fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state, struct nfs4_label *ilabel,
  87                             struct nfs4_label *olabel);
  88 #ifdef CONFIG_NFS_V4_1
  89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  90                 struct rpc_cred *);
  91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  92                 struct rpc_cred *);
  93 #endif
  94
  95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  96 static inline struct nfs4_label *
  97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  98         struct iattr *sattr, struct nfs4_label *label)
  99 {
 100         int err;
 101
 102         if (label == NULL)
 103                 return NULL;
 104
 105         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 106                 return NULL;
 107
 108         err = security_dentry_init_security(dentry, sattr->ia_mode,
 109                                 &dentry->d_name, (void **)&label->label, &label->len);
 110         if (err == 0)
 111                 return label;
 112
 113         return NULL;
 114 }
 115 static inline void
 116 nfs4_label_release_security(struct nfs4_label *label)
 117 {
 118         if (label)
 119                 security_release_secctx(label->label, label->len);
 120 }
 121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 122 {
 123         if (label)
 124                 return server->attr_bitmask;
 125
 126         return server->attr_bitmask_nl;
 127 }
 128 #else
 129 static inline struct nfs4_label *
 130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 131         struct iattr *sattr, struct nfs4_label *l)
 132 { return NULL; }
 133 static inline void
 134 nfs4_label_release_security(struct nfs4_label *label)
 135 { return; }
 136 static inline u32 *
 137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 { return server->attr_bitmask; }
 139 #endif
 140
 141 /* Prevent leaks of NFSv4 errors into userland */
 142 static int nfs4_map_errors(int err)
 143 {
 144         if (err >= -1000)
 145                 return err;
 146         switch (err) {
 147         case -NFS4ERR_RESOURCE:
 148         case -NFS4ERR_LAYOUTTRYLATER:
 149         case -NFS4ERR_RECALLCONFLICT:
 150                 return -EREMOTEIO;
 151         case -NFS4ERR_WRONGSEC:
 152         case -NFS4ERR_WRONG_CRED:
 153                 return -EPERM;
 154         case -NFS4ERR_BADOWNER:
 155         case -NFS4ERR_BADNAME:
 156                 return -EINVAL;
 157         case -NFS4ERR_SHARE_DENIED:
 158                 return -EACCES;
 159         case -NFS4ERR_MINOR_VERS_MISMATCH:
 160                 return -EPROTONOSUPPORT;
 161         case -NFS4ERR_FILE_OPEN:
 162                 return -EBUSY;
 163         default:
 164                 dprintk("%s could not handle NFSv4 error %d\n",
 165                                 __func__, -err);
 166                 break;
 167         }
 168         return -EIO;
 169 }
 170
 171 /*
 172  * This is our standard bitmap for GETATTR requests.
 173  */
 174 const u32 nfs4_fattr_bitmap[3] = {
 175         FATTR4_WORD0_TYPE
 176         | FATTR4_WORD0_CHANGE
 177         | FATTR4_WORD0_SIZE
 178         | FATTR4_WORD0_FSID
 179         | FATTR4_WORD0_FILEID,
 180         FATTR4_WORD1_MODE
 181         | FATTR4_WORD1_NUMLINKS
 182         | FATTR4_WORD1_OWNER
 183         | FATTR4_WORD1_OWNER_GROUP
 184         | FATTR4_WORD1_RAWDEV
 185         | FATTR4_WORD1_SPACE_USED
 186         | FATTR4_WORD1_TIME_ACCESS
 187         | FATTR4_WORD1_TIME_METADATA
 188         | FATTR4_WORD1_TIME_MODIFY
 189         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 190 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 191         FATTR4_WORD2_SECURITY_LABEL
 192 #endif
 193 };
 194
 195 static const u32 nfs4_pnfs_open_bitmap[3] = {
 196         FATTR4_WORD0_TYPE
 197         | FATTR4_WORD0_CHANGE
 198         | FATTR4_WORD0_SIZE
 199         | FATTR4_WORD0_FSID
 200         | FATTR4_WORD0_FILEID,
 201         FATTR4_WORD1_MODE
 202         | FATTR4_WORD1_NUMLINKS
 203         | FATTR4_WORD1_OWNER
 204         | FATTR4_WORD1_OWNER_GROUP
 205         | FATTR4_WORD1_RAWDEV
 206         | FATTR4_WORD1_SPACE_USED
 207         | FATTR4_WORD1_TIME_ACCESS
 208         | FATTR4_WORD1_TIME_METADATA
 209         | FATTR4_WORD1_TIME_MODIFY,
 210         FATTR4_WORD2_MDSTHRESHOLD
 211 };
 212
 213 static const u32 nfs4_open_noattr_bitmap[3] = {
 214         FATTR4_WORD0_TYPE
 215         | FATTR4_WORD0_CHANGE
 216         | FATTR4_WORD0_FILEID,
 217 };
 218
 219 const u32 nfs4_statfs_bitmap[3] = {
 220         FATTR4_WORD0_FILES_AVAIL
 221         | FATTR4_WORD0_FILES_FREE
 222         | FATTR4_WORD0_FILES_TOTAL,
 223         FATTR4_WORD1_SPACE_AVAIL
 224         | FATTR4_WORD1_SPACE_FREE
 225         | FATTR4_WORD1_SPACE_TOTAL
 226 };
 227
 228 const u32 nfs4_pathconf_bitmap[3] = {
 229         FATTR4_WORD0_MAXLINK
 230         | FATTR4_WORD0_MAXNAME,
 231         0
 232 };
 233
 234 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 235                         | FATTR4_WORD0_MAXREAD
 236                         | FATTR4_WORD0_MAXWRITE
 237                         | FATTR4_WORD0_LEASE_TIME,
 238                         FATTR4_WORD1_TIME_DELTA
 239                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 240                         FATTR4_WORD2_LAYOUT_BLKSIZE
 241 };
 242
 243 const u32 nfs4_fs_locations_bitmap[3] = {
 244         FATTR4_WORD0_TYPE
 245         | FATTR4_WORD0_CHANGE
 246         | FATTR4_WORD0_SIZE
 247         | FATTR4_WORD0_FSID
 248         | FATTR4_WORD0_FILEID
 249         | FATTR4_WORD0_FS_LOCATIONS,
 250         FATTR4_WORD1_MODE
 251         | FATTR4_WORD1_NUMLINKS
 252         | FATTR4_WORD1_OWNER
 253         | FATTR4_WORD1_OWNER_GROUP
 254         | FATTR4_WORD1_RAWDEV
 255         | FATTR4_WORD1_SPACE_USED
 256         | FATTR4_WORD1_TIME_ACCESS
 257         | FATTR4_WORD1_TIME_METADATA
 258         | FATTR4_WORD1_TIME_MODIFY
 259         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 260 };
 261
 262 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 263                 struct nfs4_readdir_arg *readdir)
 264 {
 265         __be32 *start, *p;
 266
 267         if (cookie > 2) {
 268                 readdir->cookie = cookie;
 269                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 270                 return;
 271         }
 272
 273         readdir->cookie = 0;
 274         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 275         if (cookie == 2)
 276                 return;
 277
 278         /*
 279          * NFSv4 servers do not return entries for '.' and '..'
 280          * Therefore, we fake these entries here.  We let '.'
 281          * have cookie 0 and '..' have cookie 1.  Note that
 282          * when talking to the server, we always send cookie 0
 283          * instead of 1 or 2.
 284          */
 285         start = p = kmap_atomic(*readdir->pages);
 286
 287         if (cookie == 0) {
 288                 *p++ = xdr_one;                                  /* next */
 289                 *p++ = xdr_zero;                   /* cookie, first word */
 290                 *p++ = xdr_one;                   /* cookie, second word */
 291                 *p++ = xdr_one;                             /* entry len */
 292                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 293                 p++;
 294                 *p++ = xdr_one;                         /* bitmap length */
 295                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 296                 *p++ = htonl(8);              /* attribute buffer length */
 297                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 298         }
 299
 300         *p++ = xdr_one;                                  /* next */
 301         *p++ = xdr_zero;                   /* cookie, first word */
 302         *p++ = xdr_two;                   /* cookie, second word */
 303         *p++ = xdr_two;                             /* entry len */
 304         memcpy(p, "..\0\0", 4);                         /* entry */
 305         p++;
 306         *p++ = xdr_one;                         /* bitmap length */
 307         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 308         *p++ = htonl(8);              /* attribute buffer length */
 309         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 310
 311         readdir->pgbase = (char *)p - (char *)start;
 312         readdir->count -= readdir->pgbase;
 313         kunmap_atomic(start);
 314 }
 315
 316 static long nfs4_update_delay(long *timeout)
 317 {
 318         long ret;
 319         if (!timeout)
 320                 return NFS4_POLL_RETRY_MAX;
 321         if (*timeout <= 0)
 322                 *timeout = NFS4_POLL_RETRY_MIN;
 323         if (*timeout > NFS4_POLL_RETRY_MAX)
 324                 *timeout = NFS4_POLL_RETRY_MAX;
 325         ret = *timeout;
 326         *timeout <<= 1;
 327         return ret;
 328 }
 329
 330 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 331 {
 332         int res = 0;
 333
 334         might_sleep();
 335
 336         freezable_schedule_timeout_killable_unsafe(
 337                 nfs4_update_delay(timeout));
 338         if (fatal_signal_pending(current))
 339                 res = -ERESTARTSYS;
 340         return res;
 341 }
 342
 343 /* This is the error handling routine for processes that are allowed
 344  * to sleep.
 345  */
 346 static int nfs4_do_handle_exception(struct nfs_server *server,
 347                 int errorcode, struct nfs4_exception *exception)
 348 {
 349         struct nfs_client *clp = server->nfs_client;
 350         struct nfs4_state *state = exception->state;
 351         struct inode *inode = exception->inode;
 352         int ret = errorcode;
 353
 354         exception->delay = 0;
 355         exception->recovering = 0;
 356         exception->retry = 0;
 357         switch(errorcode) {
 358                 case 0:
 359                         return 0;
 360                 case -NFS4ERR_OPENMODE:
 361                 case -NFS4ERR_DELEG_REVOKED:
 362                 case -NFS4ERR_ADMIN_REVOKED:
 363                 case -NFS4ERR_BAD_STATEID:
 364                         if (inode && nfs_async_inode_return_delegation(inode,
 365                                                 NULL) == 0)
 366                                 goto wait_on_recovery;
 367                         if (state == NULL)
 368                                 break;
 369                         ret = nfs4_schedule_stateid_recovery(server, state);
 370                         if (ret < 0)
 371                                 break;
 372                         goto wait_on_recovery;
 373                 case -NFS4ERR_EXPIRED:
 374                         if (state != NULL) {
 375                                 ret = nfs4_schedule_stateid_recovery(server, state);
 376                                 if (ret < 0)
 377                                         break;
 378                         }
 379                 case -NFS4ERR_STALE_STATEID:
 380                 case -NFS4ERR_STALE_CLIENTID:
 381                         nfs4_schedule_lease_recovery(clp);
 382                         goto wait_on_recovery;
 383                 case -NFS4ERR_MOVED:
 384                         ret = nfs4_schedule_migration_recovery(server);
 385                         if (ret < 0)
 386                                 break;
 387                         goto wait_on_recovery;
 388                 case -NFS4ERR_LEASE_MOVED:
 389                         nfs4_schedule_lease_moved_recovery(clp);
 390                         goto wait_on_recovery;
 391 #if defined(CONFIG_NFS_V4_1)
 392                 case -NFS4ERR_BADSESSION:
 393                 case -NFS4ERR_BADSLOT:
 394                 case -NFS4ERR_BAD_HIGH_SLOT:
 395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 396                 case -NFS4ERR_DEADSESSION:
 397                 case -NFS4ERR_SEQ_FALSE_RETRY:
 398                 case -NFS4ERR_SEQ_MISORDERED:
 399                         dprintk("%s ERROR: %d Reset session\n", __func__,
 400                                 errorcode);
 401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 402                         goto wait_on_recovery;
 403 #endif /* defined(CONFIG_NFS_V4_1) */
 404                 case -NFS4ERR_FILE_OPEN:
 405                         if (exception->timeout > HZ) {
 406                                 /* We have retried a decent amount, time to
 407                                  * fail
 408                                  */
 409                                 ret = -EBUSY;
 410                                 break;
 411                         }
 412                 case -NFS4ERR_DELAY:
 413                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 414                 case -NFS4ERR_GRACE:
 415                         exception->delay = 1;
 416                         return 0;
 417
 418                 case -NFS4ERR_RETRY_UNCACHED_REP:
 419                 case -NFS4ERR_OLD_STATEID:
 420                         exception->retry = 1;
 421                         break;
 422                 case -NFS4ERR_BADOWNER:
 423                         /* The following works around a Linux server bug! */
 424                 case -NFS4ERR_BADNAME:
 425                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 426                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 427                                 exception->retry = 1;
 428                                 printk(KERN_WARNING "NFS: v4 server %s "
 429                                                 "does not accept raw "
 430                                                 "uid/gids. "
 431                                                 "Reenabling the idmapper.\n",
 432                                                 server->nfs_client->cl_hostname);
 433                         }
 434         }
 435         /* We failed to handle the error */
 436         return nfs4_map_errors(ret);
 437 wait_on_recovery:
 438         exception->recovering = 1;
 439         return 0;
 440 }
 441
 442 /* This is the error handling routine for processes that are allowed
 443  * to sleep.
 444  */
 445 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 446 {
 447         struct nfs_client *clp = server->nfs_client;
 448         int ret;
 449
 450         ret = nfs4_do_handle_exception(server, errorcode, exception);
 451         if (exception->delay) {
 452                 ret = nfs4_delay(server->client, &exception->timeout);
 453                 goto out_retry;
 454         }
 455         if (exception->recovering) {
 456                 ret = nfs4_wait_clnt_recover(clp);
 457                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 458                         return -EIO;
 459                 goto out_retry;
 460         }
 461         return ret;
 462 out_retry:
 463         if (ret == 0)
 464                 exception->retry = 1;
 465         return ret;
 466 }
 467
 468 static int
 469 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 470                 int errorcode, struct nfs4_exception *exception)
 471 {
 472         struct nfs_client *clp = server->nfs_client;
 473         int ret;
 474
 475         ret = nfs4_do_handle_exception(server, errorcode, exception);
 476         if (exception->delay) {
 477                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 478                 goto out_retry;
 479         }
 480         if (exception->recovering) {
 481                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 482                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 483                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 484                 goto out_retry;
 485         }
 486         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 487                 ret = -EIO;
 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_error(struct rpc_task *task, struct nfs_server *server,
 497                         struct nfs4_state *state, long *timeout)
 498 {
 499         struct nfs4_exception exception = {
 500                 .state = state,
 501         };
 502
 503         if (task->tk_status >= 0)
 504                 return 0;
 505         if (timeout)
 506                 exception.timeout = *timeout;
 507         task->tk_status = nfs4_async_handle_exception(task, server,
 508                         task->tk_status,
 509                         &exception);
 510         if (exception.delay && timeout)
 511                 *timeout = exception.timeout;
 512         if (exception.retry)
 513                 return -EAGAIN;
 514         return 0;
 515 }
 516
 517 /*
 518  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 519  * or 'false' otherwise.
 520  */
 521 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 522 {
 523         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 524
 525         if (flavor == RPC_AUTH_GSS_KRB5I ||
 526             flavor == RPC_AUTH_GSS_KRB5P)
 527                 return true;
 528
 529         return false;
 530 }
 531
 532 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 533 {
 534         spin_lock(&clp->cl_lock);
 535         if (time_before(clp->cl_last_renewal,timestamp))
 536                 clp->cl_last_renewal = timestamp;
 537         spin_unlock(&clp->cl_lock);
 538 }
 539
 540 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 541 {
 542         struct nfs_client *clp = server->nfs_client;
 543
 544         if (!nfs4_has_session(clp))
 545                 do_renew_lease(clp, timestamp);
 546 }
 547
 548 struct nfs4_call_sync_data {
 549         const struct nfs_server *seq_server;
 550         struct nfs4_sequence_args *seq_args;
 551         struct nfs4_sequence_res *seq_res;
 552 };
 553
 554 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 555                         struct nfs4_sequence_res *res, int cache_reply)
 556 {
 557         args->sa_slot = NULL;
 558         args->sa_cache_this = cache_reply;
 559         args->sa_privileged = 0;
 560
 561         res->sr_slot = NULL;
 562 }
 563
 564 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 565 {
 566         args->sa_privileged = 1;
 567 }
 568
 569 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 570                          struct nfs4_sequence_args *args,
 571                          struct nfs4_sequence_res *res,
 572                          struct rpc_task *task)
 573 {
 574         struct nfs4_slot *slot;
 575
 576         /* slot already allocated? */
 577         if (res->sr_slot != NULL)
 578                 goto out_start;
 579
 580         spin_lock(&tbl->slot_tbl_lock);
 581         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 582                 goto out_sleep;
 583
 584         slot = nfs4_alloc_slot(tbl);
 585         if (IS_ERR(slot)) {
 586                 if (slot == ERR_PTR(-ENOMEM))
 587                         task->tk_timeout = HZ >> 2;
 588                 goto out_sleep;
 589         }
 590         spin_unlock(&tbl->slot_tbl_lock);
 591
 592         args->sa_slot = slot;
 593         res->sr_slot = slot;
 594
 595 out_start:
 596         rpc_call_start(task);
 597         return 0;
 598
 599 out_sleep:
 600         if (args->sa_privileged)
 601                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 602                                 NULL, RPC_PRIORITY_PRIVILEGED);
 603         else
 604                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 605         spin_unlock(&tbl->slot_tbl_lock);
 606         return -EAGAIN;
 607 }
 608 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 609
 610 static int nfs40_sequence_done(struct rpc_task *task,
 611                                struct nfs4_sequence_res *res)
 612 {
 613         struct nfs4_slot *slot = res->sr_slot;
 614         struct nfs4_slot_table *tbl;
 615
 616         if (slot == NULL)
 617                 goto out;
 618
 619         tbl = slot->table;
 620         spin_lock(&tbl->slot_tbl_lock);
 621         if (!nfs41_wake_and_assign_slot(tbl, slot))
 622                 nfs4_free_slot(tbl, slot);
 623         spin_unlock(&tbl->slot_tbl_lock);
 624
 625         res->sr_slot = NULL;
 626 out:
 627         return 1;
 628 }
 629
 630 #if defined(CONFIG_NFS_V4_1)
 631
 632 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 633 {
 634         struct nfs4_session *session;
 635         struct nfs4_slot_table *tbl;
 636         struct nfs4_slot *slot = res->sr_slot;
 637         bool send_new_highest_used_slotid = false;
 638
 639         tbl = slot->table;
 640         session = tbl->session;
 641
 642         spin_lock(&tbl->slot_tbl_lock);
 643         /* Be nice to the server: try to ensure that the last transmitted
 644          * value for highest_user_slotid <= target_highest_slotid
 645          */
 646         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 647                 send_new_highest_used_slotid = true;
 648
 649         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 650                 send_new_highest_used_slotid = false;
 651                 goto out_unlock;
 652         }
 653         nfs4_free_slot(tbl, slot);
 654
 655         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 656                 send_new_highest_used_slotid = false;
 657 out_unlock:
 658         spin_unlock(&tbl->slot_tbl_lock);
 659         res->sr_slot = NULL;
 660         if (send_new_highest_used_slotid)
 661                 nfs41_notify_server(session->clp);
 662 }
 663
 664 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 665 {
 666         struct nfs4_session *session;
 667         struct nfs4_slot *slot = res->sr_slot;
 668         struct nfs_client *clp;
 669         bool interrupted = false;
 670         int ret = 1;
 671
 672         if (slot == NULL)
 673                 goto out_noaction;
 674         /* don't increment the sequence number if the task wasn't sent */
 675         if (!RPC_WAS_SENT(task))
 676                 goto out;
 677
 678         session = slot->table->session;
 679
 680         if (slot->interrupted) {
 681                 slot->interrupted = 0;
 682                 interrupted = true;
 683         }
 684
 685         trace_nfs4_sequence_done(session, res);
 686         /* Check the SEQUENCE operation status */
 687         switch (res->sr_status) {
 688         case 0:
 689                 /* Update the slot's sequence and clientid lease timer */
 690                 ++slot->seq_nr;
 691                 clp = session->clp;
 692                 do_renew_lease(clp, res->sr_timestamp);
 693                 /* Check sequence flags */
 694                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 695                 nfs41_update_target_slotid(slot->table, slot, res);
 696                 break;
 697         case 1:
 698                 /*
 699                  * sr_status remains 1 if an RPC level error occurred.
 700                  * The server may or may not have processed the sequence
 701                  * operation..
 702                  * Mark the slot as having hosted an interrupted RPC call.
 703                  */
 704                 slot->interrupted = 1;
 705                 goto out;
 706         case -NFS4ERR_DELAY:
 707                 /* The server detected a resend of the RPC call and
 708                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 709                  * of RFC5661.
 710                  */
 711                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 712                         __func__,
 713                         slot->slot_nr,
 714                         slot->seq_nr);
 715                 goto out_retry;
 716         case -NFS4ERR_BADSLOT:
 717                 /*
 718                  * The slot id we used was probably retired. Try again
 719                  * using a different slot id.
 720                  */
 721                 goto retry_nowait;
 722         case -NFS4ERR_SEQ_MISORDERED:
 723                 /*
 724                  * Was the last operation on this sequence interrupted?
 725                  * If so, retry after bumping the sequence number.
 726                  */
 727                 if (interrupted) {
 728                         ++slot->seq_nr;
 729                         goto retry_nowait;
 730                 }
 731                 /*
 732                  * Could this slot have been previously retired?
 733                  * If so, then the server may be expecting seq_nr = 1!
 734                  */
 735                 if (slot->seq_nr != 1) {
 736                         slot->seq_nr = 1;
 737                         goto retry_nowait;
 738                 }
 739                 break;
 740         case -NFS4ERR_SEQ_FALSE_RETRY:
 741                 ++slot->seq_nr;
 742                 goto retry_nowait;
 743         default:
 744                 /* Just update the slot sequence no. */
 745                 ++slot->seq_nr;
 746         }
 747 out:
 748         /* The session may be reset by one of the error handlers. */
 749         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 750         nfs41_sequence_free_slot(res);
 751 out_noaction:
 752         return ret;
 753 retry_nowait:
 754         if (rpc_restart_call_prepare(task)) {
 755                 task->tk_status = 0;
 756                 ret = 0;
 757         }
 758         goto out;
 759 out_retry:
 760         if (!rpc_restart_call(task))
 761                 goto out;
 762         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 763         return 0;
 764 }
 765 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 766
 767 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 768 {
 769         if (res->sr_slot == NULL)
 770                 return 1;
 771         if (!res->sr_slot->table->session)
 772                 return nfs40_sequence_done(task, res);
 773         return nfs41_sequence_done(task, res);
 774 }
 775 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 776
 777 int nfs41_setup_sequence(struct nfs4_session *session,
 778                                 struct nfs4_sequence_args *args,
 779                                 struct nfs4_sequence_res *res,
 780                                 struct rpc_task *task)
 781 {
 782         struct nfs4_slot *slot;
 783         struct nfs4_slot_table *tbl;
 784
 785         dprintk("--> %s\n", __func__);
 786         /* slot already allocated? */
 787         if (res->sr_slot != NULL)
 788                 goto out_success;
 789
 790         tbl = &session->fc_slot_table;
 791
 792         task->tk_timeout = 0;
 793
 794         spin_lock(&tbl->slot_tbl_lock);
 795         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 796             !args->sa_privileged) {
 797                 /* The state manager will wait until the slot table is empty */
 798                 dprintk("%s session is draining\n", __func__);
 799                 goto out_sleep;
 800         }
 801
 802         slot = nfs4_alloc_slot(tbl);
 803         if (IS_ERR(slot)) {
 804                 /* If out of memory, try again in 1/4 second */
 805                 if (slot == ERR_PTR(-ENOMEM))
 806                         task->tk_timeout = HZ >> 2;
 807                 dprintk("<-- %s: no free slots\n", __func__);
 808                 goto out_sleep;
 809         }
 810         spin_unlock(&tbl->slot_tbl_lock);
 811
 812         args->sa_slot = slot;
 813
 814         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 815                         slot->slot_nr, slot->seq_nr);
 816
 817         res->sr_slot = slot;
 818         res->sr_timestamp = jiffies;
 819         res->sr_status_flags = 0;
 820         /*
 821          * sr_status is only set in decode_sequence, and so will remain
 822          * set to 1 if an rpc level failure occurs.
 823          */
 824         res->sr_status = 1;
 825         trace_nfs4_setup_sequence(session, args);
 826 out_success:
 827         rpc_call_start(task);
 828         return 0;
 829 out_sleep:
 830         /* Privileged tasks are queued with top priority */
 831         if (args->sa_privileged)
 832                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 833                                 NULL, RPC_PRIORITY_PRIVILEGED);
 834         else
 835                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 836         spin_unlock(&tbl->slot_tbl_lock);
 837         return -EAGAIN;
 838 }
 839 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 840
 841 static int nfs4_setup_sequence(const struct nfs_server *server,
 842                                struct nfs4_sequence_args *args,
 843                                struct nfs4_sequence_res *res,
 844                                struct rpc_task *task)
 845 {
 846         struct nfs4_session *session = nfs4_get_session(server);
 847         int ret = 0;
 848
 849         if (!session)
 850                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 851                                             args, res, task);
 852
 853         dprintk("--> %s clp %p session %p sr_slot %u\n",
 854                 __func__, session->clp, session, res->sr_slot ?
