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                         | FATTR4_WORD2_CLONE_BLKSIZE
 242 };
 243
 244 const u32 nfs4_fs_locations_bitmap[3] = {
 245         FATTR4_WORD0_TYPE
 246         | FATTR4_WORD0_CHANGE
 247         | FATTR4_WORD0_SIZE
 248         | FATTR4_WORD0_FSID
 249         | FATTR4_WORD0_FILEID
 250         | FATTR4_WORD0_FS_LOCATIONS,
 251         FATTR4_WORD1_MODE
 252         | FATTR4_WORD1_NUMLINKS
 253         | FATTR4_WORD1_OWNER
 254         | FATTR4_WORD1_OWNER_GROUP
 255         | FATTR4_WORD1_RAWDEV
 256         | FATTR4_WORD1_SPACE_USED
 257         | FATTR4_WORD1_TIME_ACCESS
 258         | FATTR4_WORD1_TIME_METADATA
 259         | FATTR4_WORD1_TIME_MODIFY
 260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 261 };
 262
 263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 264                 struct nfs4_readdir_arg *readdir)
 265 {
 266         __be32 *start, *p;
 267
 268         if (cookie > 2) {
 269                 readdir->cookie = cookie;
 270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 271                 return;
 272         }
 273
 274         readdir->cookie = 0;
 275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 276         if (cookie == 2)
 277                 return;
 278
 279         /*
 280          * NFSv4 servers do not return entries for '.' and '..'
 281          * Therefore, we fake these entries here.  We let '.'
 282          * have cookie 0 and '..' have cookie 1.  Note that
 283          * when talking to the server, we always send cookie 0
 284          * instead of 1 or 2.
 285          */
 286         start = p = kmap_atomic(*readdir->pages);
 287
 288         if (cookie == 0) {
 289                 *p++ = xdr_one;                                  /* next */
 290                 *p++ = xdr_zero;                   /* cookie, first word */
 291                 *p++ = xdr_one;                   /* cookie, second word */
 292                 *p++ = xdr_one;                             /* entry len */
 293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 294                 p++;
 295                 *p++ = xdr_one;                         /* bitmap length */
 296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 297                 *p++ = htonl(8);              /* attribute buffer length */
 298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 299         }
 300
 301         *p++ = xdr_one;                                  /* next */
 302         *p++ = xdr_zero;                   /* cookie, first word */
 303         *p++ = xdr_two;                   /* cookie, second word */
 304         *p++ = xdr_two;                             /* entry len */
 305         memcpy(p, "..\0\0", 4);                         /* entry */
 306         p++;
 307         *p++ = xdr_one;                         /* bitmap length */
 308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309         *p++ = htonl(8);              /* attribute buffer length */
 310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 311
 312         readdir->pgbase = (char *)p - (char *)start;
 313         readdir->count -= readdir->pgbase;
 314         kunmap_atomic(start);
 315 }
 316
 317 static long nfs4_update_delay(long *timeout)
 318 {
 319         long ret;
 320         if (!timeout)
 321                 return NFS4_POLL_RETRY_MAX;
 322         if (*timeout <= 0)
 323                 *timeout = NFS4_POLL_RETRY_MIN;
 324         if (*timeout > NFS4_POLL_RETRY_MAX)
 325                 *timeout = NFS4_POLL_RETRY_MAX;
 326         ret = *timeout;
 327         *timeout <<= 1;
 328         return ret;
 329 }
 330
 331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 332 {
 333         int res = 0;
 334
 335         might_sleep();
 336
 337         freezable_schedule_timeout_killable_unsafe(
 338                 nfs4_update_delay(timeout));
 339         if (fatal_signal_pending(current))
 340                 res = -ERESTARTSYS;
 341         return res;
 342 }
 343
 344 /* This is the error handling routine for processes that are allowed
 345  * to sleep.
 346  */
 347 static int nfs4_do_handle_exception(struct nfs_server *server,
 348                 int errorcode, struct nfs4_exception *exception)
 349 {
 350         struct nfs_client *clp = server->nfs_client;
 351         struct nfs4_state *state = exception->state;
 352         struct inode *inode = exception->inode;
 353         int ret = errorcode;
 354
 355         exception->delay = 0;
 356         exception->recovering = 0;
 357         exception->retry = 0;
 358         switch(errorcode) {
 359                 case 0:
 360                         return 0;
 361                 case -NFS4ERR_OPENMODE:
 362                 case -NFS4ERR_DELEG_REVOKED:
 363                 case -NFS4ERR_ADMIN_REVOKED:
 364                 case -NFS4ERR_BAD_STATEID:
 365                         if (inode && nfs_async_inode_return_delegation(inode,
 366                                                 NULL) == 0)
 367                                 goto wait_on_recovery;
 368                         if (state == NULL)
 369                                 break;
 370                         ret = nfs4_schedule_stateid_recovery(server, state);
 371                         if (ret < 0)
 372                                 break;
 373                         goto wait_on_recovery;
 374                 case -NFS4ERR_EXPIRED:
 375                         if (state != NULL) {
 376                                 ret = nfs4_schedule_stateid_recovery(server, state);
 377                                 if (ret < 0)
 378                                         break;
 379                         }
 380                 case -NFS4ERR_STALE_STATEID:
 381                 case -NFS4ERR_STALE_CLIENTID:
 382                         nfs4_schedule_lease_recovery(clp);
 383                         goto wait_on_recovery;
 384                 case -NFS4ERR_MOVED:
 385                         ret = nfs4_schedule_migration_recovery(server);
 386                         if (ret < 0)
 387                                 break;
 388                         goto wait_on_recovery;
 389                 case -NFS4ERR_LEASE_MOVED:
 390                         nfs4_schedule_lease_moved_recovery(clp);
 391                         goto wait_on_recovery;
 392 #if defined(CONFIG_NFS_V4_1)
 393                 case -NFS4ERR_BADSESSION:
 394                 case -NFS4ERR_BADSLOT:
 395                 case -NFS4ERR_BAD_HIGH_SLOT:
 396                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 397                 case -NFS4ERR_DEADSESSION:
 398                 case -NFS4ERR_SEQ_FALSE_RETRY:
 399                 case -NFS4ERR_SEQ_MISORDERED:
 400                         dprintk("%s ERROR: %d Reset session\n", __func__,
 401                                 errorcode);
 402                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 403                         goto wait_on_recovery;
 404 #endif /* defined(CONFIG_NFS_V4_1) */
 405                 case -NFS4ERR_FILE_OPEN:
 406                         if (exception->timeout > HZ) {
 407                                 /* We have retried a decent amount, time to
 408                                  * fail
 409                                  */
 410                                 ret = -EBUSY;
 411                                 break;
 412                         }
 413                 case -NFS4ERR_DELAY:
 414                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 415                 case -NFS4ERR_GRACE:
 416                         exception->delay = 1;
 417                         return 0;
 418
 419                 case -NFS4ERR_RETRY_UNCACHED_REP:
 420                 case -NFS4ERR_OLD_STATEID:
 421                         exception->retry = 1;
 422                         break;
 423                 case -NFS4ERR_BADOWNER:
 424                         /* The following works around a Linux server bug! */
 425                 case -NFS4ERR_BADNAME:
 426                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 427                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 428                                 exception->retry = 1;
 429                                 printk(KERN_WARNING "NFS: v4 server %s "
 430                                                 "does not accept raw "
 431                                                 "uid/gids. "
 432                                                 "Reenabling the idmapper.\n",
 433                                                 server->nfs_client->cl_hostname);
 434                         }
 435         }
 436         /* We failed to handle the error */
 437         return nfs4_map_errors(ret);
 438 wait_on_recovery:
 439         exception->recovering = 1;
 440         return 0;
 441 }
 442
 443 /* This is the error handling routine for processes that are allowed
 444  * to sleep.
 445  */
 446 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 447 {
 448         struct nfs_client *clp = server->nfs_client;
 449         int ret;
 450
 451         ret = nfs4_do_handle_exception(server, errorcode, exception);
 452         if (exception->delay) {
 453                 ret = nfs4_delay(server->client, &exception->timeout);
 454                 goto out_retry;
 455         }
 456         if (exception->recovering) {
 457                 ret = nfs4_wait_clnt_recover(clp);
 458                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 459                         return -EIO;
 460                 goto out_retry;
 461         }
 462         return ret;
 463 out_retry:
 464         if (ret == 0)
 465                 exception->retry = 1;
 466         return ret;
 467 }
 468
 469 static int
 470 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 471                 int errorcode, struct nfs4_exception *exception)
 472 {
 473         struct nfs_client *clp = server->nfs_client;
 474         int ret;
 475
 476         ret = nfs4_do_handle_exception(server, errorcode, exception);
 477         if (exception->delay) {
 478                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 479                 goto out_retry;
 480         }
 481         if (exception->recovering) {
 482                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 483                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 484                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 485                 goto out_retry;
 486         }
 487         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 488                 ret = -EIO;
 489         return ret;
 490 out_retry:
 491         if (ret == 0)
 492                 exception->retry = 1;
 493         return ret;
 494 }
 495
 496 static int
 497 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 498                         struct nfs4_state *state, long *timeout)
 499 {
 500         struct nfs4_exception exception = {
 501                 .state = state,
 502         };
 503
 504         if (task->tk_status >= 0)
 505                 return 0;
 506         if (timeout)
 507                 exception.timeout = *timeout;
 508         task->tk_status = nfs4_async_handle_exception(task, server,
 509                         task->tk_status,
 510                         &exception);
 511         if (exception.delay && timeout)
 512                 *timeout = exception.timeout;
 513         if (exception.retry)
 514                 return -EAGAIN;
 515         return 0;
 516 }
 517
 518 /*
 519  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 520  * or 'false' otherwise.
 521  */
 522 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 523 {
 524         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 525
 526         if (flavor == RPC_AUTH_GSS_KRB5I ||
 527             flavor == RPC_AUTH_GSS_KRB5P)
 528                 return true;
 529
 530         return false;
 531 }
 532
 533 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 534 {
 535         spin_lock(&clp->cl_lock);
 536         if (time_before(clp->cl_last_renewal,timestamp))
 537                 clp->cl_last_renewal = timestamp;
 538         spin_unlock(&clp->cl_lock);
 539 }
 540
 541 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 542 {
 543         struct nfs_client *clp = server->nfs_client;
 544
 545         if (!nfs4_has_session(clp))
 546                 do_renew_lease(clp, timestamp);
 547 }
 548
 549 struct nfs4_call_sync_data {
 550         const struct nfs_server *seq_server;
 551         struct nfs4_sequence_args *seq_args;
 552         struct nfs4_sequence_res *seq_res;
 553 };
 554
 555 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 556                         struct nfs4_sequence_res *res, int cache_reply)
 557 {
 558         args->sa_slot = NULL;
 559         args->sa_cache_this = cache_reply;
 560         args->sa_privileged = 0;
 561
 562         res->sr_slot = NULL;
 563 }
 564
 565 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 566 {
 567         args->sa_privileged = 1;
 568 }
 569
 570 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 571                          struct nfs4_sequence_args *args,
 572                          struct nfs4_sequence_res *res,
 573                          struct rpc_task *task)
 574 {
 575         struct nfs4_slot *slot;
 576
 577         /* slot already allocated? */
 578         if (res->sr_slot != NULL)
 579                 goto out_start;
 580
 581         spin_lock(&tbl->slot_tbl_lock);
 582         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 583                 goto out_sleep;
 584
 585         slot = nfs4_alloc_slot(tbl);
 586         if (IS_ERR(slot)) {
 587                 if (slot == ERR_PTR(-ENOMEM))
 588                         task->tk_timeout = HZ >> 2;
 589                 goto out_sleep;
 590         }
 591         spin_unlock(&tbl->slot_tbl_lock);
 592
 593         args->sa_slot = slot;
 594         res->sr_slot = slot;
 595
 596 out_start:
 597         rpc_call_start(task);
 598         return 0;
 599
 600 out_sleep:
 601         if (args->sa_privileged)
 602                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 603                                 NULL, RPC_PRIORITY_PRIVILEGED);
 604         else
 605                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 606         spin_unlock(&tbl->slot_tbl_lock);
 607         return -EAGAIN;
 608 }
 609 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 610
 611 static int nfs40_sequence_done(struct rpc_task *task,
 612                                struct nfs4_sequence_res *res)
 613 {
 614         struct nfs4_slot *slot = res->sr_slot;
 615         struct nfs4_slot_table *tbl;
 616
 617         if (slot == NULL)
 618                 goto out;
 619
 620         tbl = slot->table;
 621         spin_lock(&tbl->slot_tbl_lock);
 622         if (!nfs41_wake_and_assign_slot(tbl, slot))
 623                 nfs4_free_slot(tbl, slot);
 624         spin_unlock(&tbl->slot_tbl_lock);
 625
 626         res->sr_slot = NULL;
 627 out:
 628         return 1;
 629 }
 630
 631 #if defined(CONFIG_NFS_V4_1)
 632
 633 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 634 {
 635         struct nfs4_session *session;
 636         struct nfs4_slot_table *tbl;
 637         struct nfs4_slot *slot = res->sr_slot;
 638         bool send_new_highest_used_slotid = false;
 639
 640         tbl = slot->table;
 641         session = tbl->session;
 642
 643         spin_lock(&tbl->slot_tbl_lock);
 644         /* Be nice to the server: try to ensure that the last transmitted
 645          * value for highest_user_slotid <= target_highest_slotid
 646          */
 647         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 648                 send_new_highest_used_slotid = true;
 649
 650         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 651                 send_new_highest_used_slotid = false;
 652                 goto out_unlock;
 653         }
 654         nfs4_free_slot(tbl, slot);
 655
 656         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 657                 send_new_highest_used_slotid = false;
 658 out_unlock:
 659         spin_unlock(&tbl->slot_tbl_lock);
 660         res->sr_slot = NULL;
 661         if (send_new_highest_used_slotid)
 662                 nfs41_notify_server(session->clp);
 663 }
 664
 665 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 666 {
 667         struct nfs4_session *session;
 668         struct nfs4_slot *slot = res->sr_slot;
 669         struct nfs_client *clp;
 670         bool interrupted = false;
 671         int ret = 1;
 672
 673         if (slot == NULL)
 674                 goto out_noaction;
 675         /* don't increment the sequence number if the task wasn't sent */
 676         if (!RPC_WAS_SENT(task))
 677                 goto out;
 678
 679         session = slot->table->session;
 680
 681         if (slot->interrupted) {
 682                 slot->interrupted = 0;
 683                 interrupted = true;
 684         }
 685
 686         trace_nfs4_sequence_done(session, res);
 687         /* Check the SEQUENCE operation status */
 688         switch (res->sr_status) {
 689         case 0:
 690                 /* Update the slot's sequence and clientid lease timer */
 691                 ++slot->seq_nr;
 692                 clp = session->clp;
 693                 do_renew_lease(clp, res->sr_timestamp);
 694                 /* Check sequence flags */
 695                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 696                 nfs41_update_target_slotid(slot->table, slot, res);
 697                 break;
 698         case 1:
 699                 /*
 700                  * sr_status remains 1 if an RPC level error occurred.
 701                  * The server may or may not have processed the sequence
 702                  * operation..
 703                  * Mark the slot as having hosted an interrupted RPC call.
 704                  */
 705                 slot->interrupted = 1;
 706                 goto out;
 707         case -NFS4ERR_DELAY:
 708                 /* The server detected a resend of the RPC call and
 709                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 710                  * of RFC5661.
 711                  */
 712                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 713                         __func__,
 714                         slot->slot_nr,
 715                         slot->seq_nr);
 716                 goto out_retry;
 717         case -NFS4ERR_BADSLOT:
 718                 /*
 719                  * The slot id we used was probably retired. Try again
 720                  * using a different slot id.
 721                  */
 722                 goto retry_nowait;
 723         case -NFS4ERR_SEQ_MISORDERED:
 724                 /*
 725                  * Was the last operation on this sequence interrupted?
 726                  * If so, retry after bumping the sequence number.
 727                  */
 728                 if (interrupted) {
 729                         ++slot->seq_nr;
 730                         goto retry_nowait;
 731                 }
 732                 /*
 733                  * Could this slot have been previously retired?
 734                  * If so, then the server may be expecting seq_nr = 1!
 735                  */
 736                 if (slot->seq_nr != 1) {
 737                         slot->seq_nr = 1;
 738                         goto retry_nowait;
 739                 }
 740                 break;
 741         case -NFS4ERR_SEQ_FALSE_RETRY:
 742                 ++slot->seq_nr;
 743                 goto retry_nowait;
 744         default:
 745                 /* Just update the slot sequence no. */
 746                 ++slot->seq_nr;
 747         }
 748 out:
 749         /* The session may be reset by one of the error handlers. */
 750         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 751         nfs41_sequence_free_slot(res);
 752 out_noaction:
 753         return ret;
 754 retry_nowait:
 755         if (rpc_restart_call_prepare(task)) {
 756                 task->tk_status = 0;
 757                 ret = 0;
 758         }
 759         goto out;
 760 out_retry:
 761         if (!rpc_restart_call(task))
 762                 goto out;
 763         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 764         return 0;
 765 }
 766 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 767
 768 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 769 {
 770         if (res->sr_slot == NULL)
 771                 return 1;
 772         if (!res->sr_slot->table->session)
 773                 return nfs40_sequence_done(task, res);
 774         return nfs41_sequence_done(task, res);
 775 }
 776 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 777
 778 int nfs41_setup_sequence(struct nfs4_session *session,
 779                                 struct nfs4_sequence_args *args,
 780                                 struct nfs4_sequence_res *res,
 781                                 struct rpc_task *task)
 782 {
 783         struct nfs4_slot *slot;
 784         struct nfs4_slot_table *tbl;
 785
 786         dprintk("--> %s\n", __func__);
 787         /* slot already allocated? */
 788         if (res->sr_slot != NULL)
 789                 goto out_success;
 790
 791         tbl = &session->fc_slot_table;
 792
 793         task->tk_timeout = 0;
 794
 795         spin_lock(&tbl->slot_tbl_lock);
 796         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 797             !args->sa_privileged) {
 798                 /* The state manager will wait until the slot table is empty */
 799                 dprintk("%s session is draining\n", __func__);
 800                 goto out_sleep;
 801         }
 802
 803         slot = nfs4_alloc_slot(tbl);
 804         if (IS_ERR(slot)) {
 805                 /* If out of memory, try again in 1/4 second */
 806                 if (slot == ERR_PTR(-ENOMEM))
 807                         task->tk_timeout = HZ >> 2;
 808                 dprintk("<-- %s: no free slots\n", __func__);
 809                 goto out_sleep;
 810         }
 811         spin_unlock(&tbl->slot_tbl_lock);
 812
 813         args->sa_slot = slot;
 814
 815         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 816                         slot->slot_nr, slot->seq_nr);
 817
 818         res->sr_slot = slot;
 819         res->sr_timestamp = jiffies;
 820         res->sr_status_flags = 0;
 821         /*
 822          * sr_status is only set in decode_sequence, and so will remain
 823          * set to 1 if an rpc level failure occurs.
