switch (task->tk_status) {
8866         case -NFS4ERR_DELAY:
8867                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8868                         rpc_restart_call_prepare(task);
8869         }
8870 }
8871
8872 static void nfs41_free_stateid_release(void *calldata)
8873 {
8874         kfree(calldata);
8875 }
8876
8877 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8878         .rpc_call_prepare = nfs41_free_stateid_prepare,
8879         .rpc_call_done = nfs41_free_stateid_done,
8880         .rpc_release = nfs41_free_stateid_release,
8881 };
8882
8883 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8884                 nfs4_stateid *stateid,
8885                 struct rpc_cred *cred,
8886                 bool privileged)
8887 {
8888         struct rpc_message msg = {
8889                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8890                 .rpc_cred = cred,
8891         };
8892         struct rpc_task_setup task_setup = {
8893                 .rpc_client = server->client,
8894                 .rpc_message = &msg,
8895                 .callback_ops = &nfs41_free_stateid_ops,
8896                 .flags = RPC_TASK_ASYNC,
8897         };
8898         struct nfs_free_stateid_data *data;
8899
8900         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8901                 &task_setup.rpc_client, &msg);
8902
8903         dprintk("NFS call  free_stateid %p\n", stateid);
8904         data = kmalloc(sizeof(*data), GFP_NOFS);
8905         if (!data)
8906                 return ERR_PTR(-ENOMEM);
8907         data->server = server;
8908         nfs4_stateid_copy(&data->args.stateid, stateid);
8909
8910         task_setup.callback_data = data;
8911
8912         msg.rpc_argp = &data->args;
8913         msg.rpc_resp = &data->res;
8914         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8915         if (privileged)
8916                 nfs4_set_sequence_privileged(&data->args.seq_args);
8917
8918         return rpc_run_task(&task_setup);
8919 }
8920
8921 /**
8922  * nfs41_free_stateid - perform a FREE_STATEID operation
8923  *
8924  * @server: server / transport on which to perform the operation
8925  * @stateid: state ID to release
8926  * @cred: credential
8927  *
8928  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8929  * negative NFS4ERR value is returned.
8930  */
8931 static int nfs41_free_stateid(struct nfs_server *server,
8932                 nfs4_stateid *stateid,
8933                 struct rpc_cred *cred)
8934 {
8935         struct rpc_task *task;
8936         int ret;
8937
8938         task = _nfs41_free_stateid(server, stateid, cred, true);
8939         if (IS_ERR(task))
8940                 return PTR_ERR(task);
8941         ret = rpc_wait_for_completion_task(task);
8942         if (!ret)
8943                 ret = task->tk_status;
8944         rpc_put_task(task);
8945         return ret;
8946 }
8947
8948 static void
8949 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8950 {
8951         struct rpc_task *task;
8952         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8953
8954         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8955         nfs4_free_lock_state(server, lsp);
8956         if (IS_ERR(task))
8957                 return;
8958         rpc_put_task(task);
8959 }
8960
8961 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8962                 const nfs4_stateid *s2)
8963 {
8964         if (s1->type != s2->type)
8965                 return false;
8966
8967         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8968                 return false;
8969
8970         if (s1->seqid == s2->seqid)
8971                 return true;
8972         if (s1->seqid == 0 || s2->seqid == 0)
8973                 return true;
8974
8975         return false;
8976 }
8977
8978 #endif /* CONFIG_NFS_V4_1 */
8979
8980 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8981                 const nfs4_stateid *s2)
8982 {
8983         return nfs4_stateid_match(s1, s2);
8984 }
8985
8986
8987 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8988         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8989         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8990         .recover_open   = nfs4_open_reclaim,
8991         .recover_lock   = nfs4_lock_reclaim,
8992         .establish_clid = nfs4_init_clientid,
8993         .detect_trunking = nfs40_discover_server_trunking,
8994 };
8995
8996 #if defined(CONFIG_NFS_V4_1)
8997 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8998         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8999         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9000         .recover_open   = nfs4_open_reclaim,
9001         .recover_lock   = nfs4_lock_reclaim,
9002         .establish_clid = nfs41_init_clientid,
9003         .reclaim_complete = nfs41_proc_reclaim_complete,
9004         .detect_trunking = nfs41_discover_server_trunking,
9005 };
9006 #endif /* CONFIG_NFS_V4_1 */
9007
9008 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9009         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9010         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9011         .recover_open   = nfs40_open_expired,