 855                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 856
 857         ret = nfs41_setup_sequence(session, args, res, task);
 858
 859         dprintk("<-- %s status=%d\n", __func__, ret);
 860         return ret;
 861 }
 862
 863 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 864 {
 865         struct nfs4_call_sync_data *data = calldata;
 866         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 867
 868         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 869
 870         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 871 }
 872
 873 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 874 {
 875         struct nfs4_call_sync_data *data = calldata;
 876
 877         nfs41_sequence_done(task, data->seq_res);
 878 }
 879
 880 static const struct rpc_call_ops nfs41_call_sync_ops = {
 881         .rpc_call_prepare = nfs41_call_sync_prepare,
 882         .rpc_call_done = nfs41_call_sync_done,
 883 };
 884
 885 #else   /* !CONFIG_NFS_V4_1 */
 886
 887 static int nfs4_setup_sequence(const struct nfs_server *server,
 888                                struct nfs4_sequence_args *args,
 889                                struct nfs4_sequence_res *res,
 890                                struct rpc_task *task)
 891 {
 892         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 893                                     args, res, task);
 894 }
 895
 896 int nfs4_sequence_done(struct rpc_task *task,
 897                        struct nfs4_sequence_res *res)
 898 {
 899         return nfs40_sequence_done(task, res);
 900 }
 901 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 902
 903 #endif  /* !CONFIG_NFS_V4_1 */
 904
 905 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 906 {
 907         struct nfs4_call_sync_data *data = calldata;
 908         nfs4_setup_sequence(data->seq_server,
 909                                 data->seq_args, data->seq_res, task);
 910 }
 911
 912 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 913 {
 914         struct nfs4_call_sync_data *data = calldata;
 915         nfs4_sequence_done(task, data->seq_res);
 916 }
 917
 918 static const struct rpc_call_ops nfs40_call_sync_ops = {
 919         .rpc_call_prepare = nfs40_call_sync_prepare,
 920         .rpc_call_done = nfs40_call_sync_done,
 921 };
 922
 923 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 924                                    struct nfs_server *server,
 925                                    struct rpc_message *msg,
 926                                    struct nfs4_sequence_args *args,
 927                                    struct nfs4_sequence_res *res)
 928 {
 929         int ret;
 930         struct rpc_task *task;
 931         struct nfs_client *clp = server->nfs_client;
 932         struct nfs4_call_sync_data data = {
 933                 .seq_server = server,
 934                 .seq_args = args,
 935                 .seq_res = res,
 936         };
 937         struct rpc_task_setup task_setup = {
 938                 .rpc_client = clnt,
 939                 .rpc_message = msg,
 940                 .callback_ops = clp->cl_mvops->call_sync_ops,
 941                 .callback_data = &data
 942         };
 943
 944         task = rpc_run_task(&task_setup);
 945         if (IS_ERR(task))
 946                 ret = PTR_ERR(task);
 947         else {
 948                 ret = task->tk_status;
 949                 rpc_put_task(task);
 950         }
 951         return ret;
 952 }
 953
 954 int nfs4_call_sync(struct rpc_clnt *clnt,
 955                    struct nfs_server *server,
 956                    struct rpc_message *msg,
 957                    struct nfs4_sequence_args *args,
 958                    struct nfs4_sequence_res *res,
 959                    int cache_reply)
 960 {
 961         nfs4_init_sequence(args, res, cache_reply);
 962         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 963 }
 964
 965 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 966 {
 967         struct nfs_inode *nfsi = NFS_I(dir);
 968
 969         spin_lock(&dir->i_lock);
 970         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 971         if (!cinfo->atomic || cinfo->before != dir->i_version)
 972                 nfs_force_lookup_revalidate(dir);
 973         dir->i_version = cinfo->after;
 974         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 975         nfs_fscache_invalidate(dir);
 976         spin_unlock(&dir->i_lock);
 977 }
 978
 979 struct nfs4_opendata {
 980         struct kref kref;
 981         struct nfs_openargs o_arg;
 982         struct nfs_openres o_res;
 983         struct nfs_open_confirmargs c_arg;
 984         struct nfs_open_confirmres c_res;
 985         struct nfs4_string owner_name;
 986         struct nfs4_string group_name;
 987         struct nfs4_label *a_label;
 988         struct nfs_fattr f_attr;
 989         struct nfs4_label *f_label;
 990         struct dentry *dir;
 991         struct dentry *dentry;
 992         struct nfs4_state_owner *owner;
 993         struct nfs4_state *state;
 994         struct iattr attrs;
 995         unsigned long timestamp;
 996         unsigned int rpc_done : 1;
 997         unsigned int file_created : 1;
 998         unsigned int is_recover : 1;
 999         int rpc_status;
1000         int cancelled;
1001 };
1002
1003 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1004                 int err, struct nfs4_exception *exception)
1005 {
1006         if (err != -EINVAL)
1007                 return false;
1008         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1009                 return false;
1010         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1011         exception->retry = 1;
1012         return true;
1013 }
1014
1015 static u32
1016 nfs4_map_atomic_open_share(struct nfs_server *server,
1017                 fmode_t fmode, int openflags)
1018 {
1019         u32 res = 0;
1020
1021         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1022         case FMODE_READ:
1023                 res = NFS4_SHARE_ACCESS_READ;
1024                 break;
1025         case FMODE_WRITE:
1026                 res = NFS4_SHARE_ACCESS_WRITE;
1027                 break;
1028         case FMODE_READ|FMODE_WRITE:
1029                 res = NFS4_SHARE_ACCESS_BOTH;
1030         }
1031         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1032                 goto out;
1033         /* Want no delegation if we're using O_DIRECT */
1034         if (openflags & O_DIRECT)
1035                 res |= NFS4_SHARE_WANT_NO_DELEG;
1036 out:
1037         return res;
1038 }
1039
1040 static enum open_claim_type4
1041 nfs4_map_atomic_open_claim(struct nfs_server *server,
1042                 enum open_claim_type4 claim)
1043 {
1044         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1045                 return claim;
1046         switch (claim) {
1047         default:
1048                 return claim;
1049         case NFS4_OPEN_CLAIM_FH:
1050                 return NFS4_OPEN_CLAIM_NULL;
1051         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1052                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1053         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1054                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1055         }
1056 }
1057
1058 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1059 {
1060         p->o_res.f_attr = &p->f_attr;
1061         p->o_res.f_label = p->f_label;
1062         p->o_res.seqid = p->o_arg.seqid;
1063         p->c_res.seqid = p->c_arg.seqid;
1064         p->o_res.server = p->o_arg.server;
1065         p->o_res.access_request = p->o_arg.access;
1066         nfs_fattr_init(&p->f_attr);
1067         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1068 }
1069
1070 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1071                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1072                 const struct iattr *attrs,
1073                 struct nfs4_label *label,
1074                 enum open_claim_type4 claim,
1075                 gfp_t gfp_mask)
1076 {
1077         struct dentry *parent = dget_parent(dentry);
1078         struct inode *dir = d_inode(parent);
1079         struct nfs_server *server = NFS_SERVER(dir);
1080         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1081         struct nfs4_opendata *p;
1082
1083         p = kzalloc(sizeof(*p), gfp_mask);
1084         if (p == NULL)
1085                 goto err;
1086
1087         p->f_label = nfs4_label_alloc(server, gfp_mask);
1088         if (IS_ERR(p->f_label))
1089                 goto err_free_p;
1090
1091         p->a_label = nfs4_label_alloc(server, gfp_mask);
1092         if (IS_ERR(p->a_label))
1093                 goto err_free_f;
1094
1095         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1096         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1097         if (IS_ERR(p->o_arg.seqid))
1098                 goto err_free_label;
1099         nfs_sb_active(dentry->d_sb);
1100         p->dentry = dget(dentry);
1101         p->dir = parent;
1102         p->owner = sp;
1103         atomic_inc(&sp->so_count);
1104         p->o_arg.open_flags = flags;
1105         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1106         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1107                         fmode, flags);
1108         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1109          * will return permission denied for all bits until close */
1110         if (!(flags & O_EXCL)) {
1111                 /* ask server to check for all possible rights as results
1112                  * are cached */
1113                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1114                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1115         }
1116         p->o_arg.clientid = server->nfs_client->cl_clientid;
1117         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1118         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1119         p->o_arg.name = &dentry->d_name;
1120         p->o_arg.server = server;
1121         p->o_arg.bitmask = nfs4_bitmask(server, label);
1122         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1123         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1124         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1125         switch (p->o_arg.claim) {
1126         case NFS4_OPEN_CLAIM_NULL:
1127         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1128         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1129                 p->o_arg.fh = NFS_FH(dir);
1130                 break;
1131         case NFS4_OPEN_CLAIM_PREVIOUS:
1132         case NFS4_OPEN_CLAIM_FH:
1133         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1134         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1135                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1136         }
1137         if (attrs != NULL && attrs->ia_valid != 0) {
1138                 __u32 verf[2];
1139
1140                 p->o_arg.u.attrs = &p->attrs;
1141                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1142
1143                 verf[0] = jiffies;
1144                 verf[1] = current->pid;
1145                 memcpy(p->o_arg.u.verifier.data, verf,
1146                                 sizeof(p->o_arg.u.verifier.data));
1147         }
1148         p->c_arg.fh = &p->o_res.fh;
1149         p->c_arg.stateid = &p->o_res.stateid;
1150         p->c_arg.seqid = p->o_arg.seqid;
1151         nfs4_init_opendata_res(p);
1152         kref_init(&p->kref);
1153         return p;
1154
1155 err_free_label:
1156         nfs4_label_free(p->a_label);
1157 err_free_f:
1158         nfs4_label_free(p->f_label);
1159 err_free_p:
1160         kfree(p);
1161 err:
1162         dput(parent);
1163         return NULL;
1164 }
1165
1166 static void nfs4_opendata_free(struct kref *kref)
1167 {
1168         struct nfs4_opendata *p = container_of(kref,
1169                         struct nfs4_opendata, kref);
1170         struct super_block *sb = p->dentry->d_sb;
1171
1172         nfs_free_seqid(p->o_arg.seqid);
1173         if (p->state != NULL)
1174                 nfs4_put_open_state(p->state);
1175         nfs4_put_state_owner(p->owner);
1176
1177         nfs4_label_free(p->a_label);
1178         nfs4_label_free(p->f_label);
1179
1180         dput(p->dir);
1181         dput(p->dentry);
1182         nfs_sb_deactive(sb);
1183         nfs_fattr_free_names(&p->f_attr);
1184         kfree(p->f_attr.mdsthreshold);
1185         kfree(p);
1186 }
1187
1188 static void nfs4_opendata_put(struct nfs4_opendata *p)
1189 {
1190         if (p != NULL)
1191                 kref_put(&p->kref, nfs4_opendata_free);
1192 }
1193
1194 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1195 {
1196         int ret;
1197
1198         ret = rpc_wait_for_completion_task(task);
1199         return ret;
1200 }
1201
1202 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1203                 fmode_t fmode)
1204 {
1205         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1206         case FMODE_READ|FMODE_WRITE:
1207                 return state->n_rdwr != 0;
1208         case FMODE_WRITE:
1209                 return state->n_wronly != 0;
1210         case FMODE_READ:
1211                 return state->n_rdonly != 0;
1212         }
1213         WARN_ON_ONCE(1);
1214         return false;
1215 }
1216
1217 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1218 {
1219         int ret = 0;
1220
1221         if (open_mode & (O_EXCL|O_TRUNC))
1222                 goto out;
1223         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1224                 case FMODE_READ:
1225                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1226                                 && state->n_rdonly != 0;
1227                         break;
1228                 case FMODE_WRITE:
1229                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1230                                 && state->n_wronly != 0;
1231                         break;
1232                 case FMODE_READ|FMODE_WRITE:
1233                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1234                                 && state->n_rdwr != 0;
1235         }
1236 out:
1237         return ret;
1238 }
1239
1240 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1241                 enum open_claim_type4 claim)
1242 {
1243         if (delegation == NULL)
1244                 return 0;
1245         if ((delegation->type & fmode) != fmode)
1246                 return 0;
1247         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1248                 return 0;
1249         switch (claim) {
1250         case NFS4_OPEN_CLAIM_NULL:
1251         case NFS4_OPEN_CLAIM_FH:
1252                 break;
1253         case NFS4_OPEN_CLAIM_PREVIOUS:
1254                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1255                         break;
1256         default:
1257                 return 0;
1258         }
1259         nfs_mark_delegation_referenced(delegation);
1260         return 1;
1261 }
1262
1263 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1264 {
1265         switch (fmode) {
1266                 case FMODE_WRITE:
1267                         state->n_wronly++;
1268                         break;
1269                 case FMODE_READ:
1270                         state->n_rdonly++;
1271                         break;
1272                 case FMODE_READ|FMODE_WRITE:
1273                         state->n_rdwr++;
1274         }
1275         nfs4_state_set_mode_locked(state, state->state | fmode);
1276 }
1277
1278 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1279 {
1280         struct nfs_client *clp = state->owner->so_server->nfs_client;
1281         bool need_recover = false;
1282
1283         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1284                 need_recover = true;
1285         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1286                 need_recover = true;
1287         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1288                 need_recover = true;
1289         if (need_recover)
1290                 nfs4_state_mark_reclaim_nograce(clp, state);
1291 }
1292
1293 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1294                 nfs4_stateid *stateid)
1295 {
1296         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1297                 return true;
1298         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1299                 nfs_test_and_clear_all_open_stateid(state);
1300                 return true;
1301         }
1302         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1303                 return true;
1304         return false;
1305 }
1306
1307 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1308 {
1309         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1310                 return;
1311         if (state->n_wronly)
1312                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1313         if (state->n_rdonly)
1314                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1315         if (state->n_rdwr)
1316                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1317         set_bit(NFS_OPEN_STATE, &state->flags);
1318 }
1319
1320 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1321                 nfs4_stateid *arg_stateid,
1322                 nfs4_stateid *stateid, fmode_t fmode)
1323 {
1324         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1325         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1326         case FMODE_WRITE:
1327                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1328                 break;
1329         case FMODE_READ:
1330                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1331                 break;
1332         case 0:
1333                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1334                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1335                 clear_bit(NFS_OPEN_STATE, &state->flags);
1336         }
1337         if (stateid == NULL)
1338                 return;
1339         /* Handle races with OPEN */
1340         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1341             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1342             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1343                 nfs_resync_open_stateid_locked(state);
1344                 return;
1345         }
1346         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1347                 nfs4_stateid_copy(&state->stateid, stateid);
1348         nfs4_stateid_copy(&state->open_stateid, stateid);
1349 }
1350
1351 static void nfs_clear_open_stateid(struct nfs4_state *state,
1352         nfs4_stateid *arg_stateid,
1353         nfs4_stateid *stateid, fmode_t fmode)
1354 {
1355         write_seqlock(&state->seqlock);
1356         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1357         write_sequnlock(&state->seqlock);
1358         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1359                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1360 }
1361
1362 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1363 {
1364         switch (fmode) {
1365                 case FMODE_READ:
1366                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1367                         break;
1368                 case FMODE_WRITE:
1369                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1370                         break;
1371                 case FMODE_READ|FMODE_WRITE:
1372                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1373         }
1374         if (!nfs_need_update_open_stateid(state, stateid))
1375                 return;
1376         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1377                 nfs4_stateid_copy(&state->stateid, stateid);
1378         nfs4_stateid_copy(&state->open_stateid, stateid);
1379 }
1380
1381 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1382 {
1383         /*
1384          * Protect the call to nfs4_state_set_mode_locked and
1385          * serialise the stateid update
1386          */
1387         write_seqlock(&state->seqlock);
1388         if (deleg_stateid != NULL) {
1389                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1390                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1391         }
1392         if (open_stateid != NULL)
1393                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1394         write_sequnlock(&state->seqlock);
1395         spin_lock(&state->owner->so_lock);
1396         update_open_stateflags(state, fmode);
1397         spin_unlock(&state->owner->so_lock);
1398 }
1399
1400 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1401 {
1402         struct nfs_inode *nfsi = NFS_I(state->inode);
1403         struct nfs_delegation *deleg_cur;
1404         int ret = 0;
1405
1406         fmode &= (FMODE_READ|FMODE_WRITE);
1407
1408         rcu_read_lock();
1409         deleg_cur = rcu_dereference(nfsi->delegation);
1410         if (deleg_cur == NULL)
1411                 goto no_delegation;
1412
1413         spin_lock(&deleg_cur->lock);
1414         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1415            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1416             (deleg_cur->type & fmode) != fmode)
1417                 goto no_delegation_unlock;
1418
1419         if (delegation == NULL)
1420                 delegation = &deleg_cur->stateid;
1421         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1422                 goto no_delegation_unlock;
1423
1424         nfs_mark_delegation_referenced(deleg_cur);
1425         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1426         ret = 1;
1427 no_delegation_unlock:
1428         spin_unlock(&deleg_cur->lock);
1429 no_delegation:
1430         rcu_read_unlock();
1431
1432         if (!ret && open_stateid != NULL) {
1433                 __update_open_stateid(state, open_stateid, NULL, fmode);
1434                 ret = 1;
1435         }
1436         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1437                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1438
1439         return ret;
1440 }
1441
1442 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1443                 const nfs4_stateid *stateid)
1444 {
1445         struct nfs4_state *state = lsp->ls_state;
1446         bool ret = false;
1447
1448         spin_lock(&state->state_lock);
1449         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1450                 goto out_noupdate;
1451         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1452                 goto out_noupdate;
1453         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1454         ret = true;
1455 out_noupdate:
1456         spin_unlock(&state->state_lock);
1457         return ret;
1458 }
1459
1460 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1461 {
1462         struct nfs_delegation *delegation;
1463
1464         rcu_read_lock();
1465         delegation = rcu_dereference(NFS_I(inode)->delegation);
1466         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1467                 rcu_read_unlock();
1468                 return;
1469         }
1470         rcu_read_unlock();
1471         nfs4_inode_return_delegation(inode);
1472 }
1473
1474 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1475 {
1476         struct nfs4_state *state = opendata->state;
1477         struct nfs_inode *nfsi = NFS_I(state->inode);
1478         struct nfs_delegation *delegation;
1479         int open_mode = opendata->o_arg.open_flags;
1480         fmode_t fmode = opendata->o_arg.fmode;
1481         enum open_claim_type4 claim = opendata->o_arg.claim;
1482         nfs4_stateid stateid;
1483         int ret = -EAGAIN;
1484
1485         for (;;) {
1486                 spin_lock(&state->owner->so_lock);
1487                 if (can_open_cached(state, fmode, open_mode)) {
1488                         update_open_stateflags(state, fmode);
1489                         spin_unlock(&state->owner->so_lock);
1490                         goto out_return_state;
1491                 }
1492                 spin_unlock(&state->owner->so_lock);
1493                 rcu_read_lock();
1494                 delegation = rcu_dereference(nfsi->delegation);
1495                 if (!can_open_delegated(delegation, fmode, claim)) {
1496                         rcu_read_unlock();
1497                         break;
1498                 }
1499                 /* Save the delegation */
1500                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1501                 rcu_read_unlock();
1502                 nfs_release_seqid(opendata->o_arg.seqid);
1503                 if (!opendata->is_recover) {
1504                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1505                         if (ret != 0)
1506                                 goto out;
1507                 }
1508                 ret = -EAGAIN;
1509
1510                 /* Try to update the stateid using the delegation */
1511                 if (update_open_stateid(state, NULL, &stateid, fmode))
1512                         goto out_return_state;
1513         }
1514 out:
1515         return ERR_PTR(ret);
1516 out_return_state:
1517         atomic_inc(&state->count);
1518         return state;
1519 }
1520
1521 static void
1522 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1523 {
1524         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1525         struct nfs_delegation *delegation;
1526         int delegation_flags = 0;
1527
1528         rcu_read_lock();
1529         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1530         if (delegation)
1531                 delegation_flags = delegation->flags;
1532         rcu_read_unlock();
1533         switch (data->o_arg.claim) {
1534         default:
1535                 break;
1536         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1537         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1538                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1539                                    "returning a delegation for "
1540                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1541                                    clp->cl_hostname);
1542                 return;
1543         }
1544         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1545                 nfs_inode_set_delegation(state->inode,
1546                                          data->owner->so_cred,
1547                                          &data->o_res);
1548         else
1549                 nfs_inode_reclaim_delegation(state->inode,
1550                                              data->owner->so_cred,
1551                                              &data->o_res);
1552 }
1553
1554 /*
1555  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1556  * and update the nfs4_state.
1557  */
1558 static struct nfs4_state *
1559 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1560 {
1561         struct inode *inode = data->state->inode;
1562         struct nfs4_state *state = data->state;
1563         int ret;
1564
1565         if (!data->rpc_done) {
1566                 if (data->rpc_status) {
1567                         ret = data->rpc_status;
1568                         goto err;
1569                 }
1570                 /* cached opens have already been processed */
1571                 goto update;
1572         }
1573
1574         ret = nfs_refresh_inode(inode, &data->f_attr);
1575         if (ret)
1576                 goto err;
1577
1578         if (data->o_res.delegation_type != 0)
1579                 nfs4_opendata_check_deleg(data, state);
1580 update:
1581         update_open_stateid(state, &data->o_res.stateid, NULL,
1582                             data->o_arg.fmode);
1583         atomic_inc(&state->count);
1584
1585         return state;
1586 err:
1587         return ERR_PTR(ret);
1588
1589 }
1590
1591 static struct nfs4_state *
1592 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1593 {
1594         struct inode *inode;
1595         struct nfs4_state *state = NULL;
1596         int ret;
1597
1598         if (!data->rpc_done) {
1599                 state = nfs4_try_open_cached(data);
1600                 goto out;
1601         }
1602
1603         ret = -EAGAIN;
1604         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1605                 goto err;
1606         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1607         ret = PTR_ERR(inode);
1608         if (IS_ERR(inode))
1609                 goto err;
1610         ret = -ENOMEM;
1611         state = nfs4_get_open_state(inode, data->owner);
1612         if (state == NULL)
1613                 goto err_put_inode;
1614         if (data->o_res.delegation_type != 0)
1615                 nfs4_opendata_check_deleg(data, state);
1616         update_open_stateid(state, &data->o_res.stateid, NULL,
1617                         data->o_arg.fmode);
1618         iput(inode);
1619 out:
1620         nfs_release_seqid(data->o_arg.seqid);
1621         return state;
1622 err_put_inode:
1623         iput(inode);
1624 err:
1625         return ERR_PTR(ret);
1626 }
1627
1628 static struct nfs4_state *
1629 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1630 {
1631         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1632                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1633         return _nfs4_opendata_to_nfs4_state(data);
1634 }
1635
1636 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1637 {
1638         struct nfs_inode *nfsi = NFS_I(state->inode);
1639         struct nfs_open_context *ctx;
1640
1641         spin_lock(&state->inode->i_lock);
1642         list_for_each_entry(ctx, &nfsi->open_files, list) {
1643                 if (ctx->state != state)
1644                         continue;
1645                 get_nfs_open_context(ctx);
1646                 spin_unlock(&state->inode->i_lock);
1647                 return ctx;
1648         }
1649         spin_unlock(&state->inode->i_lock);
1650         return ERR_PTR(-ENOENT);
1651 }
1652
1653 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1654                 struct nfs4_state *state, enum open_claim_type4 claim)
1655 {
1656         struct nfs4_opendata *opendata;
1657
1658         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1659                         NULL, NULL, claim, GFP_NOFS);
1660         if (opendata == NULL)
1661                 return ERR_PTR(-ENOMEM);
1662         opendata->state = state;
1663         atomic_inc(&state->count);
1664         return opendata;
1665 }
1666
1667 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1668                 fmode_t fmode)
1669 {
1670         struct nfs4_state *newstate;
1671         int ret;
1672
1673         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1674                 return 0;
1675         opendata->o_arg.open_flags = 0;
1676         opendata->o_arg.fmode = fmode;
1677         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1678                         NFS_SB(opendata->dentry->d_sb),
1679                         fmode, 0);
1680         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1681         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1682         nfs4_init_opendata_res(opendata);
1683         ret = _nfs4_recover_proc_open(opendata);
1684         if (ret != 0)
1685                 return ret;
1686         newstate = nfs4_opendata_to_nfs4_state(opendata);
1687         if (IS_ERR(newstate))
1688                 return PTR_ERR(newstate);
1689         if (newstate != opendata->state)
1690                 ret = -ESTALE;
1691         nfs4_close_state(newstate, fmode);
1692         return ret;
1693 }
1694
1695 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1696 {
1697         int ret;
1698
1699         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1700         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1701         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1702         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1703         /* memory barrier prior to reading state->n_* */
1704         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1705         clear_bit(NFS_OPEN_STATE, &state->flags);
1706         smp_rmb();
1707         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1708         if (ret != 0)
1709                 return ret;
1710         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1711         if (ret != 0)
1712                 return ret;
1713         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1714         if (ret != 0)
1715                 return ret;
1716         /*
1717          * We may have performed cached opens for all three recoveries.
1718          * Check if we need to update the current stateid.
1719          */
1720         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1721             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1722                 write_seqlock(&state->seqlock);
1723                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1724                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1725                 write_sequnlock(&state->seqlock);
1726         }
1727         return 0;
1728 }
1729
1730 /*
1731  * OPEN_RECLAIM:
1732  *      reclaim state on the server after a reboot.
1733  */
1734 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1735 {
1736         struct nfs_delegation *delegation;
1737         struct nfs4_opendata *opendata;
1738         fmode_t delegation_type = 0;
1739         int status;
1740
1741         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1742                         NFS4_OPEN_CLAIM_PREVIOUS);
1743         if (IS_ERR(opendata))
1744                 return PTR_ERR(opendata);
1745         rcu_read_lock();
1746         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1747         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1748                 delegation_type = delegation->type;
1749         rcu_read_unlock();
1750         opendata->o_arg.u.delegation_type = delegation_type;
1751         status = nfs4_open_recover(opendata, state);
1752         nfs4_opendata_put(opendata);
1753         return status;
1754 }
1755
1756 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1757 {
1758         struct nfs_server *server = NFS_SERVER(state->inode);
1759         struct nfs4_exception exception = { };
1760         int err;
1761         do {
1762                 err = _nfs4_do_open_reclaim(ctx, state);
1763                 trace_nfs4_open_reclaim(ctx, 0, err);
1764                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1765                         continue;
1766                 if (err != -NFS4ERR_DELAY)
1767                         break;
1768                 nfs4_handle_exception(server, err, &exception);
1769         } while (exception.retry);
1770         return err;
1771 }
1772
1773 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1774 {
1775         struct nfs_open_context *ctx;
1776         int ret;
1777
1778         ctx = nfs4_state_find_open_context(state);
1779         if (IS_ERR(ctx))
1780                 return -EAGAIN;
1781         ret = nfs4_do_open_reclaim(ctx, state);
1782         put_nfs_open_context(ctx);
1783         return ret;
1784 }
1785
1786 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1787 {
1788         switch (err) {
1789                 default:
1790                         printk(KERN_ERR "NFS: %s: unhandled error "
1791                                         "%d.\n", __func__, err);
1792                 case 0:
1793                 case -ENOENT:
1794                 case -EAGAIN:
1795                 case -ESTALE:
1796                         break;
1797                 case -NFS4ERR_BADSESSION:
1798                 case -NFS4ERR_BADSLOT:
1799                 case -NFS4ERR_BAD_HIGH_SLOT:
1800                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1801                 case -NFS4ERR_DEADSESSION:
1802                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1803                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1804                         return -EAGAIN;
1805                 case -NFS4ERR_STALE_CLIENTID:
1806                 case -NFS4ERR_STALE_STATEID:
1807                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1808                 case -NFS4ERR_EXPIRED:
1809                         /* Don't recall a delegation if it was lost */
1810                         nfs4_schedule_lease_recovery(server->nfs_client);
1811                         return -EAGAIN;
1812                 case -NFS4ERR_MOVED:
1813                         nfs4_schedule_migration_recovery(server);
1814                         return -EAGAIN;
1815                 case -NFS4ERR_LEASE_MOVED:
1816                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1817                         return -EAGAIN;
1818                 case -NFS4ERR_DELEG_REVOKED:
1819                 case -NFS4ERR_ADMIN_REVOKED:
1820                 case -NFS4ERR_BAD_STATEID:
1821                 case -NFS4ERR_OPENMODE:
1822                         nfs_inode_find_state_and_recover(state->inode,
1823                                         stateid);
1824                         nfs4_schedule_stateid_recovery(server, state);
1825                         return -EAGAIN;
1826                 case -NFS4ERR_DELAY:
1827                 case -NFS4ERR_GRACE:
1828                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1829                         ssleep(1);
1830                         return -EAGAIN;
1831                 case -ENOMEM:
1832                 case -NFS4ERR_DENIED:
1833                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1834                         return 0;
1835         }
1836         return err;
1837 }
1838
1839 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1840                 struct nfs4_state *state, const nfs4_stateid *stateid,
1841                 fmode_t type)
1842 {
1843         struct nfs_server *server = NFS_SERVER(state->inode);
1844         struct nfs4_opendata *opendata;
1845         int err = 0;
1846
1847         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1848                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1849         if (IS_ERR(opendata))
1850                 return PTR_ERR(opendata);
1851         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1852         write_seqlock(&state->seqlock);
1853         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1854         write_sequnlock(&state->seqlock);
1855         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1856         switch (type & (FMODE_READ|FMODE_WRITE)) {
1857         case FMODE_READ|FMODE_WRITE:
1858         case FMODE_WRITE:
1859                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1860                 if (err)
1861                         break;
1862                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1863                 if (err)
1864                         break;
1865         case FMODE_READ:
1866                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1867         }
1868         nfs4_opendata_put(opendata);
1869         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1870 }
1871
1872 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1873 {
1874         struct nfs4_opendata *data = calldata;
1875
1876         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1877                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1878 }
1879
1880 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1881 {
1882         struct nfs4_opendata *data = calldata;
1883
1884         nfs40_sequence_done(task, &data->c_res.seq_res);
1885
1886         data->rpc_status = task->tk_status;
1887         if (data->rpc_status == 0) {
1888                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1889                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1890                 renew_lease(data->o_res.server, data->timestamp);
1891                 data->rpc_done = 1;
1892         }
1893 }
1894
1895 static void nfs4_open_confirm_release(void *calldata)
1896 {
1897         struct nfs4_opendata *data = calldata;
1898         struct nfs4_state *state = NULL;
1899
1900         /* If this request hasn't been cancelled, do nothing */
1901         if (data->cancelled == 0)
1902                 goto out_free;
1903         /* In case of error, no cleanup! */
1904         if (!data->rpc_done)
1905                 goto out_free;
1906         state = nfs4_opendata_to_nfs4_state(data);
1907         if (!IS_ERR(state))
1908                 nfs4_close_state(state, data->o_arg.fmode);
1909 out_free:
1910         nfs4_opendata_put(data);
1911 }
1912
1913 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1914         .rpc_call_prepare = nfs4_open_confirm_prepare,
1915         .rpc_call_done = nfs4_open_confirm_done,
1916         .rpc_release = nfs4_open_confirm_release,
1917 };
1918
1919 /*
1920  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1921  */
1922 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1923 {
1924         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1925         struct rpc_task *task;
1926         struct  rpc_message msg = {
1927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1928                 .rpc_argp = &data->c_arg,
1929                 .rpc_resp = &data->c_res,
1930                 .rpc_cred = data->owner->so_cred,
1931         };
1932         struct rpc_task_setup task_setup_data = {
1933                 .rpc_client = server->client,
1934                 .rpc_message = &msg,
1935                 .callback_ops = &nfs4_open_confirm_ops,
1936                 .callback_data = data,
1937                 .workqueue = nfsiod_workqueue,
1938                 .flags = RPC_TASK_ASYNC,
1939         };
1940         int status;
1941
1942         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1943         kref_get(&data->kref);
1944         data->rpc_done = 0;
1945         data->rpc_status = 0;
1946         data->timestamp = jiffies;
1947         if (data->is_recover)
1948                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1949         task = rpc_run_task(&task_setup_data);
1950         if (IS_ERR(task))
1951                 return PTR_ERR(task);
1952         status = nfs4_wait_for_completion_rpc_task(task);
1953         if (status != 0) {
1954                 data->cancelled = 1;
1955                 smp_wmb();
1956         } else
1957                 status = data->rpc_status;
1958         rpc_put_task(task);
1959         return status;
1960 }
1961
1962 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1963 {
1964         struct nfs4_opendata *data = calldata;
1965         struct nfs4_state_owner *sp = data->owner;
1966         struct nfs_client *clp = sp->so_server->nfs_client;
1967         enum open_claim_type4 claim = data->o_arg.claim;
1968
1969         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1970                 goto out_wait;
1971         /*
1972          * Check if we still need to send an OPEN call, or if we can use
1973          * a delegation instead.
1974          */
1975         if (data->state != NULL) {
1976                 struct nfs_delegation *delegation;
1977
1978                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1979                         goto out_no_action;
1980                 rcu_read_lock();
1981                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1982                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1983                         goto unlock_no_action;
1984                 rcu_read_unlock();
1985         }
1986         /* Update client id. */
1987         data->o_arg.clientid = clp->cl_clientid;
1988         switch (claim) {
1989         default:
1990                 break;
1991         case NFS4_OPEN_CLAIM_PREVIOUS:
1992         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1993         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1994                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1995         case NFS4_OPEN_CLAIM_FH:
1996                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1997                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1998         }
1999         data->timestamp = jiffies;
2000         if (nfs4_setup_sequence(data->o_arg.server,
2001                                 &data->o_arg.seq_args,
2002                                 &data->o_res.seq_res,
2003                                 task) != 0)
2004                 nfs_release_seqid(data->o_arg.seqid);
2005
2006         /* Set the create mode (note dependency on the session type) */
2007         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2008         if (data->o_arg.open_flags & O_EXCL) {
2009                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2010                 if (nfs4_has_persistent_session(clp))
2011                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2012                 else if (clp->cl_mvops->minor_version > 0)
2013                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2014         }
2015         return;
2016 unlock_no_action:
2017         rcu_read_unlock();
2018 out_no_action:
2019         task->tk_action = NULL;
2020 out_wait:
2021         nfs4_sequence_done(task, &data->o_res.seq_res);
2022 }
2023
2024 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2025 {
2026         struct nfs4_opendata *data = calldata;
2027
2028         data->rpc_status = task->tk_status;
2029
2030         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2031                 return;
2032
2033         if (task->tk_status == 0) {
2034                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2035                         switch (data->o_res.f_attr->mode & S_IFMT) {
2036                         case S_IFREG:
2037                                 break;
2038                         case S_IFLNK:
2039                                 data->rpc_status = -ELOOP;
2040                                 break;
2041                         case S_IFDIR:
2042                                 data->rpc_status = -EISDIR;
2043                                 break;
2044                         default:
2045                                 data->rpc_status = -ENOTDIR;
2046                         }
2047                 }
2048                 renew_lease(data->o_res.server, data->timestamp);
2049                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2050                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2051         }
2052         data->rpc_done = 1;
2053 }
2054
2055 static void nfs4_open_release(void *calldata)
2056 {
2057         struct nfs4_opendata *data = calldata;
2058         struct nfs4_state *state = NULL;
2059
2060         /* If this request hasn't been cancelled, do nothing */
2061         if (data->cancelled == 0)
2062                 goto out_free;
2063         /* In case of error, no cleanup! */
2064         if (data->rpc_status != 0 || !data->rpc_done)
2065                 goto out_free;
2066         /* In case we need an open_confirm, no cleanup! */
2067         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2068                 goto out_free;
2069         state = nfs4_opendata_to_nfs4_state(data);
2070         if (!IS_ERR(state))
2071                 nfs4_close_state(state, data->o_arg.fmode);
2072 out_free:
2073         nfs4_opendata_put(data);
2074 }
2075
2076 static const struct rpc_call_ops nfs4_open_ops = {
2077         .rpc_call_prepare = nfs4_open_prepare,
2078         .rpc_call_done = nfs4_open_done,
2079         .rpc_release = nfs4_open_release,
2080 };
2081
2082 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2083 {
2084         struct inode *dir = d_inode(data->dir);
2085         struct nfs_server *server = NFS_SERVER(dir);
2086         struct nfs_openargs *o_arg = &data->o_arg;
2087         struct nfs_openres *o_res = &data->o_res;
2088         struct rpc_task *task;
2089         struct rpc_message msg = {
2090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2091                 .rpc_argp = o_arg,
2092                 .rpc_resp = o_res,
2093                 .rpc_cred = data->owner->so_cred,
2094         };
2095         struct rpc_task_setup task_setup_data = {
2096                 .rpc_client = server->client,
2097                 .rpc_message = &msg,
2098                 .callback_ops = &nfs4_open_ops,
2099                 .callback_data = data,
2100                 .workqueue = nfsiod_workqueue,
2101                 .flags = RPC_TASK_ASYNC,
2102         };
2103         int status;
2104
2105         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2106         kref_get(&data->kref);
2107         data->rpc_done = 0;
2108         data->rpc_status = 0;
2109         data->cancelled = 0;
2110         data->is_recover = 0;
2111         if (isrecover) {
2112                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2113                 data->is_recover = 1;
2114         }
2115         task = rpc_run_task(&task_setup_data);
2116         if (IS_ERR(task))
2117                 return PTR_ERR(task);
2118         status = nfs4_wait_for_completion_rpc_task(task);
2119         if (status != 0) {
2120                 data->cancelled = 1;
2121                 smp_wmb();
2122         } else
2123                 status = data->rpc_status;
2124         rpc_put_task(task);
2125
2126         return status;
2127 }
2128
2129 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2130 {
2131         struct inode *dir = d_inode(data->dir);
2132         struct nfs_openres *o_res = &data->o_res;
2133         int status;
2134
2135         status = nfs4_run_open_task(data, 1);
2136         if (status != 0 || !data->rpc_done)
2137                 return status;
2138
2139         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2140
2141         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2142                 status = _nfs4_proc_open_confirm(data);
2143                 if (status != 0)
2144                         return status;
2145         }
2146
2147         return status;
2148 }
2149
2150 /*
2151  * Additional permission checks in order to distinguish between an
2152  * open for read, and an open for execute. This works around the
2153  * fact that NFSv4 OPEN treats read and execute permissions as being
2154  * the same.
2155  * Note that in the non-execute case, we want to turn off permission
2156  * checking if we just created a new file (POSIX open() semantics).
2157  */
2158 static int nfs4_opendata_access(struct rpc_cred *cred,
2159                                 struct nfs4_opendata *opendata,
2160                                 struct nfs4_state *state, fmode_t fmode,
2161                                 int openflags)
2162 {
2163         struct nfs_access_entry cache;
2164         u32 mask;
2165
2166         /* access call failed or for some reason the server doesn't
2167          * support any access modes -- defer access call until later */
2168         if (opendata->o_res.access_supported == 0)
2169                 return 0;
2170
2171         mask = 0;
2172         /*
2173          * Use openflags to check for exec, because fmode won't
2174          * always have FMODE_EXEC set when file open for exec.
2175          */
2176         if (openflags & __FMODE_EXEC) {
2177                 /* ONLY check for exec rights */
2178                 mask = MAY_EXEC;
2179         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2180                 mask = MAY_READ;
2181
2182         cache.cred = cred;
2183         cache.jiffies = jiffies;
2184         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2185         nfs_access_add_cache(state->inode, &cache);
2186
2187         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2188                 return 0;
2189
2190         /* even though OPEN succeeded, access is denied. Close the file */
2191         nfs4_close_state(state, fmode);
2192         return -EACCES;
2193 }
2194
2195 /*
2196  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2197  */
2198 static int _nfs4_proc_open(struct nfs4_opendata *data)
2199 {
2200         struct inode *dir = d_inode(data->dir);
2201         struct nfs_server *server = NFS_SERVER(dir);
2202         struct nfs_openargs *o_arg = &data->o_arg;
2203         struct nfs_openres *o_res = &data->o_res;
2204         int status;
2205
2206         status = nfs4_run_open_task(data, 0);
2207         if (!data->rpc_done)
2208                 return status;
2209         if (status != 0) {
2210                 if (status == -NFS4ERR_BADNAME &&
2211                                 !(o_arg->open_flags & O_CREAT))
2212                         return -ENOENT;
2213                 return status;
2214         }
2215
2216         nfs_fattr_map_and_free_names(server, &data->f_attr);
2217
2218         if (o_arg->open_flags & O_CREAT) {
2219                 update_changeattr(dir, &o_res->cinfo);
2220                 if (o_arg->open_flags & O_EXCL)
2221                         data->file_created = 1;
2222                 else if (o_res->cinfo.before != o_res->cinfo.after)
2223                         data->file_created = 1;
2224         }
2225         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2226                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2227         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2228                 status = _nfs4_proc_open_confirm(data);
2229                 if (status != 0)
2230                         return status;
2231         }
2232         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2233                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2234         return 0;
2235 }
2236
2237 static int nfs4_recover_expired_lease(struct nfs_server *server)
2238 {
2239         return nfs4_client_recover_expired_lease(server->nfs_client);
2240 }
2241
2242 /*
2243  * OPEN_EXPIRED:
2244  *      reclaim state on the server after a network partition.