 824          */
 825         res->sr_status = 1;
 826         trace_nfs4_setup_sequence(session, args);
 827 out_success:
 828         rpc_call_start(task);
 829         return 0;
 830 out_sleep:
 831         /* Privileged tasks are queued with top priority */
 832         if (args->sa_privileged)
 833                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 834                                 NULL, RPC_PRIORITY_PRIVILEGED);
 835         else
 836                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 837         spin_unlock(&tbl->slot_tbl_lock);
 838         return -EAGAIN;
 839 }
 840 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 841
 842 static int nfs4_setup_sequence(const struct nfs_server *server,
 843                                struct nfs4_sequence_args *args,
 844                                struct nfs4_sequence_res *res,
 845                                struct rpc_task *task)
 846 {
 847         struct nfs4_session *session = nfs4_get_session(server);
 848         int ret = 0;
 849
 850         if (!session)
 851                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 852                                             args, res, task);
 853
 854         dprintk("--> %s clp %p session %p sr_slot %u\n",
 855                 __func__, session->clp, session, res->sr_slot ?
 856                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 857
 858         ret = nfs41_setup_sequence(session, args, res, task);
 859
 860         dprintk("<-- %s status=%d\n", __func__, ret);
 861         return ret;
 862 }
 863
 864 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 865 {
 866         struct nfs4_call_sync_data *data = calldata;
 867         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 868
 869         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 870
 871         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 872 }
 873
 874 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 875 {
 876         struct nfs4_call_sync_data *data = calldata;
 877
 878         nfs41_sequence_done(task, data->seq_res);
 879 }
 880
 881 static const struct rpc_call_ops nfs41_call_sync_ops = {
 882         .rpc_call_prepare = nfs41_call_sync_prepare,
 883         .rpc_call_done = nfs41_call_sync_done,
 884 };
 885
 886 #else   /* !CONFIG_NFS_V4_1 */
 887
 888 static int nfs4_setup_sequence(const struct nfs_server *server,
 889                                struct nfs4_sequence_args *args,
 890                                struct nfs4_sequence_res *res,
 891                                struct rpc_task *task)
 892 {
 893         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 894                                     args, res, task);
 895 }
 896
 897 int nfs4_sequence_done(struct rpc_task *task,
 898                        struct nfs4_sequence_res *res)
 899 {
 900         return nfs40_sequence_done(task, res);
 901 }
 902 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 903
 904 #endif  /* !CONFIG_NFS_V4_1 */
 905
 906 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 907 {
 908         struct nfs4_call_sync_data *data = calldata;
 909         nfs4_setup_sequence(data->seq_server,
 910                                 data->seq_args, data->seq_res, task);
 911 }
 912
 913 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 914 {
 915         struct nfs4_call_sync_data *data = calldata;
 916         nfs4_sequence_done(task, data->seq_res);
 917 }
 918
 919 static const struct rpc_call_ops nfs40_call_sync_ops = {
 920         .rpc_call_prepare = nfs40_call_sync_prepare,
 921         .rpc_call_done = nfs40_call_sync_done,
 922 };
 923
 924 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 925                                    struct nfs_server *server,
 926                                    struct rpc_message *msg,
 927                                    struct nfs4_sequence_args *args,
 928                                    struct nfs4_sequence_res *res)
 929 {
 930         int ret;
 931         struct rpc_task *task;
 932         struct nfs_client *clp = server->nfs_client;
 933         struct nfs4_call_sync_data data = {
 934                 .seq_server = server,
 935                 .seq_args = args,
 936                 .seq_res = res,
 937         };
 938         struct rpc_task_setup task_setup = {
 939                 .rpc_client = clnt,
 940                 .rpc_message = msg,
 941                 .callback_ops = clp->cl_mvops->call_sync_ops,
 942                 .callback_data = &data
 943         };
 944
 945         task = rpc_run_task(&task_setup);
 946         if (IS_ERR(task))
 947                 ret = PTR_ERR(task);
 948         else {
 949                 ret = task->tk_status;
 950                 rpc_put_task(task);
 951         }
 952         return ret;
 953 }
 954
 955 int nfs4_call_sync(struct rpc_clnt *clnt,
 956                    struct nfs_server *server,
 957                    struct rpc_message *msg,
 958                    struct nfs4_sequence_args *args,
 959                    struct nfs4_sequence_res *res,
 960                    int cache_reply)
 961 {
 962         nfs4_init_sequence(args, res, cache_reply);
 963         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 964 }
 965
 966 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 967 {
 968         struct nfs_inode *nfsi = NFS_I(dir);
 969
 970         spin_lock(&dir->i_lock);
 971         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 972         if (!cinfo->atomic || cinfo->before != dir->i_version)
 973                 nfs_force_lookup_revalidate(dir);
 974         dir->i_version = cinfo->after;
 975         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 976         nfs_fscache_invalidate(dir);
 977         spin_unlock(&dir->i_lock);
 978 }
 979
 980 struct nfs4_opendata {
 981         struct kref kref;
 982         struct nfs_openargs o_arg;
 983         struct nfs_openres o_res;
 984         struct nfs_open_confirmargs c_arg;
 985         struct nfs_open_confirmres c_res;
 986         struct nfs4_string owner_name;
 987         struct nfs4_string group_name;
 988         struct nfs4_label *a_label;
 989         struct nfs_fattr f_attr;
 990         struct nfs4_label *f_label;
 991         struct dentry *dir;
 992         struct dentry *dentry;
 993         struct nfs4_state_owner *owner;
 994         struct nfs4_state *state;
 995         struct iattr attrs;
 996         unsigned long timestamp;
 997         unsigned int rpc_done : 1;
 998         unsigned int file_created : 1;
 999         unsigned int is_recover : 1;
1000         int rpc_status;
1001         int cancelled;
1002 };
1003
1004 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1005                 int err, struct nfs4_exception *exception)
1006 {
1007         if (err != -EINVAL)
1008                 return false;
1009         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1010                 return false;
1011         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1012         exception->retry = 1;
1013         return true;
1014 }
1015
1016 static u32
1017 nfs4_map_atomic_open_share(struct nfs_server *server,
1018                 fmode_t fmode, int openflags)
1019 {
1020         u32 res = 0;
1021
1022         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1023         case FMODE_READ:
1024                 res = NFS4_SHARE_ACCESS_READ;
1025                 break;
1026         case FMODE_WRITE:
1027                 res = NFS4_SHARE_ACCESS_WRITE;
1028                 break;
1029         case FMODE_READ|FMODE_WRITE:
1030                 res = NFS4_SHARE_ACCESS_BOTH;
1031         }
1032         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1033                 goto out;
1034         /* Want no delegation if we're using O_DIRECT */
1035         if (openflags & O_DIRECT)
1036                 res |= NFS4_SHARE_WANT_NO_DELEG;
1037 out:
1038         return res;
1039 }
1040
1041 static enum open_claim_type4
1042 nfs4_map_atomic_open_claim(struct nfs_server *server,
1043                 enum open_claim_type4 claim)
1044 {
1045         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1046                 return claim;
1047         switch (claim) {
1048         default:
1049                 return claim;
1050         case NFS4_OPEN_CLAIM_FH:
1051                 return NFS4_OPEN_CLAIM_NULL;
1052         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1053                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1054         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1055                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1056         }
1057 }
1058
1059 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1060 {
1061         p->o_res.f_attr = &p->f_attr;
1062         p->o_res.f_label = p->f_label;
1063         p->o_res.seqid = p->o_arg.seqid;
1064         p->c_res.seqid = p->c_arg.seqid;
1065         p->o_res.server = p->o_arg.server;
1066         p->o_res.access_request = p->o_arg.access;
1067         nfs_fattr_init(&p->f_attr);
1068         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1069 }
1070
1071 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1072                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1073                 const struct iattr *attrs,
1074                 struct nfs4_label *label,
1075                 enum open_claim_type4 claim,
1076                 gfp_t gfp_mask)
1077 {
1078         struct dentry *parent = dget_parent(dentry);
1079         struct inode *dir = d_inode(parent);
1080         struct nfs_server *server = NFS_SERVER(dir);
1081         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1082         struct nfs4_opendata *p;
1083
1084         p = kzalloc(sizeof(*p), gfp_mask);
1085         if (p == NULL)
1086                 goto err;
1087
1088         p->f_label = nfs4_label_alloc(server, gfp_mask);
1089         if (IS_ERR(p->f_label))
1090                 goto err_free_p;
1091
1092         p->a_label = nfs4_label_alloc(server, gfp_mask);
1093         if (IS_ERR(p->a_label))
1094                 goto err_free_f;
1095
1096         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1097         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1098         if (IS_ERR(p->o_arg.seqid))
1099                 goto err_free_label;
1100         nfs_sb_active(dentry->d_sb);
1101         p->dentry = dget(dentry);
1102         p->dir = parent;
1103         p->owner = sp;
1104         atomic_inc(&sp->so_count);
1105         p->o_arg.open_flags = flags;
1106         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1107         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1108                         fmode, flags);
1109         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1110          * will return permission denied for all bits until close */
1111         if (!(flags & O_EXCL)) {
1112                 /* ask server to check for all possible rights as results
1113                  * are cached */
1114                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1115                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1116         }
1117         p->o_arg.clientid = server->nfs_client->cl_clientid;
1118         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1119         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1120         p->o_arg.name = &dentry->d_name;
1121         p->o_arg.server = server;
1122         p->o_arg.bitmask = nfs4_bitmask(server, label);
1123         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1124         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1125         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1126         switch (p->o_arg.claim) {
1127         case NFS4_OPEN_CLAIM_NULL:
1128         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1129         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1130                 p->o_arg.fh = NFS_FH(dir);
1131                 break;
1132         case NFS4_OPEN_CLAIM_PREVIOUS:
1133         case NFS4_OPEN_CLAIM_FH:
1134         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1135         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1136                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1137         }
1138         if (attrs != NULL && attrs->ia_valid != 0) {
1139                 __u32 verf[2];
1140
1141                 p->o_arg.u.attrs = &p->attrs;
1142                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1143
1144                 verf[0] = jiffies;
1145                 verf[1] = current->pid;
1146                 memcpy(p->o_arg.u.verifier.data, verf,
1147                                 sizeof(p->o_arg.u.verifier.data));
1148         }
1149         p->c_arg.fh = &p->o_res.fh;
1150         p->c_arg.stateid = &p->o_res.stateid;
1151         p->c_arg.seqid = p->o_arg.seqid;
1152         nfs4_init_opendata_res(p);
1153         kref_init(&p->kref);
1154         return p;
1155
1156 err_free_label:
1157         nfs4_label_free(p->a_label);
1158 err_free_f:
1159         nfs4_label_free(p->f_label);
1160 err_free_p:
1161         kfree(p);
1162 err:
1163         dput(parent);
1164         return NULL;
1165 }
1166
1167 static void nfs4_opendata_free(struct kref *kref)
1168 {
1169         struct nfs4_opendata *p = container_of(kref,
1170                         struct nfs4_opendata, kref);
1171         struct super_block *sb = p->dentry->d_sb;
1172
1173         nfs_free_seqid(p->o_arg.seqid);
1174         if (p->state != NULL)
1175                 nfs4_put_open_state(p->state);
1176         nfs4_put_state_owner(p->owner);
1177
1178         nfs4_label_free(p->a_label);
1179         nfs4_label_free(p->f_label);
1180
1181         dput(p->dir);
1182         dput(p->dentry);
1183         nfs_sb_deactive(sb);
1184         nfs_fattr_free_names(&p->f_attr);
1185         kfree(p->f_attr.mdsthreshold);
1186         kfree(p);
1187 }
1188
1189 static void nfs4_opendata_put(struct nfs4_opendata *p)
1190 {
1191         if (p != NULL)
1192                 kref_put(&p->kref, nfs4_opendata_free);
1193 }
1194
1195 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1196 {
1197         int ret;
1198
1199         ret = rpc_wait_for_completion_task(task);
1200         return ret;
1201 }
1202
1203 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1204                 fmode_t fmode)
1205 {
1206         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1207         case FMODE_READ|FMODE_WRITE:
1208                 return state->n_rdwr != 0;
1209         case FMODE_WRITE:
1210                 return state->n_wronly != 0;
1211         case FMODE_READ:
1212                 return state->n_rdonly != 0;
1213         }
1214         WARN_ON_ONCE(1);
1215         return false;
1216 }
1217
1218 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1219 {
1220         int ret = 0;
1221
1222         if (open_mode & (O_EXCL|O_TRUNC))
1223                 goto out;
1224         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1225                 case FMODE_READ:
1226                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1227                                 && state->n_rdonly != 0;
1228                         break;
1229                 case FMODE_WRITE:
1230                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1231                                 && state->n_wronly != 0;
1232                         break;
1233                 case FMODE_READ|FMODE_WRITE:
1234                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1235                                 && state->n_rdwr != 0;
1236         }
1237 out:
1238         return ret;
1239 }
1240
1241 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1242                 enum open_claim_type4 claim)
1243 {
1244         if (delegation == NULL)
1245                 return 0;
1246         if ((delegation->type & fmode) != fmode)
1247                 return 0;
1248         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1249                 return 0;
1250         switch (claim) {
1251         case NFS4_OPEN_CLAIM_NULL:
1252         case NFS4_OPEN_CLAIM_FH:
1253                 break;
1254         case NFS4_OPEN_CLAIM_PREVIOUS:
1255                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1256                         break;
1257         default:
1258                 return 0;
1259         }
1260         nfs_mark_delegation_referenced(delegation);
1261         return 1;
1262 }
1263
1264 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1265 {
1266         switch (fmode) {
1267                 case FMODE_WRITE:
1268                         state->n_wronly++;
1269                         break;
1270                 case FMODE_READ:
1271                         state->n_rdonly++;
1272                         break;
1273                 case FMODE_READ|FMODE_WRITE:
1274                         state->n_rdwr++;
1275         }
1276         nfs4_state_set_mode_locked(state, state->state | fmode);
1277 }
1278
1279 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1280 {
1281         struct nfs_client *clp = state->owner->so_server->nfs_client;
1282         bool need_recover = false;
1283
1284         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1285                 need_recover = true;
1286         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1287                 need_recover = true;
1288         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1289                 need_recover = true;
1290         if (need_recover)
1291                 nfs4_state_mark_reclaim_nograce(clp, state);
1292 }
1293
1294 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1295                 nfs4_stateid *stateid)
1296 {
1297         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1298                 return true;
1299         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1300                 nfs_test_and_clear_all_open_stateid(state);
1301                 return true;
1302         }
1303         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1304                 return true;
1305         return false;
1306 }
1307
1308 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1309 {
1310         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1311                 return;
1312         if (state->n_wronly)
1313                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1314         if (state->n_rdonly)
1315                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1316         if (state->n_rdwr)
1317                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1318         set_bit(NFS_OPEN_STATE, &state->flags);
1319 }
1320
1321 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1322                 nfs4_stateid *arg_stateid,
1323                 nfs4_stateid *stateid, fmode_t fmode)
1324 {
1325         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1326         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1327         case FMODE_WRITE:
1328                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1329                 break;
1330         case FMODE_READ:
1331                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1332                 break;
1333         case 0:
1334                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1335                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1336                 clear_bit(NFS_OPEN_STATE, &state->flags);
1337         }
1338         if (stateid == NULL)
1339                 return;
1340         /* Handle races with OPEN */
1341         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1342             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1343             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1344                 nfs_resync_open_stateid_locked(state);
1345                 return;
1346         }
1347         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1348                 nfs4_stateid_copy(&state->stateid, stateid);
1349         nfs4_stateid_copy(&state->open_stateid, stateid);
1350 }
1351
1352 static void nfs_clear_open_stateid(struct nfs4_state *state,
1353         nfs4_stateid *arg_stateid,
1354         nfs4_stateid *stateid, fmode_t fmode)
1355 {
1356         write_seqlock(&state->seqlock);
1357         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1358         write_sequnlock(&state->seqlock);
1359         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1360                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1361 }
1362
1363 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1364 {
1365         switch (fmode) {
1366                 case FMODE_READ:
1367                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1368                         break;
1369                 case FMODE_WRITE:
1370                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1371                         break;
1372                 case FMODE_READ|FMODE_WRITE:
1373                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1374         }
1375         if (!nfs_need_update_open_stateid(state, stateid))
1376                 return;
1377         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1378                 nfs4_stateid_copy(&state->stateid, stateid);
1379         nfs4_stateid_copy(&state->open_stateid, stateid);
1380 }
1381
1382 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1383 {
1384         /*
1385          * Protect the call to nfs4_state_set_mode_locked and
1386          * serialise the stateid update
1387          */
1388         spin_lock(&state->owner->so_lock);
1389         write_seqlock(&state->seqlock);
1390         if (deleg_stateid != NULL) {
1391                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1392                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1393         }
1394         if (open_stateid != NULL)
1395                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1396         write_sequnlock(&state->seqlock);
1397         update_open_stateflags(state, fmode);
1398         spin_unlock(&state->owner->so_lock);
1399 }
1400
1401 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1402 {
1403         struct nfs_inode *nfsi = NFS_I(state->inode);
1404         struct nfs_delegation *deleg_cur;
1405         int ret = 0;
1406
1407         fmode &= (FMODE_READ|FMODE_WRITE);
1408
1409         rcu_read_lock();
1410         deleg_cur = rcu_dereference(nfsi->delegation);
1411         if (deleg_cur == NULL)
1412                 goto no_delegation;
1413
1414         spin_lock(&deleg_cur->lock);
1415         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1416            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1417             (deleg_cur->type & fmode) != fmode)
1418                 goto no_delegation_unlock;
1419
1420         if (delegation == NULL)
1421                 delegation = &deleg_cur->stateid;
1422         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1423                 goto no_delegation_unlock;
1424
1425         nfs_mark_delegation_referenced(deleg_cur);
1426         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1427         ret = 1;
1428 no_delegation_unlock:
1429         spin_unlock(&deleg_cur->lock);
1430 no_delegation:
1431         rcu_read_unlock();
1432
1433         if (!ret && open_stateid != NULL) {
1434                 __update_open_stateid(state, open_stateid, NULL, fmode);
1435                 ret = 1;
1436         }
1437         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1438                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1439
1440         return ret;
1441 }
1442
1443 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1444                 const nfs4_stateid *stateid)
1445 {
1446         struct nfs4_state *state = lsp->ls_state;
1447         bool ret = false;
1448
1449         spin_lock(&state->state_lock);
1450         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1451                 goto out_noupdate;
1452         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1453                 goto out_noupdate;
1454         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1455         ret = true;
1456 out_noupdate:
1457         spin_unlock(&state->state_lock);
1458         return ret;
1459 }
1460
1461 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1462 {
1463         struct nfs_delegation *delegation;
1464
1465         rcu_read_lock();
1466         delegation = rcu_dereference(NFS_I(inode)->delegation);
1467         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1468                 rcu_read_unlock();
1469                 return;
1470         }
1471         rcu_read_unlock();
1472         nfs4_inode_return_delegation(inode);
1473 }
1474
1475 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1476 {
1477         struct nfs4_state *state = opendata->state;
1478         struct nfs_inode *nfsi = NFS_I(state->inode);
1479         struct nfs_delegation *delegation;
1480         int open_mode = opendata->o_arg.open_flags;
1481         fmode_t fmode = opendata->o_arg.fmode;
1482         enum open_claim_type4 claim = opendata->o_arg.claim;
1483         nfs4_stateid stateid;
1484         int ret = -EAGAIN;
1485
1486         for (;;) {
1487                 spin_lock(&state->owner->so_lock);
1488                 if (can_open_cached(state, fmode, open_mode)) {
1489                         update_open_stateflags(state, fmode);
1490                         spin_unlock(&state->owner->so_lock);
1491                         goto out_return_state;
1492                 }
1493                 spin_unlock(&state->owner->so_lock);
1494                 rcu_read_lock();
1495                 delegation = rcu_dereference(nfsi->delegation);
1496                 if (!can_open_delegated(delegation, fmode, claim)) {
1497                         rcu_read_unlock();
1498                         break;
1499                 }
1500                 /* Save the delegation */
1501                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1502                 rcu_read_unlock();
1503                 nfs_release_seqid(opendata->o_arg.seqid);
1504                 if (!opendata->is_recover) {
1505                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1506                         if (ret != 0)
1507                                 goto out;
1508                 }
1509                 ret = -EAGAIN;
1510
1511                 /* Try to update the stateid using the delegation */
1512                 if (update_open_stateid(state, NULL, &stateid, fmode))
1513                         goto out_return_state;
1514         }
1515 out:
1516         return ERR_PTR(ret);
1517 out_return_state:
1518         atomic_inc(&state->count);
1519         return state;
1520 }
1521
1522 static void
1523 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1524 {
1525         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1526         struct nfs_delegation *delegation;
1527         int delegation_flags = 0;
1528
1529         rcu_read_lock();
1530         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1531         if (delegation)
1532                 delegation_flags = delegation->flags;
1533         rcu_read_unlock();
1534         switch (data->o_arg.claim) {
1535         default:
1536                 break;
1537         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1538         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1539                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1540                                    "returning a delegation for "
1541                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1542                                    clp->cl_hostname);
1543                 return;
1544         }
1545         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1546                 nfs_inode_set_delegation(state->inode,
1547                                          data->owner->so_cred,
1548                                          &data->o_res);
1549         else
1550                 nfs_inode_reclaim_delegation(state->inode,
1551                                              data->owner->so_cred,
1552                                              &data->o_res);
1553 }
1554
1555 /*
1556  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1557  * and update the nfs4_state.