9012         .recover_lock   = nfs4_lock_expired,
9013         .establish_clid = nfs4_init_clientid,
9014 };
9015
9016 #if defined(CONFIG_NFS_V4_1)
9017 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9018         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9019         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9020         .recover_open   = nfs41_open_expired,
9021         .recover_lock   = nfs41_lock_expired,
9022         .establish_clid = nfs41_init_clientid,
9023 };
9024 #endif /* CONFIG_NFS_V4_1 */
9025
9026 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9027         .sched_state_renewal = nfs4_proc_async_renew,
9028         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9029         .renew_lease = nfs4_proc_renew,
9030 };
9031
9032 #if defined(CONFIG_NFS_V4_1)
9033 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9034         .sched_state_renewal = nfs41_proc_async_sequence,
9035         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9036         .renew_lease = nfs4_proc_sequence,
9037 };
9038 #endif
9039
9040 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9041         .get_locations = _nfs40_proc_get_locations,
9042         .fsid_present = _nfs40_proc_fsid_present,
9043 };
9044
9045 #if defined(CONFIG_NFS_V4_1)
9046 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9047         .get_locations = _nfs41_proc_get_locations,
9048         .fsid_present = _nfs41_proc_fsid_present,
9049 };
9050 #endif  /* CONFIG_NFS_V4_1 */
9051
9052 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9053         .minor_version = 0,
9054         .init_caps = NFS_CAP_READDIRPLUS
9055                 | NFS_CAP_ATOMIC_OPEN
9056                 | NFS_CAP_POSIX_LOCK,
9057         .init_client = nfs40_init_client,
9058         .shutdown_client = nfs40_shutdown_client,
9059         .match_stateid = nfs4_match_stateid,
9060         .find_root_sec = nfs4_find_root_sec,
9061         .free_lock_state = nfs4_release_lockowner,
9062         .alloc_seqid = nfs_alloc_seqid,
9063         .call_sync_ops = &nfs40_call_sync_ops,
9064         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9065         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9066         .state_renewal_ops = &nfs40_state_renewal_ops,
9067         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9068 };
9069
9070 #if defined(CONFIG_NFS_V4_1)
9071 static struct nfs_seqid *
9072 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9073 {
9074         return NULL;
9075 }
9076
9077 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9078         .minor_version = 1,
9079         .init_caps = NFS_CAP_READDIRPLUS
9080                 | NFS_CAP_ATOMIC_OPEN
9081                 | NFS_CAP_POSIX_LOCK
9082                 | NFS_CAP_STATEID_NFSV41
9083                 | NFS_CAP_ATOMIC_OPEN_V1,
9084         .init_client = nfs41_init_client,
9085         .shutdown_client = nfs41_shutdown_client,
9086         .match_stateid = nfs41_match_stateid,
9087         .find_root_sec = nfs41_find_root_sec,
9088         .free_lock_state = nfs41_free_lock_state,
9089         .alloc_seqid = nfs_alloc_no_seqid,
9090         .session_trunk = nfs4_test_session_trunk,
9091         .call_sync_ops = &nfs41_call_sync_ops,
9092         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9093         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9094         .state_renewal_ops = &nfs41_state_renewal_ops,
9095         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9096 };
9097 #endif
9098
9099 #if defined(CONFIG_NFS_V4_2)
9100 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9101         .minor_version = 2,
9102         .init_caps = NFS_CAP_READDIRPLUS
9103                 | NFS_CAP_ATOMIC_OPEN
9104                 | NFS_CAP_POSIX_LOCK
9105                 | NFS_CAP_STATEID_NFSV41
9106                 | NFS_CAP_ATOMIC_OPEN_V1
9107                 | NFS_CAP_ALLOCATE
9108                 | NFS_CAP_COPY
9109                 | NFS_CAP_DEALLOCATE
9110                 | NFS_CAP_SEEK
9111                 | NFS_CAP_LAYOUTSTATS
9112                 | NFS_CAP_CLONE,
9113         .init_client = nfs41_init_client,
9114         .shutdown_client = nfs41_shutdown_client,
9115         .match_stateid = nfs41_match_stateid,
9116         .find_root_sec = nfs41_find_root_sec,
9117         .free_lock_state = nfs41_free_lock_state,
9118         .call_sync_ops = &nfs41_call_sync_ops,
9119         .alloc_seqid = nfs_alloc_no_seqid,
9120         .session_trunk = nfs4_test_session_trunk,
9121         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9122         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9123         .state_renewal_ops = &nfs41_state_renewal_ops,
9124         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9125 };
9126 #endif
9127
9128 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9129         [0] = &nfs_v4_0_minor_ops,
9130 #if defined(CONFIG_NFS_V4_1)
9131         [1] = &nfs_v4_1_minor_ops,
9132 #endif
9133 #if defined(CONFIG_NFS_V4_2)
9134         [2] = &nfs_v4_2_minor_ops,