2245  *      Assumes caller holds the appropriate lock
2246  */
2247 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2248 {
2249         struct nfs4_opendata *opendata;
2250         int ret;
2251
2252         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2253                         NFS4_OPEN_CLAIM_FH);
2254         if (IS_ERR(opendata))
2255                 return PTR_ERR(opendata);
2256         ret = nfs4_open_recover(opendata, state);
2257         if (ret == -ESTALE)
2258                 d_drop(ctx->dentry);
2259         nfs4_opendata_put(opendata);
2260         return ret;
2261 }
2262
2263 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2264 {
2265         struct nfs_server *server = NFS_SERVER(state->inode);
2266         struct nfs4_exception exception = { };
2267         int err;
2268
2269         do {
2270                 err = _nfs4_open_expired(ctx, state);
2271                 trace_nfs4_open_expired(ctx, 0, err);
2272                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2273                         continue;
2274                 switch (err) {
2275                 default:
2276                         goto out;
2277                 case -NFS4ERR_GRACE:
2278                 case -NFS4ERR_DELAY:
2279                         nfs4_handle_exception(server, err, &exception);
2280                         err = 0;
2281                 }
2282         } while (exception.retry);
2283 out:
2284         return err;
2285 }
2286
2287 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2288 {
2289         struct nfs_open_context *ctx;
2290         int ret;
2291
2292         ctx = nfs4_state_find_open_context(state);
2293         if (IS_ERR(ctx))
2294                 return -EAGAIN;
2295         ret = nfs4_do_open_expired(ctx, state);
2296         put_nfs_open_context(ctx);
2297         return ret;
2298 }
2299
2300 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2301 {
2302         nfs_remove_bad_delegation(state->inode);
2303         write_seqlock(&state->seqlock);
2304         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2305         write_sequnlock(&state->seqlock);
2306         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2307 }
2308
2309 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2310 {
2311         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2312                 nfs_finish_clear_delegation_stateid(state);
2313 }
2314
2315 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2316 {
2317         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2318         nfs40_clear_delegation_stateid(state);
2319         return nfs4_open_expired(sp, state);
2320 }
2321
2322 #if defined(CONFIG_NFS_V4_1)
2323 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2324 {
2325         struct nfs_server *server = NFS_SERVER(state->inode);
2326         nfs4_stateid stateid;
2327         struct nfs_delegation *delegation;
2328         struct rpc_cred *cred;
2329         int status;
2330
2331         /* Get the delegation credential for use by test/free_stateid */
2332         rcu_read_lock();
2333         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2334         if (delegation == NULL) {
2335                 rcu_read_unlock();
2336                 return;
2337         }
2338
2339         nfs4_stateid_copy(&stateid, &delegation->stateid);
2340         cred = get_rpccred(delegation->cred);
2341         rcu_read_unlock();
2342         status = nfs41_test_stateid(server, &stateid, cred);
2343         trace_nfs4_test_delegation_stateid(state, NULL, status);
2344
2345         if (status != NFS_OK) {
2346                 /* Free the stateid unless the server explicitly
2347                  * informs us the stateid is unrecognized. */
2348                 if (status != -NFS4ERR_BAD_STATEID)
2349                         nfs41_free_stateid(server, &stateid, cred);
2350                 nfs_finish_clear_delegation_stateid(state);
2351         }
2352
2353         put_rpccred(cred);
2354 }
2355
2356 /**
2357  * nfs41_check_open_stateid - possibly free an open stateid
2358  *
2359  * @state: NFSv4 state for an inode
2360  *
2361  * Returns NFS_OK if recovery for this stateid is now finished.
2362  * Otherwise a negative NFS4ERR value is returned.
2363  */
2364 static int nfs41_check_open_stateid(struct nfs4_state *state)
2365 {
2366         struct nfs_server *server = NFS_SERVER(state->inode);
2367         nfs4_stateid *stateid = &state->open_stateid;
2368         struct rpc_cred *cred = state->owner->so_cred;
2369         int status;
2370
2371         /* If a state reset has been done, test_stateid is unneeded */
2372         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2373             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2374             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2375                 return -NFS4ERR_BAD_STATEID;
2376
2377         status = nfs41_test_stateid(server, stateid, cred);
2378         trace_nfs4_test_open_stateid(state, NULL, status);
2379         if (status != NFS_OK) {
2380                 /* Free the stateid unless the server explicitly
2381                  * informs us the stateid is unrecognized. */
2382                 if (status != -NFS4ERR_BAD_STATEID)
2383                         nfs41_free_stateid(server, stateid, cred);
2384
2385                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2386                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2387                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2388                 clear_bit(NFS_OPEN_STATE, &state->flags);
2389         }
2390         return status;
2391 }
2392
2393 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2394 {
2395         int status;
2396
2397         nfs41_check_delegation_stateid(state);
2398         status = nfs41_check_open_stateid(state);
2399         if (status != NFS_OK)
2400                 status = nfs4_open_expired(sp, state);
2401         return status;
2402 }
2403 #endif
2404
2405 /*
2406  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2407  * fields corresponding to attributes that were used to store the verifier.
2408  * Make sure we clobber those fields in the later setattr call
2409  */
2410 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2411                                 struct iattr *sattr, struct nfs4_label **label)
2412 {
2413         const u32 *attrset = opendata->o_res.attrset;
2414
2415         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2416             !(sattr->ia_valid & ATTR_ATIME_SET))
2417                 sattr->ia_valid |= ATTR_ATIME;
2418
2419         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2420             !(sattr->ia_valid & ATTR_MTIME_SET))
2421                 sattr->ia_valid |= ATTR_MTIME;
2422
2423         /* Except MODE, it seems harmless of setting twice. */
2424         if ((attrset[1] & FATTR4_WORD1_MODE))
2425                 sattr->ia_valid &= ~ATTR_MODE;
2426
2427         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2428                 *label = NULL;
2429 }
2430
2431 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2432                 fmode_t fmode,
2433                 int flags,
2434                 struct nfs_open_context *ctx)
2435 {
2436         struct nfs4_state_owner *sp = opendata->owner;
2437         struct nfs_server *server = sp->so_server;
2438         struct dentry *dentry;
2439         struct nfs4_state *state;
2440         unsigned int seq;
2441         int ret;
2442
2443         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2444
2445         ret = _nfs4_proc_open(opendata);
2446         if (ret != 0)
2447                 goto out;
2448
2449         state = nfs4_opendata_to_nfs4_state(opendata);
2450         ret = PTR_ERR(state);
2451         if (IS_ERR(state))
2452                 goto out;
2453         if (server->caps & NFS_CAP_POSIX_LOCK)
2454                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2455
2456         dentry = opendata->dentry;
2457         if (d_really_is_negative(dentry)) {
2458                 /* FIXME: Is this d_drop() ever needed? */
2459                 d_drop(dentry);
2460                 dentry = d_add_unique(dentry, igrab(state->inode));
2461                 if (dentry == NULL) {
2462                         dentry = opendata->dentry;
2463                 } else if (dentry != ctx->dentry) {
2464                         dput(ctx->dentry);
2465                         ctx->dentry = dget(dentry);
2466                 }
2467                 nfs_set_verifier(dentry,
2468                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2469         }
2470
2471         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2472         if (ret != 0)
2473                 goto out;
2474
2475         ctx->state = state;
2476         if (d_inode(dentry) == state->inode) {
2477                 nfs_inode_attach_open_context(ctx);
2478                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2479                         nfs4_schedule_stateid_recovery(server, state);
2480         }
2481 out:
2482         return ret;
2483 }
2484
2485 /*
2486  * Returns a referenced nfs4_state
2487  */
2488 static int _nfs4_do_open(struct inode *dir,
2489                         struct nfs_open_context *ctx,
2490                         int flags,
2491                         struct iattr *sattr,
2492                         struct nfs4_label *label,
2493                         int *opened)
2494 {
2495         struct nfs4_state_owner  *sp;
2496         struct nfs4_state     *state = NULL;
2497         struct nfs_server       *server = NFS_SERVER(dir);
2498         struct nfs4_opendata *opendata;
2499         struct dentry *dentry = ctx->dentry;
2500         struct rpc_cred *cred = ctx->cred;
2501         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2502         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2503         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2504         struct nfs4_label *olabel = NULL;
2505         int status;
2506
2507         /* Protect against reboot recovery conflicts */
2508         status = -ENOMEM;
2509         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2510         if (sp == NULL) {
2511                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2512                 goto out_err;
2513         }
2514         status = nfs4_recover_expired_lease(server);
2515         if (status != 0)
2516                 goto err_put_state_owner;
2517         if (d_really_is_positive(dentry))
2518                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2519         status = -ENOMEM;
2520         if (d_really_is_positive(dentry))
2521                 claim = NFS4_OPEN_CLAIM_FH;
2522         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2523                         label, claim, GFP_KERNEL);
2524         if (opendata == NULL)
2525                 goto err_put_state_owner;
2526
2527         if (label) {
2528                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2529                 if (IS_ERR(olabel)) {
2530                         status = PTR_ERR(olabel);
2531                         goto err_opendata_put;
2532                 }
2533         }
2534
2535         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2536                 if (!opendata->f_attr.mdsthreshold) {
2537                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2538                         if (!opendata->f_attr.mdsthreshold)
2539                                 goto err_free_label;
2540                 }
2541                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2542         }
2543         if (d_really_is_positive(dentry))
2544                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2545
2546         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2547         if (status != 0)
2548                 goto err_free_label;
2549         state = ctx->state;
2550
2551         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2552             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2553                 nfs4_exclusive_attrset(opendata, sattr, &label);
2554
2555                 nfs_fattr_init(opendata->o_res.f_attr);
2556                 status = nfs4_do_setattr(state->inode, cred,
2557                                 opendata->o_res.f_attr, sattr,
2558                                 state, label, olabel);
2559                 if (status == 0) {
2560                         nfs_setattr_update_inode(state->inode, sattr,
2561                                         opendata->o_res.f_attr);
2562                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2563                 }
2564         }
2565         if (opened && opendata->file_created)
2566                 *opened |= FILE_CREATED;
2567
2568         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2569                 *ctx_th = opendata->f_attr.mdsthreshold;
2570                 opendata->f_attr.mdsthreshold = NULL;
2571         }
2572
2573         nfs4_label_free(olabel);
2574
2575         nfs4_opendata_put(opendata);
2576         nfs4_put_state_owner(sp);
2577         return 0;
2578 err_free_label:
2579         nfs4_label_free(olabel);
2580 err_opendata_put:
2581         nfs4_opendata_put(opendata);
2582 err_put_state_owner:
2583         nfs4_put_state_owner(sp);
2584 out_err:
2585         return status;
2586 }
2587
2588
2589 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2590                                         struct nfs_open_context *ctx,
2591                                         int flags,
2592                                         struct iattr *sattr,
2593                                         struct nfs4_label *label,
2594                                         int *opened)
2595 {
2596         struct nfs_server *server = NFS_SERVER(dir);
2597         struct nfs4_exception exception = { };
2598         struct nfs4_state *res;
2599         int status;
2600
2601         do {
2602                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2603                 res = ctx->state;
2604                 trace_nfs4_open_file(ctx, flags, status);
2605                 if (status == 0)
2606                         break;
2607                 /* NOTE: BAD_SEQID means the server and client disagree about the
2608                  * book-keeping w.r.t. state-changing operations
2609                  * (OPEN/CLOSE/LOCK/LOCKU...)
2610                  * It is actually a sign of a bug on the client or on the server.
2611                  *
2612                  * If we receive a BAD_SEQID error in the particular case of
2613                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2614                  * have unhashed the old state_owner for us, and that we can
2615                  * therefore safely retry using a new one. We should still warn
2616                  * the user though...
2617                  */
2618                 if (status == -NFS4ERR_BAD_SEQID) {
2619                         pr_warn_ratelimited("NFS: v4 server %s "
2620                                         " returned a bad sequence-id error!\n",
2621                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2622                         exception.retry = 1;
2623                         continue;
2624                 }
2625                 /*
2626                  * BAD_STATEID on OPEN means that the server cancelled our
2627                  * state before it received the OPEN_CONFIRM.
2628                  * Recover by retrying the request as per the discussion
2629                  * on Page 181 of RFC3530.
2630                  */
2631                 if (status == -NFS4ERR_BAD_STATEID) {
2632                         exception.retry = 1;
2633                         continue;
2634                 }
2635                 if (status == -EAGAIN) {
2636                         /* We must have found a delegation */
2637                         exception.retry = 1;
2638                         continue;
2639                 }
2640                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2641                         continue;
2642                 res = ERR_PTR(nfs4_handle_exception(server,
2643                                         status, &exception));
2644         } while (exception.retry);
2645         return res;
2646 }
2647
2648 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2649                             struct nfs_fattr *fattr, struct iattr *sattr,
2650                             struct nfs4_state *state, struct nfs4_label *ilabel,
2651                             struct nfs4_label *olabel)
2652 {
2653         struct nfs_server *server = NFS_SERVER(inode);
2654         struct nfs_setattrargs  arg = {
2655                 .fh             = NFS_FH(inode),
2656                 .iap            = sattr,
2657                 .server         = server,
2658                 .bitmask = server->attr_bitmask,
2659                 .label          = ilabel,
2660         };
2661         struct nfs_setattrres  res = {
2662                 .fattr          = fattr,
2663                 .label          = olabel,
2664                 .server         = server,
2665         };
2666         struct rpc_message msg = {
2667                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2668                 .rpc_argp       = &arg,
2669                 .rpc_resp       = &res,
2670                 .rpc_cred       = cred,
2671         };
2672         unsigned long timestamp = jiffies;
2673         fmode_t fmode;
2674         bool truncate;
2675         int status;
2676
2677         arg.bitmask = nfs4_bitmask(server, ilabel);
2678         if (ilabel)
2679                 arg.bitmask = nfs4_bitmask(server, olabel);
2680
2681         nfs_fattr_init(fattr);
2682
2683         /* Servers should only apply open mode checks for file size changes */
2684         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2685         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2686
2687         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2688                 /* Use that stateid */
2689         } else if (truncate && state != NULL) {
2690                 struct nfs_lockowner lockowner = {
2691                         .l_owner = current->files,
2692                         .l_pid = current->tgid,
2693                 };
2694                 if (!nfs4_valid_open_stateid(state))
2695                         return -EBADF;
2696                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2697                                 &lockowner) == -EIO)
2698                         return -EBADF;
2699         } else
2700                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2701
2702         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2703         if (status == 0 && state != NULL)
2704                 renew_lease(server, timestamp);
2705         return status;
2706 }
2707
2708 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2709                            struct nfs_fattr *fattr, struct iattr *sattr,
2710                            struct nfs4_state *state, struct nfs4_label *ilabel,
2711                            struct nfs4_label *olabel)
2712 {
2713         struct nfs_server *server = NFS_SERVER(inode);
2714         struct nfs4_exception exception = {
2715                 .state = state,
2716                 .inode = inode,
2717         };
2718         int err;
2719         do {
2720                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2721                 trace_nfs4_setattr(inode, err);
2722                 switch (err) {
2723                 case -NFS4ERR_OPENMODE:
2724                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2725                                 pr_warn_once("NFSv4: server %s is incorrectly "
2726                                                 "applying open mode checks to "
2727                                                 "a SETATTR that is not "
2728                                                 "changing file size.\n",
2729                                                 server->nfs_client->cl_hostname);
2730                         }
2731                         if (state && !(state->state & FMODE_WRITE)) {
2732                                 err = -EBADF;
2733                                 if (sattr->ia_valid & ATTR_OPEN)
2734                                         err = -EACCES;
2735                                 goto out;
2736                         }
2737                 }
2738                 err = nfs4_handle_exception(server, err, &exception);
2739         } while (exception.retry);
2740 out:
2741         return err;
2742 }
2743
2744 static bool
2745 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2746 {
2747         if (inode == NULL || !nfs_have_layout(inode))
2748                 return false;
2749
2750         return pnfs_wait_on_layoutreturn(inode, task);
2751 }
2752
2753 struct nfs4_closedata {
2754         struct inode *inode;
2755         struct nfs4_state *state;
2756         struct nfs_closeargs arg;
2757         struct nfs_closeres res;
2758         struct nfs_fattr fattr;
2759         unsigned long timestamp;
2760         bool roc;
2761         u32 roc_barrier;
2762 };
2763
2764 static void nfs4_free_closedata(void *data)
2765 {
2766         struct nfs4_closedata *calldata = data;
2767         struct nfs4_state_owner *sp = calldata->state->owner;
2768         struct super_block *sb = calldata->state->inode->i_sb;
2769
2770         if (calldata->roc)
2771                 pnfs_roc_release(calldata->state->inode);
2772         nfs4_put_open_state(calldata->state);
2773         nfs_free_seqid(calldata->arg.seqid);
2774         nfs4_put_state_owner(sp);
2775         nfs_sb_deactive(sb);
2776         kfree(calldata);
2777 }
2778
2779 static void nfs4_close_done(struct rpc_task *task, void *data)
2780 {
2781         struct nfs4_closedata *calldata = data;
2782         struct nfs4_state *state = calldata->state;
2783         struct nfs_server *server = NFS_SERVER(calldata->inode);
2784         nfs4_stateid *res_stateid = NULL;
2785
2786         dprintk("%s: begin!\n", __func__);
2787         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2788                 return;
2789         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2790         /* hmm. we are done with the inode, and in the process of freeing
2791          * the state_owner. we keep this around to process errors
2792          */
2793         switch (task->tk_status) {
2794                 case 0:
2795                         res_stateid = &calldata->res.stateid;
2796                         if (calldata->roc)
2797                                 pnfs_roc_set_barrier(state->inode,
2798                                                      calldata->roc_barrier);
2799                         renew_lease(server, calldata->timestamp);
2800                         break;
2801                 case -NFS4ERR_ADMIN_REVOKED:
2802                 case -NFS4ERR_STALE_STATEID:
2803                 case -NFS4ERR_OLD_STATEID:
2804                 case -NFS4ERR_BAD_STATEID:
2805                 case -NFS4ERR_EXPIRED:
2806                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2807                                                 &state->open_stateid)) {
2808                                 rpc_restart_call_prepare(task);
2809                                 goto out_release;
2810                         }
2811                         if (calldata->arg.fmode == 0)
2812                                 break;
2813                 default:
2814                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2815                                 rpc_restart_call_prepare(task);
2816                                 goto out_release;
2817                         }
2818         }
2819         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2820                         res_stateid, calldata->arg.fmode);
2821 out_release:
2822         nfs_release_seqid(calldata->arg.seqid);
2823         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2824         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2825 }
2826
2827 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2828 {
2829         struct nfs4_closedata *calldata = data;
2830         struct nfs4_state *state = calldata->state;
2831         struct inode *inode = calldata->inode;
2832         bool is_rdonly, is_wronly, is_rdwr;
2833         int call_close = 0;
2834
2835         dprintk("%s: begin!\n", __func__);
2836         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2837                 goto out_wait;
2838
2839         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2840         spin_lock(&state->owner->so_lock);
2841         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2842         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2843         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2844         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2845         /* Calculate the change in open mode */
2846         calldata->arg.fmode = 0;
2847         if (state->n_rdwr == 0) {
2848                 if (state->n_rdonly == 0)
2849                         call_close |= is_rdonly;
2850                 else if (is_rdonly)
2851                         calldata->arg.fmode |= FMODE_READ;
2852                 if (state->n_wronly == 0)
2853                         call_close |= is_wronly;
2854                 else if (is_wronly)
2855                         calldata->arg.fmode |= FMODE_WRITE;
2856         } else if (is_rdwr)
2857                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2858
2859         if (calldata->arg.fmode == 0)
2860                 call_close |= is_rdwr;
2861
2862         if (!nfs4_valid_open_stateid(state))
2863                 call_close = 0;
2864         spin_unlock(&state->owner->so_lock);
2865
2866         if (!call_close) {
2867                 /* Note: exit _without_ calling nfs4_close_done */
2868                 goto out_no_action;
2869         }
2870
2871         if (nfs4_wait_on_layoutreturn(inode, task)) {
2872                 nfs_release_seqid(calldata->arg.seqid);
2873                 goto out_wait;
2874         }
2875
2876         if (calldata->arg.fmode == 0)
2877                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2878         if (calldata->roc)
2879                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2880
2881         calldata->arg.share_access =
2882                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2883                                 calldata->arg.fmode, 0);
2884
2885         nfs_fattr_init(calldata->res.fattr);
2886         calldata->timestamp = jiffies;
2887         if (nfs4_setup_sequence(NFS_SERVER(inode),
2888                                 &calldata->arg.seq_args,
2889                                 &calldata->res.seq_res,
2890                                 task) != 0)
2891                 nfs_release_seqid(calldata->arg.seqid);
2892         dprintk("%s: done!\n", __func__);
2893         return;
2894 out_no_action:
2895         task->tk_action = NULL;
2896 out_wait:
2897         nfs4_sequence_done(task, &calldata->res.seq_res);
2898 }
2899
2900 static const struct rpc_call_ops nfs4_close_ops = {
2901         .rpc_call_prepare = nfs4_close_prepare,
2902         .rpc_call_done = nfs4_close_done,
2903         .rpc_release = nfs4_free_closedata,
2904 };
2905
2906 static bool nfs4_roc(struct inode *inode)
2907 {
2908         if (!nfs_have_layout(inode))
2909                 return false;
2910         return pnfs_roc(inode);
2911 }
2912
2913 /*
2914  * It is possible for data to be read/written from a mem-mapped file
2915  * after the sys_close call (which hits the vfs layer as a flush).
2916  * This means that we can't safely call nfsv4 close on a file until
2917  * the inode is cleared. This in turn means that we are not good
2918  * NFSv4 citizens - we do not indicate to the server to update the file's
2919  * share state even when we are done with one of the three share
2920  * stateid's in the inode.
2921  *
2922  * NOTE: Caller must be holding the sp->so_owner semaphore!
2923  */
2924 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2925 {
2926         struct nfs_server *server = NFS_SERVER(state->inode);
2927         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2928         struct nfs4_closedata *calldata;
2929         struct nfs4_state_owner *sp = state->owner;
2930         struct rpc_task *task;
2931         struct rpc_message msg = {
2932                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2933                 .rpc_cred = state->owner->so_cred,
2934         };
2935         struct rpc_task_setup task_setup_data = {
2936                 .rpc_client = server->client,
2937                 .rpc_message = &msg,
2938                 .callback_ops = &nfs4_close_ops,
2939                 .workqueue = nfsiod_workqueue,
2940                 .flags = RPC_TASK_ASYNC,
2941         };
2942         int status = -ENOMEM;
2943
2944         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2945                 &task_setup_data.rpc_client, &msg);
2946
2947         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2948         if (calldata == NULL)
2949                 goto out;
2950         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2951         calldata->inode = state->inode;
2952         calldata->state = state;
2953         calldata->arg.fh = NFS_FH(state->inode);
2954         /* Serialization for the sequence id */
2955         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2956         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2957         if (IS_ERR(calldata->arg.seqid))
2958                 goto out_free_calldata;
2959         calldata->arg.fmode = 0;
2960         calldata->arg.bitmask = server->cache_consistency_bitmask;
2961         calldata->res.fattr = &calldata->fattr;
2962         calldata->res.seqid = calldata->arg.seqid;
2963         calldata->res.server = server;
2964         calldata->roc = nfs4_roc(state->inode);
2965         nfs_sb_active(calldata->inode->i_sb);
2966
2967         msg.rpc_argp = &calldata->arg;
2968         msg.rpc_resp = &calldata->res;
2969         task_setup_data.callback_data = calldata;
2970         task = rpc_run_task(&task_setup_data);
2971         if (IS_ERR(task))
2972                 return PTR_ERR(task);
2973         status = 0;
2974         if (wait)
2975                 status = rpc_wait_for_completion_task(task);
2976         rpc_put_task(task);
2977         return status;
2978 out_free_calldata:
2979         kfree(calldata);
2980 out:
2981         nfs4_put_open_state(state);
2982         nfs4_put_state_owner(sp);
2983         return status;
2984 }
2985
2986 static struct inode *
2987 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2988                 int open_flags, struct iattr *attr, int *opened)
2989 {
2990         struct nfs4_state *state;
2991         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2992
2993         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2994
2995         /* Protect against concurrent sillydeletes */
2996         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2997
2998         nfs4_label_release_security(label);
2999
3000         if (IS_ERR(state))
3001                 return ERR_CAST(state);
3002         return state->inode;
3003 }
3004
3005 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3006 {
3007         if (ctx->state == NULL)
3008                 return;
3009         if (is_sync)
3010                 nfs4_close_sync(ctx->state, ctx->mode);
3011         else
3012                 nfs4_close_state(ctx->state, ctx->mode);
3013 }
3014
3015 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3016 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3017 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3018
3019 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3020 {
3021         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3022         struct nfs4_server_caps_arg args = {
3023                 .fhandle = fhandle,
3024                 .bitmask = bitmask,
3025         };
3026         struct nfs4_server_caps_res res = {};
3027         struct rpc_message msg = {
3028                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3029                 .rpc_argp = &args,
3030                 .rpc_resp = &res,
3031         };
3032         int status;
3033
3034         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3035                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3036                      FATTR4_WORD0_LINK_SUPPORT |
3037                      FATTR4_WORD0_SYMLINK_SUPPORT |
3038                      FATTR4_WORD0_ACLSUPPORT;
3039         if (minorversion)
3040                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3041
3042         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3043         if (status == 0) {
3044                 /* Sanity check the server answers */
3045                 switch (minorversion) {
3046                 case 0:
3047                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3048                         res.attr_bitmask[2] = 0;
3049                         break;
3050                 case 1:
3051                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3052                         break;
3053                 case 2:
3054                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3055                 }
3056                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3057                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3058                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3059                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3060                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3061                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3062                                 NFS_CAP_SECURITY_LABEL);
3063                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3064                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3065                         server->caps |= NFS_CAP_ACLS;
3066                 if (res.has_links != 0)
3067                         server->caps |= NFS_CAP_HARDLINKS;
3068                 if (res.has_symlinks != 0)
3069                         server->caps |= NFS_CAP_SYMLINKS;
3070                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3071                         server->caps |= NFS_CAP_FILEID;
3072                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3073                         server->caps |= NFS_CAP_MODE;
3074                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3075                         server->caps |= NFS_CAP_NLINK;
3076                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3077                         server->caps |= NFS_CAP_OWNER;
3078                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3079                         server->caps |= NFS_CAP_OWNER_GROUP;
3080                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3081                         server->caps |= NFS_CAP_ATIME;
3082                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3083                         server->caps |= NFS_CAP_CTIME;
3084                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3085                         server->caps |= NFS_CAP_MTIME;
3086 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3087                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3088                         server->caps |= NFS_CAP_SECURITY_LABEL;
3089 #endif
3090                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3091                                 sizeof(server->attr_bitmask));
3092                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3093
3094                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3095                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3096                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3097                 server->cache_consistency_bitmask[2] = 0;
3098                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3099                         sizeof(server->exclcreat_bitmask));
3100                 server->acl_bitmask = res.acl_bitmask;
3101                 server->fh_expire_type = res.fh_expire_type;
3102         }
3103
3104         return status;
3105 }
3106
3107 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3108 {
3109         struct nfs4_exception exception = { };
3110         int err;
3111         do {
3112                 err = nfs4_handle_exception(server,
3113                                 _nfs4_server_capabilities(server, fhandle),
3114                                 &exception);
3115         } while (exception.retry);
3116         return err;
3117 }
3118
3119 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3120                 struct nfs_fsinfo *info)
3121 {
3122         u32 bitmask[3];
3123         struct nfs4_lookup_root_arg args = {
3124                 .bitmask = bitmask,
3125         };
3126         struct nfs4_lookup_res res = {
3127                 .server = server,
3128                 .fattr = info->fattr,
3129                 .fh = fhandle,
3130         };
3131         struct rpc_message msg = {
3132                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3133                 .rpc_argp = &args,
3134                 .rpc_resp = &res,
3135         };
3136
3137         bitmask[0] = nfs4_fattr_bitmap[0];
3138         bitmask[1] = nfs4_fattr_bitmap[1];
3139         /*
3140          * Process the label in the upcoming getfattr
3141          */
3142         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3143
3144         nfs_fattr_init(info->fattr);
3145         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3146 }
3147
3148 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3149                 struct nfs_fsinfo *info)
3150 {
3151         struct nfs4_exception exception = { };
3152         int err;
3153         do {
3154                 err = _nfs4_lookup_root(server, fhandle, info);
3155                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3156                 switch (err) {
3157                 case 0:
3158                 case -NFS4ERR_WRONGSEC:
3159                         goto out;
3160                 default:
3161                         err = nfs4_handle_exception(server, err, &exception);
3162                 }
3163         } while (exception.retry);
3164 out:
3165         return err;
3166 }
3167
3168 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3169                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3170 {
3171         struct rpc_auth_create_args auth_args = {
3172                 .pseudoflavor = flavor,
3173         };
3174         struct rpc_auth *auth;
3175         int ret;
3176
3177         auth = rpcauth_create(&auth_args, server->client);
3178         if (IS_ERR(auth)) {
3179                 ret = -EACCES;
3180                 goto out;
3181         }
3182         ret = nfs4_lookup_root(server, fhandle, info);
3183 out:
3184         return ret;
3185 }
3186
3187 /*
3188  * Retry pseudoroot lookup with various security flavors.  We do this when:
3189  *
3190  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3191  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3192  *
3193  * Returns zero on success, or a negative NFS4ERR value, or a
3194  * negative errno value.
3195  */
3196 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3197                               struct nfs_fsinfo *info)
3198 {
3199         /* Per 3530bis 15.33.5 */
3200         static const rpc_authflavor_t flav_array[] = {
3201                 RPC_AUTH_GSS_KRB5P,
3202                 RPC_AUTH_GSS_KRB5I,
3203                 RPC_AUTH_GSS_KRB5,
3204                 RPC_AUTH_UNIX,                  /* courtesy */
3205                 RPC_AUTH_NULL,
3206         };
3207         int status = -EPERM;
3208         size_t i;
3209
3210         if (server->auth_info.flavor_len > 0) {
3211                 /* try each flavor specified by user */
3212                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3213                         status = nfs4_lookup_root_sec(server, fhandle, info,
3214                                                 server->auth_info.flavors[i]);
3215                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3216                                 continue;
3217                         break;
3218                 }
3219         } else {
3220                 /* no flavors specified by user, try default list */
3221                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3222                         status = nfs4_lookup_root_sec(server, fhandle, info,
3223                                                       flav_array[i]);
3224                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3225                                 continue;
3226                         break;
3227                 }
3228         }
3229
3230         /*
3231          * -EACCESS could mean that the user doesn't have correct permissions
3232          * to access the mount.  It could also mean that we tried to mount
3233          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3234          * existing mount programs don't handle -EACCES very well so it should
3235          * be mapped to -EPERM instead.
3236          */
3237         if (status == -EACCES)
3238                 status = -EPERM;
3239         return status;
3240 }
3241
3242 static int nfs4_do_find_root_sec(struct nfs_server *server,
3243                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3244 {
3245         int mv = server->nfs_client->cl_minorversion;
3246         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3247 }
3248
3249 /**
3250  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3251  * @server: initialized nfs_server handle
3252  * @fhandle: we fill in the pseudo-fs root file handle
3253  * @info: we fill in an FSINFO struct
3254  * @auth_probe: probe the auth flavours
3255  *
3256  * Returns zero on success, or a negative errno.
3257  */
3258 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3259                          struct nfs_fsinfo *info,
3260                          bool auth_probe)
3261 {
3262         int status = 0;
3263
3264         if (!auth_probe)
3265                 status = nfs4_lookup_root(server, fhandle, info);
3266
3267         if (auth_probe || status == NFS4ERR_WRONGSEC)
3268                 status = nfs4_do_find_root_sec(server, fhandle, info);
3269
3270         if (status == 0)
3271                 status = nfs4_server_capabilities(server, fhandle);
3272         if (status == 0)
3273                 status = nfs4_do_fsinfo(server, fhandle, info);
3274
3275         return nfs4_map_errors(status);
3276 }
3277
3278 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3279                               struct nfs_fsinfo *info)
3280 {
3281         int error;
3282         struct nfs_fattr *fattr = info->fattr;
3283         struct nfs4_label *label = NULL;
3284
3285         error = nfs4_server_capabilities(server, mntfh);
3286         if (error < 0) {
3287                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3288                 return error;
3289         }
3290
3291         label = nfs4_label_alloc(server, GFP_KERNEL);
3292         if (IS_ERR(label))
3293                 return PTR_ERR(label);
3294
3295         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3296         if (error < 0) {
3297                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3298                 goto err_free_label;
3299         }
3300
3301         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3302             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3303                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3304
3305 err_free_label:
3306         nfs4_label_free(label);
3307
3308         return error;
3309 }
3310
3311 /*
3312  * Get locations and (maybe) other attributes of a referral.