1558  */
1559 static struct nfs4_state *
1560 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1561 {
1562         struct inode *inode = data->state->inode;
1563         struct nfs4_state *state = data->state;
1564         int ret;
1565
1566         if (!data->rpc_done) {
1567                 if (data->rpc_status) {
1568                         ret = data->rpc_status;
1569                         goto err;
1570                 }
1571                 /* cached opens have already been processed */
1572                 goto update;
1573         }
1574
1575         ret = nfs_refresh_inode(inode, &data->f_attr);
1576         if (ret)
1577                 goto err;
1578
1579         if (data->o_res.delegation_type != 0)
1580                 nfs4_opendata_check_deleg(data, state);
1581 update:
1582         update_open_stateid(state, &data->o_res.stateid, NULL,
1583                             data->o_arg.fmode);
1584         atomic_inc(&state->count);
1585
1586         return state;
1587 err:
1588         return ERR_PTR(ret);
1589
1590 }
1591
1592 static struct nfs4_state *
1593 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1594 {
1595         struct inode *inode;
1596         struct nfs4_state *state = NULL;
1597         int ret;
1598
1599         if (!data->rpc_done) {
1600                 state = nfs4_try_open_cached(data);
1601                 goto out;
1602         }
1603
1604         ret = -EAGAIN;
1605         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1606                 goto err;
1607         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1608         ret = PTR_ERR(inode);
1609         if (IS_ERR(inode))
1610                 goto err;
1611         ret = -ENOMEM;
1612         state = nfs4_get_open_state(inode, data->owner);
1613         if (state == NULL)
1614                 goto err_put_inode;
1615         if (data->o_res.delegation_type != 0)
1616                 nfs4_opendata_check_deleg(data, state);
1617         update_open_stateid(state, &data->o_res.stateid, NULL,
1618                         data->o_arg.fmode);
1619         iput(inode);
1620 out:
1621         nfs_release_seqid(data->o_arg.seqid);
1622         return state;
1623 err_put_inode:
1624         iput(inode);
1625 err:
1626         return ERR_PTR(ret);
1627 }
1628
1629 static struct nfs4_state *
1630 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1631 {
1632         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1633                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1634         return _nfs4_opendata_to_nfs4_state(data);
1635 }
1636
1637 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1638 {
1639         struct nfs_inode *nfsi = NFS_I(state->inode);
1640         struct nfs_open_context *ctx;
1641
1642         spin_lock(&state->inode->i_lock);
1643         list_for_each_entry(ctx, &nfsi->open_files, list) {
1644                 if (ctx->state != state)
1645                         continue;
1646                 get_nfs_open_context(ctx);
1647                 spin_unlock(&state->inode->i_lock);
1648                 return ctx;
1649         }
1650         spin_unlock(&state->inode->i_lock);
1651         return ERR_PTR(-ENOENT);
1652 }
1653
1654 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1655                 struct nfs4_state *state, enum open_claim_type4 claim)
1656 {
1657         struct nfs4_opendata *opendata;
1658
1659         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1660                         NULL, NULL, claim, GFP_NOFS);
1661         if (opendata == NULL)
1662                 return ERR_PTR(-ENOMEM);
1663         opendata->state = state;
1664         atomic_inc(&state->count);
1665         return opendata;
1666 }
1667
1668 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1669                 fmode_t fmode)
1670 {
1671         struct nfs4_state *newstate;
1672         int ret;
1673
1674         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1675                 return 0;
1676         opendata->o_arg.open_flags = 0;
1677         opendata->o_arg.fmode = fmode;
1678         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1679                         NFS_SB(opendata->dentry->d_sb),
1680                         fmode, 0);
1681         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1682         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1683         nfs4_init_opendata_res(opendata);
1684         ret = _nfs4_recover_proc_open(opendata);
1685         if (ret != 0)
1686                 return ret;
1687         newstate = nfs4_opendata_to_nfs4_state(opendata);
1688         if (IS_ERR(newstate))
1689                 return PTR_ERR(newstate);
1690         if (newstate != opendata->state)
1691                 ret = -ESTALE;
1692         nfs4_close_state(newstate, fmode);
1693         return ret;
1694 }
1695
1696 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1697 {
1698         int ret;
1699
1700         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1701         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1702         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1703         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1704         /* memory barrier prior to reading state->n_* */
1705         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1706         clear_bit(NFS_OPEN_STATE, &state->flags);
1707         smp_rmb();
1708         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1709         if (ret != 0)
1710                 return ret;
1711         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1712         if (ret != 0)
1713                 return ret;
1714         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1715         if (ret != 0)
1716                 return ret;
1717         /*
1718          * We may have performed cached opens for all three recoveries.
1719          * Check if we need to update the current stateid.
1720          */
1721         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1722             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1723                 write_seqlock(&state->seqlock);
1724                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1725                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1726                 write_sequnlock(&state->seqlock);
1727         }
1728         return 0;
1729 }
1730
1731 /*
1732  * OPEN_RECLAIM:
1733  *      reclaim state on the server after a reboot.
1734  */
1735 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1736 {
1737         struct nfs_delegation *delegation;
1738         struct nfs4_opendata *opendata;
1739         fmode_t delegation_type = 0;
1740         int status;
1741
1742         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1743                         NFS4_OPEN_CLAIM_PREVIOUS);
1744         if (IS_ERR(opendata))
1745                 return PTR_ERR(opendata);
1746         rcu_read_lock();
1747         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1748         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1749                 delegation_type = delegation->type;
1750         rcu_read_unlock();
1751         opendata->o_arg.u.delegation_type = delegation_type;
1752         status = nfs4_open_recover(opendata, state);
1753         nfs4_opendata_put(opendata);
1754         return status;
1755 }
1756
1757 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1758 {
1759         struct nfs_server *server = NFS_SERVER(state->inode);
1760         struct nfs4_exception exception = { };
1761         int err;
1762         do {
1763                 err = _nfs4_do_open_reclaim(ctx, state);
1764                 trace_nfs4_open_reclaim(ctx, 0, err);
1765                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1766                         continue;
1767                 if (err != -NFS4ERR_DELAY)
1768                         break;
1769                 nfs4_handle_exception(server, err, &exception);
1770         } while (exception.retry);
1771         return err;
1772 }
1773
1774 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1775 {
1776         struct nfs_open_context *ctx;
1777         int ret;
1778
1779         ctx = nfs4_state_find_open_context(state);
1780         if (IS_ERR(ctx))
1781                 return -EAGAIN;
1782         ret = nfs4_do_open_reclaim(ctx, state);
1783         put_nfs_open_context(ctx);
1784         return ret;
1785 }
1786
1787 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1788 {
1789         switch (err) {
1790                 default:
1791                         printk(KERN_ERR "NFS: %s: unhandled error "
1792                                         "%d.\n", __func__, err);
1793                 case 0:
1794                 case -ENOENT:
1795                 case -EAGAIN:
1796                 case -ESTALE:
1797                         break;
1798                 case -NFS4ERR_BADSESSION:
1799                 case -NFS4ERR_BADSLOT:
1800                 case -NFS4ERR_BAD_HIGH_SLOT:
1801                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1802                 case -NFS4ERR_DEADSESSION:
1803                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1804                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1805                         return -EAGAIN;
1806                 case -NFS4ERR_STALE_CLIENTID:
1807                 case -NFS4ERR_STALE_STATEID:
1808                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1809                 case -NFS4ERR_EXPIRED:
1810                         /* Don't recall a delegation if it was lost */
1811                         nfs4_schedule_lease_recovery(server->nfs_client);
1812                         return -EAGAIN;
1813                 case -NFS4ERR_MOVED:
1814                         nfs4_schedule_migration_recovery(server);
1815                         return -EAGAIN;
1816                 case -NFS4ERR_LEASE_MOVED:
1817                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1818                         return -EAGAIN;
1819                 case -NFS4ERR_DELEG_REVOKED:
1820                 case -NFS4ERR_ADMIN_REVOKED:
1821                 case -NFS4ERR_BAD_STATEID:
1822                 case -NFS4ERR_OPENMODE:
1823                         nfs_inode_find_state_and_recover(state->inode,
1824                                         stateid);
1825                         nfs4_schedule_stateid_recovery(server, state);
1826                         return -EAGAIN;
1827                 case -NFS4ERR_DELAY:
1828                 case -NFS4ERR_GRACE:
1829                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1830                         ssleep(1);
1831                         return -EAGAIN;
1832                 case -ENOMEM:
1833                 case -NFS4ERR_DENIED:
1834                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1835                         return 0;
1836         }
1837         return err;
1838 }
1839
1840 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1841                 struct nfs4_state *state, const nfs4_stateid *stateid,
1842                 fmode_t type)
1843 {
1844         struct nfs_server *server = NFS_SERVER(state->inode);
1845         struct nfs4_opendata *opendata;
1846         int err = 0;
1847
1848         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1849                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1850         if (IS_ERR(opendata))
1851                 return PTR_ERR(opendata);
1852         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1853         write_seqlock(&state->seqlock);
1854         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1855         write_sequnlock(&state->seqlock);
1856         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1857         switch (type & (FMODE_READ|FMODE_WRITE)) {
1858         case FMODE_READ|FMODE_WRITE:
1859         case FMODE_WRITE:
1860                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1861                 if (err)
1862                         break;
1863                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1864                 if (err)
1865                         break;
1866         case FMODE_READ:
1867                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1868         }
1869         nfs4_opendata_put(opendata);
1870         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1871 }
1872
1873 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1874 {
1875         struct nfs4_opendata *data = calldata;
1876
1877         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1878                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1879 }
1880
1881 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1882 {
1883         struct nfs4_opendata *data = calldata;
1884
1885         nfs40_sequence_done(task, &data->c_res.seq_res);
1886
1887         data->rpc_status = task->tk_status;
1888         if (data->rpc_status == 0) {
1889                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1890                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1891                 renew_lease(data->o_res.server, data->timestamp);
1892                 data->rpc_done = 1;
1893         }
1894 }
1895
1896 static void nfs4_open_confirm_release(void *calldata)
1897 {
1898         struct nfs4_opendata *data = calldata;
1899         struct nfs4_state *state = NULL;
1900
1901         /* If this request hasn't been cancelled, do nothing */
1902         if (data->cancelled == 0)
1903                 goto out_free;
1904         /* In case of error, no cleanup! */
1905         if (!data->rpc_done)
1906                 goto out_free;
1907         state = nfs4_opendata_to_nfs4_state(data);
1908         if (!IS_ERR(state))
1909                 nfs4_close_state(state, data->o_arg.fmode);
1910 out_free:
1911         nfs4_opendata_put(data);
1912 }
1913
1914 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1915         .rpc_call_prepare = nfs4_open_confirm_prepare,
1916         .rpc_call_done = nfs4_open_confirm_done,
1917         .rpc_release = nfs4_open_confirm_release,
1918 };
1919
1920 /*
1921  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1922  */
1923 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1924 {
1925         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1926         struct rpc_task *task;
1927         struct  rpc_message msg = {
1928                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1929                 .rpc_argp = &data->c_arg,
1930                 .rpc_resp = &data->c_res,
1931                 .rpc_cred = data->owner->so_cred,
1932         };
1933         struct rpc_task_setup task_setup_data = {
1934                 .rpc_client = server->client,
1935                 .rpc_message = &msg,
1936                 .callback_ops = &nfs4_open_confirm_ops,
1937                 .callback_data = data,
1938                 .workqueue = nfsiod_workqueue,
1939                 .flags = RPC_TASK_ASYNC,
1940         };
1941         int status;
1942
1943         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1944         kref_get(&data->kref);
1945         data->rpc_done = 0;
1946         data->rpc_status = 0;
1947         data->timestamp = jiffies;
1948         if (data->is_recover)
1949                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1950         task = rpc_run_task(&task_setup_data);
1951         if (IS_ERR(task))
1952                 return PTR_ERR(task);
1953         status = nfs4_wait_for_completion_rpc_task(task);
1954         if (status != 0) {
1955                 data->cancelled = 1;
1956                 smp_wmb();
1957         } else
1958                 status = data->rpc_status;
1959         rpc_put_task(task);
1960         return status;
1961 }
1962
1963 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1964 {
1965         struct nfs4_opendata *data = calldata;
1966         struct nfs4_state_owner *sp = data->owner;
1967         struct nfs_client *clp = sp->so_server->nfs_client;
1968         enum open_claim_type4 claim = data->o_arg.claim;
1969
1970         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1971                 goto out_wait;
1972         /*
1973          * Check if we still need to send an OPEN call, or if we can use
1974          * a delegation instead.