9135 #endif
9136 };
9137
9138 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9139 {
9140         ssize_t error, error2;
9141
9142         error = generic_listxattr(dentry, list, size);
9143         if (error < 0)
9144                 return error;
9145         if (list) {
9146                 list += error;
9147                 size -= error;
9148         }
9149
9150         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9151         if (error2 < 0)
9152                 return error2;
9153         return error + error2;
9154 }
9155
9156 static const struct inode_operations nfs4_dir_inode_operations = {
9157         .create         = nfs_create,
9158         .lookup         = nfs_lookup,
9159         .atomic_open    = nfs_atomic_open,
9160         .link           = nfs_link,
9161         .unlink         = nfs_unlink,
9162         .symlink        = nfs_symlink,
9163         .mkdir          = nfs_mkdir,
9164         .rmdir          = nfs_rmdir,
9165         .mknod          = nfs_mknod,
9166         .rename         = nfs_rename,
9167         .permission     = nfs_permission,
9168         .getattr        = nfs_getattr,
9169         .setattr        = nfs_setattr,
9170         .getxattr       = generic_getxattr,
9171         .setxattr       = generic_setxattr,
9172         .listxattr      = nfs4_listxattr,
9173         .removexattr    = generic_removexattr,
9174 };
9175
9176 static const struct inode_operations nfs4_file_inode_operations = {
9177         .permission     = nfs_permission,
9178         .getattr        = nfs_getattr,
9179         .setattr        = nfs_setattr,
9180         .getxattr       = generic_getxattr,
9181         .setxattr       = generic_setxattr,
9182         .listxattr      = nfs4_listxattr,
9183         .removexattr    = generic_removexattr,
9184 };
9185
9186 const struct nfs_rpc_ops nfs_v4_clientops = {
9187         .version        = 4,                    /* protocol version */
9188         .dentry_ops     = &nfs4_dentry_operations,
9189         .dir_inode_ops  = &nfs4_dir_inode_operations,
9190         .file_inode_ops = &nfs4_file_inode_operations,
9191         .file_ops       = &nfs4_file_operations,
9192         .getroot        = nfs4_proc_get_root,
9193         .submount       = nfs4_submount,
9194         .try_mount      = nfs4_try_mount,
9195         .getattr        = nfs4_proc_getattr,
9196         .setattr        = nfs4_proc_setattr,
9197         .lookup         = nfs4_proc_lookup,
9198         .access         = nfs4_proc_access,
9199         .readlink       = nfs4_proc_readlink,
9200         .create         = nfs4_proc_create,
9201         .remove         = nfs4_proc_remove,
9202         .unlink_setup   = nfs4_proc_unlink_setup,
9203         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9204         .unlink_done    = nfs4_proc_unlink_done,
9205         .rename_setup   = nfs4_proc_rename_setup,
9206         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9207         .rename_done    = nfs4_proc_rename_done,
9208         .link           = nfs4_proc_link,
9209         .symlink        = nfs4_proc_symlink,
9210         .mkdir          = nfs4_proc_mkdir,
9211         .rmdir          = nfs4_proc_remove,
9212         .readdir        = nfs4_proc_readdir,
9213         .mknod          = nfs4_proc_mknod,
9214         .statfs         = nfs4_proc_statfs,
9215         .fsinfo         = nfs4_proc_fsinfo,
9216         .pathconf       = nfs4_proc_pathconf,
9217         .set_capabilities = nfs4_server_capabilities,
9218         .decode_dirent  = nfs4_decode_dirent,
9219         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9220         .read_setup     = nfs4_proc_read_setup,
9221         .read_done      = nfs4_read_done,
9222         .write_setup    = nfs4_proc_write_setup,
9223         .write_done     = nfs4_write_done,
9224         .commit_setup   = nfs4_proc_commit_setup,
9225         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9226         .commit_done    = nfs4_commit_done,
9227         .lock           = nfs4_proc_lock,
9228         .clear_acl_cache = nfs4_zap_acl_attr,
9229         .close_context  = nfs4_close_context,
9230         .open_context   = nfs4_atomic_open,
9231         .have_delegation = nfs4_have_delegation,
9232         .return_delegation = nfs4_inode_return_delegation,
9233         .alloc_client   = nfs4_alloc_client,
9234         .init_client    = nfs4_init_client,
9235         .free_client    = nfs4_free_client,
9236         .create_server  = nfs4_create_server,
9237         .clone_server   = nfs_clone_server,
9238 };
9239
9240 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9241         .name   = XATTR_NAME_NFSV4_ACL,
9242         .list   = nfs4_xattr_list_nfs4_acl,
9243         .get    = nfs4_xattr_get_nfs4_acl,
9244         .set    = nfs4_xattr_set_nfs4_acl,
9245 };
9246
9247 const struct xattr_handler *nfs4_xattr_handlers[] = {
9248         &nfs4_xattr_nfs4_acl_handler,
9249 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9250         &nfs4_xattr_nfs4_label_handler,
9251 #endif
9252         NULL
9253 };
9254
9255 /*
9256  * Local variables:
9257  *  c-basic-offset: 8
9258  * End:
9259  */