3313  * Note that we'll actually follow the referral later when
3314  * we detect fsid mismatch in inode revalidation
3315  */
3316 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3317                              const struct qstr *name, struct nfs_fattr *fattr,
3318                              struct nfs_fh *fhandle)
3319 {
3320         int status = -ENOMEM;
3321         struct page *page = NULL;
3322         struct nfs4_fs_locations *locations = NULL;
3323
3324         page = alloc_page(GFP_KERNEL);
3325         if (page == NULL)
3326                 goto out;
3327         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3328         if (locations == NULL)
3329                 goto out;
3330
3331         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3332         if (status != 0)
3333                 goto out;
3334
3335         /*
3336          * If the fsid didn't change, this is a migration event, not a
3337          * referral.  Cause us to drop into the exception handler, which
3338          * will kick off migration recovery.
3339          */
3340         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3341                 dprintk("%s: server did not return a different fsid for"
3342                         " a referral at %s\n", __func__, name->name);
3343                 status = -NFS4ERR_MOVED;
3344                 goto out;
3345         }
3346         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3347         nfs_fixup_referral_attributes(&locations->fattr);
3348
3349         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3350         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3351         memset(fhandle, 0, sizeof(struct nfs_fh));
3352 out:
3353         if (page)
3354                 __free_page(page);
3355         kfree(locations);
3356         return status;
3357 }
3358
3359 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3360                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3361 {
3362         struct nfs4_getattr_arg args = {
3363                 .fh = fhandle,
3364                 .bitmask = server->attr_bitmask,
3365         };
3366         struct nfs4_getattr_res res = {
3367                 .fattr = fattr,
3368                 .label = label,
3369                 .server = server,
3370         };
3371         struct rpc_message msg = {
3372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3373                 .rpc_argp = &args,
3374                 .rpc_resp = &res,
3375         };
3376
3377         args.bitmask = nfs4_bitmask(server, label);
3378
3379         nfs_fattr_init(fattr);
3380         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3381 }
3382
3383 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3384                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3385 {
3386         struct nfs4_exception exception = { };
3387         int err;
3388         do {
3389                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3390                 trace_nfs4_getattr(server, fhandle, fattr, err);
3391                 err = nfs4_handle_exception(server, err,
3392                                 &exception);
3393         } while (exception.retry);
3394         return err;
3395 }
3396
3397 /*
3398  * The file is not closed if it is opened due to the a request to change
3399  * the size of the file. The open call will not be needed once the
3400  * VFS layer lookup-intents are implemented.
3401  *
3402  * Close is called when the inode is destroyed.
3403  * If we haven't opened the file for O_WRONLY, we
3404  * need to in the size_change case to obtain a stateid.
3405  *
3406  * Got race?
3407  * Because OPEN is always done by name in nfsv4, it is
3408  * possible that we opened a different file by the same
3409  * name.  We can recognize this race condition, but we
3410  * can't do anything about it besides returning an error.
3411  *
3412  * This will be fixed with VFS changes (lookup-intent).
3413  */
3414 static int
3415 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3416                   struct iattr *sattr)
3417 {
3418         struct inode *inode = d_inode(dentry);
3419         struct rpc_cred *cred = NULL;
3420         struct nfs4_state *state = NULL;
3421         struct nfs4_label *label = NULL;
3422         int status;
3423
3424         if (pnfs_ld_layoutret_on_setattr(inode) &&
3425             sattr->ia_valid & ATTR_SIZE &&
3426             sattr->ia_size < i_size_read(inode))
3427                 pnfs_commit_and_return_layout(inode);
3428
3429         nfs_fattr_init(fattr);
3430
3431         /* Deal with open(O_TRUNC) */
3432         if (sattr->ia_valid & ATTR_OPEN)
3433                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3434
3435         /* Optimization: if the end result is no change, don't RPC */
3436         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3437                 return 0;
3438
3439         /* Search for an existing open(O_WRITE) file */
3440         if (sattr->ia_valid & ATTR_FILE) {
3441                 struct nfs_open_context *ctx;
3442
3443                 ctx = nfs_file_open_context(sattr->ia_file);
3444                 if (ctx) {
3445                         cred = ctx->cred;
3446                         state = ctx->state;
3447                 }
3448         }
3449
3450         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3451         if (IS_ERR(label))
3452                 return PTR_ERR(label);
3453
3454         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3455         if (status == 0) {
3456                 nfs_setattr_update_inode(inode, sattr, fattr);
3457                 nfs_setsecurity(inode, fattr, label);
3458         }
3459         nfs4_label_free(label);
3460         return status;
3461 }
3462
3463 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3464                 const struct qstr *name, struct nfs_fh *fhandle,
3465                 struct nfs_fattr *fattr, struct nfs4_label *label)
3466 {
3467         struct nfs_server *server = NFS_SERVER(dir);
3468         int                    status;
3469         struct nfs4_lookup_arg args = {
3470                 .bitmask = server->attr_bitmask,
3471                 .dir_fh = NFS_FH(dir),
3472                 .name = name,
3473         };
3474         struct nfs4_lookup_res res = {
3475                 .server = server,
3476                 .fattr = fattr,
3477                 .label = label,
3478                 .fh = fhandle,
3479         };
3480         struct rpc_message msg = {
3481                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3482                 .rpc_argp = &args,
3483                 .rpc_resp = &res,
3484         };
3485
3486         args.bitmask = nfs4_bitmask(server, label);
3487
3488         nfs_fattr_init(fattr);
3489
3490         dprintk("NFS call  lookup %s\n", name->name);
3491         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3492         dprintk("NFS reply lookup: %d\n", status);
3493         return status;
3494 }
3495
3496 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3497 {
3498         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3499                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3500         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3501         fattr->nlink = 2;
3502 }
3503
3504 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3505                                    struct qstr *name, struct nfs_fh *fhandle,
3506                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3507 {
3508         struct nfs4_exception exception = { };
3509         struct rpc_clnt *client = *clnt;
3510         int err;
3511         do {
3512                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3513                 trace_nfs4_lookup(dir, name, err);
3514                 switch (err) {
3515                 case -NFS4ERR_BADNAME:
3516                         err = -ENOENT;
3517                         goto out;
3518                 case -NFS4ERR_MOVED:
3519                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3520                         if (err == -NFS4ERR_MOVED)
3521                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3522                         goto out;
3523                 case -NFS4ERR_WRONGSEC:
3524                         err = -EPERM;
3525                         if (client != *clnt)
3526                                 goto out;
3527                         client = nfs4_negotiate_security(client, dir, name);
3528                         if (IS_ERR(client))
3529                                 return PTR_ERR(client);
3530
3531                         exception.retry = 1;
3532                         break;
3533                 default:
3534                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3535                 }
3536         } while (exception.retry);
3537
3538 out:
3539         if (err == 0)
3540                 *clnt = client;
3541         else if (client != *clnt)
3542                 rpc_shutdown_client(client);
3543
3544         return err;
3545 }
3546
3547 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3548                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3549                             struct nfs4_label *label)
3550 {
3551         int status;
3552         struct rpc_clnt *client = NFS_CLIENT(dir);
3553
3554         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3555         if (client != NFS_CLIENT(dir)) {
3556                 rpc_shutdown_client(client);
3557                 nfs_fixup_secinfo_attributes(fattr);
3558         }
3559         return status;
3560 }
3561
3562 struct rpc_clnt *
3563 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3564                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3565 {
3566         struct rpc_clnt *client = NFS_CLIENT(dir);
3567         int status;
3568
3569         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3570         if (status < 0)
3571                 return ERR_PTR(status);
3572         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3573 }
3574
3575 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3576 {
3577         struct nfs_server *server = NFS_SERVER(inode);
3578         struct nfs4_accessargs args = {
3579                 .fh = NFS_FH(inode),
3580                 .bitmask = server->cache_consistency_bitmask,
3581         };
3582         struct nfs4_accessres res = {
3583                 .server = server,
3584         };
3585         struct rpc_message msg = {
3586                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3587                 .rpc_argp = &args,
3588                 .rpc_resp = &res,
3589                 .rpc_cred = entry->cred,
3590         };
3591         int mode = entry->mask;
3592         int status = 0;
3593
3594         /*
3595          * Determine which access bits we want to ask for...
3596          */
3597         if (mode & MAY_READ)
3598                 args.access |= NFS4_ACCESS_READ;
3599         if (S_ISDIR(inode->i_mode)) {
3600                 if (mode & MAY_WRITE)
3601                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3602                 if (mode & MAY_EXEC)
3603                         args.access |= NFS4_ACCESS_LOOKUP;
3604         } else {
3605                 if (mode & MAY_WRITE)
3606                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3607                 if (mode & MAY_EXEC)
3608                         args.access |= NFS4_ACCESS_EXECUTE;
3609         }
3610
3611         res.fattr = nfs_alloc_fattr();
3612         if (res.fattr == NULL)
3613                 return -ENOMEM;
3614
3615         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3616         if (!status) {
3617                 nfs_access_set_mask(entry, res.access);
3618                 nfs_refresh_inode(inode, res.fattr);
3619         }
3620         nfs_free_fattr(res.fattr);
3621         return status;
3622 }
3623
3624 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3625 {
3626         struct nfs4_exception exception = { };
3627         int err;
3628         do {
3629                 err = _nfs4_proc_access(inode, entry);
3630                 trace_nfs4_access(inode, err);
3631                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3632                                 &exception);
3633         } while (exception.retry);
3634         return err;
3635 }
3636
3637 /*
3638  * TODO: For the time being, we don't try to get any attributes
3639  * along with any of the zero-copy operations READ, READDIR,
3640  * READLINK, WRITE.
3641  *
3642  * In the case of the first three, we want to put the GETATTR
3643  * after the read-type operation -- this is because it is hard
3644  * to predict the length of a GETATTR response in v4, and thus
3645  * align the READ data correctly.  This means that the GETATTR
3646  * may end up partially falling into the page cache, and we should
3647  * shift it into the 'tail' of the xdr_buf before processing.
3648  * To do this efficiently, we need to know the total length
3649  * of data received, which doesn't seem to be available outside
3650  * of the RPC layer.
3651  *
3652  * In the case of WRITE, we also want to put the GETATTR after
3653  * the operation -- in this case because we want to make sure
3654  * we get the post-operation mtime and size.
3655  *
3656  * Both of these changes to the XDR layer would in fact be quite
3657  * minor, but I decided to leave them for a subsequent patch.
3658  */
3659 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3660                 unsigned int pgbase, unsigned int pglen)
3661 {
3662         struct nfs4_readlink args = {
3663                 .fh       = NFS_FH(inode),
3664                 .pgbase   = pgbase,
3665                 .pglen    = pglen,
3666                 .pages    = &page,
3667         };
3668         struct nfs4_readlink_res res;
3669         struct rpc_message msg = {
3670                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3671                 .rpc_argp = &args,
3672                 .rpc_resp = &res,
3673         };
3674
3675         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3676 }
3677
3678 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3679                 unsigned int pgbase, unsigned int pglen)
3680 {
3681         struct nfs4_exception exception = { };
3682         int err;
3683         do {
3684                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3685                 trace_nfs4_readlink(inode, err);
3686                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3687                                 &exception);
3688         } while (exception.retry);
3689         return err;
3690 }
3691
3692 /*
3693  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3694  */
3695 static int
3696 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3697                  int flags)
3698 {
3699         struct nfs4_label l, *ilabel = NULL;
3700         struct nfs_open_context *ctx;
3701         struct nfs4_state *state;
3702         int status = 0;
3703
3704         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3705         if (IS_ERR(ctx))
3706                 return PTR_ERR(ctx);
3707
3708         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3709
3710         sattr->ia_mode &= ~current_umask();
3711         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3712         if (IS_ERR(state)) {
3713                 status = PTR_ERR(state);
3714                 goto out;
3715         }
3716 out:
3717         nfs4_label_release_security(ilabel);
3718         put_nfs_open_context(ctx);
3719         return status;
3720 }
3721
3722 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3723 {
3724         struct nfs_server *server = NFS_SERVER(dir);
3725         struct nfs_removeargs args = {
3726                 .fh = NFS_FH(dir),
3727                 .name = *name,
3728         };
3729         struct nfs_removeres res = {
3730                 .server = server,
3731         };
3732         struct rpc_message msg = {
3733                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3734                 .rpc_argp = &args,
3735                 .rpc_resp = &res,
3736         };
3737         int status;
3738
3739         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3740         if (status == 0)
3741                 update_changeattr(dir, &res.cinfo);
3742         return status;
3743 }
3744
3745 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3746 {
3747         struct nfs4_exception exception = { };
3748         int err;
3749         do {
3750                 err = _nfs4_proc_remove(dir, name);
3751                 trace_nfs4_remove(dir, name, err);
3752                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3753                                 &exception);
3754         } while (exception.retry);
3755         return err;
3756 }
3757
3758 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3759 {
3760         struct nfs_server *server = NFS_SERVER(dir);
3761         struct nfs_removeargs *args = msg->rpc_argp;
3762         struct nfs_removeres *res = msg->rpc_resp;
3763
3764         res->server = server;
3765         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3766         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3767
3768         nfs_fattr_init(res->dir_attr);
3769 }
3770
3771 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3772 {
3773         nfs4_setup_sequence(NFS_SERVER(data->dir),
3774                         &data->args.seq_args,
3775                         &data->res.seq_res,
3776                         task);
3777 }
3778
3779 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3780 {
3781         struct nfs_unlinkdata *data = task->tk_calldata;
3782         struct nfs_removeres *res = &data->res;
3783
3784         if (!nfs4_sequence_done(task, &res->seq_res))
3785                 return 0;
3786         if (nfs4_async_handle_error(task, res->server, NULL,
3787                                     &data->timeout) == -EAGAIN)
3788                 return 0;
3789         update_changeattr(dir, &res->cinfo);
3790         return 1;
3791 }
3792
3793 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3794 {
3795         struct nfs_server *server = NFS_SERVER(dir);
3796         struct nfs_renameargs *arg = msg->rpc_argp;
3797         struct nfs_renameres *res = msg->rpc_resp;
3798
3799         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3800         res->server = server;
3801         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3802 }
3803
3804 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3805 {
3806         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3807                         &data->args.seq_args,
3808                         &data->res.seq_res,
3809                         task);
3810 }
3811
3812 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3813                                  struct inode *new_dir)
3814 {
3815         struct nfs_renamedata *data = task->tk_calldata;
3816         struct nfs_renameres *res = &data->res;
3817
3818         if (!nfs4_sequence_done(task, &res->seq_res))
3819                 return 0;
3820         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3821                 return 0;
3822
3823         update_changeattr(old_dir, &res->old_cinfo);
3824         update_changeattr(new_dir, &res->new_cinfo);
3825         return 1;
3826 }
3827
3828 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3829 {
3830         struct nfs_server *server = NFS_SERVER(inode);
3831         struct nfs4_link_arg arg = {
3832                 .fh     = NFS_FH(inode),
3833                 .dir_fh = NFS_FH(dir),
3834                 .name   = name,
3835                 .bitmask = server->attr_bitmask,
3836         };
3837         struct nfs4_link_res res = {
3838                 .server = server,
3839                 .label = NULL,
3840         };
3841         struct rpc_message msg = {
3842                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3843                 .rpc_argp = &arg,
3844                 .rpc_resp = &res,
3845         };
3846         int status = -ENOMEM;
3847
3848         res.fattr = nfs_alloc_fattr();
3849         if (res.fattr == NULL)
3850                 goto out;
3851
3852         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3853         if (IS_ERR(res.label)) {
3854                 status = PTR_ERR(res.label);
3855                 goto out;
3856         }
3857         arg.bitmask = nfs4_bitmask(server, res.label);
3858
3859         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3860         if (!status) {
3861                 update_changeattr(dir, &res.cinfo);
3862                 status = nfs_post_op_update_inode(inode, res.fattr);
3863                 if (!status)
3864                         nfs_setsecurity(inode, res.fattr, res.label);
3865         }
3866
3867
3868         nfs4_label_free(res.label);
3869
3870 out:
3871         nfs_free_fattr(res.fattr);
3872         return status;
3873 }
3874
3875 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3876 {
3877         struct nfs4_exception exception = { };
3878         int err;
3879         do {
3880                 err = nfs4_handle_exception(NFS_SERVER(inode),
3881                                 _nfs4_proc_link(inode, dir, name),
3882                                 &exception);
3883         } while (exception.retry);
3884         return err;
3885 }
3886
3887 struct nfs4_createdata {
3888         struct rpc_message msg;
3889         struct nfs4_create_arg arg;
3890         struct nfs4_create_res res;
3891         struct nfs_fh fh;
3892         struct nfs_fattr fattr;
3893         struct nfs4_label *label;
3894 };
3895
3896 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3897                 struct qstr *name, struct iattr *sattr, u32 ftype)
3898 {
3899         struct nfs4_createdata *data;
3900
3901         data = kzalloc(sizeof(*data), GFP_KERNEL);
3902         if (data != NULL) {
3903                 struct nfs_server *server = NFS_SERVER(dir);
3904
3905                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3906                 if (IS_ERR(data->label))
3907                         goto out_free;
3908
3909                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3910                 data->msg.rpc_argp = &data->arg;
3911                 data->msg.rpc_resp = &data->res;
3912                 data->arg.dir_fh = NFS_FH(dir);
3913                 data->arg.server = server;
3914                 data->arg.name = name;
3915                 data->arg.attrs = sattr;
3916                 data->arg.ftype = ftype;
3917                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3918                 data->res.server = server;
3919                 data->res.fh = &data->fh;
3920                 data->res.fattr = &data->fattr;
3921                 data->res.label = data->label;
3922                 nfs_fattr_init(data->res.fattr);
3923         }
3924         return data;
3925 out_free:
3926         kfree(data);
3927         return NULL;
3928 }
3929
3930 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3931 {
3932         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3933                                     &data->arg.seq_args, &data->res.seq_res, 1);
3934         if (status == 0) {
3935                 update_changeattr(dir, &data->res.dir_cinfo);
3936                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3937         }
3938         return status;
3939 }
3940
3941 static void nfs4_free_createdata(struct nfs4_createdata *data)
3942 {
3943         nfs4_label_free(data->label);
3944         kfree(data);
3945 }
3946
3947 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3948                 struct page *page, unsigned int len, struct iattr *sattr,
3949                 struct nfs4_label *label)
3950 {
3951         struct nfs4_createdata *data;
3952         int status = -ENAMETOOLONG;
3953
3954         if (len > NFS4_MAXPATHLEN)
3955                 goto out;
3956
3957         status = -ENOMEM;
3958         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3959         if (data == NULL)
3960                 goto out;
3961
3962         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3963         data->arg.u.symlink.pages = &page;
3964         data->arg.u.symlink.len = len;
3965         data->arg.label = label;
3966
3967         status = nfs4_do_create(dir, dentry, data);
3968
3969         nfs4_free_createdata(data);
3970 out:
3971         return status;
3972 }
3973
3974 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3975                 struct page *page, unsigned int len, struct iattr *sattr)
3976 {
3977         struct nfs4_exception exception = { };
3978         struct nfs4_label l, *label = NULL;
3979         int err;
3980
3981         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3982
3983         do {
3984                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3985                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3986                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3987                                 &exception);
3988         } while (exception.retry);
3989
3990         nfs4_label_release_security(label);
3991         return err;
3992 }
3993
3994 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3995                 struct iattr *sattr, struct nfs4_label *label)
3996 {
3997         struct nfs4_createdata *data;
3998         int status = -ENOMEM;
3999
4000         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4001         if (data == NULL)
4002                 goto out;
4003
4004         data->arg.label = label;
4005         status = nfs4_do_create(dir, dentry, data);
4006
4007         nfs4_free_createdata(data);
4008 out:
4009         return status;
4010 }
4011
4012 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4013                 struct iattr *sattr)
4014 {
4015         struct nfs4_exception exception = { };
4016         struct nfs4_label l, *label = NULL;
4017         int err;
4018
4019         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4020
4021         sattr->ia_mode &= ~current_umask();
4022         do {
4023                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4024                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4025                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4026                                 &exception);
4027         } while (exception.retry);
4028         nfs4_label_release_security(label);
4029
4030         return err;
4031 }
4032
4033 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4034                 u64 cookie, struct page **pages, unsigned int count, int plus)
4035 {
4036         struct inode            *dir = d_inode(dentry);
4037         struct nfs4_readdir_arg args = {
4038                 .fh = NFS_FH(dir),
4039                 .pages = pages,
4040                 .pgbase = 0,
4041                 .count = count,
4042                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4043                 .plus = plus,
4044         };
4045         struct nfs4_readdir_res res;
4046         struct rpc_message msg = {
4047                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4048                 .rpc_argp = &args,
4049                 .rpc_resp = &res,
4050                 .rpc_cred = cred,
4051         };
4052         int                     status;
4053
4054         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4055                         dentry,
4056                         (unsigned long long)cookie);
4057         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4058         res.pgbase = args.pgbase;
4059         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4060         if (status >= 0) {
4061                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4062                 status += args.pgbase;
4063         }
4064
4065         nfs_invalidate_atime(dir);
4066
4067         dprintk("%s: returns %d\n", __func__, status);
4068         return status;
4069 }
4070
4071 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4072                 u64 cookie, struct page **pages, unsigned int count, int plus)
4073 {
4074         struct nfs4_exception exception = { };
4075         int err;
4076         do {
4077                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4078                                 pages, count, plus);
4079                 trace_nfs4_readdir(d_inode(dentry), err);
4080                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4081                                 &exception);
4082         } while (exception.retry);
4083         return err;
4084 }
4085
4086 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4087                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4088 {
4089         struct nfs4_createdata *data;
4090         int mode = sattr->ia_mode;
4091         int status = -ENOMEM;
4092
4093         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4094         if (data == NULL)
4095                 goto out;
4096
4097         if (S_ISFIFO(mode))
4098                 data->arg.ftype = NF4FIFO;
4099         else if (S_ISBLK(mode)) {
4100                 data->arg.ftype = NF4BLK;
4101                 data->arg.u.device.specdata1 = MAJOR(rdev);
4102                 data->arg.u.device.specdata2 = MINOR(rdev);
4103         }
4104         else if (S_ISCHR(mode)) {
4105                 data->arg.ftype = NF4CHR;
4106                 data->arg.u.device.specdata1 = MAJOR(rdev);
4107                 data->arg.u.device.specdata2 = MINOR(rdev);
4108         } else if (!S_ISSOCK(mode)) {
4109                 status = -EINVAL;
4110                 goto out_free;
4111         }
4112
4113         data->arg.label = label;
4114         status = nfs4_do_create(dir, dentry, data);
4115 out_free:
4116         nfs4_free_createdata(data);
4117 out:
4118         return status;
4119 }
4120
4121 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4122                 struct iattr *sattr, dev_t rdev)
4123 {
4124         struct nfs4_exception exception = { };
4125         struct nfs4_label l, *label = NULL;
4126         int err;
4127
4128         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4129
4130         sattr->ia_mode &= ~current_umask();
4131         do {
4132                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4133                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4134                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4135                                 &exception);
4136         } while (exception.retry);
4137
4138         nfs4_label_release_security(label);
4139
4140         return err;
4141 }
4142
4143 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4144                  struct nfs_fsstat *fsstat)
4145 {
4146         struct nfs4_statfs_arg args = {
4147                 .fh = fhandle,
4148                 .bitmask = server->attr_bitmask,
4149         };
4150         struct nfs4_statfs_res res = {
4151                 .fsstat = fsstat,
4152         };
4153         struct rpc_message msg = {
4154                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4155                 .rpc_argp = &args,
4156                 .rpc_resp = &res,
4157         };
4158
4159         nfs_fattr_init(fsstat->fattr);
4160         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4161 }
4162
4163 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4164 {
4165         struct nfs4_exception exception = { };
4166         int err;
4167         do {
4168                 err = nfs4_handle_exception(server,
4169                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4170                                 &exception);
4171         } while (exception.retry);
4172         return err;
4173 }
4174
4175 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4176                 struct nfs_fsinfo *fsinfo)
4177 {
4178         struct nfs4_fsinfo_arg args = {
4179                 .fh = fhandle,
4180                 .bitmask = server->attr_bitmask,
4181         };
4182         struct nfs4_fsinfo_res res = {
4183                 .fsinfo = fsinfo,
4184         };
4185         struct rpc_message msg = {
4186                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4187                 .rpc_argp = &args,
4188                 .rpc_resp = &res,
4189         };
4190
4191         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4192 }
4193
4194 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4195 {
4196         struct nfs4_exception exception = { };
4197         unsigned long now = jiffies;
4198         int err;
4199
4200         do {
4201                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4202                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4203                 if (err == 0) {
4204                         struct nfs_client *clp = server->nfs_client;
4205
4206                         spin_lock(&clp->cl_lock);
4207                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4208                         clp->cl_last_renewal = now;
4209                         spin_unlock(&clp->cl_lock);
4210                         break;
4211                 }
4212                 err = nfs4_handle_exception(server, err, &exception);
4213         } while (exception.retry);
4214         return err;
4215 }
4216
4217 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4218 {
4219         int error;
4220
4221         nfs_fattr_init(fsinfo->fattr);
4222         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4223         if (error == 0) {
4224                 /* block layout checks this! */
4225                 server->pnfs_blksize = fsinfo->blksize;
4226                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4227         }
4228
4229         return error;
4230 }
4231
4232 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4233                 struct nfs_pathconf *pathconf)
4234 {
4235         struct nfs4_pathconf_arg args = {
4236                 .fh = fhandle,
4237                 .bitmask = server->attr_bitmask,
4238         };
4239         struct nfs4_pathconf_res res = {
4240                 .pathconf = pathconf,
4241         };
4242         struct rpc_message msg = {
4243                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4244                 .rpc_argp = &args,
4245                 .rpc_resp = &res,
4246         };
4247
4248         /* None of the pathconf attributes are mandatory to implement */
4249         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4250                 memset(pathconf, 0, sizeof(*pathconf));
4251                 return 0;
4252         }
4253
4254         nfs_fattr_init(pathconf->fattr);
4255         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4256 }
4257
4258 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4259                 struct nfs_pathconf *pathconf)
4260 {
4261         struct nfs4_exception exception = { };
4262         int err;
4263
4264         do {
4265                 err = nfs4_handle_exception(server,
4266                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4267                                 &exception);
4268         } while (exception.retry);
4269         return err;
4270 }
4271
4272 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4273                 const struct nfs_open_context *ctx,
4274                 const struct nfs_lock_context *l_ctx,
4275                 fmode_t fmode)
4276 {
4277         const struct nfs_lockowner *lockowner = NULL;
4278
4279         if (l_ctx != NULL)
4280                 lockowner = &l_ctx->lockowner;
4281         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4282 }
4283 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4284
4285 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4286                 const struct nfs_open_context *ctx,
4287                 const struct nfs_lock_context *l_ctx,
4288                 fmode_t fmode)
4289 {
4290         nfs4_stateid current_stateid;
4291
4292         /* If the current stateid represents a lost lock, then exit */
4293         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4294                 return true;
4295         return nfs4_stateid_match(stateid, &current_stateid);
4296 }
4297
4298 static bool nfs4_error_stateid_expired(int err)
4299 {
4300         switch (err) {
4301         case -NFS4ERR_DELEG_REVOKED:
4302         case -NFS4ERR_ADMIN_REVOKED:
4303         case -NFS4ERR_BAD_STATEID:
4304         case -NFS4ERR_STALE_STATEID:
4305         case -NFS4ERR_OLD_STATEID:
4306         case -NFS4ERR_OPENMODE:
4307         case -NFS4ERR_EXPIRED:
4308                 return true;
4309         }
4310         return false;
4311 }
4312
4313 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4314 {
4315         nfs_invalidate_atime(hdr->inode);
4316 }
4317
4318 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4319 {
4320         struct nfs_server *server = NFS_SERVER(hdr->inode);
4321
4322         trace_nfs4_read(hdr, task->tk_status);
4323         if (nfs4_async_handle_error(task, server,
4324                                     hdr->args.context->state,
4325                                     NULL) == -EAGAIN) {
4326                 rpc_restart_call_prepare(task);
4327                 return -EAGAIN;
4328         }
4329
4330         __nfs4_read_done_cb(hdr);
4331         if (task->tk_status > 0)
4332                 renew_lease(server, hdr->timestamp);
4333         return 0;
4334 }
4335
4336 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4337                 struct nfs_pgio_args *args)
4338 {
4339
4340         if (!nfs4_error_stateid_expired(task->tk_status) ||
4341                 nfs4_stateid_is_current(&args->stateid,
4342                                 args->context,
4343                                 args->lock_context,
4344                                 FMODE_READ))
4345                 return false;
4346         rpc_restart_call_prepare(task);
4347         return true;
4348 }
4349
4350 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4351 {
4352
4353         dprintk("--> %s\n", __func__);
4354
4355         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4356                 return -EAGAIN;
4357         if (nfs4_read_stateid_changed(task, &hdr->args))
4358                 return -EAGAIN;
4359         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4360                                     nfs4_read_done_cb(task, hdr);
4361 }
4362
4363 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4364                                  struct rpc_message *msg)
4365 {
4366         hdr->timestamp   = jiffies;
4367         hdr->pgio_done_cb = nfs4_read_done_cb;
4368         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4369         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4370 }
4371
4372 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4373                                       struct nfs_pgio_header *hdr)
4374 {
4375         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4376                         &hdr->args.seq_args,
4377                         &hdr->res.seq_res,
4378                         task))
4379                 return 0;
4380         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4381                                 hdr->args.lock_context,
4382                                 hdr->rw_ops->rw_mode) == -EIO)
4383                 return -EIO;
4384         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4385                 return -EIO;
4386         return 0;
4387 }
4388
4389 static int nfs4_write_done_cb(struct rpc_task *task,
4390                               struct nfs_pgio_header *hdr)
4391 {
4392         struct inode *inode = hdr->inode;
4393
4394         trace_nfs4_write(hdr, task->tk_status);
4395         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4396                                     hdr->args.context->state,
4397                                     NULL) == -EAGAIN) {
4398                 rpc_restart_call_prepare(task);
4399                 return -EAGAIN;
4400         }
4401         if (task->tk_status >= 0) {
4402                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4403                 nfs_writeback_update_inode(hdr);
4404         }
4405         return 0;
4406 }
4407
4408 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4409                 struct nfs_pgio_args *args)
4410 {
4411
4412         if (!nfs4_error_stateid_expired(task->tk_status) ||
4413                 nfs4_stateid_is_current(&args->stateid,
4414                                 args->context,
4415                                 args->lock_context,
4416                                 FMODE_WRITE))
4417                 return false;
4418         rpc_restart_call_prepare(task);
4419         return true;
4420 }
4421
4422 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4423 {
4424         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4425                 return -EAGAIN;
4426         if (nfs4_write_stateid_changed(task, &hdr->args))
4427                 return -EAGAIN;
4428         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4429                 nfs4_write_done_cb(task, hdr);
4430 }
4431
4432 static
4433 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4434 {
4435         /* Don't request attributes for pNFS or O_DIRECT writes */
4436         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4437                 return false;
4438         /* Otherwise, request attributes if and only if we don't hold
4439          * a delegation
4440          */
4441         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4442 }
4443
4444 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4445                                   struct rpc_message *msg)
4446 {
4447         struct nfs_server *server = NFS_SERVER(hdr->inode);
4448
4449         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4450                 hdr->args.bitmask = NULL;
4451                 hdr->res.fattr = NULL;
4452         } else
4453                 hdr->args.bitmask = server->cache_consistency_bitmask;
4454
4455         if (!hdr->pgio_done_cb)
4456                 hdr->pgio_done_cb = nfs4_write_done_cb;
4457         hdr->res.server = server;
4458         hdr->timestamp   = jiffies;
4459
4460         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4461         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4462 }
4463
4464 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4465 {
4466         nfs4_setup_sequence(NFS_SERVER(data->inode),
4467                         &data->args.seq_args,
4468                         &data->res.seq_res,
4469                         task);
4470 }
4471
4472 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4473 {
4474         struct inode *inode = data->inode;
4475
4476         trace_nfs4_commit(data, task->tk_status);
4477         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4478                                     NULL, NULL) == -EAGAIN) {
4479                 rpc_restart_call_prepare(task);
4480                 return -EAGAIN;
4481         }
4482         return 0;
4483 }
4484
4485 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4486 {
4487         if (!nfs4_sequence_done(task, &data->res.seq_res))
4488                 return -EAGAIN;
4489         return data->commit_done_cb(task, data);
4490 }
4491
4492 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4493 {
4494         struct nfs_server *server = NFS_SERVER(data->inode);
4495
4496         if (data->commit_done_cb == NULL)
4497                 data->commit_done_cb = nfs4_commit_done_cb;
4498         data->res.server = server;
4499         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4500         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4501 }
4502
4503 struct nfs4_renewdata {
4504         struct nfs_client       *client;
4505         unsigned long           timestamp;
4506 };
4507
4508 /*
4509  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4510  * standalone procedure for queueing an asynchronous RENEW.