1975          */
1976         if (data->state != NULL) {
1977                 struct nfs_delegation *delegation;
1978
1979                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1980                         goto out_no_action;
1981                 rcu_read_lock();
1982                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1983                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1984                         goto unlock_no_action;
1985                 rcu_read_unlock();
1986         }
1987         /* Update client id. */
1988         data->o_arg.clientid = clp->cl_clientid;
1989         switch (claim) {
1990         default:
1991                 break;
1992         case NFS4_OPEN_CLAIM_PREVIOUS:
1993         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1994         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1995                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1996         case NFS4_OPEN_CLAIM_FH:
1997                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1998                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1999         }
2000         data->timestamp = jiffies;
2001         if (nfs4_setup_sequence(data->o_arg.server,
2002                                 &data->o_arg.seq_args,
2003                                 &data->o_res.seq_res,
2004                                 task) != 0)
2005                 nfs_release_seqid(data->o_arg.seqid);
2006
2007         /* Set the create mode (note dependency on the session type) */
2008         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2009         if (data->o_arg.open_flags & O_EXCL) {
2010                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2011                 if (nfs4_has_persistent_session(clp))
2012                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2013                 else if (clp->cl_mvops->minor_version > 0)
2014                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2015         }
2016         return;
2017 unlock_no_action:
2018         rcu_read_unlock();
2019 out_no_action:
2020         task->tk_action = NULL;
2021 out_wait:
2022         nfs4_sequence_done(task, &data->o_res.seq_res);
2023 }
2024
2025 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2026 {
2027         struct nfs4_opendata *data = calldata;
2028
2029         data->rpc_status = task->tk_status;
2030
2031         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2032                 return;
2033
2034         if (task->tk_status == 0) {
2035                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2036                         switch (data->o_res.f_attr->mode & S_IFMT) {
2037                         case S_IFREG:
2038                                 break;
2039                         case S_IFLNK:
2040                                 data->rpc_status = -ELOOP;
2041                                 break;
2042                         case S_IFDIR:
2043                                 data->rpc_status = -EISDIR;
2044                                 break;
2045                         default:
2046                                 data->rpc_status = -ENOTDIR;
2047                         }
2048                 }
2049                 renew_lease(data->o_res.server, data->timestamp);
2050                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2051                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2052         }
2053         data->rpc_done = 1;
2054 }
2055
2056 static void nfs4_open_release(void *calldata)
2057 {
2058         struct nfs4_opendata *data = calldata;
2059         struct nfs4_state *state = NULL;
2060
2061         /* If this request hasn't been cancelled, do nothing */
2062         if (data->cancelled == 0)
2063                 goto out_free;
2064         /* In case of error, no cleanup! */
2065         if (data->rpc_status != 0 || !data->rpc_done)
2066                 goto out_free;
2067         /* In case we need an open_confirm, no cleanup! */
2068         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2069                 goto out_free;
2070         state = nfs4_opendata_to_nfs4_state(data);
2071         if (!IS_ERR(state))
2072                 nfs4_close_state(state, data->o_arg.fmode);
2073 out_free:
2074         nfs4_opendata_put(data);
2075 }
2076
2077 static const struct rpc_call_ops nfs4_open_ops = {
2078         .rpc_call_prepare = nfs4_open_prepare,
2079         .rpc_call_done = nfs4_open_done,
2080         .rpc_release = nfs4_open_release,
2081 };
2082
2083 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2084 {
2085         struct inode *dir = d_inode(data->dir);
2086         struct nfs_server *server = NFS_SERVER(dir);
2087         struct nfs_openargs *o_arg = &data->o_arg;
2088         struct nfs_openres *o_res = &data->o_res;
2089         struct rpc_task *task;
2090         struct rpc_message msg = {
2091                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2092                 .rpc_argp = o_arg,
2093                 .rpc_resp = o_res,
2094                 .rpc_cred = data->owner->so_cred,
2095         };
2096         struct rpc_task_setup task_setup_data = {
2097                 .rpc_client = server->client,
2098                 .rpc_message = &msg,
2099                 .callback_ops = &nfs4_open_ops,
2100                 .callback_data = data,
2101                 .workqueue = nfsiod_workqueue,
2102                 .flags = RPC_TASK_ASYNC,
2103         };
2104         int status;
2105
2106         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2107         kref_get(&data->kref);
2108         data->rpc_done = 0;
2109         data->rpc_status = 0;
2110         data->cancelled = 0;
2111         data->is_recover = 0;
2112         if (isrecover) {
2113                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2114                 data->is_recover = 1;
2115         }
2116         task = rpc_run_task(&task_setup_data);
2117         if (IS_ERR(task))
2118                 return PTR_ERR(task);
2119         status = nfs4_wait_for_completion_rpc_task(task);
2120         if (status != 0) {
2121                 data->cancelled = 1;
2122                 smp_wmb();
2123         } else
2124                 status = data->rpc_status;
2125         rpc_put_task(task);
2126
2127         return status;
2128 }
2129
2130 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2131 {
2132         struct inode *dir = d_inode(data->dir);
2133         struct nfs_openres *o_res = &data->o_res;
2134         int status;
2135
2136         status = nfs4_run_open_task(data, 1);
2137         if (status != 0 || !data->rpc_done)
2138                 return status;
2139
2140         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2141
2142         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2143                 status = _nfs4_proc_open_confirm(data);
2144                 if (status != 0)
2145                         return status;
2146         }
2147
2148         return status;
2149 }
2150
2151 /*
2152  * Additional permission checks in order to distinguish between an
2153  * open for read, and an open for execute. This works around the
2154  * fact that NFSv4 OPEN treats read and execute permissions as being
2155  * the same.
2156  * Note that in the non-execute case, we want to turn off permission
2157  * checking if we just created a new file (POSIX open() semantics).
2158  */
2159 static int nfs4_opendata_access(struct rpc_cred *cred,
2160                                 struct nfs4_opendata *opendata,
2161                                 struct nfs4_state *state, fmode_t fmode,
2162                                 int openflags)
2163 {
2164         struct nfs_access_entry cache;
2165         u32 mask;
2166
2167         /* access call failed or for some reason the server doesn't
2168          * support any access modes -- defer access call until later */
2169         if (opendata->o_res.access_supported == 0)
2170                 return 0;
2171
2172         mask = 0;
2173         /*
2174          * Use openflags to check for exec, because fmode won't
2175          * always have FMODE_EXEC set when file open for exec.
2176          */
2177         if (openflags & __FMODE_EXEC) {
2178                 /* ONLY check for exec rights */
2179                 mask = MAY_EXEC;
2180         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2181                 mask = MAY_READ;
2182
2183         cache.cred = cred;
2184         cache.jiffies = jiffies;
2185         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2186         nfs_access_add_cache(state->inode, &cache);
2187
2188         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2189                 return 0;
2190
2191         /* even though OPEN succeeded, access is denied. Close the file */
2192         nfs4_close_state(state, fmode);
2193         return -EACCES;
2194 }
2195
2196 /*
2197  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2198  */
2199 static int _nfs4_proc_open(struct nfs4_opendata *data)
2200 {
2201         struct inode *dir = d_inode(data->dir);
2202         struct nfs_server *server = NFS_SERVER(dir);
2203         struct nfs_openargs *o_arg = &data->o_arg;
2204         struct nfs_openres *o_res = &data->o_res;
2205         int status;
2206
2207         status = nfs4_run_open_task(data, 0);
2208         if (!data->rpc_done)
2209                 return status;
2210         if (status != 0) {
2211                 if (status == -NFS4ERR_BADNAME &&
2212                                 !(o_arg->open_flags & O_CREAT))
2213                         return -ENOENT;
2214                 return status;
2215         }
2216
2217         nfs_fattr_map_and_free_names(server, &data->f_attr);
2218
2219         if (o_arg->open_flags & O_CREAT) {
2220                 update_changeattr(dir, &o_res->cinfo);
2221                 if (o_arg->open_flags & O_EXCL)
2222                         data->file_created = 1;
2223                 else if (o_res->cinfo.before != o_res->cinfo.after)
2224                         data->file_created = 1;
2225         }
2226         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2227                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2228         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2229                 status = _nfs4_proc_open_confirm(data);
2230                 if (status != 0)
2231                         return status;
2232         }
2233         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2234                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2235         return 0;
2236 }
2237
2238 static int nfs4_recover_expired_lease(struct nfs_server *server)
2239 {
2240         return nfs4_client_recover_expired_lease(server->nfs_client);
2241 }
2242
2243 /*
2244  * OPEN_EXPIRED:
2245  *      reclaim state on the server after a network partition.
2246  *      Assumes caller holds the appropriate lock
2247  */
2248 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2249 {
2250         struct nfs4_opendata *opendata;
2251         int ret;
2252
2253         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2254                         NFS4_OPEN_CLAIM_FH);
2255         if (IS_ERR(opendata))
2256                 return PTR_ERR(opendata);
2257         ret = nfs4_open_recover(opendata, state);
2258         if (ret == -ESTALE)
2259                 d_drop(ctx->dentry);
2260         nfs4_opendata_put(opendata);
2261         return ret;
2262 }
2263
2264 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2265 {
2266         struct nfs_server *server = NFS_SERVER(state->inode);
2267         struct nfs4_exception exception = { };
2268         int err;
2269
2270         do {
2271                 err = _nfs4_open_expired(ctx, state);
2272                 trace_nfs4_open_expired(ctx, 0, err);
2273                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2274                         continue;
2275                 switch (err) {
2276                 default:
2277                         goto out;
2278                 case -NFS4ERR_GRACE:
2279                 case -NFS4ERR_DELAY:
2280                         nfs4_handle_exception(server, err, &exception);
2281                         err = 0;
2282                 }
2283         } while (exception.retry);
2284 out:
2285         return err;
2286 }
2287
2288 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2289 {
2290         struct nfs_open_context *ctx;
2291         int ret;
2292
2293         ctx = nfs4_state_find_open_context(state);
2294         if (IS_ERR(ctx))
2295                 return -EAGAIN;
2296         ret = nfs4_do_open_expired(ctx, state);
2297         put_nfs_open_context(ctx);
2298         return ret;
2299 }
2300
2301 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2302 {
2303         nfs_remove_bad_delegation(state->inode);
2304         write_seqlock(&state->seqlock);
2305         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2306         write_sequnlock(&state->seqlock);
2307         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2308 }
2309
2310 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2311 {
2312         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2313                 nfs_finish_clear_delegation_stateid(state);
2314 }
2315
2316 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2317 {
2318         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2319         nfs40_clear_delegation_stateid(state);
2320         return nfs4_open_expired(sp, state);
2321 }
2322
2323 #if defined(CONFIG_NFS_V4_1)
2324 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2325 {
2326         struct nfs_server *server = NFS_SERVER(state->inode);
2327         nfs4_stateid stateid;
2328         struct nfs_delegation *delegation;
2329         struct rpc_cred *cred;
2330         int status;
2331
2332         /* Get the delegation credential for use by test/free_stateid */
2333         rcu_read_lock();
2334         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2335         if (delegation == NULL) {
2336                 rcu_read_unlock();
2337                 return;
2338         }
2339
2340         nfs4_stateid_copy(&stateid, &delegation->stateid);
2341         cred = get_rpccred(delegation->cred);
2342         rcu_read_unlock();
2343         status = nfs41_test_stateid(server, &stateid, cred);
2344         trace_nfs4_test_delegation_stateid(state, NULL, status);
2345
2346         if (status != NFS_OK) {
2347                 /* Free the stateid unless the server explicitly
2348                  * informs us the stateid is unrecognized. */
2349                 if (status != -NFS4ERR_BAD_STATEID)
2350                         nfs41_free_stateid(server, &stateid, cred);
2351                 nfs_finish_clear_delegation_stateid(state);
2352         }
2353
2354         put_rpccred(cred);
2355 }
2356
2357 /**
2358  * nfs41_check_open_stateid - possibly free an open stateid
2359  *
2360  * @state: NFSv4 state for an inode
2361  *
2362  * Returns NFS_OK if recovery for this stateid is now finished.
2363  * Otherwise a negative NFS4ERR value is returned.
2364  */
2365 static int nfs41_check_open_stateid(struct nfs4_state *state)
2366 {
2367         struct nfs_server *server = NFS_SERVER(state->inode);
2368         nfs4_stateid *stateid = &state->open_stateid;
2369         struct rpc_cred *cred = state->owner->so_cred;
2370         int status;
2371
2372         /* If a state reset has been done, test_stateid is unneeded */
2373         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2374             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2375             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2376                 return -NFS4ERR_BAD_STATEID;
2377
2378         status = nfs41_test_stateid(server, stateid, cred);
2379         trace_nfs4_test_open_stateid(state, NULL, status);
2380         if (status != NFS_OK) {
2381                 /* Free the stateid unless the server explicitly
2382                  * informs us the stateid is unrecognized. */
2383                 if (status != -NFS4ERR_BAD_STATEID)
2384                         nfs41_free_stateid(server, stateid, cred);
2385
2386                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2387                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2388                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2389                 clear_bit(NFS_OPEN_STATE, &state->flags);
2390         }
2391         return status;
2392 }
2393
2394 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2395 {
2396         int status;
2397
2398         nfs41_check_delegation_stateid(state);
2399         status = nfs41_check_open_stateid(state);
2400         if (status != NFS_OK)
2401                 status = nfs4_open_expired(sp, state);
2402         return status;
2403 }
2404 #endif
2405
2406 /*
2407  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2408  * fields corresponding to attributes that were used to store the verifier.
2409  * Make sure we clobber those fields in the later setattr call
2410  */
2411 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2412                                 struct iattr *sattr, struct nfs4_label **label)
2413 {
2414         const u32 *attrset = opendata->o_res.attrset;
2415
2416         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2417             !(sattr->ia_valid & ATTR_ATIME_SET))
2418                 sattr->ia_valid |= ATTR_ATIME;
2419
2420         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2421             !(sattr->ia_valid & ATTR_MTIME_SET))
2422                 sattr->ia_valid |= ATTR_MTIME;
2423
2424         /* Except MODE, it seems harmless of setting twice. */
2425         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2426                 attrset[1] & FATTR4_WORD1_MODE)
2427                 sattr->ia_valid &= ~ATTR_MODE;
2428
2429         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2430                 *label = NULL;
2431 }
2432
2433 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2434                 fmode_t fmode,
2435                 int flags,
2436                 struct nfs_open_context *ctx)
2437 {
2438         struct nfs4_state_owner *sp = opendata->owner;
2439         struct nfs_server *server = sp->so_server;
2440         struct dentry *dentry;
2441         struct nfs4_state *state;
2442         unsigned int seq;
2443         int ret;
2444
2445         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2446
2447         ret = _nfs4_proc_open(opendata);
2448         if (ret != 0)
2449                 goto out;
2450
2451         state = nfs4_opendata_to_nfs4_state(opendata);
2452         ret = PTR_ERR(state);
2453         if (IS_ERR(state))
2454                 goto out;
2455         ctx->state = state;
2456         if (server->caps & NFS_CAP_POSIX_LOCK)
2457                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2458
2459         dentry = opendata->dentry;
2460         if (d_really_is_negative(dentry)) {
2461                 /* FIXME: Is this d_drop() ever needed? */
2462                 d_drop(dentry);
2463                 dentry = d_add_unique(dentry, igrab(state->inode));
2464                 if (dentry == NULL) {
2465                         dentry = opendata->dentry;
2466                 } else {
2467                         dput(ctx->dentry);
2468                         ctx->dentry = dentry;
2469                 }
2470                 nfs_set_verifier(dentry,
2471                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2472         }
2473
2474         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2475         if (ret != 0)
2476                 goto out;
2477
2478         if (d_inode(dentry) == state->inode) {
2479                 nfs_inode_attach_open_context(ctx);
2480                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2481                         nfs4_schedule_stateid_recovery(server, state);
2482         }
2483 out:
2484         return ret;
2485 }
2486
2487 /*
2488  * Returns a referenced nfs4_state
2489  */
2490 static int _nfs4_do_open(struct inode *dir,
2491                         struct nfs_open_context *ctx,
2492                         int flags,
2493                         struct iattr *sattr,
2494                         struct nfs4_label *label,
2495                         int *opened)
2496 {
2497         struct nfs4_state_owner  *sp;
2498         struct nfs4_state     *state = NULL;
2499         struct nfs_server       *server = NFS_SERVER(dir);
2500         struct nfs4_opendata *opendata;
2501         struct dentry *dentry = ctx->dentry;
2502         struct rpc_cred *cred = ctx->cred;
2503         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2504         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2505         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2506         struct nfs4_label *olabel = NULL;
2507         int status;
2508
2509         /* Protect against reboot recovery conflicts */
2510         status = -ENOMEM;
2511         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2512         if (sp == NULL) {
2513                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2514                 goto out_err;
2515         }
2516         status = nfs4_recover_expired_lease(server);
2517         if (status != 0)
2518                 goto err_put_state_owner;
2519         if (d_really_is_positive(dentry))
2520                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2521         status = -ENOMEM;
2522         if (d_really_is_positive(dentry))
2523                 claim = NFS4_OPEN_CLAIM_FH;
2524         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2525                         label, claim, GFP_KERNEL);
2526         if (opendata == NULL)
2527                 goto err_put_state_owner;
2528
2529         if (label) {
2530                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2531                 if (IS_ERR(olabel)) {
2532                         status = PTR_ERR(olabel);
2533                         goto err_opendata_put;
2534                 }
2535         }
2536
2537         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2538                 if (!opendata->f_attr.mdsthreshold) {
2539                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2540                         if (!opendata->f_attr.mdsthreshold)
2541                                 goto err_free_label;
2542                 }
2543                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2544         }
2545         if (d_really_is_positive(dentry))
2546                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2547
2548         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2549         if (status != 0)
2550                 goto err_free_label;
2551         state = ctx->state;
2552
2553         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2554             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2555                 nfs4_exclusive_attrset(opendata, sattr, &label);
2556
2557                 nfs_fattr_init(opendata->o_res.f_attr);
2558                 status = nfs4_do_setattr(state->inode, cred,
2559                                 opendata->o_res.f_attr, sattr,
2560                                 state, label, olabel);
2561                 if (status == 0) {
2562                         nfs_setattr_update_inode(state->inode, sattr,
2563                                         opendata->o_res.f_attr);
2564                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2565                 }
2566         }
2567         if (opened && opendata->file_created)
2568                 *opened |= FILE_CREATED;
2569
2570         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2571                 *ctx_th = opendata->f_attr.mdsthreshold;
2572                 opendata->f_attr.mdsthreshold = NULL;
2573         }
2574
2575         nfs4_label_free(olabel);
2576
2577         nfs4_opendata_put(opendata);
2578         nfs4_put_state_owner(sp);
2579         return 0;
2580 err_free_label:
2581         nfs4_label_free(olabel);
2582 err_opendata_put:
2583         nfs4_opendata_put(opendata);
2584 err_put_state_owner:
2585         nfs4_put_state_owner(sp);
2586 out_err:
2587         return status;
2588 }
2589
2590
2591 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2592                                         struct nfs_open_context *ctx,
2593                                         int flags,
2594                                         struct iattr *sattr,
2595                                         struct nfs4_label *label,
2596                                         int *opened)
2597 {
2598         struct nfs_server *server = NFS_SERVER(dir);
2599         struct nfs4_exception exception = { };
2600         struct nfs4_state *res;
2601         int status;
2602
2603         do {
2604                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2605                 res = ctx->state;
2606                 trace_nfs4_open_file(ctx, flags, status);
2607                 if (status == 0)
2608                         break;
2609                 /* NOTE: BAD_SEQID means the server and client disagree about the
2610                  * book-keeping w.r.t. state-changing operations
2611                  * (OPEN/CLOSE/LOCK/LOCKU...)
2612                  * It is actually a sign of a bug on the client or on the server.
2613                  *
2614                  * If we receive a BAD_SEQID error in the particular case of
2615                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2616                  * have unhashed the old state_owner for us, and that we can
2617                  * therefore safely retry using a new one. We should still warn
2618                  * the user though...
2619                  */
2620                 if (status == -NFS4ERR_BAD_SEQID) {
2621                         pr_warn_ratelimited("NFS: v4 server %s "
2622                                         " returned a bad sequence-id error!\n",
2623                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2624                         exception.retry = 1;
2625                         continue;
2626                 }
2627                 /*
2628                  * BAD_STATEID on OPEN means that the server cancelled our
2629                  * state before it received the OPEN_CONFIRM.
2630                  * Recover by retrying the request as per the discussion
2631                  * on Page 181 of RFC3530.
2632                  */
2633                 if (status == -NFS4ERR_BAD_STATEID) {
2634                         exception.retry = 1;
2635                         continue;
2636                 }
2637                 if (status == -EAGAIN) {
2638                         /* We must have found a delegation */
2639                         exception.retry = 1;
2640                         continue;
2641                 }
2642                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2643                         continue;
2644                 res = ERR_PTR(nfs4_handle_exception(server,
2645                                         status, &exception));
2646         } while (exception.retry);
2647         return res;
2648 }
2649
2650 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2651                             struct nfs_fattr *fattr, struct iattr *sattr,
2652                             struct nfs4_state *state, struct nfs4_label *ilabel,
2653                             struct nfs4_label *olabel)
2654 {
2655         struct nfs_server *server = NFS_SERVER(inode);
2656         struct nfs_setattrargs  arg = {
2657                 .fh             = NFS_FH(inode),
2658                 .iap            = sattr,
2659                 .server         = server,
2660                 .bitmask = server->attr_bitmask,
2661                 .label          = ilabel,
2662         };
2663         struct nfs_setattrres  res = {
2664                 .fattr          = fattr,
2665                 .label          = olabel,
2666                 .server         = server,
2667         };
2668         struct rpc_message msg = {
2669                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2670                 .rpc_argp       = &arg,
2671                 .rpc_resp       = &res,
2672                 .rpc_cred       = cred,
2673         };
2674         unsigned long timestamp = jiffies;
2675         fmode_t fmode;
2676         bool truncate;
2677         int status;
2678
2679         arg.bitmask = nfs4_bitmask(server, ilabel);
2680         if (ilabel)
2681                 arg.bitmask = nfs4_bitmask(server, olabel);
2682
2683         nfs_fattr_init(fattr);
2684
2685         /* Servers should only apply open mode checks for file size changes */
2686         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2687         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2688
2689         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2690                 /* Use that stateid */
2691         } else if (truncate && state != NULL) {
2692                 struct nfs_lockowner lockowner = {
2693                         .l_owner = current->files,
2694                         .l_pid = current->tgid,
2695                 };
2696                 if (!nfs4_valid_open_stateid(state))
2697                         return -EBADF;
2698                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2699                                 &lockowner) == -EIO)
2700                         return -EBADF;
2701         } else
2702                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2703
2704         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2705         if (status == 0 && state != NULL)
2706                 renew_lease(server, timestamp);
2707         return status;
2708 }
2709
2710 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2711                            struct nfs_fattr *fattr, struct iattr *sattr,
2712                            struct nfs4_state *state, struct nfs4_label *ilabel,
2713                            struct nfs4_label *olabel)
2714 {
2715         struct nfs_server *server = NFS_SERVER(inode);
2716         struct nfs4_exception exception = {
2717                 .state = state,
2718                 .inode = inode,
2719         };
2720         int err;
2721         do {
2722                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2723                 trace_nfs4_setattr(inode, err);
2724                 switch (err) {
2725                 case -NFS4ERR_OPENMODE:
2726                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2727                                 pr_warn_once("NFSv4: server %s is incorrectly "
2728                                                 "applying open mode checks to "
2729                                                 "a SETATTR that is not "
2730                                                 "changing file size.\n",
2731                                                 server->nfs_client->cl_hostname);
2732                         }
2733                         if (state && !(state->state & FMODE_WRITE)) {
2734                                 err = -EBADF;
2735                                 if (sattr->ia_valid & ATTR_OPEN)
2736                                         err = -EACCES;
2737                                 goto out;
2738                         }
2739                 }
2740                 err = nfs4_handle_exception(server, err, &exception);
2741         } while (exception.retry);
2742 out:
2743         return err;
2744 }
2745
2746 static bool
2747 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2748 {
2749         if (inode == NULL || !nfs_have_layout(inode))
2750                 return false;
2751
2752         return pnfs_wait_on_layoutreturn(inode, task);
2753 }
2754
2755 struct nfs4_closedata {
2756         struct inode *inode;
2757         struct nfs4_state *state;
2758         struct nfs_closeargs arg;
2759         struct nfs_closeres res;
2760         struct nfs_fattr fattr;
2761         unsigned long timestamp;
2762         bool roc;
2763         u32 roc_barrier;
2764 };
2765
2766 static void nfs4_free_closedata(void *data)
2767 {
2768         struct nfs4_closedata *calldata = data;
2769         struct nfs4_state_owner *sp = calldata->state->owner;
2770         struct super_block *sb = calldata->state->inode->i_sb;
2771
2772         if (calldata->roc)
2773                 pnfs_roc_release(calldata->state->inode);
2774         nfs4_put_open_state(calldata->state);
2775         nfs_free_seqid(calldata->arg.seqid);
2776         nfs4_put_state_owner(sp);
2777         nfs_sb_deactive(sb);
2778         kfree(calldata);
2779 }
2780
2781 static void nfs4_close_done(struct rpc_task *task, void *data)
2782 {
2783         struct nfs4_closedata *calldata = data;
2784         struct nfs4_state *state = calldata->state;
2785         struct nfs_server *server = NFS_SERVER(calldata->inode);
2786         nfs4_stateid *res_stateid = NULL;
2787
2788         dprintk("%s: begin!\n", __func__);
2789         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2790                 return;
2791         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2792         /* hmm. we are done with the inode, and in the process of freeing
2793          * the state_owner. we keep this around to process errors
2794          */
2795         switch (task->tk_status) {
2796                 case 0:
2797                         res_stateid = &calldata->res.stateid;
2798                         if (calldata->roc)
2799                                 pnfs_roc_set_barrier(state->inode,
2800                                                      calldata->roc_barrier);
2801                         renew_lease(server, calldata->timestamp);
2802                         break;
2803                 case -NFS4ERR_ADMIN_REVOKED:
2804                 case -NFS4ERR_STALE_STATEID:
2805                 case -NFS4ERR_OLD_STATEID:
2806                 case -NFS4ERR_BAD_STATEID:
2807                 case -NFS4ERR_EXPIRED:
2808                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2809                                                 &state->open_stateid)) {
2810                                 rpc_restart_call_prepare(task);
2811                                 goto out_release;
2812                         }
2813                         if (calldata->arg.fmode == 0)
2814                                 break;
2815                 default:
2816                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2817                                 rpc_restart_call_prepare(task);
2818                                 goto out_release;
2819                         }
2820         }
2821         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2822                         res_stateid, calldata->arg.fmode);
2823 out_release:
2824         nfs_release_seqid(calldata->arg.seqid);
2825         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2826         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2827 }
2828
2829 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2830 {
2831         struct nfs4_closedata *calldata = data;
2832         struct nfs4_state *state = calldata->state;
2833         struct inode *inode = calldata->inode;
2834         bool is_rdonly, is_wronly, is_rdwr;
2835         int call_close = 0;
2836
2837         dprintk("%s: begin!\n", __func__);
2838         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2839                 goto out_wait;
2840
2841         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2842         spin_lock(&state->owner->so_lock);
2843         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2844         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2845         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2846         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2847         /* Calculate the change in open mode */
2848         calldata->arg.fmode = 0;
2849         if (state->n_rdwr == 0) {
2850                 if (state->n_rdonly == 0)
2851                         call_close |= is_rdonly;
2852                 else if (is_rdonly)
2853                         calldata->arg.fmode |= FMODE_READ;
2854                 if (state->n_wronly == 0)
2855                         call_close |= is_wronly;
2856                 else if (is_wronly)
2857                         calldata->arg.fmode |= FMODE_WRITE;
2858                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2859                         call_close |= is_rdwr;
2860         } else if (is_rdwr)
2861                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2862
2863         if (!nfs4_valid_open_stateid(state))
2864                 call_close = 0;
2865         spin_unlock(&state->owner->so_lock);
2866
2867         if (!call_close) {
2868                 /* Note: exit _without_ calling nfs4_close_done */
2869                 goto out_no_action;
2870         }
2871
2872         if (nfs4_wait_on_layoutreturn(inode, task)) {
2873                 nfs_release_seqid(calldata->arg.seqid);
2874                 goto out_wait;
2875         }
2876
2877         if (calldata->arg.fmode == 0)
2878                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2879         if (calldata->roc)
2880                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2881
2882         calldata->arg.share_access =
2883                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2884                                 calldata->arg.fmode, 0);
2885
2886         nfs_fattr_init(calldata->res.fattr);
2887         calldata->timestamp = jiffies;
2888         if (nfs4_setup_sequence(NFS_SERVER(inode),
2889                                 &calldata->arg.seq_args,
2890                                 &calldata->res.seq_res,
2891                                 task) != 0)
2892                 nfs_release_seqid(calldata->arg.seqid);
2893         dprintk("%s: done!\n", __func__);
2894         return;
2895 out_no_action:
2896         task->tk_action = NULL;
2897 out_wait:
2898         nfs4_sequence_done(task, &calldata->res.seq_res);
2899 }
2900
2901 static const struct rpc_call_ops nfs4_close_ops = {
2902         .rpc_call_prepare = nfs4_close_prepare,
2903         .rpc_call_done = nfs4_close_done,
2904         .rpc_release = nfs4_free_closedata,
2905 };
2906
2907 static bool nfs4_roc(struct inode *inode)
2908 {
2909         if (!nfs_have_layout(inode))
2910                 return false;
2911         return pnfs_roc(inode);
2912 }
2913
2914 /*
2915  * It is possible for data to be read/written from a mem-mapped file
2916  * after the sys_close call (which hits the vfs layer as a flush).
2917  * This means that we can't safely call nfsv4 close on a file until
2918  * the inode is cleared. This in turn means that we are not good
2919  * NFSv4 citizens - we do not indicate to the server to update the file's
2920  * share state even when we are done with one of the three share
2921  * stateid's in the inode.
2922  *
2923  * NOTE: Caller must be holding the sp->so_owner semaphore!
2924  */
2925 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2926 {
2927         struct nfs_server *server = NFS_SERVER(state->inode);
2928         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2929         struct nfs4_closedata *calldata;
2930         struct nfs4_state_owner *sp = state->owner;
2931         struct rpc_task *task;
2932         struct rpc_message msg = {
2933                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2934                 .rpc_cred = state->owner->so_cred,
2935         };
2936         struct rpc_task_setup task_setup_data = {
2937                 .rpc_client = server->client,
2938                 .rpc_message = &msg,
2939                 .callback_ops = &nfs4_close_ops,
2940                 .workqueue = nfsiod_workqueue,
2941                 .flags = RPC_TASK_ASYNC,
2942         };
2943         int status = -ENOMEM;
2944
2945         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2946                 &task_setup_data.rpc_client, &msg);
2947
2948         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2949         if (calldata == NULL)
2950                 goto out;
2951         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2952         calldata->inode = state->inode;
2953         calldata->state = state;
2954         calldata->arg.fh = NFS_FH(state->inode);
2955         /* Serialization for the sequence id */
2956         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2957         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2958         if (IS_ERR(calldata->arg.seqid))
2959                 goto out_free_calldata;
2960         calldata->arg.fmode = 0;
2961         calldata->arg.bitmask = server->cache_consistency_bitmask;
2962         calldata->res.fattr = &calldata->fattr;
2963         calldata->res.seqid = calldata->arg.seqid;
2964         calldata->res.server = server;
2965         calldata->roc = nfs4_roc(state->inode);
2966         nfs_sb_active(calldata->inode->i_sb);
2967
2968         msg.rpc_argp = &calldata->arg;
2969         msg.rpc_resp = &calldata->res;
2970         task_setup_data.callback_data = calldata;
2971         task = rpc_run_task(&task_setup_data);
2972         if (IS_ERR(task))
2973                 return PTR_ERR(task);
2974         status = 0;
2975         if (wait)
2976                 status = rpc_wait_for_completion_task(task);
2977         rpc_put_task(task);
2978         return status;
2979 out_free_calldata:
2980         kfree(calldata);
2981 out:
2982         nfs4_put_open_state(state);
2983         nfs4_put_state_owner(sp);
2984         return status;
2985 }
2986
2987 static struct inode *
2988 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2989                 int open_flags, struct iattr *attr, int *opened)
2990 {
2991         struct nfs4_state *state;
2992         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2993
2994         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2995
2996         /* Protect against concurrent sillydeletes */
2997         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2998
2999         nfs4_label_release_security(label);
3000
3001         if (IS_ERR(state))
3002                 return ERR_CAST(state);
3003         return state->inode;
3004 }
3005
3006 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3007 {
3008         if (ctx->state == NULL)
3009                 return;
3010         if (is_sync)
3011                 nfs4_close_sync(ctx->state, ctx->mode);
3012         else
3013                 nfs4_close_state(ctx->state, ctx->mode);
3014 }
3015
3016 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3017 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3018 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3019
3020 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3021 {
3022         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3023         struct nfs4_server_caps_arg args = {
3024                 .fhandle = fhandle,
3025                 .bitmask = bitmask,
3026         };
3027         struct nfs4_server_caps_res res = {};
3028         struct rpc_message msg = {
3029                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3030                 .rpc_argp = &args,
3031                 .rpc_resp = &res,
3032         };
3033         int status;
3034
3035         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3036                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3037                      FATTR4_WORD0_LINK_SUPPORT |
3038                      FATTR4_WORD0_SYMLINK_SUPPORT |
3039                      FATTR4_WORD0_ACLSUPPORT;
3040         if (minorversion)
3041                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3042
3043         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3044         if (status == 0) {
3045                 /* Sanity check the server answers */
3046                 switch (minorversion) {
3047                 case 0:
3048                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3049                         res.attr_bitmask[2] = 0;
3050                         break;
3051                 case 1:
3052                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3053                         break;
3054                 case 2:
3055                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3056                 }
3057                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3058                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3059                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3060                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3061                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3062                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3063                                 NFS_CAP_SECURITY_LABEL);
3064                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3065                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3066                         server->caps |= NFS_CAP_ACLS;
3067                 if (res.has_links != 0)
3068                         server->caps |= NFS_CAP_HARDLINKS;
3069                 if (res.has_symlinks != 0)
3070                         server->caps |= NFS_CAP_SYMLINKS;
3071                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3072                         server->caps |= NFS_CAP_FILEID;
3073                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3074                         server->caps |= NFS_CAP_MODE;
3075                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3076                         server->caps |= NFS_CAP_NLINK;
3077                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3078                         server->caps |= NFS_CAP_OWNER;
3079                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3080                         server->caps |= NFS_CAP_OWNER_GROUP;
3081                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3082                         server->caps |= NFS_CAP_ATIME;
3083                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3084                         server->caps |= NFS_CAP_CTIME;
3085                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3086                         server->caps |= NFS_CAP_MTIME;
3087 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3088                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3089                         server->caps |= NFS_CAP_SECURITY_LABEL;
3090 #endif
3091                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3092                                 sizeof(server->attr_bitmask));
3093                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3094
3095                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3096                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3097                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3098                 server->cache_consistency_bitmask[2] = 0;
3099                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3100                         sizeof(server->exclcreat_bitmask));
3101                 server->acl_bitmask = res.acl_bitmask;
3102                 server->fh_expire_type = res.fh_expire_type;
3103         }
3104
3105         return status;
3106 }
3107
3108 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3109 {
3110         struct nfs4_exception exception = { };
3111         int err;
3112         do {
3113                 err = nfs4_handle_exception(server,
3114                                 _nfs4_server_capabilities(server, fhandle),
3115                                 &exception);
3116         } while (exception.retry);
3117         return err;
3118 }
3119
3120 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3121                 struct nfs_fsinfo *info)
3122 {
3123         u32 bitmask[3];
3124         struct nfs4_lookup_root_arg args = {
3125                 .bitmask = bitmask,
3126         };
3127         struct nfs4_lookup_res res = {
3128                 .server = server,
3129                 .fattr = info->fattr,
3130                 .fh = fhandle,
3131         };
3132         struct rpc_message msg = {
3133                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3134                 .rpc_argp = &args,
3135                 .rpc_resp = &res,
3136         };
3137
3138         bitmask[0] = nfs4_fattr_bitmap[0];
3139         bitmask[1] = nfs4_fattr_bitmap[1];
3140         /*
3141          * Process the label in the upcoming getfattr
3142          */
3143         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3144
3145         nfs_fattr_init(info->fattr);
3146         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3147 }
3148
3149 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3150                 struct nfs_fsinfo *info)
3151 {
3152         struct nfs4_exception exception = { };
3153         int err;
3154         do {
3155                 err = _nfs4_lookup_root(server, fhandle, info);
3156                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3157                 switch (err) {
3158                 case 0:
3159                 case -NFS4ERR_WRONGSEC:
3160                         goto out;
3161                 default:
3162                         err = nfs4_handle_exception(server, err, &exception);
3163                 }
3164         } while (exception.retry);
3165 out:
3166         return err;
3167 }
3168
3169 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3170                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3171 {
3172         struct rpc_auth_create_args auth_args = {
3173                 .pseudoflavor = flavor,
3174         };
3175         struct rpc_auth *auth;
3176         int ret;
3177
3178         auth = rpcauth_create(&auth_args, server->client);
3179         if (IS_ERR(auth)) {
3180                 ret = -EACCES;
3181                 goto out;
3182         }
3183         ret = nfs4_lookup_root(server, fhandle, info);
3184 out:
3185         return ret;
3186 }
3187
3188 /*
3189  * Retry pseudoroot lookup with various security flavors.  We do this when:
3190  *
3191  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3192  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3193  *
3194  * Returns zero on success, or a negative NFS4ERR value, or a
3195  * negative errno value.