4511  */
4512 static void nfs4_renew_release(void *calldata)
4513 {
4514         struct nfs4_renewdata *data = calldata;
4515         struct nfs_client *clp = data->client;
4516
4517         if (atomic_read(&clp->cl_count) > 1)
4518                 nfs4_schedule_state_renewal(clp);
4519         nfs_put_client(clp);
4520         kfree(data);
4521 }
4522
4523 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4524 {
4525         struct nfs4_renewdata *data = calldata;
4526         struct nfs_client *clp = data->client;
4527         unsigned long timestamp = data->timestamp;
4528
4529         trace_nfs4_renew_async(clp, task->tk_status);
4530         switch (task->tk_status) {
4531         case 0:
4532                 break;
4533         case -NFS4ERR_LEASE_MOVED:
4534                 nfs4_schedule_lease_moved_recovery(clp);
4535                 break;
4536         default:
4537                 /* Unless we're shutting down, schedule state recovery! */
4538                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4539                         return;
4540                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4541                         nfs4_schedule_lease_recovery(clp);
4542                         return;
4543                 }
4544                 nfs4_schedule_path_down_recovery(clp);
4545         }
4546         do_renew_lease(clp, timestamp);
4547 }
4548
4549 static const struct rpc_call_ops nfs4_renew_ops = {
4550         .rpc_call_done = nfs4_renew_done,
4551         .rpc_release = nfs4_renew_release,
4552 };
4553
4554 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4555 {
4556         struct rpc_message msg = {
4557                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4558                 .rpc_argp       = clp,
4559                 .rpc_cred       = cred,
4560         };
4561         struct nfs4_renewdata *data;
4562
4563         if (renew_flags == 0)
4564                 return 0;
4565         if (!atomic_inc_not_zero(&clp->cl_count))
4566                 return -EIO;
4567         data = kmalloc(sizeof(*data), GFP_NOFS);
4568         if (data == NULL)
4569                 return -ENOMEM;
4570         data->client = clp;
4571         data->timestamp = jiffies;
4572         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4573                         &nfs4_renew_ops, data);
4574 }
4575
4576 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4577 {
4578         struct rpc_message msg = {
4579                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4580                 .rpc_argp       = clp,
4581                 .rpc_cred       = cred,
4582         };
4583         unsigned long now = jiffies;
4584         int status;
4585
4586         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4587         if (status < 0)
4588                 return status;
4589         do_renew_lease(clp, now);
4590         return 0;
4591 }
4592
4593 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4594 {
4595         return server->caps & NFS_CAP_ACLS;
4596 }
4597
4598 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4599  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4600  * the stack.
4601  */
4602 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4603
4604 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4605                 struct page **pages, unsigned int *pgbase)
4606 {
4607         struct page *newpage, **spages;
4608         int rc = 0;
4609         size_t len;
4610         spages = pages;
4611
4612         do {
4613                 len = min_t(size_t, PAGE_SIZE, buflen);
4614                 newpage = alloc_page(GFP_KERNEL);
4615
4616                 if (newpage == NULL)
4617                         goto unwind;
4618                 memcpy(page_address(newpage), buf, len);
4619                 buf += len;
4620                 buflen -= len;
4621                 *pages++ = newpage;
4622                 rc++;
4623         } while (buflen != 0);
4624
4625         return rc;
4626
4627 unwind:
4628         for(; rc > 0; rc--)
4629                 __free_page(spages[rc-1]);
4630         return -ENOMEM;
4631 }
4632
4633 struct nfs4_cached_acl {
4634         int cached;
4635         size_t len;
4636         char data[0];
4637 };
4638
4639 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4640 {
4641         struct nfs_inode *nfsi = NFS_I(inode);
4642
4643         spin_lock(&inode->i_lock);
4644         kfree(nfsi->nfs4_acl);
4645         nfsi->nfs4_acl = acl;
4646         spin_unlock(&inode->i_lock);
4647 }
4648
4649 static void nfs4_zap_acl_attr(struct inode *inode)
4650 {
4651         nfs4_set_cached_acl(inode, NULL);
4652 }
4653
4654 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4655 {
4656         struct nfs_inode *nfsi = NFS_I(inode);
4657         struct nfs4_cached_acl *acl;
4658         int ret = -ENOENT;
4659
4660         spin_lock(&inode->i_lock);
4661         acl = nfsi->nfs4_acl;
4662         if (acl == NULL)
4663                 goto out;
4664         if (buf == NULL) /* user is just asking for length */
4665                 goto out_len;
4666         if (acl->cached == 0)
4667                 goto out;
4668         ret = -ERANGE; /* see getxattr(2) man page */
4669         if (acl->len > buflen)
4670                 goto out;
4671         memcpy(buf, acl->data, acl->len);
4672 out_len:
4673         ret = acl->len;
4674 out:
4675         spin_unlock(&inode->i_lock);
4676         return ret;
4677 }
4678
4679 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4680 {
4681         struct nfs4_cached_acl *acl;
4682         size_t buflen = sizeof(*acl) + acl_len;
4683
4684         if (buflen <= PAGE_SIZE) {
4685                 acl = kmalloc(buflen, GFP_KERNEL);
4686                 if (acl == NULL)
4687                         goto out;
4688                 acl->cached = 1;
4689                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4690         } else {
4691                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4692                 if (acl == NULL)
4693                         goto out;
4694                 acl->cached = 0;
4695         }
4696         acl->len = acl_len;
4697 out:
4698         nfs4_set_cached_acl(inode, acl);
4699 }
4700
4701 /*
4702  * The getxattr API returns the required buffer length when called with a
4703  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4704  * the required buf.  On a NULL buf, we send a page of data to the server
4705  * guessing that the ACL request can be serviced by a page. If so, we cache
4706  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4707  * the cache. If not so, we throw away the page, and cache the required
4708  * length. The next getxattr call will then produce another round trip to
4709  * the server, this time with the input buf of the required size.
4710  */
4711 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4712 {
4713         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4714         struct nfs_getaclargs args = {
4715                 .fh = NFS_FH(inode),
4716                 .acl_pages = pages,
4717                 .acl_len = buflen,
4718         };
4719         struct nfs_getaclres res = {
4720                 .acl_len = buflen,
4721         };
4722         struct rpc_message msg = {
4723                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4724                 .rpc_argp = &args,
4725                 .rpc_resp = &res,
4726         };
4727         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4728         int ret = -ENOMEM, i;
4729
4730         /* As long as we're doing a round trip to the server anyway,
4731          * let's be prepared for a page of acl data. */
4732         if (npages == 0)
4733                 npages = 1;
4734         if (npages > ARRAY_SIZE(pages))
4735                 return -ERANGE;
4736
4737         for (i = 0; i < npages; i++) {
4738                 pages[i] = alloc_page(GFP_KERNEL);
4739                 if (!pages[i])
4740                         goto out_free;
4741         }
4742
4743         /* for decoding across pages */
4744         res.acl_scratch = alloc_page(GFP_KERNEL);
4745         if (!res.acl_scratch)
4746                 goto out_free;
4747
4748         args.acl_len = npages * PAGE_SIZE;
4749         args.acl_pgbase = 0;
4750
4751         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4752                 __func__, buf, buflen, npages, args.acl_len);
4753         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4754                              &msg, &args.seq_args, &res.seq_res, 0);
4755         if (ret)
4756                 goto out_free;
4757
4758         /* Handle the case where the passed-in buffer is too short */
4759         if (res.acl_flags & NFS4_ACL_TRUNC) {
4760                 /* Did the user only issue a request for the acl length? */
4761                 if (buf == NULL)
4762                         goto out_ok;
4763                 ret = -ERANGE;
4764                 goto out_free;
4765         }
4766         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4767         if (buf) {
4768                 if (res.acl_len > buflen) {
4769                         ret = -ERANGE;
4770                         goto out_free;
4771                 }
4772                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4773         }
4774 out_ok:
4775         ret = res.acl_len;
4776 out_free:
4777         for (i = 0; i < npages; i++)
4778                 if (pages[i])
4779                         __free_page(pages[i]);
4780         if (res.acl_scratch)
4781                 __free_page(res.acl_scratch);
4782         return ret;
4783 }
4784
4785 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4786 {
4787         struct nfs4_exception exception = { };
4788         ssize_t ret;
4789         do {
4790                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4791                 trace_nfs4_get_acl(inode, ret);
4792                 if (ret >= 0)
4793                         break;
4794                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4795         } while (exception.retry);
4796         return ret;
4797 }
4798
4799 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4800 {
4801         struct nfs_server *server = NFS_SERVER(inode);
4802         int ret;
4803
4804         if (!nfs4_server_supports_acls(server))
4805                 return -EOPNOTSUPP;
4806         ret = nfs_revalidate_inode(server, inode);
4807         if (ret < 0)
4808                 return ret;
4809         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4810                 nfs_zap_acl_cache(inode);
4811         ret = nfs4_read_cached_acl(inode, buf, buflen);
4812         if (ret != -ENOENT)
4813                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4814                  * but no cached acl data, just the acl length */
4815                 return ret;
4816         return nfs4_get_acl_uncached(inode, buf, buflen);
4817 }
4818
4819 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4820 {
4821         struct nfs_server *server = NFS_SERVER(inode);
4822         struct page *pages[NFS4ACL_MAXPAGES];
4823         struct nfs_setaclargs arg = {
4824                 .fh             = NFS_FH(inode),
4825                 .acl_pages      = pages,
4826                 .acl_len        = buflen,
4827         };
4828         struct nfs_setaclres res;
4829         struct rpc_message msg = {
4830                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4831                 .rpc_argp       = &arg,
4832                 .rpc_resp       = &res,
4833         };
4834         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4835         int ret, i;
4836
4837         if (!nfs4_server_supports_acls(server))
4838                 return -EOPNOTSUPP;
4839         if (npages > ARRAY_SIZE(pages))
4840                 return -ERANGE;
4841         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4842         if (i < 0)
4843                 return i;
4844         nfs4_inode_return_delegation(inode);
4845         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4846
4847         /*
4848          * Free each page after tx, so the only ref left is
4849          * held by the network stack
4850          */
4851         for (; i > 0; i--)
4852                 put_page(pages[i-1]);
4853
4854         /*
4855          * Acl update can result in inode attribute update.
4856          * so mark the attribute cache invalid.
4857          */
4858         spin_lock(&inode->i_lock);
4859         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4860         spin_unlock(&inode->i_lock);
4861         nfs_access_zap_cache(inode);
4862         nfs_zap_acl_cache(inode);
4863         return ret;
4864 }
4865
4866 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4867 {
4868         struct nfs4_exception exception = { };
4869         int err;
4870         do {
4871                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4872                 trace_nfs4_set_acl(inode, err);
4873                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4874                                 &exception);
4875         } while (exception.retry);
4876         return err;
4877 }
4878
4879 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4880 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4881                                         size_t buflen)
4882 {
4883         struct nfs_server *server = NFS_SERVER(inode);
4884         struct nfs_fattr fattr;
4885         struct nfs4_label label = {0, 0, buflen, buf};
4886
4887         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4888         struct nfs4_getattr_arg arg = {
4889                 .fh             = NFS_FH(inode),
4890                 .bitmask        = bitmask,
4891         };
4892         struct nfs4_getattr_res res = {
4893                 .fattr          = &fattr,
4894                 .label          = &label,
4895                 .server         = server,
4896         };
4897         struct rpc_message msg = {
4898                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4899                 .rpc_argp       = &arg,
4900                 .rpc_resp       = &res,
4901         };
4902         int ret;
4903
4904         nfs_fattr_init(&fattr);
4905
4906         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4907         if (ret)
4908                 return ret;
4909         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4910                 return -ENOENT;
4911         if (buflen < label.len)
4912                 return -ERANGE;
4913         return 0;
4914 }
4915
4916 static int nfs4_get_security_label(struct inode *inode, void *buf,
4917                                         size_t buflen)
4918 {
4919         struct nfs4_exception exception = { };
4920         int err;
4921
4922         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4923                 return -EOPNOTSUPP;
4924
4925         do {
4926                 err = _nfs4_get_security_label(inode, buf, buflen);
4927                 trace_nfs4_get_security_label(inode, err);
4928                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4929                                 &exception);
4930         } while (exception.retry);
4931         return err;
4932 }
4933
4934 static int _nfs4_do_set_security_label(struct inode *inode,
4935                 struct nfs4_label *ilabel,
4936                 struct nfs_fattr *fattr,
4937                 struct nfs4_label *olabel)
4938 {
4939
4940         struct iattr sattr = {0};
4941         struct nfs_server *server = NFS_SERVER(inode);
4942         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4943         struct nfs_setattrargs arg = {
4944                 .fh             = NFS_FH(inode),
4945                 .iap            = &sattr,
4946                 .server         = server,
4947                 .bitmask        = bitmask,
4948                 .label          = ilabel,
4949         };
4950         struct nfs_setattrres res = {
4951                 .fattr          = fattr,
4952                 .label          = olabel,
4953                 .server         = server,
4954         };
4955         struct rpc_message msg = {
4956                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4957                 .rpc_argp       = &arg,
4958                 .rpc_resp       = &res,
4959         };
4960         int status;
4961
4962         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4963
4964         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4965         if (status)
4966                 dprintk("%s failed: %d\n", __func__, status);
4967
4968         return status;
4969 }
4970
4971 static int nfs4_do_set_security_label(struct inode *inode,
4972                 struct nfs4_label *ilabel,
4973                 struct nfs_fattr *fattr,
4974                 struct nfs4_label *olabel)
4975 {
4976         struct nfs4_exception exception = { };
4977         int err;
4978
4979         do {
4980                 err = _nfs4_do_set_security_label(inode, ilabel,
4981                                 fattr, olabel);
4982                 trace_nfs4_set_security_label(inode, err);
4983                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4984                                 &exception);
4985         } while (exception.retry);
4986         return err;
4987 }
4988
4989 static int
4990 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4991 {
4992         struct nfs4_label ilabel, *olabel = NULL;
4993         struct nfs_fattr fattr;
4994         struct rpc_cred *cred;
4995         struct inode *inode = d_inode(dentry);
4996         int status;
4997
4998         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4999                 return -EOPNOTSUPP;
5000
5001         nfs_fattr_init(&fattr);
5002
5003         ilabel.pi = 0;
5004         ilabel.lfs = 0;
5005         ilabel.label = (char *)buf;
5006         ilabel.len = buflen;
5007
5008         cred = rpc_lookup_cred();
5009         if (IS_ERR(cred))
5010                 return PTR_ERR(cred);
5011
5012         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5013         if (IS_ERR(olabel)) {
5014                 status = -PTR_ERR(olabel);
5015                 goto out;
5016         }
5017
5018         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5019         if (status == 0)
5020                 nfs_setsecurity(inode, &fattr, olabel);
5021
5022         nfs4_label_free(olabel);
5023 out:
5024         put_rpccred(cred);
5025         return status;
5026 }
5027 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5028
5029
5030 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5031                                     nfs4_verifier *bootverf)
5032 {
5033         __be32 verf[2];
5034
5035         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5036                 /* An impossible timestamp guarantees this value
5037                  * will never match a generated boot time. */
5038                 verf[0] = 0;
5039                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5040         } else {
5041                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5042                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5043                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5044         }
5045         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5046 }
5047
5048 static int
5049 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5050 {
5051         int result;
5052         size_t len;
5053         char *str;
5054
5055         if (clp->cl_owner_id != NULL)
5056                 return 0;
5057
5058         rcu_read_lock();
5059         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5060                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5061                 1 +
5062                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5063                 1;
5064         rcu_read_unlock();
5065
5066         if (len > NFS4_OPAQUE_LIMIT + 1)
5067                 return -EINVAL;
5068
5069         /*
5070          * Since this string is allocated at mount time, and held until the
5071          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5072          * about a memory-reclaim deadlock.
5073          */
5074         str = kmalloc(len, GFP_KERNEL);
5075         if (!str)
5076                 return -ENOMEM;
5077
5078         rcu_read_lock();
5079         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5080                         clp->cl_ipaddr,
5081                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5082                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5083         rcu_read_unlock();
5084
5085         clp->cl_owner_id = str;
5086         return 0;
5087 }
5088
5089 static int
5090 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5091 {
5092         int result;
5093         size_t len;
5094         char *str;
5095
5096         len = 10 + 10 + 1 + 10 + 1 +
5097                 strlen(nfs4_client_id_uniquifier) + 1 +
5098                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5099
5100         if (len > NFS4_OPAQUE_LIMIT + 1)
5101                 return -EINVAL;
5102
5103         /*
5104          * Since this string is allocated at mount time, and held until the
5105          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5106          * about a memory-reclaim deadlock.
5107          */
5108         str = kmalloc(len, GFP_KERNEL);
5109         if (!str)
5110                 return -ENOMEM;
5111
5112         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5113                         clp->rpc_ops->version, clp->cl_minorversion,
5114                         nfs4_client_id_uniquifier,
5115                         clp->cl_rpcclient->cl_nodename);
5116         clp->cl_owner_id = str;
5117         return 0;
5118 }
5119
5120 static int
5121 nfs4_init_uniform_client_string(struct nfs_client *clp)
5122 {
5123         int result;
5124         size_t len;
5125         char *str;
5126
5127         if (clp->cl_owner_id != NULL)
5128                 return 0;
5129
5130         if (nfs4_client_id_uniquifier[0] != '\0')
5131                 return nfs4_init_uniquifier_client_string(clp);
5132
5133         len = 10 + 10 + 1 + 10 + 1 +
5134                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5135
5136         if (len > NFS4_OPAQUE_LIMIT + 1)
5137                 return -EINVAL;
5138
5139         /*
5140          * Since this string is allocated at mount time, and held until the
5141          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5142          * about a memory-reclaim deadlock.
5143          */
5144         str = kmalloc(len, GFP_KERNEL);
5145         if (!str)
5146                 return -ENOMEM;
5147
5148         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5149                         clp->rpc_ops->version, clp->cl_minorversion,
5150                         clp->cl_rpcclient->cl_nodename);
5151         clp->cl_owner_id = str;
5152         return 0;
5153 }
5154
5155 /*
5156  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5157  * services.  Advertise one based on the address family of the
5158  * clientaddr.
5159  */
5160 static unsigned int
5161 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5162 {
5163         if (strchr(clp->cl_ipaddr, ':') != NULL)
5164                 return scnprintf(buf, len, "tcp6");
5165         else
5166                 return scnprintf(buf, len, "tcp");
5167 }
5168
5169 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5170 {
5171         struct nfs4_setclientid *sc = calldata;
5172
5173         if (task->tk_status == 0)
5174                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5175 }
5176
5177 static const struct rpc_call_ops nfs4_setclientid_ops = {
5178         .rpc_call_done = nfs4_setclientid_done,
5179 };
5180
5181 /**
5182  * nfs4_proc_setclientid - Negotiate client ID
5183  * @clp: state data structure
5184  * @program: RPC program for NFSv4 callback service
5185  * @port: IP port number for NFS4 callback service
5186  * @cred: RPC credential to use for this call
5187  * @res: where to place the result
5188  *
5189  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5190  */
5191 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5192                 unsigned short port, struct rpc_cred *cred,
5193                 struct nfs4_setclientid_res *res)
5194 {
5195         nfs4_verifier sc_verifier;
5196         struct nfs4_setclientid setclientid = {
5197                 .sc_verifier = &sc_verifier,
5198                 .sc_prog = program,
5199                 .sc_clnt = clp,
5200         };
5201         struct rpc_message msg = {
5202                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5203                 .rpc_argp = &setclientid,
5204                 .rpc_resp = res,
5205                 .rpc_cred = cred,
5206         };
5207         struct rpc_task *task;
5208         struct rpc_task_setup task_setup_data = {
5209                 .rpc_client = clp->cl_rpcclient,
5210                 .rpc_message = &msg,
5211                 .callback_ops = &nfs4_setclientid_ops,
5212                 .callback_data = &setclientid,
5213                 .flags = RPC_TASK_TIMEOUT,
5214         };
5215         int status;
5216
5217         /* nfs_client_id4 */
5218         nfs4_init_boot_verifier(clp, &sc_verifier);
5219
5220         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5221                 status = nfs4_init_uniform_client_string(clp);
5222         else
5223                 status = nfs4_init_nonuniform_client_string(clp);
5224
5225         if (status)
5226                 goto out;
5227
5228         /* cb_client4 */
5229         setclientid.sc_netid_len =
5230                                 nfs4_init_callback_netid(clp,
5231                                                 setclientid.sc_netid,
5232                                                 sizeof(setclientid.sc_netid));
5233         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5234                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5235                                 clp->cl_ipaddr, port >> 8, port & 255);
5236
5237         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5238                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5239                 clp->cl_owner_id);
5240         task = rpc_run_task(&task_setup_data);
5241         if (IS_ERR(task)) {
5242                 status = PTR_ERR(task);
5243                 goto out;
5244         }
5245         status = task->tk_status;
5246         if (setclientid.sc_cred) {
5247                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5248                 put_rpccred(setclientid.sc_cred);
5249         }
5250         rpc_put_task(task);
5251 out:
5252         trace_nfs4_setclientid(clp, status);
5253         dprintk("NFS reply setclientid: %d\n", status);
5254         return status;
5255 }
5256
5257 /**
5258  * nfs4_proc_setclientid_confirm - Confirm client ID
5259  * @clp: state data structure
5260  * @res: result of a previous SETCLIENTID
5261  * @cred: RPC credential to use for this call
5262  *
5263  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5264  */
5265 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5266                 struct nfs4_setclientid_res *arg,
5267                 struct rpc_cred *cred)
5268 {
5269         struct rpc_message msg = {
5270                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5271                 .rpc_argp = arg,
5272                 .rpc_cred = cred,
5273         };
5274         int status;
5275
5276         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5277                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5278                 clp->cl_clientid);
5279         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5280         trace_nfs4_setclientid_confirm(clp, status);
5281         dprintk("NFS reply setclientid_confirm: %d\n", status);
5282         return status;
5283 }
5284
5285 struct nfs4_delegreturndata {
5286         struct nfs4_delegreturnargs args;
5287         struct nfs4_delegreturnres res;
5288         struct nfs_fh fh;
5289         nfs4_stateid stateid;
5290         unsigned long timestamp;
5291         struct nfs_fattr fattr;
5292         int rpc_status;
5293         struct inode *inode;
5294         bool roc;
5295         u32 roc_barrier;
5296 };
5297
5298 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5299 {
5300         struct nfs4_delegreturndata *data = calldata;
5301
5302         if (!nfs4_sequence_done(task, &data->res.seq_res))
5303                 return;
5304
5305         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5306         switch (task->tk_status) {
5307         case 0:
5308                 renew_lease(data->res.server, data->timestamp);
5309         case -NFS4ERR_ADMIN_REVOKED:
5310         case -NFS4ERR_DELEG_REVOKED:
5311         case -NFS4ERR_BAD_STATEID:
5312         case -NFS4ERR_OLD_STATEID:
5313         case -NFS4ERR_STALE_STATEID:
5314         case -NFS4ERR_EXPIRED:
5315                 task->tk_status = 0;
5316                 if (data->roc)
5317                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5318                 break;
5319         default:
5320                 if (nfs4_async_handle_error(task, data->res.server,
5321                                             NULL, NULL) == -EAGAIN) {
5322                         rpc_restart_call_prepare(task);
5323                         return;
5324                 }
5325         }
5326         data->rpc_status = task->tk_status;
5327 }
5328
5329 static void nfs4_delegreturn_release(void *calldata)
5330 {
5331         struct nfs4_delegreturndata *data = calldata;
5332         struct inode *inode = data->inode;
5333
5334         if (inode) {
5335                 if (data->roc)
5336                         pnfs_roc_release(inode);
5337                 nfs_iput_and_deactive(inode);
5338         }
5339         kfree(calldata);
5340 }
5341
5342 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5343 {
5344         struct nfs4_delegreturndata *d_data;
5345
5346         d_data = (struct nfs4_delegreturndata *)data;
5347
5348         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5349                 return;
5350
5351         if (d_data->roc)
5352                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5353
5354         nfs4_setup_sequence(d_data->res.server,
5355                         &d_data->args.seq_args,
5356                         &d_data->res.seq_res,
5357                         task);
5358 }
5359
5360 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5361         .rpc_call_prepare = nfs4_delegreturn_prepare,
5362         .rpc_call_done = nfs4_delegreturn_done,
5363         .rpc_release = nfs4_delegreturn_release,
5364 };
5365
5366 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5367 {
5368         struct nfs4_delegreturndata *data;
5369         struct nfs_server *server = NFS_SERVER(inode);
5370         struct rpc_task *task;
5371         struct rpc_message msg = {
5372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5373                 .rpc_cred = cred,
5374         };
5375         struct rpc_task_setup task_setup_data = {
5376                 .rpc_client = server->client,
5377                 .rpc_message = &msg,
5378                 .callback_ops = &nfs4_delegreturn_ops,
5379                 .flags = RPC_TASK_ASYNC,
5380         };
5381         int status = 0;
5382
5383         data = kzalloc(sizeof(*data), GFP_NOFS);
5384         if (data == NULL)
5385                 return -ENOMEM;
5386         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5387         data->args.fhandle = &data->fh;
5388         data->args.stateid = &data->stateid;
5389         data->args.bitmask = server->cache_consistency_bitmask;
5390         nfs_copy_fh(&data->fh, NFS_FH(inode));
5391         nfs4_stateid_copy(&data->stateid, stateid);
5392         data->res.fattr = &data->fattr;
5393         data->res.server = server;
5394         nfs_fattr_init(data->res.fattr);
5395         data->timestamp = jiffies;
5396         data->rpc_status = 0;
5397         data->inode = nfs_igrab_and_active(inode);
5398         if (data->inode)
5399                 data->roc = nfs4_roc(inode);
5400
5401         task_setup_data.callback_data = data;
5402         msg.rpc_argp = &data->args;
5403         msg.rpc_resp = &data->res;
5404         task = rpc_run_task(&task_setup_data);
5405         if (IS_ERR(task))
5406                 return PTR_ERR(task);
5407         if (!issync)
5408                 goto out;
5409         status = nfs4_wait_for_completion_rpc_task(task);
5410         if (status != 0)
5411                 goto out;
5412         status = data->rpc_status;
5413         if (status == 0)
5414                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5415         else
5416                 nfs_refresh_inode(inode, &data->fattr);
5417 out:
5418         rpc_put_task(task);
5419         return status;
5420 }
5421
5422 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5423 {
5424         struct nfs_server *server = NFS_SERVER(inode);
5425         struct nfs4_exception exception = { };
5426         int err;
5427         do {
5428                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5429                 trace_nfs4_delegreturn(inode, err);
5430                 switch (err) {
5431                         case -NFS4ERR_STALE_STATEID:
5432                         case -NFS4ERR_EXPIRED:
5433                         case 0:
5434                                 return 0;
5435                 }
5436                 err = nfs4_handle_exception(server, err, &exception);
5437         } while (exception.retry);
5438         return err;
5439 }
5440
5441 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5442 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5443
5444 /*
5445  * sleep, with exponential backoff, and retry the LOCK operation.
5446  */
5447 static unsigned long
5448 nfs4_set_lock_task_retry(unsigned long timeout)
5449 {
5450         freezable_schedule_timeout_killable_unsafe(timeout);
5451         timeout <<= 1;
5452         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5453                 return NFS4_LOCK_MAXTIMEOUT;
5454         return timeout;
5455 }
5456
5457 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5458 {
5459         struct inode *inode = state->inode;
5460         struct nfs_server *server = NFS_SERVER(inode);
5461         struct nfs_client *clp = server->nfs_client;
5462         struct nfs_lockt_args arg = {
5463                 .fh = NFS_FH(inode),
5464                 .fl = request,
5465         };
5466         struct nfs_lockt_res res = {
5467                 .denied = request,
5468         };
5469         struct rpc_message msg = {
5470                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5471                 .rpc_argp       = &arg,
5472                 .rpc_resp       = &res,
5473                 .rpc_cred       = state->owner->so_cred,
5474         };
5475         struct nfs4_lock_state *lsp;
5476         int status;
5477
5478         arg.lock_owner.clientid = clp->cl_clientid;
5479         status = nfs4_set_lock_state(state, request);
5480         if (status != 0)
5481                 goto out;
5482         lsp = request->fl_u.nfs4_fl.owner;
5483         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5484         arg.lock_owner.s_dev = server->s_dev;
5485         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5486         switch (status) {
5487                 case 0:
5488                         request->fl_type = F_UNLCK;
5489                         break;
5490                 case -NFS4ERR_DENIED:
5491                         status = 0;
5492         }
5493         request->fl_ops->fl_release_private(request);
5494         request->fl_ops = NULL;
5495 out:
5496         return status;
5497 }
5498
5499 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5500 {
5501         struct nfs4_exception exception = { };
5502         int err;
5503
5504         do {
5505                 err = _nfs4_proc_getlk(state, cmd, request);
5506                 trace_nfs4_get_lock(request, state, cmd, err);
5507                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5508                                 &exception);
5509         } while (exception.retry);
5510         return err;
5511 }
5512
5513 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5514 {
5515         int res = 0;
5516         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5517                 case FL_POSIX:
5518                         res = posix_lock_inode_wait(inode, fl);
5519                         break;
5520                 case FL_FLOCK:
5521                         res = flock_lock_inode_wait(inode, fl);
5522                         break;
5523                 default:
5524                         BUG();
5525         }
5526         return res;
5527 }
5528
5529 struct nfs4_unlockdata {
5530         struct nfs_locku_args arg;
5531         struct nfs_locku_res res;
5532         struct nfs4_lock_state *lsp;
5533         struct nfs_open_context *ctx;
5534         struct file_lock fl;
5535         struct nfs_server *server;
5536         unsigned long timestamp;
5537 };
5538
5539 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5540                 struct nfs_open_context *ctx,
5541                 struct nfs4_lock_state *lsp,
5542                 struct nfs_seqid *seqid)
5543 {
5544         struct nfs4_unlockdata *p;
5545         struct inode *inode = lsp->ls_state->inode;
5546
5547         p = kzalloc(sizeof(*p), GFP_NOFS);
5548         if (p == NULL)
5549                 return NULL;
5550         p->arg.fh = NFS_FH(inode);
5551         p->arg.fl = &p->fl;
5552         p->arg.seqid = seqid;
5553         p->res.seqid = seqid;
5554         p->lsp = lsp;
5555         atomic_inc(&lsp->ls_count);
5556         /* Ensure we don't close file until we're done freeing locks! */
5557         p->ctx = get_nfs_open_context(ctx);
5558         memcpy(&p->fl, fl, sizeof(p->fl));
5559         p->server = NFS_SERVER(inode);
5560         return p;
5561 }
5562
5563 static void nfs4_locku_release_calldata(void *data)
5564 {
5565         struct nfs4_unlockdata *calldata = data;
5566         nfs_free_seqid(calldata->arg.seqid);
5567         nfs4_put_lock_state(calldata->lsp);
5568         put_nfs_open_context(calldata->ctx);
5569         kfree(calldata);
5570 }
5571
5572 static void nfs4_locku_done(struct rpc_task *task, void *data)
5573 {
5574         struct nfs4_unlockdata *calldata = data;
5575
5576         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5577                 return;
5578         switch (task->tk_status) {
5579                 case 0:
5580                         renew_lease(calldata->server, calldata->timestamp);
5581                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5582                         if (nfs4_update_lock_stateid(calldata->lsp,
5583                                         &calldata->res.stateid))
5584                                 break;
5585                 case -NFS4ERR_BAD_STATEID:
5586                 case -NFS4ERR_OLD_STATEID:
5587                 case -NFS4ERR_STALE_STATEID:
5588                 case -NFS4ERR_EXPIRED:
5589                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5590                                                 &calldata->lsp->ls_stateid))
5591                                 rpc_restart_call_prepare(task);
5592                         break;
5593                 default:
5594                         if (nfs4_async_handle_error(task, calldata->server,
5595                                                     NULL, NULL) == -EAGAIN)
5596                                 rpc_restart_call_prepare(task);
5597         }
5598         nfs_release_seqid(calldata->arg.seqid);
5599 }
5600
5601 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5602 {
5603         struct nfs4_unlockdata *calldata = data;
5604
5605         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5606                 goto out_wait;
5607         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5608         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5609                 /* Note: exit _without_ running nfs4_locku_done */
5610                 goto out_no_action;
5611         }
5612         calldata->timestamp = jiffies;
5613         if (nfs4_setup_sequence(calldata->server,
5614                                 &calldata->arg.seq_args,
5615                                 &calldata->res.seq_res,
5616                                 task) != 0)
5617                 nfs_release_seqid(calldata->arg.seqid);
5618         return;
5619 out_no_action:
5620         task->tk_action = NULL;
5621 out_wait:
5622         nfs4_sequence_done(task, &calldata->res.seq_res);
5623 }
5624
5625 static const struct rpc_call_ops nfs4_locku_ops = {
5626         .rpc_call_prepare = nfs4_locku_prepare,
5627         .rpc_call_done = nfs4_locku_done,
5628         .rpc_release = nfs4_locku_release_calldata,
5629 };
5630
5631 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5632                 struct nfs_open_context *ctx,
5633                 struct nfs4_lock_state *lsp,
5634                 struct nfs_seqid *seqid)
5635 {
5636         struct nfs4_unlockdata *data;
5637         struct rpc_message msg = {
5638                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5639                 .rpc_cred = ctx->cred,
5640         };
5641         struct rpc_task_setup task_setup_data = {
5642                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5643                 .rpc_message = &msg,
5644                 .callback_ops = &nfs4_locku_ops,
5645                 .workqueue = nfsiod_workqueue,
5646                 .flags = RPC_TASK_ASYNC,
5647         };
5648
5649         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5650                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5651
5652         /* Ensure this is an unlock - when canceling a lock, the
5653          * canceled lock is passed in, and it won't be an unlock.