3196  */
3197 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3198                               struct nfs_fsinfo *info)
3199 {
3200         /* Per 3530bis 15.33.5 */
3201         static const rpc_authflavor_t flav_array[] = {
3202                 RPC_AUTH_GSS_KRB5P,
3203                 RPC_AUTH_GSS_KRB5I,
3204                 RPC_AUTH_GSS_KRB5,
3205                 RPC_AUTH_UNIX,                  /* courtesy */
3206                 RPC_AUTH_NULL,
3207         };
3208         int status = -EPERM;
3209         size_t i;
3210
3211         if (server->auth_info.flavor_len > 0) {
3212                 /* try each flavor specified by user */
3213                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3214                         status = nfs4_lookup_root_sec(server, fhandle, info,
3215                                                 server->auth_info.flavors[i]);
3216                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3217                                 continue;
3218                         break;
3219                 }
3220         } else {
3221                 /* no flavors specified by user, try default list */
3222                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3223                         status = nfs4_lookup_root_sec(server, fhandle, info,
3224                                                       flav_array[i]);
3225                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3226                                 continue;
3227                         break;
3228                 }
3229         }
3230
3231         /*
3232          * -EACCESS could mean that the user doesn't have correct permissions
3233          * to access the mount.  It could also mean that we tried to mount
3234          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3235          * existing mount programs don't handle -EACCES very well so it should
3236          * be mapped to -EPERM instead.
3237          */
3238         if (status == -EACCES)
3239                 status = -EPERM;
3240         return status;
3241 }
3242
3243 static int nfs4_do_find_root_sec(struct nfs_server *server,
3244                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3245 {
3246         int mv = server->nfs_client->cl_minorversion;
3247         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3248 }
3249
3250 /**
3251  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3252  * @server: initialized nfs_server handle
3253  * @fhandle: we fill in the pseudo-fs root file handle
3254  * @info: we fill in an FSINFO struct
3255  * @auth_probe: probe the auth flavours
3256  *
3257  * Returns zero on success, or a negative errno.
3258  */
3259 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3260                          struct nfs_fsinfo *info,
3261                          bool auth_probe)
3262 {
3263         int status = 0;
3264
3265         if (!auth_probe)
3266                 status = nfs4_lookup_root(server, fhandle, info);
3267
3268         if (auth_probe || status == NFS4ERR_WRONGSEC)
3269                 status = nfs4_do_find_root_sec(server, fhandle, info);
3270
3271         if (status == 0)
3272                 status = nfs4_server_capabilities(server, fhandle);
3273         if (status == 0)
3274                 status = nfs4_do_fsinfo(server, fhandle, info);
3275
3276         return nfs4_map_errors(status);
3277 }
3278
3279 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3280                               struct nfs_fsinfo *info)
3281 {
3282         int error;
3283         struct nfs_fattr *fattr = info->fattr;
3284         struct nfs4_label *label = NULL;
3285
3286         error = nfs4_server_capabilities(server, mntfh);
3287         if (error < 0) {
3288                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3289                 return error;
3290         }
3291
3292         label = nfs4_label_alloc(server, GFP_KERNEL);
3293         if (IS_ERR(label))
3294                 return PTR_ERR(label);
3295
3296         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3297         if (error < 0) {
3298                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3299                 goto err_free_label;
3300         }
3301
3302         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3303             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3304                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3305
3306 err_free_label:
3307         nfs4_label_free(label);
3308
3309         return error;
3310 }
3311
3312 /*
3313  * Get locations and (maybe) other attributes of a referral.
3314  * Note that we'll actually follow the referral later when
3315  * we detect fsid mismatch in inode revalidation
3316  */
3317 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3318                              const struct qstr *name, struct nfs_fattr *fattr,
3319                              struct nfs_fh *fhandle)
3320 {
3321         int status = -ENOMEM;
3322         struct page *page = NULL;
3323         struct nfs4_fs_locations *locations = NULL;
3324
3325         page = alloc_page(GFP_KERNEL);
3326         if (page == NULL)
3327                 goto out;
3328         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3329         if (locations == NULL)
3330                 goto out;
3331
3332         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3333         if (status != 0)
3334                 goto out;
3335
3336         /*
3337          * If the fsid didn't change, this is a migration event, not a
3338          * referral.  Cause us to drop into the exception handler, which
3339          * will kick off migration recovery.
3340          */
3341         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3342                 dprintk("%s: server did not return a different fsid for"
3343                         " a referral at %s\n", __func__, name->name);
3344                 status = -NFS4ERR_MOVED;
3345                 goto out;
3346         }
3347         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3348         nfs_fixup_referral_attributes(&locations->fattr);
3349
3350         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3351         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3352         memset(fhandle, 0, sizeof(struct nfs_fh));
3353 out:
3354         if (page)
3355                 __free_page(page);
3356         kfree(locations);
3357         return status;
3358 }
3359
3360 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3361                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3362 {
3363         struct nfs4_getattr_arg args = {
3364                 .fh = fhandle,
3365                 .bitmask = server->attr_bitmask,
3366         };
3367         struct nfs4_getattr_res res = {
3368                 .fattr = fattr,
3369                 .label = label,
3370                 .server = server,
3371         };
3372         struct rpc_message msg = {
3373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3374                 .rpc_argp = &args,
3375                 .rpc_resp = &res,
3376         };
3377
3378         args.bitmask = nfs4_bitmask(server, label);
3379
3380         nfs_fattr_init(fattr);
3381         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3382 }
3383
3384 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3385                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3386 {
3387         struct nfs4_exception exception = { };
3388         int err;
3389         do {
3390                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3391                 trace_nfs4_getattr(server, fhandle, fattr, err);
3392                 err = nfs4_handle_exception(server, err,
3393                                 &exception);
3394         } while (exception.retry);
3395         return err;
3396 }
3397
3398 /*
3399  * The file is not closed if it is opened due to the a request to change
3400  * the size of the file. The open call will not be needed once the
3401  * VFS layer lookup-intents are implemented.
3402  *
3403  * Close is called when the inode is destroyed.
3404  * If we haven't opened the file for O_WRONLY, we
3405  * need to in the size_change case to obtain a stateid.
3406  *
3407  * Got race?
3408  * Because OPEN is always done by name in nfsv4, it is
3409  * possible that we opened a different file by the same
3410  * name.  We can recognize this race condition, but we
3411  * can't do anything about it besides returning an error.
3412  *
3413  * This will be fixed with VFS changes (lookup-intent).
3414  */
3415 static int
3416 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3417                   struct iattr *sattr)
3418 {
3419         struct inode *inode = d_inode(dentry);
3420         struct rpc_cred *cred = NULL;
3421         struct nfs4_state *state = NULL;
3422         struct nfs4_label *label = NULL;
3423         int status;
3424
3425         if (pnfs_ld_layoutret_on_setattr(inode) &&
3426             sattr->ia_valid & ATTR_SIZE &&
3427             sattr->ia_size < i_size_read(inode))
3428                 pnfs_commit_and_return_layout(inode);
3429
3430         nfs_fattr_init(fattr);
3431
3432         /* Deal with open(O_TRUNC) */
3433         if (sattr->ia_valid & ATTR_OPEN)
3434                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3435
3436         /* Optimization: if the end result is no change, don't RPC */
3437         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3438                 return 0;
3439
3440         /* Search for an existing open(O_WRITE) file */
3441         if (sattr->ia_valid & ATTR_FILE) {
3442                 struct nfs_open_context *ctx;
3443
3444                 ctx = nfs_file_open_context(sattr->ia_file);
3445                 if (ctx) {
3446                         cred = ctx->cred;
3447                         state = ctx->state;
3448                 }
3449         }
3450
3451         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3452         if (IS_ERR(label))
3453                 return PTR_ERR(label);
3454
3455         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3456         if (status == 0) {
3457                 nfs_setattr_update_inode(inode, sattr, fattr);
3458                 nfs_setsecurity(inode, fattr, label);
3459         }
3460         nfs4_label_free(label);
3461         return status;
3462 }
3463
3464 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3465                 const struct qstr *name, struct nfs_fh *fhandle,
3466                 struct nfs_fattr *fattr, struct nfs4_label *label)
3467 {
3468         struct nfs_server *server = NFS_SERVER(dir);
3469         int                    status;
3470         struct nfs4_lookup_arg args = {
3471                 .bitmask = server->attr_bitmask,
3472                 .dir_fh = NFS_FH(dir),
3473                 .name = name,
3474         };
3475         struct nfs4_lookup_res res = {
3476                 .server = server,
3477                 .fattr = fattr,
3478                 .label = label,
3479                 .fh = fhandle,
3480         };
3481         struct rpc_message msg = {
3482                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3483                 .rpc_argp = &args,
3484                 .rpc_resp = &res,
3485         };
3486
3487         args.bitmask = nfs4_bitmask(server, label);
3488
3489         nfs_fattr_init(fattr);
3490
3491         dprintk("NFS call  lookup %s\n", name->name);
3492         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3493         dprintk("NFS reply lookup: %d\n", status);
3494         return status;
3495 }
3496
3497 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3498 {
3499         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3500                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3501         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3502         fattr->nlink = 2;
3503 }
3504
3505 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3506                                    struct qstr *name, struct nfs_fh *fhandle,
3507                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3508 {
3509         struct nfs4_exception exception = { };
3510         struct rpc_clnt *client = *clnt;
3511         int err;
3512         do {
3513                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3514                 trace_nfs4_lookup(dir, name, err);
3515                 switch (err) {
3516                 case -NFS4ERR_BADNAME:
3517                         err = -ENOENT;
3518                         goto out;
3519                 case -NFS4ERR_MOVED:
3520                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3521                         if (err == -NFS4ERR_MOVED)
3522                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3523                         goto out;
3524                 case -NFS4ERR_WRONGSEC:
3525                         err = -EPERM;
3526                         if (client != *clnt)
3527                                 goto out;
3528                         client = nfs4_negotiate_security(client, dir, name);
3529                         if (IS_ERR(client))
3530                                 return PTR_ERR(client);
3531
3532                         exception.retry = 1;
3533                         break;
3534                 default:
3535                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3536                 }
3537         } while (exception.retry);
3538
3539 out:
3540         if (err == 0)
3541                 *clnt = client;
3542         else if (client != *clnt)
3543                 rpc_shutdown_client(client);
3544
3545         return err;
3546 }
3547
3548 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3549                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3550                             struct nfs4_label *label)
3551 {
3552         int status;
3553         struct rpc_clnt *client = NFS_CLIENT(dir);
3554
3555         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3556         if (client != NFS_CLIENT(dir)) {
3557                 rpc_shutdown_client(client);
3558                 nfs_fixup_secinfo_attributes(fattr);
3559         }
3560         return status;
3561 }
3562
3563 struct rpc_clnt *
3564 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3565                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3566 {
3567         struct rpc_clnt *client = NFS_CLIENT(dir);
3568         int status;
3569
3570         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3571         if (status < 0)
3572                 return ERR_PTR(status);
3573         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3574 }
3575
3576 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3577 {
3578         struct nfs_server *server = NFS_SERVER(inode);
3579         struct nfs4_accessargs args = {
3580                 .fh = NFS_FH(inode),
3581                 .bitmask = server->cache_consistency_bitmask,
3582         };
3583         struct nfs4_accessres res = {
3584                 .server = server,
3585         };
3586         struct rpc_message msg = {
3587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3588                 .rpc_argp = &args,
3589                 .rpc_resp = &res,
3590                 .rpc_cred = entry->cred,
3591         };
3592         int mode = entry->mask;
3593         int status = 0;
3594
3595         /*
3596          * Determine which access bits we want to ask for...
3597          */
3598         if (mode & MAY_READ)
3599                 args.access |= NFS4_ACCESS_READ;
3600         if (S_ISDIR(inode->i_mode)) {
3601                 if (mode & MAY_WRITE)
3602                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3603                 if (mode & MAY_EXEC)
3604                         args.access |= NFS4_ACCESS_LOOKUP;
3605         } else {
3606                 if (mode & MAY_WRITE)
3607                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3608                 if (mode & MAY_EXEC)
3609                         args.access |= NFS4_ACCESS_EXECUTE;
3610         }
3611
3612         res.fattr = nfs_alloc_fattr();
3613         if (res.fattr == NULL)
3614                 return -ENOMEM;
3615
3616         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3617         if (!status) {
3618                 nfs_access_set_mask(entry, res.access);
3619                 nfs_refresh_inode(inode, res.fattr);
3620         }
3621         nfs_free_fattr(res.fattr);
3622         return status;
3623 }
3624
3625 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3626 {
3627         struct nfs4_exception exception = { };
3628         int err;
3629         do {
3630                 err = _nfs4_proc_access(inode, entry);
3631                 trace_nfs4_access(inode, err);
3632                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3633                                 &exception);
3634         } while (exception.retry);
3635         return err;
3636 }
3637
3638 /*
3639  * TODO: For the time being, we don't try to get any attributes
3640  * along with any of the zero-copy operations READ, READDIR,
3641  * READLINK, WRITE.
3642  *
3643  * In the case of the first three, we want to put the GETATTR
3644  * after the read-type operation -- this is because it is hard
3645  * to predict the length of a GETATTR response in v4, and thus
3646  * align the READ data correctly.  This means that the GETATTR
3647  * may end up partially falling into the page cache, and we should
3648  * shift it into the 'tail' of the xdr_buf before processing.
3649  * To do this efficiently, we need to know the total length
3650  * of data received, which doesn't seem to be available outside
3651  * of the RPC layer.
3652  *
3653  * In the case of WRITE, we also want to put the GETATTR after
3654  * the operation -- in this case because we want to make sure
3655  * we get the post-operation mtime and size.
3656  *
3657  * Both of these changes to the XDR layer would in fact be quite
3658  * minor, but I decided to leave them for a subsequent patch.
3659  */
3660 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3661                 unsigned int pgbase, unsigned int pglen)
3662 {
3663         struct nfs4_readlink args = {
3664                 .fh       = NFS_FH(inode),
3665                 .pgbase   = pgbase,
3666                 .pglen    = pglen,
3667                 .pages    = &page,
3668         };
3669         struct nfs4_readlink_res res;
3670         struct rpc_message msg = {
3671                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3672                 .rpc_argp = &args,
3673                 .rpc_resp = &res,
3674         };
3675
3676         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3677 }
3678
3679 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3680                 unsigned int pgbase, unsigned int pglen)
3681 {
3682         struct nfs4_exception exception = { };
3683         int err;
3684         do {
3685                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3686                 trace_nfs4_readlink(inode, err);
3687                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3688                                 &exception);
3689         } while (exception.retry);
3690         return err;
3691 }
3692
3693 /*
3694  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3695  */
3696 static int
3697 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3698                  int flags)
3699 {
3700         struct nfs4_label l, *ilabel = NULL;
3701         struct nfs_open_context *ctx;
3702         struct nfs4_state *state;
3703         int status = 0;
3704
3705         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3706         if (IS_ERR(ctx))
3707                 return PTR_ERR(ctx);
3708
3709         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3710
3711         sattr->ia_mode &= ~current_umask();
3712         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3713         if (IS_ERR(state)) {
3714                 status = PTR_ERR(state);
3715                 goto out;
3716         }
3717 out:
3718         nfs4_label_release_security(ilabel);
3719         put_nfs_open_context(ctx);
3720         return status;
3721 }
3722
3723 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3724 {
3725         struct nfs_server *server = NFS_SERVER(dir);
3726         struct nfs_removeargs args = {
3727                 .fh = NFS_FH(dir),
3728                 .name = *name,
3729         };
3730         struct nfs_removeres res = {
3731                 .server = server,
3732         };
3733         struct rpc_message msg = {
3734                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3735                 .rpc_argp = &args,
3736                 .rpc_resp = &res,
3737         };
3738         int status;
3739
3740         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3741         if (status == 0)
3742                 update_changeattr(dir, &res.cinfo);
3743         return status;
3744 }
3745
3746 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3747 {
3748         struct nfs4_exception exception = { };
3749         int err;
3750         do {
3751                 err = _nfs4_proc_remove(dir, name);
3752                 trace_nfs4_remove(dir, name, err);
3753                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3754                                 &exception);
3755         } while (exception.retry);
3756         return err;
3757 }
3758
3759 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3760 {
3761         struct nfs_server *server = NFS_SERVER(dir);
3762         struct nfs_removeargs *args = msg->rpc_argp;
3763         struct nfs_removeres *res = msg->rpc_resp;
3764
3765         res->server = server;
3766         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3767         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3768
3769         nfs_fattr_init(res->dir_attr);
3770 }
3771
3772 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3773 {
3774         nfs4_setup_sequence(NFS_SERVER(data->dir),
3775                         &data->args.seq_args,
3776                         &data->res.seq_res,
3777                         task);
3778 }
3779
3780 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3781 {
3782         struct nfs_unlinkdata *data = task->tk_calldata;
3783         struct nfs_removeres *res = &data->res;
3784
3785         if (!nfs4_sequence_done(task, &res->seq_res))
3786                 return 0;
3787         if (nfs4_async_handle_error(task, res->server, NULL,
3788                                     &data->timeout) == -EAGAIN)
3789                 return 0;
3790         update_changeattr(dir, &res->cinfo);
3791         return 1;
3792 }
3793
3794 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3795 {
3796         struct nfs_server *server = NFS_SERVER(dir);
3797         struct nfs_renameargs *arg = msg->rpc_argp;
3798         struct nfs_renameres *res = msg->rpc_resp;
3799
3800         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3801         res->server = server;
3802         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3803 }
3804
3805 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3806 {
3807         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3808                         &data->args.seq_args,
3809                         &data->res.seq_res,
3810                         task);
3811 }
3812
3813 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3814                                  struct inode *new_dir)
3815 {
3816         struct nfs_renamedata *data = task->tk_calldata;
3817         struct nfs_renameres *res = &data->res;
3818
3819         if (!nfs4_sequence_done(task, &res->seq_res))
3820                 return 0;
3821         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3822                 return 0;
3823
3824         update_changeattr(old_dir, &res->old_cinfo);
3825         update_changeattr(new_dir, &res->new_cinfo);
3826         return 1;
3827 }
3828
3829 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3830 {
3831         struct nfs_server *server = NFS_SERVER(inode);
3832         struct nfs4_link_arg arg = {
3833                 .fh     = NFS_FH(inode),
3834                 .dir_fh = NFS_FH(dir),
3835                 .name   = name,
3836                 .bitmask = server->attr_bitmask,
3837         };
3838         struct nfs4_link_res res = {
3839                 .server = server,
3840                 .label = NULL,
3841         };
3842         struct rpc_message msg = {
3843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3844                 .rpc_argp = &arg,
3845                 .rpc_resp = &res,
3846         };
3847         int status = -ENOMEM;
3848
3849         res.fattr = nfs_alloc_fattr();
3850         if (res.fattr == NULL)
3851                 goto out;
3852
3853         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3854         if (IS_ERR(res.label)) {
3855                 status = PTR_ERR(res.label);
3856                 goto out;
3857         }
3858         arg.bitmask = nfs4_bitmask(server, res.label);
3859
3860         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3861         if (!status) {
3862                 update_changeattr(dir, &res.cinfo);
3863                 status = nfs_post_op_update_inode(inode, res.fattr);
3864                 if (!status)
3865                         nfs_setsecurity(inode, res.fattr, res.label);
3866         }
3867
3868
3869         nfs4_label_free(res.label);
3870
3871 out:
3872         nfs_free_fattr(res.fattr);
3873         return status;
3874 }
3875
3876 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3877 {
3878         struct nfs4_exception exception = { };
3879         int err;
3880         do {
3881                 err = nfs4_handle_exception(NFS_SERVER(inode),
3882                                 _nfs4_proc_link(inode, dir, name),
3883                                 &exception);
3884         } while (exception.retry);
3885         return err;
3886 }
3887
3888 struct nfs4_createdata {
3889         struct rpc_message msg;
3890         struct nfs4_create_arg arg;
3891         struct nfs4_create_res res;
3892         struct nfs_fh fh;
3893         struct nfs_fattr fattr;
3894         struct nfs4_label *label;
3895 };
3896
3897 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3898                 struct qstr *name, struct iattr *sattr, u32 ftype)
3899 {
3900         struct nfs4_createdata *data;
3901
3902         data = kzalloc(sizeof(*data), GFP_KERNEL);
3903         if (data != NULL) {
3904                 struct nfs_server *server = NFS_SERVER(dir);
3905
3906                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3907                 if (IS_ERR(data->label))
3908                         goto out_free;
3909
3910                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3911                 data->msg.rpc_argp = &data->arg;
3912                 data->msg.rpc_resp = &data->res;
3913                 data->arg.dir_fh = NFS_FH(dir);
3914                 data->arg.server = server;
3915                 data->arg.name = name;
3916                 data->arg.attrs = sattr;
3917                 data->arg.ftype = ftype;
3918                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3919                 data->res.server = server;
3920                 data->res.fh = &data->fh;
3921                 data->res.fattr = &data->fattr;
3922                 data->res.label = data->label;
3923                 nfs_fattr_init(data->res.fattr);
3924         }
3925         return data;
3926 out_free:
3927         kfree(data);
3928         return NULL;
3929 }
3930
3931 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3932 {
3933         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3934                                     &data->arg.seq_args, &data->res.seq_res, 1);
3935         if (status == 0) {
3936                 update_changeattr(dir, &data->res.dir_cinfo);
3937                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3938         }
3939         return status;
3940 }
3941
3942 static void nfs4_free_createdata(struct nfs4_createdata *data)
3943 {
3944         nfs4_label_free(data->label);
3945         kfree(data);
3946 }
3947
3948 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3949                 struct page *page, unsigned int len, struct iattr *sattr,
3950                 struct nfs4_label *label)
3951 {
3952         struct nfs4_createdata *data;
3953         int status = -ENAMETOOLONG;
3954
3955         if (len > NFS4_MAXPATHLEN)
3956                 goto out;
3957
3958         status = -ENOMEM;
3959         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3960         if (data == NULL)
3961                 goto out;
3962
3963         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3964         data->arg.u.symlink.pages = &page;
3965         data->arg.u.symlink.len = len;
3966         data->arg.label = label;
3967
3968         status = nfs4_do_create(dir, dentry, data);
3969
3970         nfs4_free_createdata(data);
3971 out:
3972         return status;
3973 }
3974
3975 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3976                 struct page *page, unsigned int len, struct iattr *sattr)
3977 {
3978         struct nfs4_exception exception = { };
3979         struct nfs4_label l, *label = NULL;
3980         int err;
3981
3982         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3983
3984         do {
3985                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3986                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3987                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3988                                 &exception);
3989         } while (exception.retry);
3990
3991         nfs4_label_release_security(label);
3992         return err;
3993 }
3994
3995 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3996                 struct iattr *sattr, struct nfs4_label *label)
3997 {
3998         struct nfs4_createdata *data;
3999         int status = -ENOMEM;
4000
4001         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4002         if (data == NULL)
4003                 goto out;
4004
4005         data->arg.label = label;
4006         status = nfs4_do_create(dir, dentry, data);
4007
4008         nfs4_free_createdata(data);
4009 out:
4010         return status;
4011 }
4012
4013 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4014                 struct iattr *sattr)
4015 {
4016         struct nfs4_exception exception = { };
4017         struct nfs4_label l, *label = NULL;
4018         int err;
4019
4020         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4021
4022         sattr->ia_mode &= ~current_umask();
4023         do {
4024                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4025                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4026                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4027                                 &exception);
4028         } while (exception.retry);
4029         nfs4_label_release_security(label);
4030
4031         return err;
4032 }
4033
4034 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4035                 u64 cookie, struct page **pages, unsigned int count, int plus)
4036 {
4037         struct inode            *dir = d_inode(dentry);
4038         struct nfs4_readdir_arg args = {
4039                 .fh = NFS_FH(dir),
4040                 .pages = pages,
4041                 .pgbase = 0,
4042                 .count = count,
4043                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4044                 .plus = plus,
4045         };
4046         struct nfs4_readdir_res res;
4047         struct rpc_message msg = {
4048                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4049                 .rpc_argp = &args,
4050                 .rpc_resp = &res,
4051                 .rpc_cred = cred,
4052         };
4053         int                     status;
4054
4055         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4056                         dentry,
4057                         (unsigned long long)cookie);
4058         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4059         res.pgbase = args.pgbase;
4060         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4061         if (status >= 0) {
4062                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4063                 status += args.pgbase;
4064         }
4065
4066         nfs_invalidate_atime(dir);
4067
4068         dprintk("%s: returns %d\n", __func__, status);
4069         return status;
4070 }
4071
4072 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4073                 u64 cookie, struct page **pages, unsigned int count, int plus)
4074 {
4075         struct nfs4_exception exception = { };
4076         int err;
4077         do {
4078                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4079                                 pages, count, plus);
4080                 trace_nfs4_readdir(d_inode(dentry), err);
4081                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4082                                 &exception);
4083         } while (exception.retry);
4084         return err;
4085 }
4086
4087 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4088                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4089 {
4090         struct nfs4_createdata *data;
4091         int mode = sattr->ia_mode;
4092         int status = -ENOMEM;
4093
4094         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4095         if (data == NULL)
4096                 goto out;
4097
4098         if (S_ISFIFO(mode))
4099                 data->arg.ftype = NF4FIFO;
4100         else if (S_ISBLK(mode)) {
4101                 data->arg.ftype = NF4BLK;
4102                 data->arg.u.device.specdata1 = MAJOR(rdev);
4103                 data->arg.u.device.specdata2 = MINOR(rdev);
4104         }
4105         else if (S_ISCHR(mode)) {
4106                 data->arg.ftype = NF4CHR;
4107                 data->arg.u.device.specdata1 = MAJOR(rdev);
4108                 data->arg.u.device.specdata2 = MINOR(rdev);
4109         } else if (!S_ISSOCK(mode)) {
4110                 status = -EINVAL;
4111                 goto out_free;
4112         }
4113
4114         data->arg.label = label;
4115         status = nfs4_do_create(dir, dentry, data);
4116 out_free:
4117         nfs4_free_createdata(data);
4118 out:
4119         return status;
4120 }
4121
4122 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4123                 struct iattr *sattr, dev_t rdev)
4124 {
4125         struct nfs4_exception exception = { };
4126         struct nfs4_label l, *label = NULL;
4127         int err;
4128
4129         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4130
4131         sattr->ia_mode &= ~current_umask();
4132         do {
4133                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4134                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4135                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4136                                 &exception);
4137         } while (exception.retry);
4138
4139         nfs4_label_release_security(label);
4140
4141         return err;
4142 }
4143
4144 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4145                  struct nfs_fsstat *fsstat)
4146 {
4147         struct nfs4_statfs_arg args = {
4148                 .fh = fhandle,
4149                 .bitmask = server->attr_bitmask,
4150         };
4151         struct nfs4_statfs_res res = {
4152                 .fsstat = fsstat,
4153         };
4154         struct rpc_message msg = {
4155                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4156                 .rpc_argp = &args,
4157                 .rpc_resp = &res,
4158         };
4159
4160         nfs_fattr_init(fsstat->fattr);
4161         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4162 }
4163
4164 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4165 {
4166         struct nfs4_exception exception = { };
4167         int err;
4168         do {
4169                 err = nfs4_handle_exception(server,
4170                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4171                                 &exception);
4172         } while (exception.retry);
4173         return err;
4174 }
4175
4176 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4177                 struct nfs_fsinfo *fsinfo)
4178 {
4179         struct nfs4_fsinfo_arg args = {
4180                 .fh = fhandle,
4181                 .bitmask = server->attr_bitmask,
4182         };
4183         struct nfs4_fsinfo_res res = {
4184                 .fsinfo = fsinfo,
4185         };
4186         struct rpc_message msg = {
4187                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4188                 .rpc_argp = &args,
4189                 .rpc_resp = &res,
4190         };
4191
4192         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4193 }
4194
4195 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4196 {
4197         struct nfs4_exception exception = { };
4198         unsigned long now = jiffies;
4199         int err;
4200
4201         do {
4202                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4203                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4204                 if (err == 0) {
4205                         struct nfs_client *clp = server->nfs_client;
4206
4207                         spin_lock(&clp->cl_lock);
4208                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4209                         clp->cl_last_renewal = now;
4210                         spin_unlock(&clp->cl_lock);
4211                         break;
4212                 }
4213                 err = nfs4_handle_exception(server, err, &exception);
4214         } while (exception.retry);
4215         return err;
4216 }
4217
4218 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4219 {
4220         int error;
4221
4222         nfs_fattr_init(fsinfo->fattr);
4223         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4224         if (error == 0) {
4225                 /* block layout checks this! */
4226                 server->pnfs_blksize = fsinfo->blksize;
4227                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4228         }
4229
4230         return error;
4231 }
4232
4233 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4234                 struct nfs_pathconf *pathconf)
4235 {
4236         struct nfs4_pathconf_arg args = {
4237                 .fh = fhandle,
4238                 .bitmask = server->attr_bitmask,
4239         };
4240         struct nfs4_pathconf_res res = {
4241                 .pathconf = pathconf,
4242         };
4243         struct rpc_message msg = {
4244                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4245                 .rpc_argp = &args,
4246                 .rpc_resp = &res,
4247         };
4248
4249         /* None of the pathconf attributes are mandatory to implement */
4250         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4251                 memset(pathconf, 0, sizeof(*pathconf));
4252                 return 0;
4253         }
4254
4255         nfs_fattr_init(pathconf->fattr);
4256         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4257 }
4258
4259 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4260                 struct nfs_pathconf *pathconf)
4261 {
4262         struct nfs4_exception exception = { };
4263         int err;
4264
4265         do {
4266                 err = nfs4_handle_exception(server,
4267                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4268                                 &exception);
4269         } while (exception.retry);
4270         return err;
4271 }
4272
4273 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4274                 const struct nfs_open_context *ctx,
4275                 const struct nfs_lock_context *l_ctx,
4276                 fmode_t fmode)
4277 {
4278         const struct nfs_lockowner *lockowner = NULL;
4279
4280         if (l_ctx != NULL)
4281                 lockowner = &l_ctx->lockowner;
4282         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4283 }
4284 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4285
4286 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4287                 const struct nfs_open_context *ctx,
4288                 const struct nfs_lock_context *l_ctx,
4289                 fmode_t fmode)
4290 {
4291         nfs4_stateid current_stateid;
4292
4293         /* If the current stateid represents a lost lock, then exit */
4294         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4295                 return true;
4296         return nfs4_stateid_match(stateid, &current_stateid);
4297 }
4298
4299 static bool nfs4_error_stateid_expired(int err)
4300 {
4301         switch (err) {
4302         case -NFS4ERR_DELEG_REVOKED:
4303         case -NFS4ERR_ADMIN_REVOKED:
4304         case -NFS4ERR_BAD_STATEID:
4305         case -NFS4ERR_STALE_STATEID:
4306         case -NFS4ERR_OLD_STATEID:
4307         case -NFS4ERR_OPENMODE:
4308         case -NFS4ERR_EXPIRED:
4309                 return true;
4310         }
4311         return false;
4312 }
4313
4314 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4315 {
4316         nfs_invalidate_atime(hdr->inode);
4317 }
4318
4319 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4320 {
4321         struct nfs_server *server = NFS_SERVER(hdr->inode);
4322
4323         trace_nfs4_read(hdr, task->tk_status);
4324         if (nfs4_async_handle_error(task, server,
4325                                     hdr->args.context->state,
4326                                     NULL) == -EAGAIN) {
4327                 rpc_restart_call_prepare(task);
4328                 return -EAGAIN;
4329         }
4330
4331         __nfs4_read_done_cb(hdr);
4332         if (task->tk_status > 0)
4333                 renew_lease(server, hdr->timestamp);
4334         return 0;
4335 }
4336
4337 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4338                 struct nfs_pgio_args *args)
4339 {
4340
4341         if (!nfs4_error_stateid_expired(task->tk_status) ||
4342                 nfs4_stateid_is_current(&args->stateid,
4343                                 args->context,
4344                                 args->lock_context,
4345                                 FMODE_READ))
4346                 return false;
4347         rpc_restart_call_prepare(task);
4348         return true;
4349 }
4350
4351 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4352 {
4353
4354         dprintk("--> %s\n", __func__);
4355
4356         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4357                 return -EAGAIN;
4358         if (nfs4_read_stateid_changed(task, &hdr->args))
4359                 return -EAGAIN;
4360         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4361                                     nfs4_read_done_cb(task, hdr);
4362 }
4363
4364 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4365                                  struct rpc_message *msg)
4366 {
4367         hdr->timestamp   = jiffies;
4368         hdr->pgio_done_cb = nfs4_read_done_cb;
4369         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4370         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4371 }
4372
4373 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4374                                       struct nfs_pgio_header *hdr)
4375 {
4376         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4377                         &hdr->args.seq_args,
4378                         &hdr->res.seq_res,
4379                         task))
4380                 return 0;
4381         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4382                                 hdr->args.lock_context,
4383                                 hdr->rw_ops->rw_mode) == -EIO)
4384                 return -EIO;
4385         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4386                 return -EIO;
4387         return 0;
4388 }
4389
4390 static int nfs4_write_done_cb(struct rpc_task *task,
4391                               struct nfs_pgio_header *hdr)
4392 {
4393         struct inode *inode = hdr->inode;
4394
4395         trace_nfs4_write(hdr, task->tk_status);
4396         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4397                                     hdr->args.context->state,
4398                                     NULL) == -EAGAIN) {
4399                 rpc_restart_call_prepare(task);
4400                 return -EAGAIN;
4401         }
4402         if (task->tk_status >= 0) {
4403                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4404                 nfs_writeback_update_inode(hdr);
4405         }
4406         return 0;
4407 }
4408
4409 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4410                 struct nfs_pgio_args *args)
4411 {
4412
4413         if (!nfs4_error_stateid_expired(task->tk_status) ||
4414                 nfs4_stateid_is_current(&args->stateid,
4415                                 args->context,
4416                                 args->lock_context,
4417                                 FMODE_WRITE))
4418                 return false;
4419         rpc_restart_call_prepare(task);
4420         return true;
4421 }
4422
4423 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4424 {
4425         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4426                 return -EAGAIN;
4427         if (nfs4_write_stateid_changed(task, &hdr->args))
4428                 return -EAGAIN;
4429         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4430                 nfs4_write_done_cb(task, hdr);
4431 }
4432
4433 static
4434 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4435 {
4436         /* Don't request attributes for pNFS or O_DIRECT writes */
4437         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4438                 return false;
4439         /* Otherwise, request attributes if and only if we don't hold
4440          * a delegation
4441          */
4442         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4443 }
4444
4445 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4446                                   struct rpc_message *msg)
4447 {
4448         struct nfs_server *server = NFS_SERVER(hdr->inode);
4449
4450         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4451                 hdr->args.bitmask = NULL;
4452                 hdr->res.fattr = NULL;
4453         } else
4454                 hdr->args.bitmask = server->cache_consistency_bitmask;
4455
4456         if (!hdr->pgio_done_cb)
4457                 hdr->pgio_done_cb = nfs4_write_done_cb;
4458         hdr->res.server = server;
4459         hdr->timestamp   = jiffies;
4460
4461         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4462         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4463 }
4464
4465 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4466 {
4467         nfs4_setup_sequence(NFS_SERVER(data->inode),
4468                         &data->args.seq_args,
4469                         &data->res.seq_res,
4470                         task);
4471 }
4472
4473 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4474 {
4475         struct inode *inode = data->inode;
4476
4477         trace_nfs4_commit(data, task->tk_status);
4478         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4479                                     NULL, NULL) == -EAGAIN) {
4480                 rpc_restart_call_prepare(task);
4481                 return -EAGAIN;
4482         }
4483         return 0;
4484 }
4485
4486 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4487 {
4488         if (!nfs4_sequence_done(task, &data->res.seq_res))
4489                 return -EAGAIN;
4490         return data->commit_done_cb(task, data);
4491 }
4492
4493 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4494 {
4495         struct nfs_server *server = NFS_SERVER(data->inode);
4496
4497         if (data->commit_done_cb == NULL)
4498                 data->commit_done_cb = nfs4_commit_done_cb;
4499         data->res.server = server;
4500         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4501         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4502 }
4503
4504 struct nfs4_renewdata {
4505         struct nfs_client       *client;
4506         unsigned long           timestamp;
4507 };
4508
4509 /*
4510  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4511  * standalone procedure for queueing an asynchronous RENEW.