5654          */
5655         fl->fl_type = F_UNLCK;
5656
5657         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5658         if (data == NULL) {
5659                 nfs_free_seqid(seqid);
5660                 return ERR_PTR(-ENOMEM);
5661         }
5662
5663         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5664         msg.rpc_argp = &data->arg;
5665         msg.rpc_resp = &data->res;
5666         task_setup_data.callback_data = data;
5667         return rpc_run_task(&task_setup_data);
5668 }
5669
5670 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5671 {
5672         struct inode *inode = state->inode;
5673         struct nfs4_state_owner *sp = state->owner;
5674         struct nfs_inode *nfsi = NFS_I(inode);
5675         struct nfs_seqid *seqid;
5676         struct nfs4_lock_state *lsp;
5677         struct rpc_task *task;
5678         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5679         int status = 0;
5680         unsigned char fl_flags = request->fl_flags;
5681
5682         status = nfs4_set_lock_state(state, request);
5683         /* Unlock _before_ we do the RPC call */
5684         request->fl_flags |= FL_EXISTS;
5685         /* Exclude nfs_delegation_claim_locks() */
5686         mutex_lock(&sp->so_delegreturn_mutex);
5687         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5688         down_read(&nfsi->rwsem);
5689         if (do_vfs_lock(inode, request) == -ENOENT) {
5690                 up_read(&nfsi->rwsem);
5691                 mutex_unlock(&sp->so_delegreturn_mutex);
5692                 goto out;
5693         }
5694         up_read(&nfsi->rwsem);
5695         mutex_unlock(&sp->so_delegreturn_mutex);
5696         if (status != 0)
5697                 goto out;
5698         /* Is this a delegated lock? */
5699         lsp = request->fl_u.nfs4_fl.owner;
5700         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5701                 goto out;
5702         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5703         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5704         status = -ENOMEM;
5705         if (IS_ERR(seqid))
5706                 goto out;
5707         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5708         status = PTR_ERR(task);
5709         if (IS_ERR(task))
5710                 goto out;
5711         status = nfs4_wait_for_completion_rpc_task(task);
5712         rpc_put_task(task);
5713 out:
5714         request->fl_flags = fl_flags;
5715         trace_nfs4_unlock(request, state, F_SETLK, status);
5716         return status;
5717 }
5718
5719 struct nfs4_lockdata {
5720         struct nfs_lock_args arg;
5721         struct nfs_lock_res res;
5722         struct nfs4_lock_state *lsp;
5723         struct nfs_open_context *ctx;
5724         struct file_lock fl;
5725         unsigned long timestamp;
5726         int rpc_status;
5727         int cancelled;
5728         struct nfs_server *server;
5729 };
5730
5731 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5732                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5733                 gfp_t gfp_mask)
5734 {
5735         struct nfs4_lockdata *p;
5736         struct inode *inode = lsp->ls_state->inode;
5737         struct nfs_server *server = NFS_SERVER(inode);
5738         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5739
5740         p = kzalloc(sizeof(*p), gfp_mask);
5741         if (p == NULL)
5742                 return NULL;
5743
5744         p->arg.fh = NFS_FH(inode);
5745         p->arg.fl = &p->fl;
5746         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5747         if (IS_ERR(p->arg.open_seqid))
5748                 goto out_free;
5749         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5750         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5751         if (IS_ERR(p->arg.lock_seqid))
5752                 goto out_free_seqid;
5753         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5754         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5755         p->arg.lock_owner.s_dev = server->s_dev;
5756         p->res.lock_seqid = p->arg.lock_seqid;
5757         p->lsp = lsp;
5758         p->server = server;
5759         atomic_inc(&lsp->ls_count);
5760         p->ctx = get_nfs_open_context(ctx);
5761         get_file(fl->fl_file);
5762         memcpy(&p->fl, fl, sizeof(p->fl));
5763         return p;
5764 out_free_seqid:
5765         nfs_free_seqid(p->arg.open_seqid);
5766 out_free:
5767         kfree(p);
5768         return NULL;
5769 }
5770
5771 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5772 {
5773         struct nfs4_lockdata *data = calldata;
5774         struct nfs4_state *state = data->lsp->ls_state;
5775
5776         dprintk("%s: begin!\n", __func__);
5777         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5778                 goto out_wait;
5779         /* Do we need to do an open_to_lock_owner? */
5780         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5781                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5782                         goto out_release_lock_seqid;
5783                 }
5784                 nfs4_stateid_copy(&data->arg.open_stateid,
5785                                 &state->open_stateid);
5786                 data->arg.new_lock_owner = 1;
5787                 data->res.open_seqid = data->arg.open_seqid;
5788         } else {
5789                 data->arg.new_lock_owner = 0;
5790                 nfs4_stateid_copy(&data->arg.lock_stateid,
5791                                 &data->lsp->ls_stateid);
5792         }
5793         if (!nfs4_valid_open_stateid(state)) {
5794                 data->rpc_status = -EBADF;
5795                 task->tk_action = NULL;
5796                 goto out_release_open_seqid;
5797         }
5798         data->timestamp = jiffies;
5799         if (nfs4_setup_sequence(data->server,
5800                                 &data->arg.seq_args,
5801                                 &data->res.seq_res,
5802                                 task) == 0)
5803                 return;
5804 out_release_open_seqid:
5805         nfs_release_seqid(data->arg.open_seqid);
5806 out_release_lock_seqid:
5807         nfs_release_seqid(data->arg.lock_seqid);
5808 out_wait:
5809         nfs4_sequence_done(task, &data->res.seq_res);
5810         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5811 }
5812
5813 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5814 {
5815         struct nfs4_lockdata *data = calldata;
5816         struct nfs4_lock_state *lsp = data->lsp;
5817
5818         dprintk("%s: begin!\n", __func__);
5819
5820         if (!nfs4_sequence_done(task, &data->res.seq_res))
5821                 return;
5822
5823         data->rpc_status = task->tk_status;
5824         switch (task->tk_status) {
5825         case 0:
5826                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5827                                 data->timestamp);
5828                 if (data->arg.new_lock) {
5829                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5830                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5831                                 rpc_restart_call_prepare(task);
5832                                 break;
5833                         }
5834                 }
5835                 if (data->arg.new_lock_owner != 0) {
5836                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5837                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5838                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5839                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5840                         rpc_restart_call_prepare(task);
5841                 break;
5842         case -NFS4ERR_BAD_STATEID:
5843         case -NFS4ERR_OLD_STATEID:
5844         case -NFS4ERR_STALE_STATEID:
5845         case -NFS4ERR_EXPIRED:
5846                 if (data->arg.new_lock_owner != 0) {
5847                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5848                                                 &lsp->ls_state->open_stateid))
5849                                 rpc_restart_call_prepare(task);
5850                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5851                                                 &lsp->ls_stateid))
5852                                 rpc_restart_call_prepare(task);
5853         }
5854         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5855 }
5856
5857 static void nfs4_lock_release(void *calldata)
5858 {
5859         struct nfs4_lockdata *data = calldata;
5860
5861         dprintk("%s: begin!\n", __func__);
5862         nfs_free_seqid(data->arg.open_seqid);
5863         if (data->cancelled != 0) {
5864                 struct rpc_task *task;
5865                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5866                                 data->arg.lock_seqid);
5867                 if (!IS_ERR(task))
5868                         rpc_put_task_async(task);
5869                 dprintk("%s: cancelling lock!\n", __func__);
5870         } else
5871                 nfs_free_seqid(data->arg.lock_seqid);
5872         nfs4_put_lock_state(data->lsp);
5873         put_nfs_open_context(data->ctx);
5874         fput(data->fl.fl_file);
5875         kfree(data);
5876         dprintk("%s: done!\n", __func__);
5877 }
5878
5879 static const struct rpc_call_ops nfs4_lock_ops = {
5880         .rpc_call_prepare = nfs4_lock_prepare,
5881         .rpc_call_done = nfs4_lock_done,
5882         .rpc_release = nfs4_lock_release,
5883 };
5884
5885 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5886 {
5887         switch (error) {
5888         case -NFS4ERR_ADMIN_REVOKED:
5889         case -NFS4ERR_BAD_STATEID:
5890                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5891                 if (new_lock_owner != 0 ||
5892                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5893                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5894                 break;
5895         case -NFS4ERR_STALE_STATEID:
5896                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5897         case -NFS4ERR_EXPIRED:
5898                 nfs4_schedule_lease_recovery(server->nfs_client);
5899         };
5900 }
5901
5902 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5903 {
5904         struct nfs4_lockdata *data;
5905         struct rpc_task *task;
5906         struct rpc_message msg = {
5907                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5908                 .rpc_cred = state->owner->so_cred,
5909         };
5910         struct rpc_task_setup task_setup_data = {
5911                 .rpc_client = NFS_CLIENT(state->inode),
5912                 .rpc_message = &msg,
5913                 .callback_ops = &nfs4_lock_ops,
5914                 .workqueue = nfsiod_workqueue,
5915                 .flags = RPC_TASK_ASYNC,
5916         };
5917         int ret;
5918
5919         dprintk("%s: begin!\n", __func__);
5920         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5921                         fl->fl_u.nfs4_fl.owner,
5922                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5923         if (data == NULL)
5924                 return -ENOMEM;
5925         if (IS_SETLKW(cmd))
5926                 data->arg.block = 1;
5927         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5928         msg.rpc_argp = &data->arg;
5929         msg.rpc_resp = &data->res;
5930         task_setup_data.callback_data = data;
5931         if (recovery_type > NFS_LOCK_NEW) {
5932                 if (recovery_type == NFS_LOCK_RECLAIM)
5933                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5934                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5935         } else
5936                 data->arg.new_lock = 1;
5937         task = rpc_run_task(&task_setup_data);
5938         if (IS_ERR(task))
5939                 return PTR_ERR(task);
5940         ret = nfs4_wait_for_completion_rpc_task(task);
5941         if (ret == 0) {
5942                 ret = data->rpc_status;
5943                 if (ret)
5944                         nfs4_handle_setlk_error(data->server, data->lsp,
5945                                         data->arg.new_lock_owner, ret);
5946         } else
5947                 data->cancelled = 1;
5948         rpc_put_task(task);
5949         dprintk("%s: done, ret = %d!\n", __func__, ret);
5950         return ret;
5951 }
5952
5953 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5954 {
5955         struct nfs_server *server = NFS_SERVER(state->inode);
5956         struct nfs4_exception exception = {
5957                 .inode = state->inode,
5958         };
5959         int err;
5960
5961         do {
5962                 /* Cache the lock if possible... */
5963                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5964                         return 0;
5965                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5966                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5967                 if (err != -NFS4ERR_DELAY)
5968                         break;
5969                 nfs4_handle_exception(server, err, &exception);
5970         } while (exception.retry);
5971         return err;
5972 }
5973
5974 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5975 {
5976         struct nfs_server *server = NFS_SERVER(state->inode);
5977         struct nfs4_exception exception = {
5978                 .inode = state->inode,
5979         };
5980         int err;
5981
5982         err = nfs4_set_lock_state(state, request);
5983         if (err != 0)
5984                 return err;
5985         if (!recover_lost_locks) {
5986                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5987                 return 0;
5988         }
5989         do {
5990                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5991                         return 0;
5992                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5993                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5994                 switch (err) {
5995                 default:
5996                         goto out;
5997                 case -NFS4ERR_GRACE:
5998                 case -NFS4ERR_DELAY:
5999                         nfs4_handle_exception(server, err, &exception);
6000                         err = 0;
6001                 }
6002         } while (exception.retry);
6003 out:
6004         return err;
6005 }
6006
6007 #if defined(CONFIG_NFS_V4_1)
6008 /**
6009  * nfs41_check_expired_locks - possibly free a lock stateid
6010  *
6011  * @state: NFSv4 state for an inode
6012  *
6013  * Returns NFS_OK if recovery for this stateid is now finished.
6014  * Otherwise a negative NFS4ERR value is returned.
6015  */
6016 static int nfs41_check_expired_locks(struct nfs4_state *state)
6017 {
6018         int status, ret = -NFS4ERR_BAD_STATEID;
6019         struct nfs4_lock_state *lsp;
6020         struct nfs_server *server = NFS_SERVER(state->inode);
6021
6022         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6023                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6024                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6025
6026                         status = nfs41_test_stateid(server,
6027                                         &lsp->ls_stateid,
6028                                         cred);
6029                         trace_nfs4_test_lock_stateid(state, lsp, status);
6030                         if (status != NFS_OK) {
6031                                 /* Free the stateid unless the server
6032                                  * informs us the stateid is unrecognized. */
6033                                 if (status != -NFS4ERR_BAD_STATEID)
6034                                         nfs41_free_stateid(server,
6035                                                         &lsp->ls_stateid,
6036                                                         cred);
6037                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6038                                 ret = status;
6039                         }
6040                 }
6041         };
6042
6043         return ret;
6044 }
6045
6046 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6047 {
6048         int status = NFS_OK;
6049
6050         if (test_bit(LK_STATE_IN_USE, &state->flags))
6051                 status = nfs41_check_expired_locks(state);
6052         if (status != NFS_OK)
6053                 status = nfs4_lock_expired(state, request);
6054         return status;
6055 }
6056 #endif
6057
6058 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6059 {
6060         struct nfs_inode *nfsi = NFS_I(state->inode);
6061         unsigned char fl_flags = request->fl_flags;
6062         int status = -ENOLCK;
6063
6064         if ((fl_flags & FL_POSIX) &&
6065                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6066                 goto out;
6067         /* Is this a delegated open? */
6068         status = nfs4_set_lock_state(state, request);
6069         if (status != 0)
6070                 goto out;
6071         request->fl_flags |= FL_ACCESS;
6072         status = do_vfs_lock(state->inode, request);
6073         if (status < 0)
6074                 goto out;
6075         down_read(&nfsi->rwsem);
6076         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6077                 /* Yes: cache locks! */
6078                 /* ...but avoid races with delegation recall... */
6079                 request->fl_flags = fl_flags & ~FL_SLEEP;
6080                 status = do_vfs_lock(state->inode, request);
6081                 up_read(&nfsi->rwsem);
6082                 goto out;
6083         }
6084         up_read(&nfsi->rwsem);
6085         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6086 out:
6087         request->fl_flags = fl_flags;
6088         return status;
6089 }
6090
6091 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6092 {
6093         struct nfs4_exception exception = {
6094                 .state = state,
6095                 .inode = state->inode,
6096         };
6097         int err;
6098
6099         do {
6100                 err = _nfs4_proc_setlk(state, cmd, request);
6101                 trace_nfs4_set_lock(request, state, cmd, err);
6102                 if (err == -NFS4ERR_DENIED)
6103                         err = -EAGAIN;
6104                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6105                                 err, &exception);
6106         } while (exception.retry);
6107         return err;
6108 }
6109
6110 static int
6111 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6112 {
6113         struct nfs_open_context *ctx;
6114         struct nfs4_state *state;
6115         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6116         int status;
6117
6118         /* verify open state */
6119         ctx = nfs_file_open_context(filp);
6120         state = ctx->state;
6121
6122         if (request->fl_start < 0 || request->fl_end < 0)
6123                 return -EINVAL;
6124
6125         if (IS_GETLK(cmd)) {
6126                 if (state != NULL)
6127                         return nfs4_proc_getlk(state, F_GETLK, request);
6128                 return 0;
6129         }
6130
6131         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6132                 return -EINVAL;
6133
6134         if (request->fl_type == F_UNLCK) {
6135                 if (state != NULL)
6136                         return nfs4_proc_unlck(state, cmd, request);
6137                 return 0;
6138         }
6139
6140         if (state == NULL)
6141                 return -ENOLCK;
6142         /*
6143          * Don't rely on the VFS having checked the file open mode,
6144          * since it won't do this for flock() locks.
6145          */
6146         switch (request->fl_type) {
6147         case F_RDLCK:
6148                 if (!(filp->f_mode & FMODE_READ))
6149                         return -EBADF;
6150                 break;
6151         case F_WRLCK:
6152                 if (!(filp->f_mode & FMODE_WRITE))
6153                         return -EBADF;
6154         }
6155
6156         do {
6157                 status = nfs4_proc_setlk(state, cmd, request);
6158                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6159                         break;
6160                 timeout = nfs4_set_lock_task_retry(timeout);
6161                 status = -ERESTARTSYS;
6162                 if (signalled())
6163                         break;
6164         } while(status < 0);
6165         return status;
6166 }
6167
6168 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6169 {
6170         struct nfs_server *server = NFS_SERVER(state->inode);
6171         int err;
6172
6173         err = nfs4_set_lock_state(state, fl);
6174         if (err != 0)
6175                 return err;
6176         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6177         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6178 }
6179
6180 struct nfs_release_lockowner_data {
6181         struct nfs4_lock_state *lsp;
6182         struct nfs_server *server;
6183         struct nfs_release_lockowner_args args;
6184         struct nfs_release_lockowner_res res;
6185         unsigned long timestamp;
6186 };
6187
6188 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6189 {
6190         struct nfs_release_lockowner_data *data = calldata;
6191         struct nfs_server *server = data->server;
6192         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6193                              &data->args.seq_args, &data->res.seq_res, task);
6194         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6195         data->timestamp = jiffies;
6196 }
6197
6198 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6199 {
6200         struct nfs_release_lockowner_data *data = calldata;
6201         struct nfs_server *server = data->server;
6202
6203         nfs40_sequence_done(task, &data->res.seq_res);
6204
6205         switch (task->tk_status) {
6206         case 0:
6207                 renew_lease(server, data->timestamp);
6208                 break;
6209         case -NFS4ERR_STALE_CLIENTID:
6210         case -NFS4ERR_EXPIRED:
6211                 nfs4_schedule_lease_recovery(server->nfs_client);
6212                 break;
6213         case -NFS4ERR_LEASE_MOVED:
6214         case -NFS4ERR_DELAY:
6215                 if (nfs4_async_handle_error(task, server,
6216                                             NULL, NULL) == -EAGAIN)
6217                         rpc_restart_call_prepare(task);
6218         }
6219 }
6220
6221 static void nfs4_release_lockowner_release(void *calldata)
6222 {
6223         struct nfs_release_lockowner_data *data = calldata;
6224         nfs4_free_lock_state(data->server, data->lsp);
6225         kfree(calldata);
6226 }
6227
6228 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6229         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6230         .rpc_call_done = nfs4_release_lockowner_done,
6231         .rpc_release = nfs4_release_lockowner_release,
6232 };
6233
6234 static void
6235 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6236 {
6237         struct nfs_release_lockowner_data *data;
6238         struct rpc_message msg = {
6239                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6240         };
6241
6242         if (server->nfs_client->cl_mvops->minor_version != 0)
6243                 return;
6244
6245         data = kmalloc(sizeof(*data), GFP_NOFS);
6246         if (!data)
6247                 return;
6248         data->lsp = lsp;
6249         data->server = server;
6250         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6251         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6252         data->args.lock_owner.s_dev = server->s_dev;
6253
6254         msg.rpc_argp = &data->args;
6255         msg.rpc_resp = &data->res;
6256         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6257         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6258 }
6259
6260 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6261
6262 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6263                                    const void *buf, size_t buflen,
6264                                    int flags, int type)
6265 {
6266         if (strcmp(key, "") != 0)
6267                 return -EINVAL;
6268
6269         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6270 }
6271
6272 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6273                                    void *buf, size_t buflen, int type)
6274 {
6275         if (strcmp(key, "") != 0)
6276                 return -EINVAL;
6277
6278         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6279 }
6280
6281 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6282                                        size_t list_len, const char *name,
6283                                        size_t name_len, int type)
6284 {
6285         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6286
6287         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6288                 return 0;
6289
6290         if (list && len <= list_len)
6291                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6292         return len;
6293 }
6294
6295 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6296 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6297 {
6298         return server->caps & NFS_CAP_SECURITY_LABEL;
6299 }
6300
6301 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6302                                    const void *buf, size_t buflen,
6303                                    int flags, int type)
6304 {
6305         if (security_ismaclabel(key))
6306                 return nfs4_set_security_label(dentry, buf, buflen);
6307
6308         return -EOPNOTSUPP;
6309 }
6310
6311 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6312                                    void *buf, size_t buflen, int type)
6313 {
6314         if (security_ismaclabel(key))
6315                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6316         return -EOPNOTSUPP;
6317 }
6318
6319 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6320                                        size_t list_len, const char *name,
6321                                        size_t name_len, int type)
6322 {
6323         size_t len = 0;
6324
6325         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6326                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6327                 if (list && len <= list_len)
6328                         security_inode_listsecurity(d_inode(dentry), list, len);
6329         }
6330         return len;
6331 }
6332
6333 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6334         .prefix = XATTR_SECURITY_PREFIX,
6335         .list   = nfs4_xattr_list_nfs4_label,
6336         .get    = nfs4_xattr_get_nfs4_label,
6337         .set    = nfs4_xattr_set_nfs4_label,
6338 };
6339 #endif
6340
6341
6342 /*
6343  * nfs_fhget will use either the mounted_on_fileid or the fileid
6344  */
6345 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6346 {
6347         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6348                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6349               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6350               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6351                 return;
6352
6353         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6354                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6355         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6356         fattr->nlink = 2;
6357 }
6358
6359 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6360                                    const struct qstr *name,
6361                                    struct nfs4_fs_locations *fs_locations,
6362                                    struct page *page)
6363 {
6364         struct nfs_server *server = NFS_SERVER(dir);
6365         u32 bitmask[3] = {
6366                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6367         };
6368         struct nfs4_fs_locations_arg args = {
6369                 .dir_fh = NFS_FH(dir),
6370                 .name = name,
6371                 .page = page,
6372                 .bitmask = bitmask,
6373         };
6374         struct nfs4_fs_locations_res res = {
6375                 .fs_locations = fs_locations,
6376         };
6377         struct rpc_message msg = {
6378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6379                 .rpc_argp = &args,
6380                 .rpc_resp = &res,
6381         };
6382         int status;
6383
6384         dprintk("%s: start\n", __func__);
6385
6386         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6387          * is not supported */
6388         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6389                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6390         else
6391                 bitmask[0] |= FATTR4_WORD0_FILEID;
6392
6393         nfs_fattr_init(&fs_locations->fattr);
6394         fs_locations->server = server;
6395         fs_locations->nlocations = 0;
6396         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6397         dprintk("%s: returned status = %d\n", __func__, status);
6398         return status;
6399 }
6400
6401 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6402                            const struct qstr *name,
6403                            struct nfs4_fs_locations *fs_locations,
6404                            struct page *page)
6405 {
6406         struct nfs4_exception exception = { };
6407         int err;
6408         do {
6409                 err = _nfs4_proc_fs_locations(client, dir, name,
6410                                 fs_locations, page);
6411                 trace_nfs4_get_fs_locations(dir, name, err);
6412                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6413                                 &exception);
6414         } while (exception.retry);
6415         return err;
6416 }
6417
6418 /*
6419  * This operation also signals the server that this client is
6420  * performing migration recovery.  The server can stop returning
6421  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6422  * appended to this compound to identify the client ID which is
6423  * performing recovery.
6424  */
6425 static int _nfs40_proc_get_locations(struct inode *inode,
6426                                      struct nfs4_fs_locations *locations,
6427                                      struct page *page, struct rpc_cred *cred)
6428 {
6429         struct nfs_server *server = NFS_SERVER(inode);
6430         struct rpc_clnt *clnt = server->client;
6431         u32 bitmask[2] = {
6432                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6433         };
6434         struct nfs4_fs_locations_arg args = {
6435                 .clientid       = server->nfs_client->cl_clientid,
6436                 .fh             = NFS_FH(inode),
6437                 .page           = page,
6438                 .bitmask        = bitmask,
6439                 .migration      = 1,            /* skip LOOKUP */
6440                 .renew          = 1,            /* append RENEW */
6441         };
6442         struct nfs4_fs_locations_res res = {
6443                 .fs_locations   = locations,
6444                 .migration      = 1,
6445                 .renew          = 1,
6446         };
6447         struct rpc_message msg = {
6448                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6449                 .rpc_argp       = &args,
6450                 .rpc_resp       = &res,
6451                 .rpc_cred       = cred,
6452         };
6453         unsigned long now = jiffies;
6454         int status;
6455
6456         nfs_fattr_init(&locations->fattr);
6457         locations->server = server;
6458         locations->nlocations = 0;
6459
6460         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6461         nfs4_set_sequence_privileged(&args.seq_args);
6462         status = nfs4_call_sync_sequence(clnt, server, &msg,
6463                                         &args.seq_args, &res.seq_res);
6464         if (status)
6465                 return status;
6466
6467         renew_lease(server, now);
6468         return 0;
6469 }
6470
6471 #ifdef CONFIG_NFS_V4_1
6472
6473 /*
6474  * This operation also signals the server that this client is
6475  * performing migration recovery.  The server can stop asserting
6476  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6477  * performing this operation is identified in the SEQUENCE
6478  * operation in this compound.
6479  *
6480  * When the client supports GETATTR(fs_locations_info), it can
6481  * be plumbed in here.
6482  */
6483 static int _nfs41_proc_get_locations(struct inode *inode,
6484                                      struct nfs4_fs_locations *locations,
6485                                      struct page *page, struct rpc_cred *cred)
6486 {
6487         struct nfs_server *server = NFS_SERVER(inode);
6488         struct rpc_clnt *clnt = server->client;
6489         u32 bitmask[2] = {
6490                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6491         };
6492         struct nfs4_fs_locations_arg args = {
6493                 .fh             = NFS_FH(inode),
6494                 .page           = page,
6495                 .bitmask        = bitmask,
6496                 .migration      = 1,            /* skip LOOKUP */
6497         };
6498         struct nfs4_fs_locations_res res = {
6499                 .fs_locations   = locations,
6500                 .migration      = 1,
6501         };
6502         struct rpc_message msg = {
6503                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6504                 .rpc_argp       = &args,
6505                 .rpc_resp       = &res,
6506                 .rpc_cred       = cred,
6507         };
6508         int status;
6509
6510         nfs_fattr_init(&locations->fattr);
6511         locations->server = server;
6512         locations->nlocations = 0;
6513
6514         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6515         nfs4_set_sequence_privileged(&args.seq_args);
6516         status = nfs4_call_sync_sequence(clnt, server, &msg,
6517                                         &args.seq_args, &res.seq_res);
6518         if (status == NFS4_OK &&
6519             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6520                 status = -NFS4ERR_LEASE_MOVED;
6521         return status;
6522 }
6523
6524 #endif  /* CONFIG_NFS_V4_1 */
6525
6526 /**
6527  * nfs4_proc_get_locations - discover locations for a migrated FSID
6528  * @inode: inode on FSID that is migrating
6529  * @locations: result of query
6530  * @page: buffer
6531  * @cred: credential to use for this operation
6532  *
6533  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6534  * operation failed, or a negative errno if a local error occurred.
6535  *
6536  * On success, "locations" is filled in, but if the server has
6537  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6538  * asserted.
6539  *
6540  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6541  * from this client that require migration recovery.
6542  */
6543 int nfs4_proc_get_locations(struct inode *inode,
6544                             struct nfs4_fs_locations *locations,
6545                             struct page *page, struct rpc_cred *cred)
6546 {
6547         struct nfs_server *server = NFS_SERVER(inode);
6548         struct nfs_client *clp = server->nfs_client;
6549         const struct nfs4_mig_recovery_ops *ops =
6550                                         clp->cl_mvops->mig_recovery_ops;
6551         struct nfs4_exception exception = { };
6552         int status;
6553
6554         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6555                 (unsigned long long)server->fsid.major,
6556                 (unsigned long long)server->fsid.minor,
6557                 clp->cl_hostname);
6558         nfs_display_fhandle(NFS_FH(inode), __func__);
6559
6560         do {
6561                 status = ops->get_locations(inode, locations, page, cred);
6562                 if (status != -NFS4ERR_DELAY)
6563                         break;
6564                 nfs4_handle_exception(server, status, &exception);
6565         } while (exception.retry);
6566         return status;
6567 }
6568
6569 /*
6570  * This operation also signals the server that this client is
6571  * performing "lease moved" recovery.  The server can stop
6572  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6573  * is appended to this compound to identify the client ID which is
6574  * performing recovery.
6575  */
6576 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6577 {
6578         struct nfs_server *server = NFS_SERVER(inode);
6579         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6580         struct rpc_clnt *clnt = server->client;
6581         struct nfs4_fsid_present_arg args = {
6582                 .fh             = NFS_FH(inode),
6583                 .clientid       = clp->cl_clientid,
6584                 .renew          = 1,            /* append RENEW */
6585         };
6586         struct nfs4_fsid_present_res res = {
6587                 .renew          = 1,
6588         };
6589         struct rpc_message msg = {
6590                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6591                 .rpc_argp       = &args,
6592                 .rpc_resp       = &res,
6593                 .rpc_cred       = cred,
6594         };
6595         unsigned long now = jiffies;
6596         int status;
6597
6598         res.fh = nfs_alloc_fhandle();
6599         if (res.fh == NULL)
6600                 return -ENOMEM;
6601
6602         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6603         nfs4_set_sequence_privileged(&args.seq_args);
6604         status = nfs4_call_sync_sequence(clnt, server, &msg,
6605                                                 &args.seq_args, &res.seq_res);
6606         nfs_free_fhandle(res.fh);
6607         if (status)
6608                 return status;
6609
6610         do_renew_lease(clp, now);
6611         return 0;
6612 }
6613
6614 #ifdef CONFIG_NFS_V4_1
6615
6616 /*
6617  * This operation also signals the server that this client is
6618  * performing "lease moved" recovery.  The server can stop asserting
6619  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6620  * this operation is identified in the SEQUENCE operation in this
6621  * compound.
6622  */
6623 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6624 {
6625         struct nfs_server *server = NFS_SERVER(inode);
6626         struct rpc_clnt *clnt = server->client;
6627         struct nfs4_fsid_present_arg args = {
6628                 .fh             = NFS_FH(inode),
6629         };
6630         struct nfs4_fsid_present_res res = {
6631         };
6632         struct rpc_message msg = {
6633                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6634                 .rpc_argp       = &args,
6635                 .rpc_resp       = &res,
6636                 .rpc_cred       = cred,
6637         };
6638         int status;
6639
6640         res.fh = nfs_alloc_fhandle();
6641         if (res.fh == NULL)
6642                 return -ENOMEM;
6643
6644         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6645         nfs4_set_sequence_privileged(&args.seq_args);
6646         status = nfs4_call_sync_sequence(clnt, server, &msg,
6647                                                 &args.seq_args, &res.seq_res);
6648         nfs_free_fhandle(res.fh);
6649         if (status == NFS4_OK &&
6650             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6651                 status = -NFS4ERR_LEASE_MOVED;
6652         return status;
6653 }
6654
6655 #endif  /* CONFIG_NFS_V4_1 */
6656
6657 /**
6658  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6659  * @inode: inode on FSID to check
6660  * @cred: credential to use for this operation
6661  *
6662  * Server indicates whether the FSID is present, moved, or not
6663  * recognized.  This operation is necessary to clear a LEASE_MOVED
6664  * condition for this client ID.
6665  *
6666  * Returns NFS4_OK if the FSID is present on this server,
6667  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6668  *  NFS4ERR code if some error occurred on the server, or a
6669  *  negative errno if a local failure occurred.
6670  */
6671 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6672 {
6673         struct nfs_server *server = NFS_SERVER(inode);
6674         struct nfs_client *clp = server->nfs_client;
6675         const struct nfs4_mig_recovery_ops *ops =
6676                                         clp->cl_mvops->mig_recovery_ops;
6677         struct nfs4_exception exception = { };
6678         int status;
6679
6680         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6681                 (unsigned long long)server->fsid.major,
6682                 (unsigned long long)server->fsid.minor,
6683                 clp->cl_hostname);
6684         nfs_display_fhandle(NFS_FH(inode), __func__);
6685
6686         do {
6687                 status = ops->fsid_present(inode, cred);
6688                 if (status != -NFS4ERR_DELAY)
6689                         break;
6690                 nfs4_handle_exception(server, status, &exception);
6691         } while (exception.retry);
6692         return status;
6693 }
6694
6695 /**
6696  * If 'use_integrity' is true and the state managment nfs_client
6697  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6698  * and the machine credential as per RFC3530bis and RFC5661 Security
6699  * Considerations sections. Otherwise, just use the user cred with the
6700  * filesystem's rpc_client.
6701  */
6702 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6703 {
6704         int status;
6705         struct nfs4_secinfo_arg args = {
6706                 .dir_fh = NFS_FH(dir),
6707                 .name   = name,
6708         };
6709         struct nfs4_secinfo_res res = {
6710                 .flavors     = flavors,
6711         };
6712         struct rpc_message msg = {
6713                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6714                 .rpc_argp = &args,
6715                 .rpc_resp = &res,
6716         };
6717         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6718         struct rpc_cred *cred = NULL;
6719
6720         if (use_integrity) {
6721                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6722                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6723                 msg.rpc_cred = cred;
6724         }
6725
6726         dprintk("NFS call  secinfo %s\n", name->name);
6727
6728         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6729                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6730
6731         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6732                                 &res.seq_res, 0);
6733         dprintk("NFS reply  secinfo: %d\n", status);
6734
6735         if (cred)
6736                 put_rpccred(cred);
6737
6738         return status;
6739 }
6740
6741 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6742                       struct nfs4_secinfo_flavors *flavors)
6743 {
6744         struct nfs4_exception exception = { };
6745         int err;
6746         do {
6747                 err = -NFS4ERR_WRONGSEC;
6748
6749                 /* try to use integrity protection with machine cred */
6750                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6751                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6752
6753                 /*
6754                  * if unable to use integrity protection, or SECINFO with
6755                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6756                  * disallowed by spec, but exists in deployed servers) use
6757                  * the current filesystem's rpc_client and the user cred.
6758                  */
6759                 if (err == -NFS4ERR_WRONGSEC)
6760                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6761
6762                 trace_nfs4_secinfo(dir, name, err);
6763                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6764                                 &exception);
6765         } while (exception.retry);
6766         return err;
6767 }
6768
6769 #ifdef CONFIG_NFS_V4_1
6770 /*
6771  * Check the exchange flags returned by the server for invalid flags, having
6772  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6773  * DS flags set.
6774  */
6775 static int nfs4_check_cl_exchange_flags(u32 flags)
6776 {
6777         if (flags & ~EXCHGID4_FLAG_MASK_R)
6778                 goto out_inval;
6779         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6780             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6781                 goto out_inval;
6782         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6783                 goto out_inval;
6784         return NFS_OK;
6785 out_inval:
6786         return -NFS4ERR_INVAL;
6787 }
6788
6789 static bool
6790 nfs41_same_server_scope(struct nfs41_server_scope *a,
6791                         struct nfs41_server_scope *b)
6792 {
6793         if (a->server_scope_sz == b->server_scope_sz &&
6794             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6795                 return true;
6796
6797         return false;
6798 }
6799
6800 /*
6801  * nfs4_proc_bind_conn_to_session()
6802  *
6803  * The 4.1 client currently uses the same TCP connection for the
6804  * fore and backchannel.
6805  */
6806 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6807 {
6808         int status;
6809         struct nfs41_bind_conn_to_session_args args = {
6810                 .client = clp,
6811                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6812         };
6813         struct nfs41_bind_conn_to_session_res res;
6814         struct rpc_message msg = {
6815                 .rpc_proc =
6816                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6817                 .rpc_argp = &args,
6818                 .rpc_resp = &res,
6819                 .rpc_cred = cred,
6820         };
6821
6822         dprintk("--> %s\n", __func__);
6823
6824         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6825         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6826                 args.dir = NFS4_CDFC4_FORE;
6827
6828         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6829         trace_nfs4_bind_conn_to_session(clp, status);
6830         if (status == 0) {
6831                 if (memcmp(res.sessionid.data,
6832                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6833                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6834                         status = -EIO;
6835                         goto out;
6836                 }
6837                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6838                         dprintk("NFS: %s: Unexpected direction from server\n",
6839                                 __func__);
6840                         status = -EIO;
6841                         goto out;
6842                 }
6843                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6844                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6845                                 __func__);
6846                         status = -EIO;
6847                         goto out;
6848                 }
6849         }
6850 out:
6851         dprintk("<-- %s status= %d\n", __func__, status);
6852         return status;
6853 }
6854
6855 /*
6856  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6857  * and operations we'd like to see to enable certain features in the allow map
6858  */
6859 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6860         .how = SP4_MACH_CRED,
6861         .enforce.u.words = {
6862                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6863                       1 << (OP_EXCHANGE_ID - 32) |
6864                       1 << (OP_CREATE_SESSION - 32) |
6865                       1 << (OP_DESTROY_SESSION - 32) |
6866                       1 << (OP_DESTROY_CLIENTID - 32)
6867         },
6868         .allow.u.words = {
6869                 [0] = 1 << (OP_CLOSE) |
6870                       1 << (OP_LOCKU) |
6871                       1 << (OP_COMMIT),
6872                 [1] = 1 << (OP_SECINFO - 32) |
6873                       1 << (OP_SECINFO_NO_NAME - 32) |
6874                       1 << (OP_TEST_STATEID - 32) |
6875                       1 << (OP_FREE_STATEID - 32) |
6876                       1 << (OP_WRITE - 32)
6877         }
6878 };
6879
6880 /*
6881  * Select the state protection mode for client `clp' given the server results
6882  * from exchange_id in `sp'.
6883  *
6884  * Returns 0 on success, negative errno otherwise.
6885  */
6886 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6887                                  struct nfs41_state_protection *sp)
6888 {
6889         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6890                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6891                       1 << (OP_EXCHANGE_ID - 32) |
6892                       1 << (OP_CREATE_SESSION - 32) |
6893                       1 << (OP_DESTROY_SESSION - 32) |
6894                       1 << (OP_DESTROY_CLIENTID - 32)
6895         };
6896         unsigned int i;
6897
6898         if (sp->how == SP4_MACH_CRED) {
6899                 /* Print state protect result */
6900                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6901                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6902                         if (test_bit(i, sp->enforce.u.longs))
6903                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6904                         if (test_bit(i, sp->allow.u.longs))
6905                                 dfprintk(MOUNT, "  allow op %d\n", i);
6906                 }
6907
6908                 /* make sure nothing is on enforce list that isn't supported */
6909                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6910                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6911                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6912                                 return -EINVAL;
6913                         }
6914                 }
6915
6916                 /*
6917                  * Minimal mode - state operations are allowed to use machine
6918                  * credential.  Note this already happens by default, so the
6919                  * client doesn't have to do anything more than the negotiation.
6920                  *
6921                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6922                  *       we're already using the machine cred for exchange_id
6923                  *       and will never use a different cred.
6924                  */
6925                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6926                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6927                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6928                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6929                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6930                         dfprintk(MOUNT, "  minimal mode enabled\n");
6931                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6932                 } else {
6933                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6934                         return -EINVAL;
6935                 }
6936
6937                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6938                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6939                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6940                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6941                 }
6942
6943                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6944                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6945                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6946                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6947                 }
6948
6949                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6950                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6951                         dfprintk(MOUNT, "  stateid mode enabled\n");
6952                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6953                 }
6954
6955                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6956                         dfprintk(MOUNT, "  write mode enabled\n");
6957                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6958                 }
6959
6960                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6961                         dfprintk(MOUNT, "  commit mode enabled\n");
6962                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6963                 }
6964         }
6965
6966         return 0;
6967 }
6968
6969 /*
6970  * _nfs4_proc_exchange_id()
6971  *
6972  * Wrapper for EXCHANGE_ID operation.
6973  */
6974 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6975         u32 sp4_how)
6976 {
6977         nfs4_verifier verifier;
6978         struct nfs41_exchange_id_args args = {
6979                 .verifier = &verifier,
6980                 .client = clp,
6981 #ifdef CONFIG_NFS_V4_1_MIGRATION
6982                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6983                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6984                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6985 #else
6986                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6987                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6988 #endif
6989         };
6990         struct nfs41_exchange_id_res res = {
6991                 0
6992         };
6993         int status;
6994         struct rpc_message msg = {
6995                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6996                 .rpc_argp = &args,
6997                 .rpc_resp = &res,
6998                 .rpc_cred = cred,
6999         };
7000
7001         nfs4_init_boot_verifier(clp, &verifier);
7002
7003         status = nfs4_init_uniform_client_string(clp);
7004         if (status)
7005                 goto out;
7006
7007         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7008                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7009                 clp->cl_owner_id);
7010
7011         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7012                                         GFP_NOFS);
7013         if (unlikely(res.server_owner == NULL)) {
7014                 status = -ENOMEM;
7015                 goto out;
7016         }
7017
7018         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7019                                         GFP_NOFS);
7020         if (unlikely(res.server_scope == NULL)) {
7021                 status = -ENOMEM;
7022                 goto out_server_owner;
7023         }
7024
7025         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7026         if (unlikely(res.impl_id == NULL)) {
7027                 status = -ENOMEM;
7028                 goto out_server_scope;
7029         }
7030
7031         switch (sp4_how) {
7032         case SP4_NONE:
7033                 args.state_protect.how = SP4_NONE;
7034                 break;
7035
7036         case SP4_MACH_CRED:
7037                 args.state_protect = nfs4_sp4_mach_cred_request;
7038                 break;
7039
7040         default:
7041                 /* unsupported! */
7042                 WARN_ON_ONCE(1);
7043                 status = -EINVAL;
7044                 goto out_impl_id;
7045         }
7046
7047         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7048         trace_nfs4_exchange_id(clp, status);
7049         if (status == 0)
7050                 status = nfs4_check_cl_exchange_flags(res.flags);
7051
7052         if (status == 0)
7053                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7054
7055         if (status == 0) {
7056                 clp->cl_clientid = res.clientid;
7057                 clp->cl_exchange_flags = res.flags;
7058                 /* Client ID is not confirmed */
7059                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7060                         clear_bit(NFS4_SESSION_ESTABLISHED,
7061                                         &clp->cl_session->session_state);
7062                         clp->cl_seqid = res.seqid;
7063                 }
7064
7065                 kfree(clp->cl_serverowner);
7066                 clp->cl_serverowner = res.server_owner;
7067                 res.server_owner = NULL;
7068
7069                 /* use the most recent implementation id */
7070                 kfree(clp->cl_implid);
7071                 clp->cl_implid = res.impl_id;
7072                 res.impl_id = NULL;
7073
7074                 if (clp->cl_serverscope != NULL &&
7075                     !nfs41_same_server_scope(clp->cl_serverscope,
7076                                              res.server_scope)) {
7077                         dprintk("%s: server_scope mismatch detected\n",
7078                                 __func__);
7079                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7080                         kfree(clp->cl_serverscope);
7081                         clp->cl_serverscope = NULL;
7082                 }
7083
7084                 if (clp->cl_serverscope == NULL) {
7085                         clp->cl_serverscope = res.server_scope;
7086                         res.server_scope = NULL;
7087                 }
7088         }
7089
7090 out_impl_id:
7091         kfree(res.impl_id);
7092 out_server_scope:
7093         kfree(res.server_scope);
7094 out_server_owner:
7095         kfree(res.server_owner);
7096 out:
7097         if (clp->cl_implid != NULL)
7098                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7099                         "domain: %s, name: %s, date: %llu,%u\n",
7100                         clp->cl_implid->domain, clp->cl_implid->name,
7101                         clp->cl_implid->date.seconds,
7102                         clp->cl_implid->date.nseconds);
7103         dprintk("NFS reply exchange_id: %d\n", status);
7104         return status;
7105 }
7106
7107 /*
7108  * nfs4_proc_exchange_id()
7109  *
7110  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7111  *
7112  * Since the clientid has expired, all compounds using sessions
7113  * associated with the stale clientid will be returning
7114  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7115  * be in some phase of session reset.
7116  *
7117  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7118  */
7119 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7120 {
7121         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7122         int status;
7123
7124         /* try SP4_MACH_CRED if krb5i/p */
7125         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7126             authflavor == RPC_AUTH_GSS_KRB5P) {
7127                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7128                 if (!status)
7129                         return 0;
7130         }
7131
7132         /* try SP4_NONE */
7133         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7134 }
7135
7136 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7137                 struct rpc_cred *cred)
7138 {
7139         struct rpc_message msg = {
7140                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7141                 .rpc_argp = clp,
7142                 .rpc_cred = cred,
7143         };
7144         int status;
7145
7146         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7147         trace_nfs4_destroy_clientid(clp, status);
7148         if (status)
7149                 dprintk("NFS: Got error %d from the server %s on "
7150                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7151         return status;
7152 }
7153
7154 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7155                 struct rpc_cred *cred)
7156 {
7157         unsigned int loop;
7158         int ret;
7159
7160         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7161                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7162                 switch (ret) {
7163                 case -NFS4ERR_DELAY:
7164                 case -NFS4ERR_CLIENTID_BUSY:
7165                         ssleep(1);
7166                         break;
7167                 default:
7168                         return ret;
7169                 }
7170         }
7171         return 0;
7172 }
7173
7174 int nfs4_destroy_clientid(struct nfs_client *clp)
7175 {
7176         struct rpc_cred *cred;
7177         int ret = 0;
7178
7179         if (clp->cl_mvops->minor_version < 1)
7180                 goto out;
7181         if (clp->cl_exchange_flags == 0)
7182                 goto out;
7183         if (clp->cl_preserve_clid)
7184                 goto out;
7185         cred = nfs4_get_clid_cred(clp);
7186         ret = nfs4_proc_destroy_clientid(clp, cred);
7187         if (cred)
7188                 put_rpccred(cred);
7189         switch (ret) {
7190         case 0:
7191         case -NFS4ERR_STALE_CLIENTID:
7192                 clp->cl_exchange_flags = 0;
7193         }
7194 out:
7195         return ret;
7196 }
7197
7198 struct nfs4_get_lease_time_data {
7199         struct nfs4_get_lease_time_args *args;
7200         struct nfs4_get_lease_time_res *res;
7201         struct nfs_client *clp;
7202 };
7203
7204 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7205                                         void *calldata)
7206 {
7207         struct nfs4_get_lease_time_data *data =
7208                         (struct nfs4_get_lease_time_data *)calldata;
7209
7210         dprintk("--> %s\n", __func__);
7211         /* just setup sequence, do not trigger session recovery
7212            since we're invoked within one */
7213         nfs41_setup_sequence(data->clp->cl_session,
7214                         &data->args->la_seq_args,
7215                         &data->res->lr_seq_res,
7216                         task);
7217         dprintk("<-- %s\n", __func__);
7218 }
7219
7220 /*
7221  * Called from nfs4_state_manager thread for session setup, so don't recover
7222  * from sequence operation or clientid errors.
7223  */
7224 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7225 {
7226         struct nfs4_get_lease_time_data *data =
7227                         (struct nfs4_get_lease_time_data *)calldata;
7228
7229         dprintk("--> %s\n", __func__);
7230         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7231                 return;
7232         switch (task->tk_status) {
7233         case -NFS4ERR_DELAY:
7234         case -NFS4ERR_GRACE:
7235                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7236                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7237                 task->tk_status = 0;
7238                 /* fall through */
7239         case -NFS4ERR_RETRY_UNCACHED_REP:
7240                 rpc_restart_call_prepare(task);
7241                 return;
7242         }
7243         dprintk("<-- %s\n", __func__);
7244 }
7245
7246 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7247         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7248         .rpc_call_done = nfs4_get_lease_time_done,
7249 };
7250
7251 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7252 {
7253         struct rpc_task *task;
7254         struct nfs4_get_lease_time_args args;
7255         struct nfs4_get_lease_time_res res = {
7256                 .lr_fsinfo = fsinfo,
7257         };
7258         struct nfs4_get_lease_time_data data = {
7259                 .args = &args,
7260                 .res = &res,
7261                 .clp = clp,
7262         };
7263         struct rpc_message msg = {
7264                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7265                 .rpc_argp = &args,
7266                 .rpc_resp = &res,
7267         };
7268         struct rpc_task_setup task_setup = {
7269                 .rpc_client = clp->cl_rpcclient,
7270                 .rpc_message = &msg,
7271                 .callback_ops = &nfs4_get_lease_time_ops,
7272                 .callback_data = &data,
7273                 .flags = RPC_TASK_TIMEOUT,
7274         };
7275         int status;
7276
7277         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7278         nfs4_set_sequence_privileged(&args.la_seq_args);
7279         dprintk("--> %s\n", __func__);
7280         task = rpc_run_task(&task_setup);
7281
7282         if (IS_ERR(task))
7283                 status = PTR_ERR(task);
7284         else {
7285                 status = task->tk_status;
7286                 rpc_put_task(task);
7287         }
7288         dprintk("<-- %s return %d\n", __func__, status);
7289
7290         return status;
7291 }
7292
7293 /*
7294  * Initialize the values to be used by the client in CREATE_SESSION
7295  * If nfs4_init_session set the fore channel request and response sizes,
7296  * use them.
7297  *
7298  * Set the back channel max_resp_sz_cached to zero to force the client to
7299  * always set csa_cachethis to FALSE because the current implementation
7300  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7301  */
7302 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7303 {
7304         unsigned int max_rqst_sz, max_resp_sz;
7305
7306         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7307         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7308
7309         /* Fore channel attributes */
7310         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7311         args->fc_attrs.max_resp_sz = max_resp_sz;
7312         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7313         args->fc_attrs.max_reqs = max_session_slots;
7314
7315         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7316                 "max_ops=%u max_reqs=%u\n",
7317                 __func__,
7318                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7319                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7320
7321         /* Back channel attributes */
7322         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7323         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7324         args->bc_attrs.max_resp_sz_cached = 0;
7325         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7326         args->bc_attrs.max_reqs = 1;
7327
7328         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7329                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7330                 __func__,
7331                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7332                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7333                 args->bc_attrs.max_reqs);
7334 }
7335
7336 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7337                 struct nfs41_create_session_res *res)
7338 {
7339         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7340         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7341
7342         if (rcvd->max_resp_sz > sent->max_resp_sz)
7343                 return -EINVAL;
7344         /*
7345          * Our requested max_ops is the minimum we need; we're not
7346          * prepared to break up compounds into smaller pieces than that.
7347          * So, no point even trying to continue if the server won't
7348          * cooperate:
7349          */
7350         if (rcvd->max_ops < sent->max_ops)
7351                 return -EINVAL;
7352         if (rcvd->max_reqs == 0)
7353                 return -EINVAL;
7354         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7355                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7356         return 0;
7357 }
7358
7359 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7360                 struct nfs41_create_session_res *res)
7361 {
7362         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7363         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7364
7365         if (!(res->flags & SESSION4_BACK_CHAN))
7366                 goto out;
7367         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7368                 return -EINVAL;
7369         if (rcvd->max_resp_sz < sent->max_resp_sz)
7370                 return -EINVAL;
7371         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7372                 return -EINVAL;
7373         /* These would render the backchannel useless: */
7374         if (rcvd->max_ops != sent->max_ops)
7375                 return -EINVAL;
7376         if (rcvd->max_reqs != sent->max_reqs)
7377                 return -EINVAL;
7378 out:
7379         return 0;
7380 }
7381
7382 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7383                                      struct nfs41_create_session_res *res)
7384 {
7385         int ret;
7386
7387         ret = nfs4_verify_fore_channel_attrs(args, res);
7388         if (ret)
7389                 return ret;
7390         return nfs4_verify_back_channel_attrs(args, res);
7391 }
7392
7393 static void nfs4_update_session(struct nfs4_session *session,
7394                 struct nfs41_create_session_res *res)
7395 {
7396         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7397         /* Mark client id and session as being confirmed */
7398         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7399         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7400         session->flags = res->flags;
7401         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7402         if (res->flags & SESSION4_BACK_CHAN)
7403                 memcpy(&session->bc_attrs, &res->bc_attrs,
7404                                 sizeof(session->bc_attrs));
7405 }
7406
7407 static int _nfs4_proc_create_session(struct nfs_client *clp,
7408                 struct rpc_cred *cred)
7409 {
7410         struct nfs4_session *session = clp->cl_session;
7411         struct nfs41_create_session_args args = {
7412                 .client = clp,
7413                 .clientid = clp->cl_clientid,
7414                 .seqid = clp->cl_seqid,
7415                 .cb_program = NFS4_CALLBACK,
7416         };
7417         struct nfs41_create_session_res res;
7418
7419         struct rpc_message msg = {
7420                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7421                 .rpc_argp = &args,
7422                 .rpc_resp = &res,
7423                 .rpc_cred = cred,
7424         };
7425         int status;
7426
7427         nfs4_init_channel_attrs(&args);
7428         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7429
7430         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7431         trace_nfs4_create_session(clp, status);
7432
7433         if (!status) {
7434                 /* Verify the session's negotiated channel_attrs values */
7435                 status = nfs4_verify_channel_attrs(&args, &res);
7436                 /* Increment the clientid slot sequence id */
7437                 if (clp->cl_seqid == res.seqid)
7438                         clp->cl_seqid++;
7439                 if (status)
7440                         goto out;
7441                 nfs4_update_session(session, &res);
7442         }
7443 out:
7444         return status;
7445 }
7446
7447 /*
7448  * Issues a CREATE_SESSION operation to the server.
7449  * It is the responsibility of the caller to verify the session is
7450  * expired before calling this routine.
7451  */
7452 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7453 {
7454         int status;
7455         unsigned *ptr;
7456         struct nfs4_session *session = clp->cl_session;
7457
7458         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7459
7460         status = _nfs4_proc_create_session(clp, cred);
7461         if (status)
7462                 goto out;
7463
7464         /* Init or reset the session slot tables */
7465         status = nfs4_setup_session_slot_tables(session);
7466         dprintk("slot table setup returned %d\n", status);
7467         if (status)
7468                 goto out;
7469
7470         ptr = (unsigned *)&session->sess_id.data[0];
7471         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7472                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7473 out:
7474         dprintk("<-- %s\n", __func__);
7475         return status;
7476 }
7477
7478 /*
7479  * Issue the over-the-wire RPC DESTROY_SESSION.
7480  * The caller must serialize access to this routine.
7481  */
7482 int nfs4_proc_destroy_session(struct nfs4_session *session,
7483                 struct rpc_cred *cred)
7484 {
7485         struct rpc_message msg = {
7486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7487                 .rpc_argp = session,
7488                 .rpc_cred = cred,
7489         };
7490         int status = 0;
7491
7492         dprintk("--> nfs4_proc_destroy_session\n");
7493
7494         /* session is still being setup */
7495         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7496                 return 0;
7497
7498         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7499         trace_nfs4_destroy_session(session->clp, status);
7500
7501         if (status)
7502                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7503                         "Session has been destroyed regardless...\n", status);
7504
7505         dprintk("<-- nfs4_proc_destroy_session\n");
7506         return status;
7507 }
7508
7509 /*
7510  * Renew the cl_session lease.
7511  */
7512 struct nfs4_sequence_data {
7513         struct nfs_client *clp;
7514         struct nfs4_sequence_args args;
7515         struct nfs4_sequence_res res;
7516 };
7517
7518 static void nfs41_sequence_release(void *data)
7519 {
7520         struct nfs4_sequence_data *calldata = data;
7521         struct nfs_client *clp = calldata->clp;
7522
7523         if (atomic_read(&clp->cl_count) > 1)
7524                 nfs4_schedule_state_renewal(clp);
7525         nfs_put_client(clp);
7526         kfree(calldata);
7527 }
7528
7529 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7530 {
7531         switch(task->tk_status) {
7532         case -NFS4ERR_DELAY:
7533                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7534                 return -EAGAIN;
7535         default:
7536                 nfs4_schedule_lease_recovery(clp);
7537         }
7538         return 0;
7539 }
7540
7541 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7542 {
7543         struct nfs4_sequence_data *calldata = data;
7544         struct nfs_client *clp = calldata->clp;
7545
7546         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7547                 return;
7548
7549         trace_nfs4_sequence(clp, task->tk_status);
7550         if (task->tk_status < 0) {
7551                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7552                 if (atomic_read(&clp->cl_count) == 1)
7553                         goto out;
7554
7555                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7556                         rpc_restart_call_prepare(task);
7557                         return;
7558                 }
7559         }
7560         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7561 out:
7562         dprintk("<-- %s\n", __func__);
7563 }
7564
7565 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7566 {
7567         struct nfs4_sequence_data *calldata = data;
7568         struct nfs_client *clp = calldata->clp;
7569         struct nfs4_sequence_args *args;
7570         struct nfs4_sequence_res *res;
7571
7572         args = task->tk_msg.rpc_argp;
7573         res = task->tk_msg.rpc_resp;
7574
7575         nfs41_setup_sequence(clp->cl_session, args, res, task);
7576 }
7577
7578 static const struct rpc_call_ops nfs41_sequence_ops = {
7579         .rpc_call_done = nfs41_sequence_call_done,
7580         .rpc_call_prepare = nfs41_sequence_prepare,
7581         .rpc_release = nfs41_sequence_release,
7582 };
7583
7584 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7585                 struct rpc_cred *cred,
7586                 bool is_privileged)
7587 {
7588         struct nfs4_sequence_data *calldata;
7589         struct rpc_message msg = {
7590                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7591                 .rpc_cred = cred,
7592         };
7593         struct rpc_task_setup task_setup_data = {
7594                 .rpc_client = clp->cl_rpcclient,
7595                 .rpc_message = &msg,
7596                 .callback_ops = &nfs41_sequence_ops,
7597                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7598         };
7599
7600         if (!atomic_inc_not_zero(&clp->cl_count))
7601                 return ERR_PTR(-EIO);
7602         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7603         if (calldata == NULL) {
7604                 nfs_put_client(clp);
7605                 return ERR_PTR(-ENOMEM);
7606         }
7607         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7608         if (is_privileged)
7609                 nfs4_set_sequence_privileged(&calldata->args);
7610         msg.rpc_argp = &calldata->args;
7611         msg.rpc_resp = &calldata->res;
7612         calldata->clp = clp;
7613         task_setup_data.callback_data = calldata;
7614
7615         return rpc_run_task(&task_setup_data);
7616 }
7617
7618 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7619 {
7620         struct rpc_task *task;
7621         int ret = 0;
7622
7623         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7624                 return -EAGAIN;
7625         task = _nfs41_proc_sequence(clp, cred, false);
7626         if (IS_ERR(task))
7627                 ret = PTR_ERR(task);
7628         else
7629                 rpc_put_task_async(task);
7630         dprintk("<-- %s status=%d\n", __func__, ret);
7631         return ret;
7632 }
7633
7634 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7635 {
7636         struct rpc_task *task;
7637         int ret;
7638
7639         task = _nfs41_proc_sequence(clp, cred, true);
7640         if (IS_ERR(task)) {
7641                 ret = PTR_ERR(task);
7642                 goto out;
7643         }
7644         ret = rpc_wait_for_completion_task(task);
7645         if (!ret)
7646                 ret = task->tk_status;
7647         rpc_put_task(task);
7648 out:
7649         dprintk("<-- %s status=%d\n", __func__, ret);
7650         return ret;
7651 }
7652
7653 struct nfs4_reclaim_complete_data {
7654         struct nfs_client *clp;
7655         struct nfs41_reclaim_complete_args arg;
7656         struct nfs41_reclaim_complete_res res;
7657 };
7658
7659 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7660 {
7661         struct nfs4_reclaim_complete_data *calldata = data;
7662
7663         nfs41_setup_sequence(calldata->clp->cl_session,
7664                         &calldata->arg.seq_args,
7665                         &calldata->res.seq_res,
7666                         task);
7667 }
7668
7669 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7670 {
7671         switch(task->tk_status) {
7672         case 0:
7673         case -NFS4ERR_COMPLETE_ALREADY:
7674         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7675                 break;
7676         case -NFS4ERR_DELAY:
7677                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7678                 /* fall through */
7679         case -NFS4ERR_RETRY_UNCACHED_REP:
7680                 return -EAGAIN;
7681         default:
7682                 nfs4_schedule_lease_recovery(clp);
7683         }
7684         return 0;
7685 }
7686
7687 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7688 {
7689         struct nfs4_reclaim_complete_data *calldata = data;
7690         struct nfs_client *clp = calldata->clp;
7691         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7692
7693         dprintk("--> %s\n", __func__);
7694         if (!nfs41_sequence_done(task, res))
7695                 return;
7696
7697         trace_nfs4_reclaim_complete(clp, task->tk_status);
7698         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7699                 rpc_restart_call_prepare(task);
7700                 return;
7701         }
7702         dprintk("<-- %s\n", __func__);
7703 }
7704
7705 static void nfs4_free_reclaim_complete_data(void *data)
7706 {
7707         struct nfs4_reclaim_complete_data *calldata = data;
7708
7709         kfree(calldata);
7710 }
7711
7712 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7713         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7714         .rpc_call_done = nfs4_reclaim_complete_done,
7715         .rpc_release = nfs4_free_reclaim_complete_data,
7716 };
7717
7718 /*
7719  * Issue a global reclaim complete.
7720  */
7721 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7722                 struct rpc_cred *cred)
7723 {
7724         struct nfs4_reclaim_complete_data *calldata;
7725         struct rpc_task *task;
7726         struct rpc_message msg = {
7727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7728                 .rpc_cred = cred,
7729         };
7730         struct rpc_task_setup task_setup_data = {
7731                 .rpc_client = clp->cl_rpcclient,
7732                 .rpc_message = &msg,
7733                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7734                 .flags = RPC_TASK_ASYNC,
7735         };
7736         int status = -ENOMEM;
7737
7738         dprintk("--> %s\n", __func__);
7739         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7740         if (calldata == NULL)
7741                 goto out;
7742         calldata->clp = clp;
7743         calldata->arg.one_fs = 0;
7744
7745         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7746         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7747         msg.rpc_argp = &calldata->arg;
7748         msg.rpc_resp = &calldata->res;
7749         task_setup_data.callback_data = calldata;
7750         task = rpc_run_task(&task_setup_data);
7751         if (IS_ERR(task)) {
7752                 status = PTR_ERR(task);
7753                 goto out;
7754         }
7755         status = nfs4_wait_for_completion_rpc_task(task);
7756         if (status == 0)
7757                 status = task->tk_status;
7758         rpc_put_task(task);
7759         return 0;
7760 out:
7761         dprintk("<-- %s status=%d\n", __func__, status);
7762         return status;
7763 }
7764
7765 static void
7766 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7767 {
7768         struct nfs4_layoutget *lgp = calldata;
7769         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7770         struct nfs4_session *session = nfs4_get_session(server);
7771
7772         dprintk("--> %s\n", __func__);
7773         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7774          * right now covering the LAYOUTGET we are about to send.
7775          * However, that is not so catastrophic, and there seems
7776          * to be no way to prevent it completely.
7777          */
7778         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7779                                 &lgp->res.seq_res, task))
7780                 return;
7781         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7782                                           NFS_I(lgp->args.inode)->layout,
7783                                           &lgp->args.range,
7784                                           lgp->args.ctx->state)) {
7785                 rpc_exit(task, NFS4_OK);
7786         }
7787 }
7788
7789 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7790 {
7791         struct nfs4_layoutget *lgp = calldata;
7792         struct inode *inode = lgp->args.inode;
7793         struct nfs_server *server = NFS_SERVER(inode);
7794         struct pnfs_layout_hdr *lo;
7795         struct nfs4_state *state = NULL;
7796         unsigned long timeo, now, giveup;
7797
7798         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7799
7800         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7801                 goto out;
7802
7803         switch (task->tk_status) {
7804         case 0:
7805                 goto out;
7806         /*
7807          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7808          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7809          */
7810         case -NFS4ERR_BADLAYOUT:
7811                 goto out_overflow;
7812         /*
7813          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7814          * (or clients) writing to the same RAID stripe except when
7815          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7816          */
7817         case -NFS4ERR_LAYOUTTRYLATER:
7818                 if (lgp->args.minlength == 0)
7819                         goto out_overflow;
7820         /*
7821          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7822          * existing layout before getting a new one).
7823          */
7824         case -NFS4ERR_RECALLCONFLICT:
7825                 timeo = rpc_get_timeout(task->tk_client);
7826                 giveup = lgp->args.timestamp + timeo;
7827                 now = jiffies;
7828                 if (time_after(giveup, now)) {
7829                         unsigned long delay;
7830
7831                         /* Delay for:
7832                          * - Not less then NFS4_POLL_RETRY_MIN.
7833                          * - One last time a jiffie before we give up
7834                          * - exponential backoff (time_now minus start_attempt)
7835                          */
7836                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7837                                     min((giveup - now - 1),
7838                                         now - lgp->args.timestamp));
7839
7840                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7841                                 __func__, delay);
7842                         rpc_delay(task, delay);
7843                         /* Do not call nfs4_async_handle_error() */
7844                         goto out_restart;
7845                 }
7846                 break;
7847         case -NFS4ERR_EXPIRED:
7848         case -NFS4ERR_BAD_STATEID:
7849                 spin_lock(&inode->i_lock);
7850                 if (nfs4_stateid_match(&lgp->args.stateid,
7851                                         &lgp->args.ctx->state->stateid)) {
7852                         spin_unlock(&inode->i_lock);
7853                         /* If the open stateid was bad, then recover it. */
7854                         state = lgp->args.ctx->state;
7855                         break;
7856                 }
7857                 lo = NFS_I(inode)->layout;
7858                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7859                                         &lo->plh_stateid)) {
7860                         LIST_HEAD(head);
7861
7862                         /*
7863                          * Mark the bad layout state as invalid, then retry
7864                          * with the current stateid.
7865                          */
7866                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7867                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7868                         spin_unlock(&inode->i_lock);
7869                         pnfs_free_lseg_list(&head);
7870                 } else
7871                         spin_unlock(&inode->i_lock);
7872                 goto out_restart;
7873         }
7874         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7875                 goto out_restart;
7876 out:
7877         dprintk("<-- %s\n", __func__);
7878         return;
7879 out_restart:
7880         task->tk_status = 0;
7881         rpc_restart_call_prepare(task);
7882         return;
7883 out_overflow:
7884         task->tk_status = -EOVERFLOW;
7885         goto out;
7886 }
7887
7888 static size_t max_response_pages(struct nfs_server *server)
7889 {
7890         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7891         return nfs_page_array_len(0, max_resp_sz);
7892 }
7893
7894 static void nfs4_free_pages(struct page **pages, size_t size)
7895 {
7896         int i;
7897
7898         if (!pages)
7899                 return;
7900
7901         for (i = 0; i < size; i++) {
7902                 if (!pages[i])
7903                         break;
7904                 __free_page(pages[i]);
7905         }
7906         kfree(pages);
7907 }
7908
7909 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7910 {
7911         struct page **pages;
7912         int i;
7913
7914         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7915         if (!pages) {
7916                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7917                 return NULL;
7918         }
7919
7920         for (i = 0; i < size; i++) {
7921                 pages[i] = alloc_page(gfp_flags);
7922                 if (!pages[i]) {
7923                         dprintk("%s: failed to allocate page\n", __func__);
7924                         nfs4_free_pages(pages, size);
7925                         return NULL;
7926                 }
7927         }
7928
7929         return pages;
7930 }
7931
7932 static void nfs4_layoutget_release(void *calldata)
7933 {
7934         struct nfs4_layoutget *lgp = calldata;
7935         struct inode *inode = lgp->args.inode;
7936         struct nfs_server *server = NFS_SERVER(inode);
7937         size_t max_pages = max_response_pages(server);
7938
7939         dprintk("--> %s\n", __func__);
7940         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7941         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7942         put_nfs_open_context(lgp->args.ctx);
7943         kfree(calldata);
7944         dprintk("<-- %s\n", __func__);
7945 }
7946
7947 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7948         .rpc_call_prepare = nfs4_layoutget_prepare,
7949         .rpc_call_done = nfs4_layoutget_done,
7950         .rpc_release = nfs4_layoutget_release,
7951 };
7952
7953 struct pnfs_layout_segment *
7954 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7955 {
7956         struct inode *inode = lgp->args.inode;
7957         struct nfs_server *server = NFS_SERVER(inode);
7958         size_t max_pages = max_response_pages(server);
7959         struct rpc_task *task;
7960         struct rpc_message msg = {
7961                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7962                 .rpc_argp = &lgp->args,
7963                 .rpc_resp = &lgp->res,
7964                 .rpc_cred = lgp->cred,
7965         };
7966         struct rpc_task_setup task_setup_data = {
7967                 .rpc_client = server->client,
7968                 .rpc_message = &msg,
7969                 .callback_ops = &nfs4_layoutget_call_ops,
7970                 .callback_data = lgp,
7971                 .flags = RPC_TASK_ASYNC,
7972         };
7973         struct pnfs_layout_segment *lseg = NULL;
7974         int status = 0;
7975
7976         dprintk("--> %s\n", __func__);
7977
7978         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7979         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7980
7981         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7982         if (!lgp->args.layout.pages) {
7983                 nfs4_layoutget_release(lgp);
7984                 return ERR_PTR(-ENOMEM);
7985         }
7986         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7987         lgp->args.timestamp = jiffies;
7988
7989         lgp->res.layoutp = &lgp->args.layout;
7990         lgp->res.seq_res.sr_slot = NULL;
7991         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7992
7993         task = rpc_run_task(&task_setup_data);
7994         if (IS_ERR(task))
7995                 return ERR_CAST(task);
7996         status = nfs4_wait_for_completion_rpc_task(task);
7997         if (status == 0)
7998                 status = task->tk_status;
7999         trace_nfs4_layoutget(lgp->args.ctx,
8000                         &lgp->args.range,
8001                         &lgp->res.range,
8002                         status);
8003         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8004         if (status == 0 && lgp->res.layoutp->len)
8005                 lseg = pnfs_layout_process(lgp);
8006         rpc_put_task(task);
8007         dprintk("<-- %s status=%d\n", __func__, status);
8008         if (status)
8009                 return ERR_PTR(status);
8010         return lseg;
8011 }
8012
8013 static void
8014 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8015 {
8016         struct nfs4_layoutreturn *lrp = calldata;
8017
8018         dprintk("--> %s\n", __func__);
8019         nfs41_setup_sequence(lrp->clp->cl_session,
8020                         &lrp->args.seq_args,
8021                         &lrp->res.seq_res,
8022                         task);
8023 }
8024
8025 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8026 {
8027         struct nfs4_layoutreturn *lrp = calldata;
8028         struct nfs_server *server;
8029
8030         dprintk("--> %s\n", __func__);
8031
8032         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8033                 return;
8034
8035         server = NFS_SERVER(lrp->args.inode);
8036         switch (task->tk_status) {
8037         default:
8038                 task->tk_status = 0;
8039         case 0:
8040                 break;
8041         case -NFS4ERR_DELAY:
8042                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8043                         break;
8044                 rpc_restart_call_prepare(task);
8045                 return;
8046         }
8047         dprintk("<-- %s\n", __func__);
8048 }
8049
8050 static void nfs4_layoutreturn_release(void *calldata)
8051 {
8052         struct nfs4_layoutreturn *lrp = calldata;
8053         struct pnfs_layout_hdr *lo = lrp->args.layout;
8054         LIST_HEAD(freeme);
8055
8056         dprintk("--> %s\n", __func__);
8057         spin_lock(&lo->plh_inode->i_lock);
8058         if (lrp->res.lrs_present)
8059                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8060         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8061         pnfs_clear_layoutreturn_waitbit(lo);
8062         lo->plh_block_lgets--;
8063         spin_unlock(&lo->plh_inode->i_lock);
8064         pnfs_free_lseg_list(&freeme);
8065         pnfs_put_layout_hdr(lrp->args.layout);
8066         nfs_iput_and_deactive(lrp->inode);
8067         kfree(calldata);
8068         dprintk("<-- %s\n", __func__);
8069 }
8070
8071 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8072         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8073         .rpc_call_done = nfs4_layoutreturn_done,
8074         .rpc_release = nfs4_layoutreturn_release,
8075 };
8076
8077 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8078 {
8079         struct rpc_task *task;
8080         struct rpc_message msg = {
8081                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8082                 .rpc_argp = &lrp->args,
8083                 .rpc_resp = &lrp->res,
8084                 .rpc_cred = lrp->cred,
8085         };
8086         struct rpc_task_setup task_setup_data = {
8087                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8088                 .rpc_message = &msg,
8089                 .callback_ops = &nfs4_layoutreturn_call_ops,
8090                 .callback_data = lrp,
8091         };
8092         int status = 0;
8093
8094         dprintk("--> %s\n", __func__);
8095         if (!sync) {
8096                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8097                 if (!lrp->inode) {
8098                         nfs4_layoutreturn_release(lrp);
8099                         return -EAGAIN;
8100                 }
8101                 task_setup_data.flags |= RPC_TASK_ASYNC;
8102         }
8103         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8104         task = rpc_run_task(&task_setup_data);
8105         if (IS_ERR(task))
8106                 return PTR_ERR(task);
8107         if (sync)
8108                 status = task->tk_status;
8109         trace_nfs4_layoutreturn(lrp->args.inode, status);
8110         dprintk("<-- %s status=%d\n", __func__, status);
8111         rpc_put_task(task);
8112         return status;
8113 }
8114
8115 static int
8116 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8117                 struct pnfs_device *pdev,
8118                 struct rpc_cred *cred)
8119 {
8120         struct nfs4_getdeviceinfo_args args = {
8121                 .pdev = pdev,
8122                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8123                         NOTIFY_DEVICEID4_DELETE,
8124         };
8125         struct nfs4_getdeviceinfo_res res = {
8126                 .pdev = pdev,
8127         };
8128         struct rpc_message msg = {
8129                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8130                 .rpc_argp = &args,
8131                 .rpc_resp = &res,
8132                 .rpc_cred = cred,
8133         };
8134         int status;
8135
8136         dprintk("--> %s\n", __func__);
8137         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8138         if (res.notification & ~args.notify_types)
8139                 dprintk("%s: unsupported notification\n", __func__);
8140         if (res.notification != args.notify_types)
8141                 pdev->nocache = 1;
8142
8143         dprintk("<-- %s status=%d\n", __func__, status);
8144
8145         return status;
8146 }
8147
8148 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8149                 struct pnfs_device *pdev,
8150                 struct rpc_cred *cred)
8151 {
8152         struct nfs4_exception exception = { };
8153         int err;
8154
8155         do {
8156                 err = nfs4_handle_exception(server,
8157                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8158                                         &exception);
8159         } while (exception.retry);
8160         return err;
8161 }
8162 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8163
8164 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8165 {
8166         struct nfs4_layoutcommit_data *data = calldata;
8167         struct nfs_server *server = NFS_SERVER(data->args.inode);
8168         struct nfs4_session *session = nfs4_get_session(server);
8169
8170         nfs41_setup_sequence(session,
8171                         &data->args.seq_args,
8172                         &data->res.seq_res,
8173                         task);
8174 }
8175
8176 static void
8177 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8178 {
8179         struct nfs4_layoutcommit_data *data = calldata;
8180         struct nfs_server *server = NFS_SERVER(data->args.inode);
8181
8182         if (!nfs41_sequence_done(task, &data->res.seq_res))
8183                 return;
8184
8185         switch (task->tk_status) { /* Just ignore these failures */
8186         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8187         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8188         case -NFS4ERR_BADLAYOUT:     /* no layout */
8189         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8190                 task->tk_status = 0;
8191         case 0:
8192                 break;
8193         default:
8194                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8195                         rpc_restart_call_prepare(task);
8196                         return;
8197                 }
8198         }
8199 }
8200
8201 static void nfs4_layoutcommit_release(void *calldata)
8202 {
8203         struct nfs4_layoutcommit_data *data = calldata;
8204
8205         pnfs_cleanup_layoutcommit(data);
8206         nfs_post_op_update_inode_force_wcc(data->args.inode,
8207                                            data->res.fattr);
8208         put_rpccred(data->cred);
8209         nfs_iput_and_deactive(data->inode);
8210         kfree(data);
8211 }
8212
8213 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8214         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8215         .rpc_call_done = nfs4_layoutcommit_done,
8216         .rpc_release = nfs4_layoutcommit_release,
8217 };
8218
8219 int
8220 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8221 {
8222         struct rpc_message msg = {
8223                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8224                 .rpc_argp = &data->args,
8225                 .rpc_resp = &data->res,
8226                 .rpc_cred = data->cred,
8227         };
8228         struct rpc_task_setup task_setup_data = {
8229                 .task = &data->task,
8230                 .rpc_client = NFS_CLIENT(data->args.inode),
8231                 .rpc_message = &msg,
8232                 .callback_ops = &nfs4_layoutcommit_ops,
8233                 .callback_data = data,
8234         };
8235         struct rpc_task *task;
8236         int status = 0;
8237
8238         dprintk("NFS: initiating layoutcommit call. sync %d "
8239                 "lbw: %llu inode %lu\n", sync,
8240                 data->args.lastbytewritten,
8241                 data->args.inode->i_ino);
8242
8243         if (!sync) {
8244                 data->inode = nfs_igrab_and_active(data->args.inode);
8245                 if (data->inode == NULL) {
8246                         nfs4_layoutcommit_release(data);
8247                         return -EAGAIN;
8248                 }
8249                 task_setup_data.flags = RPC_TASK_ASYNC;
8250         }
8251         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8252         task = rpc_run_task(&task_setup_data);
8253         if (IS_ERR(task))
8254                 return PTR_ERR(task);
8255         if (sync)
8256                 status = task->tk_status;
8257         trace_nfs4_layoutcommit(data->args.inode, status);
8258         dprintk("%s: status %d\n", __func__, status);
8259         rpc_put_task(task);
8260         return status;
8261 }
8262
8263 /**
8264  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8265  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8266  */
8267 static int
8268 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8269                     struct nfs_fsinfo *info,
8270                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8271 {
8272         struct nfs41_secinfo_no_name_args args = {
8273                 .style = SECINFO_STYLE_CURRENT_FH,
8274         };
8275         struct nfs4_secinfo_res res = {
8276                 .flavors = flavors,
8277         };
8278         struct rpc_message msg = {
8279                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8280                 .rpc_argp = &args,
8281                 .rpc_resp = &res,
8282         };
8283         struct rpc_clnt *clnt = server->client;
8284         struct rpc_cred *cred = NULL;
8285         int status;
8286
8287         if (use_integrity) {
8288                 clnt = server->nfs_client->cl_rpcclient;
8289                 cred = nfs4_get_clid_cred(server->nfs_client);
8290                 msg.rpc_cred = cred;
8291         }
8292
8293         dprintk("--> %s\n", __func__);
8294         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8295                                 &res.seq_res, 0);
8296         dprintk("<-- %s status=%d\n", __func__, status);
8297
8298         if (cred)
8299                 put_rpccred(cred);
8300
8301         return status;
8302 }
8303
8304 static int
8305 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8306                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8307 {
8308         struct nfs4_exception exception = { };
8309         int err;
8310         do {
8311                 /* first try using integrity protection */
8312                 err = -NFS4ERR_WRONGSEC;
8313
8314                 /* try to use integrity protection with machine cred */
8315                 if (_nfs4_is_integrity_protected(server->nfs_client))
8316                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8317                                                           flavors, true);
8318
8319                 /*
8320                  * if unable to use integrity protection, or SECINFO with
8321                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8322                  * disallowed by spec, but exists in deployed servers) use
8323                  * the current filesystem's rpc_client and the user cred.
8324                  */
8325                 if (err == -NFS4ERR_WRONGSEC)
8326                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8327                                                           flavors, false);
8328
8329                 switch (err) {
8330                 case 0:
8331                 case -NFS4ERR_WRONGSEC:
8332                 case -ENOTSUPP:
8333                         goto out;
8334                 default:
8335                         err = nfs4_handle_exception(server, err, &exception);
8336                 }
8337         } while (exception.retry);
8338 out:
8339         return err;
8340 }
8341
8342 static int
8343 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8344                     struct nfs_fsinfo *info)
8345 {
8346         int err;
8347         struct page *page;
8348         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8349         struct nfs4_secinfo_flavors *flavors;
8350         struct nfs4_secinfo4 *secinfo;
8351         int i;
8352
8353         page = alloc_page(GFP_KERNEL);
8354         if (!page) {
8355                 err = -ENOMEM;
8356                 goto out;
8357         }
8358
8359         flavors = page_address(page);
8360         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8361
8362         /*
8363          * Fall back on "guess and check" method if
8364          * the server doesn't support SECINFO_NO_NAME
8365          */
8366         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8367                 err = nfs4_find_root_sec(server, fhandle, info);
8368                 goto out_freepage;
8369         }
8370         if (err)
8371                 goto out_freepage;
8372
8373         for (i = 0; i < flavors->num_flavors; i++) {
8374                 secinfo = &flavors->flavors[i];
8375
8376                 switch (secinfo->flavor) {
8377                 case RPC_AUTH_NULL:
8378                 case RPC_AUTH_UNIX:
8379                 case RPC_AUTH_GSS:
8380                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8381                                         &secinfo->flavor_info);
8382                         break;
8383                 default:
8384                         flavor = RPC_AUTH_MAXFLAVOR;
8385                         break;
8386                 }
8387
8388                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8389                         flavor = RPC_AUTH_MAXFLAVOR;
8390
8391                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8392                         err = nfs4_lookup_root_sec(server, fhandle,
8393                                                    info, flavor);
8394                         if (!err)
8395                                 break;
8396                 }
8397         }
8398
8399         if (flavor == RPC_AUTH_MAXFLAVOR)
8400                 err = -EPERM;
8401
8402 out_freepage:
8403         put_page(page);
8404         if (err == -EACCES)
8405                 return -EPERM;
8406 out:
8407         return err;
8408 }
8409
8410 static int _nfs41_test_stateid(struct nfs_server *server,
8411                 nfs4_stateid *stateid,
8412                 struct rpc_cred *cred)
8413 {
8414         int status;
8415         struct nfs41_test_stateid_args args = {
8416                 .stateid = stateid,
8417         };
8418         struct nfs41_test_stateid_res res;
8419         struct rpc_message msg = {
8420                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8421                 .rpc_argp = &args,
8422                 .rpc_resp = &res,
8423                 .rpc_cred = cred,
8424         };
8425         struct rpc_clnt *rpc_client = server->client;
8426
8427         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8428                 &rpc_client, &msg);
8429
8430         dprintk("NFS call  test_stateid %p\n", stateid);
8431         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8432         nfs4_set_sequence_privileged(&args.seq_args);
8433         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8434                         &args.seq_args, &res.seq_res);
8435         if (status != NFS_OK) {
8436                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8437                 return status;
8438         }
8439         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8440         return -res.status;
8441 }
8442
8443 /**
8444  * nfs41_test_stateid - perform a TEST_STATEID operation
8445  *
8446  * @server: server / transport on which to perform the operation
8447  * @stateid: state ID to test
8448  * @cred: credential
8449  *
8450  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8451  * Otherwise a negative NFS4ERR value is returned if the operation
8452  * failed or the state ID is not currently valid.
8453  */
8454 static int nfs41_test_stateid(struct nfs_server *server,
8455                 nfs4_stateid *stateid,
8456                 struct rpc_cred *cred)
8457 {
8458         struct nfs4_exception exception = { };
8459         int err;
8460         do {
8461                 err = _nfs41_test_stateid(server, stateid, cred);
8462                 if (err != -NFS4ERR_DELAY)
8463                         break;
8464                 nfs4_handle_exception(server, err, &exception);
8465         } while (exception.retry);
8466         return err;
8467 }
8468
8469 struct nfs_free_stateid_data {
8470         struct nfs_server *server;
8471         struct nfs41_free_stateid_args args;
8472         struct nfs41_free_stateid_res res;
8473 };
8474
8475 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8476 {
8477         struct nfs_free_stateid_data *data = calldata;
8478         nfs41_setup_sequence(nfs4_get_session(data->server),
8479                         &data->args.seq_args,
8480                         &data->res.seq_res,
8481                         task);
8482 }
8483
8484 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8485 {
8486         struct nfs_free_stateid_data *data = calldata;
8487
8488         nfs41_sequence_done(task, &data->res.seq_res);
8489
8490         switch (task->tk_status) {
8491         case -NFS4ERR_DELAY:
8492                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8493                         rpc_restart_call_prepare(task);
8494         }
8495 }
8496
8497 static void nfs41_free_stateid_release(void *calldata)
8498 {
8499         kfree(calldata);
8500 }
8501
8502 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8503         .rpc_call_prepare = nfs41_free_stateid_prepare,
8504         .rpc_call_done = nfs41_free_stateid_done,
8505         .rpc_release = nfs41_free_stateid_release,
8506 };
8507
8508 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8509                 nfs4_stateid *stateid,
8510                 struct rpc_cred *cred,
8511                 bool privileged)
8512 {
8513         struct rpc_message msg = {
8514                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8515                 .rpc_cred = cred,
8516         };
8517         struct rpc_task_setup task_setup = {
8518                 .rpc_client = server->client,
8519                 .rpc_message = &msg,
8520                 .callback_ops = &nfs41_free_stateid_ops,
8521                 .flags = RPC_TASK_ASYNC,
8522         };
8523         struct nfs_free_stateid_data *data;
8524
8525         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8526                 &task_setup.rpc_client, &msg);
8527
8528         dprintk("NFS call  free_stateid %p\n", stateid);
8529         data = kmalloc(sizeof(*data), GFP_NOFS);
8530         if (!data)
8531                 return ERR_PTR(-ENOMEM);
8532         data->server = server;
8533         nfs4_stateid_copy(&data->args.stateid, stateid);
8534
8535         task_setup.callback_data = data;
8536
8537         msg.rpc_argp = &data->args;
8538         msg.rpc_resp = &data->res;
8539         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8540         if (privileged)
8541                 nfs4_set_sequence_privileged(&data->args.seq_args);
8542
8543         return rpc_run_task(&task_setup);
8544 }
8545
8546 /**
8547  * nfs41_free_stateid - perform a FREE_STATEID operation
8548  *
8549  * @server: server / transport on which to perform the operation
8550  * @stateid: state ID to release
8551  * @cred: credential
8552  *
8553  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8554  * negative NFS4ERR value is returned.
8555  */
8556 static int nfs41_free_stateid(struct nfs_server *server,
8557                 nfs4_stateid *stateid,
8558                 struct rpc_cred *cred)
8559 {
8560         struct rpc_task *task;
8561         int ret;
8562
8563         task = _nfs41_free_stateid(server, stateid, cred, true);
8564         if (IS_ERR(task))
8565                 return PTR_ERR(task);
8566         ret = rpc_wait_for_completion_task(task);
8567         if (!ret)
8568                 ret = task->tk_status;
8569         rpc_put_task(task);
8570         return ret;
8571 }
8572
8573 static void
8574 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8575 {
8576         struct rpc_task *task;
8577         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8578
8579         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8580         nfs4_free_lock_state(server, lsp);
8581         if (IS_ERR(task))
8582                 return;
8583         rpc_put_task(task);
8584 }
8585
8586 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8587                 const nfs4_stateid *s2)
8588 {
8589         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8590                 return false;
8591
8592         if (s1->seqid == s2->seqid)
8593                 return true;
8594         if (s1->seqid == 0 || s2->seqid == 0)
8595                 return true;
8596
8597         return false;
8598 }
8599
8600 #endif /* CONFIG_NFS_V4_1 */
8601
8602 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8603                 const nfs4_stateid *s2)
8604 {
8605         return nfs4_stateid_match(s1, s2);
8606 }
8607
8608
8609 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8610         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8611         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8612         .recover_open   = nfs4_open_reclaim,
8613         .recover_lock   = nfs4_lock_reclaim,
8614         .establish_clid = nfs4_init_clientid,
8615         .detect_trunking = nfs40_discover_server_trunking,
8616 };
8617
8618 #if defined(CONFIG_NFS_V4_1)
8619 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8620         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8621         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8622         .recover_open   = nfs4_open_reclaim,
8623         .recover_lock   = nfs4_lock_reclaim,
8624         .establish_clid = nfs41_init_clientid,
8625         .reclaim_complete = nfs41_proc_reclaim_complete,
8626         .detect_trunking = nfs41_discover_server_trunking,
8627 };
8628 #endif /* CONFIG_NFS_V4_1 */
8629
8630 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8631         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8632         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8633         .recover_open   = nfs40_open_expired,
8634         .recover_lock   = nfs4_lock_expired,
8635         .establish_clid = nfs4_init_clientid,
8636 };
8637
8638 #if defined(CONFIG_NFS_V4_1)
8639 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8640         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8641         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8642         .recover_open   = nfs41_open_expired,
8643         .recover_lock   = nfs41_lock_expired,
8644         .establish_clid = nfs41_init_clientid,
8645 };
8646 #endif /* CONFIG_NFS_V4_1 */
8647
8648 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8649         .sched_state_renewal = nfs4_proc_async_renew,
8650         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8651         .renew_lease = nfs4_proc_renew,
8652 };
8653
8654 #if defined(CONFIG_NFS_V4_1)
8655 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8656         .sched_state_renewal = nfs41_proc_async_sequence,
8657         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8658         .renew_lease = nfs4_proc_sequence,
8659 };
8660 #endif
8661
8662 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8663         .get_locations = _nfs40_proc_get_locations,
8664         .fsid_present = _nfs40_proc_fsid_present,
8665 };
8666
8667 #if defined(CONFIG_NFS_V4_1)
8668 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8669         .get_locations = _nfs41_proc_get_locations,
8670         .fsid_present = _nfs41_proc_fsid_present,
8671 };
8672 #endif  /* CONFIG_NFS_V4_1 */
8673
8674 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8675         .minor_version = 0,
8676         .init_caps = NFS_CAP_READDIRPLUS
8677                 | NFS_CAP_ATOMIC_OPEN
8678                 | NFS_CAP_POSIX_LOCK,
8679         .init_client = nfs40_init_client,
8680         .shutdown_client = nfs40_shutdown_client,
8681         .match_stateid = nfs4_match_stateid,
8682         .find_root_sec = nfs4_find_root_sec,
8683         .free_lock_state = nfs4_release_lockowner,
8684         .alloc_seqid = nfs_alloc_seqid,
8685         .call_sync_ops = &nfs40_call_sync_ops,
8686         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8687         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8688         .state_renewal_ops = &nfs40_state_renewal_ops,
8689         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8690 };
8691
8692 #if defined(CONFIG_NFS_V4_1)
8693 static struct nfs_seqid *
8694 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8695 {
8696         return NULL;
8697 }
8698
8699 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8700         .minor_version = 1,
8701         .init_caps = NFS_CAP_READDIRPLUS
8702                 | NFS_CAP_ATOMIC_OPEN
8703                 | NFS_CAP_POSIX_LOCK
8704                 | NFS_CAP_STATEID_NFSV41
8705                 | NFS_CAP_ATOMIC_OPEN_V1,
8706         .init_client = nfs41_init_client,
8707         .shutdown_client = nfs41_shutdown_client,
8708         .match_stateid = nfs41_match_stateid,
8709         .find_root_sec = nfs41_find_root_sec,
8710         .free_lock_state = nfs41_free_lock_state,
8711         .alloc_seqid = nfs_alloc_no_seqid,
8712         .call_sync_ops = &nfs41_call_sync_ops,
8713         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8714         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8715         .state_renewal_ops = &nfs41_state_renewal_ops,
8716         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8717 };
8718 #endif
8719
8720 #if defined(CONFIG_NFS_V4_2)
8721 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8722         .minor_version = 2,
8723         .init_caps = NFS_CAP_READDIRPLUS
8724                 | NFS_CAP_ATOMIC_OPEN
8725                 | NFS_CAP_POSIX_LOCK
8726                 | NFS_CAP_STATEID_NFSV41
8727                 | NFS_CAP_ATOMIC_OPEN_V1
8728                 | NFS_CAP_ALLOCATE
8729                 | NFS_CAP_DEALLOCATE
8730                 | NFS_CAP_SEEK
8731                 | NFS_CAP_LAYOUTSTATS,
8732         .init_client = nfs41_init_client,
8733         .shutdown_client = nfs41_shutdown_client,
8734         .match_stateid = nfs41_match_stateid,
8735         .find_root_sec = nfs41_find_root_sec,
8736         .free_lock_state = nfs41_free_lock_state,
8737         .call_sync_ops = &nfs41_call_sync_ops,
8738         .alloc_seqid = nfs_alloc_no_seqid,
8739         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8740         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8741         .state_renewal_ops = &nfs41_state_renewal_ops,
8742         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8743 };
8744 #endif
8745
8746 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8747         [0] = &nfs_v4_0_minor_ops,
8748 #if defined(CONFIG_NFS_V4_1)
8749         [1] = &nfs_v4_1_minor_ops,
8750 #endif
8751 #if defined(CONFIG_NFS_V4_2)
8752         [2] = &nfs_v4_2_minor_ops,
8753 #endif
8754 };
8755
8756 static const struct inode_operations nfs4_dir_inode_operations = {
8757         .create         = nfs_create,
8758         .lookup         = nfs_lookup,
8759         .atomic_open    = nfs_atomic_open,
8760         .link           = nfs_link,
8761         .unlink         = nfs_unlink,
8762         .symlink        = nfs_symlink,
8763         .mkdir          = nfs_mkdir,
8764         .rmdir          = nfs_rmdir,
8765         .mknod          = nfs_mknod,
8766         .rename         = nfs_rename,
8767         .permission     = nfs_permission,
8768         .getattr        = nfs_getattr,
8769         .setattr        = nfs_setattr,
8770         .getxattr       = generic_getxattr,
8771         .setxattr       = generic_setxattr,
8772         .listxattr      = generic_listxattr,
8773         .removexattr    = generic_removexattr,
8774 };
8775
8776 static const struct inode_operations nfs4_file_inode_operations = {
8777         .permission     = nfs_permission,
8778         .getattr        = nfs_getattr,
8779         .setattr        = nfs_setattr,
8780         .getxattr       = generic_getxattr,
8781         .setxattr       = generic_setxattr,
8782         .listxattr      = generic_listxattr,
8783         .removexattr    = generic_removexattr,
8784 };
8785
8786 const struct nfs_rpc_ops nfs_v4_clientops = {
8787         .version        = 4,                    /* protocol version */
8788         .dentry_ops     = &nfs4_dentry_operations,
8789         .dir_inode_ops  = &nfs4_dir_inode_operations,
8790         .file_inode_ops = &nfs4_file_inode_operations,
8791         .file_ops       = &nfs4_file_operations,
8792         .getroot        = nfs4_proc_get_root,
8793         .submount       = nfs4_submount,
8794         .try_mount      = nfs4_try_mount,
8795         .getattr        = nfs4_proc_getattr,
8796         .setattr        = nfs4_proc_setattr,
8797         .lookup         = nfs4_proc_lookup,
8798         .access         = nfs4_proc_access,
8799         .readlink       = nfs4_proc_readlink,
8800         .create         = nfs4_proc_create,
8801         .remove         = nfs4_proc_remove,
8802         .unlink_setup   = nfs4_proc_unlink_setup,
8803         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8804         .unlink_done    = nfs4_proc_unlink_done,
8805         .rename_setup   = nfs4_proc_rename_setup,
8806         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8807         .rename_done    = nfs4_proc_rename_done,
8808         .link           = nfs4_proc_link,
8809         .symlink        = nfs4_proc_symlink,
8810         .mkdir          = nfs4_proc_mkdir,
8811         .rmdir          = nfs4_proc_remove,
8812         .readdir        = nfs4_proc_readdir,
8813         .mknod          = nfs4_proc_mknod,
8814         .statfs         = nfs4_proc_statfs,
8815         .fsinfo         = nfs4_proc_fsinfo,
8816         .pathconf       = nfs4_proc_pathconf,
8817         .set_capabilities = nfs4_server_capabilities,
8818         .decode_dirent  = nfs4_decode_dirent,
8819         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8820         .read_setup     = nfs4_proc_read_setup,
8821         .read_done      = nfs4_read_done,
8822         .write_setup    = nfs4_proc_write_setup,
8823         .write_done     = nfs4_write_done,
8824         .commit_setup   = nfs4_proc_commit_setup,
8825         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8826         .commit_done    = nfs4_commit_done,
8827         .lock           = nfs4_proc_lock,
8828         .clear_acl_cache = nfs4_zap_acl_attr,
8829         .close_context  = nfs4_close_context,
8830         .open_context   = nfs4_atomic_open,
8831         .have_delegation = nfs4_have_delegation,
8832         .return_delegation = nfs4_inode_return_delegation,
8833         .alloc_client   = nfs4_alloc_client,
8834         .init_client    = nfs4_init_client,
8835         .free_client    = nfs4_free_client,
8836         .create_server  = nfs4_create_server,
8837         .clone_server   = nfs_clone_server,
8838 };
8839
8840 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8841         .prefix = XATTR_NAME_NFSV4_ACL,
8842         .list   = nfs4_xattr_list_nfs4_acl,
8843         .get    = nfs4_xattr_get_nfs4_acl,
8844         .set    = nfs4_xattr_set_nfs4_acl,
8845 };
8846
8847 const struct xattr_handler *nfs4_xattr_handlers[] = {
8848         &nfs4_xattr_nfs4_acl_handler,
8849 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8850         &nfs4_xattr_nfs4_label_handler,
8851 #endif
8852         NULL
8853 };
8854
8855 /*
8856  * Local variables:
8857  *  c-basic-offset: 8
8858  * End:
8859  */