4512  */
4513 static void nfs4_renew_release(void *calldata)
4514 {
4515         struct nfs4_renewdata *data = calldata;
4516         struct nfs_client *clp = data->client;
4517
4518         if (atomic_read(&clp->cl_count) > 1)
4519                 nfs4_schedule_state_renewal(clp);
4520         nfs_put_client(clp);
4521         kfree(data);
4522 }
4523
4524 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4525 {
4526         struct nfs4_renewdata *data = calldata;
4527         struct nfs_client *clp = data->client;
4528         unsigned long timestamp = data->timestamp;
4529
4530         trace_nfs4_renew_async(clp, task->tk_status);
4531         switch (task->tk_status) {
4532         case 0:
4533                 break;
4534         case -NFS4ERR_LEASE_MOVED:
4535                 nfs4_schedule_lease_moved_recovery(clp);
4536                 break;
4537         default:
4538                 /* Unless we're shutting down, schedule state recovery! */
4539                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4540                         return;
4541                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4542                         nfs4_schedule_lease_recovery(clp);
4543                         return;
4544                 }
4545                 nfs4_schedule_path_down_recovery(clp);
4546         }
4547         do_renew_lease(clp, timestamp);
4548 }
4549
4550 static const struct rpc_call_ops nfs4_renew_ops = {
4551         .rpc_call_done = nfs4_renew_done,
4552         .rpc_release = nfs4_renew_release,
4553 };
4554
4555 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4556 {
4557         struct rpc_message msg = {
4558                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4559                 .rpc_argp       = clp,
4560                 .rpc_cred       = cred,
4561         };
4562         struct nfs4_renewdata *data;
4563
4564         if (renew_flags == 0)
4565                 return 0;
4566         if (!atomic_inc_not_zero(&clp->cl_count))
4567                 return -EIO;
4568         data = kmalloc(sizeof(*data), GFP_NOFS);
4569         if (data == NULL)
4570                 return -ENOMEM;
4571         data->client = clp;
4572         data->timestamp = jiffies;
4573         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4574                         &nfs4_renew_ops, data);
4575 }
4576
4577 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4578 {
4579         struct rpc_message msg = {
4580                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4581                 .rpc_argp       = clp,
4582                 .rpc_cred       = cred,
4583         };
4584         unsigned long now = jiffies;
4585         int status;
4586
4587         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4588         if (status < 0)
4589                 return status;
4590         do_renew_lease(clp, now);
4591         return 0;
4592 }
4593
4594 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4595 {
4596         return server->caps & NFS_CAP_ACLS;
4597 }
4598
4599 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4600  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4601  * the stack.
4602  */
4603 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4604
4605 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4606                 struct page **pages)
4607 {
4608         struct page *newpage, **spages;
4609         int rc = 0;
4610         size_t len;
4611         spages = pages;
4612
4613         do {
4614                 len = min_t(size_t, PAGE_SIZE, buflen);
4615                 newpage = alloc_page(GFP_KERNEL);
4616
4617                 if (newpage == NULL)
4618                         goto unwind;
4619                 memcpy(page_address(newpage), buf, len);
4620                 buf += len;
4621                 buflen -= len;
4622                 *pages++ = newpage;
4623                 rc++;
4624         } while (buflen != 0);
4625
4626         return rc;
4627
4628 unwind:
4629         for(; rc > 0; rc--)
4630                 __free_page(spages[rc-1]);
4631         return -ENOMEM;
4632 }
4633
4634 struct nfs4_cached_acl {
4635         int cached;
4636         size_t len;
4637         char data[0];
4638 };
4639
4640 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4641 {
4642         struct nfs_inode *nfsi = NFS_I(inode);
4643
4644         spin_lock(&inode->i_lock);
4645         kfree(nfsi->nfs4_acl);
4646         nfsi->nfs4_acl = acl;
4647         spin_unlock(&inode->i_lock);
4648 }
4649
4650 static void nfs4_zap_acl_attr(struct inode *inode)
4651 {
4652         nfs4_set_cached_acl(inode, NULL);
4653 }
4654
4655 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4656 {
4657         struct nfs_inode *nfsi = NFS_I(inode);
4658         struct nfs4_cached_acl *acl;
4659         int ret = -ENOENT;
4660
4661         spin_lock(&inode->i_lock);
4662         acl = nfsi->nfs4_acl;
4663         if (acl == NULL)
4664                 goto out;
4665         if (buf == NULL) /* user is just asking for length */
4666                 goto out_len;
4667         if (acl->cached == 0)
4668                 goto out;
4669         ret = -ERANGE; /* see getxattr(2) man page */
4670         if (acl->len > buflen)
4671                 goto out;
4672         memcpy(buf, acl->data, acl->len);
4673 out_len:
4674         ret = acl->len;
4675 out:
4676         spin_unlock(&inode->i_lock);
4677         return ret;
4678 }
4679
4680 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4681 {
4682         struct nfs4_cached_acl *acl;
4683         size_t buflen = sizeof(*acl) + acl_len;
4684
4685         if (buflen <= PAGE_SIZE) {
4686                 acl = kmalloc(buflen, GFP_KERNEL);
4687                 if (acl == NULL)
4688                         goto out;
4689                 acl->cached = 1;
4690                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4691         } else {
4692                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4693                 if (acl == NULL)
4694                         goto out;
4695                 acl->cached = 0;
4696         }
4697         acl->len = acl_len;
4698 out:
4699         nfs4_set_cached_acl(inode, acl);
4700 }
4701
4702 /*
4703  * The getxattr API returns the required buffer length when called with a
4704  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4705  * the required buf.  On a NULL buf, we send a page of data to the server
4706  * guessing that the ACL request can be serviced by a page. If so, we cache
4707  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4708  * the cache. If not so, we throw away the page, and cache the required
4709  * length. The next getxattr call will then produce another round trip to
4710  * the server, this time with the input buf of the required size.
4711  */
4712 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4713 {
4714         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
4715         struct nfs_getaclargs args = {
4716                 .fh = NFS_FH(inode),
4717                 .acl_pages = pages,
4718                 .acl_len = buflen,
4719         };
4720         struct nfs_getaclres res = {
4721                 .acl_len = buflen,
4722         };
4723         struct rpc_message msg = {
4724                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4725                 .rpc_argp = &args,
4726                 .rpc_resp = &res,
4727         };
4728         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
4729         int ret = -ENOMEM, i;
4730
4731         if (npages > ARRAY_SIZE(pages))
4732                 return -ERANGE;
4733
4734         for (i = 0; i < npages; i++) {
4735                 pages[i] = alloc_page(GFP_KERNEL);
4736                 if (!pages[i])
4737                         goto out_free;
4738         }
4739
4740         /* for decoding across pages */
4741         res.acl_scratch = alloc_page(GFP_KERNEL);
4742         if (!res.acl_scratch)
4743                 goto out_free;
4744
4745         args.acl_len = npages * PAGE_SIZE;
4746
4747         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4748                 __func__, buf, buflen, npages, args.acl_len);
4749         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4750                              &msg, &args.seq_args, &res.seq_res, 0);
4751         if (ret)
4752                 goto out_free;
4753
4754         /* Handle the case where the passed-in buffer is too short */
4755         if (res.acl_flags & NFS4_ACL_TRUNC) {
4756                 /* Did the user only issue a request for the acl length? */
4757                 if (buf == NULL)
4758                         goto out_ok;
4759                 ret = -ERANGE;
4760                 goto out_free;
4761         }
4762         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4763         if (buf) {
4764                 if (res.acl_len > buflen) {
4765                         ret = -ERANGE;
4766                         goto out_free;
4767                 }
4768                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4769         }
4770 out_ok:
4771         ret = res.acl_len;
4772 out_free:
4773         for (i = 0; i < npages; i++)
4774                 if (pages[i])
4775                         __free_page(pages[i]);
4776         if (res.acl_scratch)
4777                 __free_page(res.acl_scratch);
4778         return ret;
4779 }
4780
4781 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4782 {
4783         struct nfs4_exception exception = { };
4784         ssize_t ret;
4785         do {
4786                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4787                 trace_nfs4_get_acl(inode, ret);
4788                 if (ret >= 0)
4789                         break;
4790                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4791         } while (exception.retry);
4792         return ret;
4793 }
4794
4795 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4796 {
4797         struct nfs_server *server = NFS_SERVER(inode);
4798         int ret;
4799
4800         if (!nfs4_server_supports_acls(server))
4801                 return -EOPNOTSUPP;
4802         ret = nfs_revalidate_inode(server, inode);
4803         if (ret < 0)
4804                 return ret;
4805         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4806                 nfs_zap_acl_cache(inode);
4807         ret = nfs4_read_cached_acl(inode, buf, buflen);
4808         if (ret != -ENOENT)
4809                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4810                  * but no cached acl data, just the acl length */
4811                 return ret;
4812         return nfs4_get_acl_uncached(inode, buf, buflen);
4813 }
4814
4815 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4816 {
4817         struct nfs_server *server = NFS_SERVER(inode);
4818         struct page *pages[NFS4ACL_MAXPAGES];
4819         struct nfs_setaclargs arg = {
4820                 .fh             = NFS_FH(inode),
4821                 .acl_pages      = pages,
4822                 .acl_len        = buflen,
4823         };
4824         struct nfs_setaclres res;
4825         struct rpc_message msg = {
4826                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4827                 .rpc_argp       = &arg,
4828                 .rpc_resp       = &res,
4829         };
4830         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4831         int ret, i;
4832
4833         if (!nfs4_server_supports_acls(server))
4834                 return -EOPNOTSUPP;
4835         if (npages > ARRAY_SIZE(pages))
4836                 return -ERANGE;
4837         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4838         if (i < 0)
4839                 return i;
4840         nfs4_inode_return_delegation(inode);
4841         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4842
4843         /*
4844          * Free each page after tx, so the only ref left is
4845          * held by the network stack
4846          */
4847         for (; i > 0; i--)
4848                 put_page(pages[i-1]);
4849
4850         /*
4851          * Acl update can result in inode attribute update.
4852          * so mark the attribute cache invalid.
4853          */
4854         spin_lock(&inode->i_lock);
4855         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4856         spin_unlock(&inode->i_lock);
4857         nfs_access_zap_cache(inode);
4858         nfs_zap_acl_cache(inode);
4859         return ret;
4860 }
4861
4862 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4863 {
4864         struct nfs4_exception exception = { };
4865         int err;
4866         do {
4867                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4868                 trace_nfs4_set_acl(inode, err);
4869                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4870                                 &exception);
4871         } while (exception.retry);
4872         return err;
4873 }
4874
4875 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4876 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4877                                         size_t buflen)
4878 {
4879         struct nfs_server *server = NFS_SERVER(inode);
4880         struct nfs_fattr fattr;
4881         struct nfs4_label label = {0, 0, buflen, buf};
4882
4883         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4884         struct nfs4_getattr_arg arg = {
4885                 .fh             = NFS_FH(inode),
4886                 .bitmask        = bitmask,
4887         };
4888         struct nfs4_getattr_res res = {
4889                 .fattr          = &fattr,
4890                 .label          = &label,
4891                 .server         = server,
4892         };
4893         struct rpc_message msg = {
4894                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4895                 .rpc_argp       = &arg,
4896                 .rpc_resp       = &res,
4897         };
4898         int ret;
4899
4900         nfs_fattr_init(&fattr);
4901
4902         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4903         if (ret)
4904                 return ret;
4905         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4906                 return -ENOENT;
4907         if (buflen < label.len)
4908                 return -ERANGE;
4909         return 0;
4910 }
4911
4912 static int nfs4_get_security_label(struct inode *inode, void *buf,
4913                                         size_t buflen)
4914 {
4915         struct nfs4_exception exception = { };
4916         int err;
4917
4918         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4919                 return -EOPNOTSUPP;
4920
4921         do {
4922                 err = _nfs4_get_security_label(inode, buf, buflen);
4923                 trace_nfs4_get_security_label(inode, err);
4924                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4925                                 &exception);
4926         } while (exception.retry);
4927         return err;
4928 }
4929
4930 static int _nfs4_do_set_security_label(struct inode *inode,
4931                 struct nfs4_label *ilabel,
4932                 struct nfs_fattr *fattr,
4933                 struct nfs4_label *olabel)
4934 {
4935
4936         struct iattr sattr = {0};
4937         struct nfs_server *server = NFS_SERVER(inode);
4938         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4939         struct nfs_setattrargs arg = {
4940                 .fh             = NFS_FH(inode),
4941                 .iap            = &sattr,
4942                 .server         = server,
4943                 .bitmask        = bitmask,
4944                 .label          = ilabel,
4945         };
4946         struct nfs_setattrres res = {
4947                 .fattr          = fattr,
4948                 .label          = olabel,
4949                 .server         = server,
4950         };
4951         struct rpc_message msg = {
4952                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4953                 .rpc_argp       = &arg,
4954                 .rpc_resp       = &res,
4955         };
4956         int status;
4957
4958         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4959
4960         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4961         if (status)
4962                 dprintk("%s failed: %d\n", __func__, status);
4963
4964         return status;
4965 }
4966
4967 static int nfs4_do_set_security_label(struct inode *inode,
4968                 struct nfs4_label *ilabel,
4969                 struct nfs_fattr *fattr,
4970                 struct nfs4_label *olabel)
4971 {
4972         struct nfs4_exception exception = { };
4973         int err;
4974
4975         do {
4976                 err = _nfs4_do_set_security_label(inode, ilabel,
4977                                 fattr, olabel);
4978                 trace_nfs4_set_security_label(inode, err);
4979                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4980                                 &exception);
4981         } while (exception.retry);
4982         return err;
4983 }
4984
4985 static int
4986 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4987 {
4988         struct nfs4_label ilabel, *olabel = NULL;
4989         struct nfs_fattr fattr;
4990         struct rpc_cred *cred;
4991         struct inode *inode = d_inode(dentry);
4992         int status;
4993
4994         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4995                 return -EOPNOTSUPP;
4996
4997         nfs_fattr_init(&fattr);
4998
4999         ilabel.pi = 0;
5000         ilabel.lfs = 0;
5001         ilabel.label = (char *)buf;
5002         ilabel.len = buflen;
5003
5004         cred = rpc_lookup_cred();
5005         if (IS_ERR(cred))
5006                 return PTR_ERR(cred);
5007
5008         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5009         if (IS_ERR(olabel)) {
5010                 status = -PTR_ERR(olabel);
5011                 goto out;
5012         }
5013
5014         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5015         if (status == 0)
5016                 nfs_setsecurity(inode, &fattr, olabel);
5017
5018         nfs4_label_free(olabel);
5019 out:
5020         put_rpccred(cred);
5021         return status;
5022 }
5023 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5024
5025
5026 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5027                                     nfs4_verifier *bootverf)
5028 {
5029         __be32 verf[2];
5030
5031         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5032                 /* An impossible timestamp guarantees this value
5033                  * will never match a generated boot time. */
5034                 verf[0] = 0;
5035                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5036         } else {
5037                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5038                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5039                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5040         }
5041         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5042 }
5043
5044 static int
5045 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5046 {
5047         int result;
5048         size_t len;
5049         char *str;
5050
5051         if (clp->cl_owner_id != NULL)
5052                 return 0;
5053
5054         rcu_read_lock();
5055         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5056                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5057                 1 +
5058                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5059                 1;
5060         rcu_read_unlock();
5061
5062         if (len > NFS4_OPAQUE_LIMIT + 1)
5063                 return -EINVAL;
5064
5065         /*
5066          * Since this string is allocated at mount time, and held until the
5067          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5068          * about a memory-reclaim deadlock.
5069          */
5070         str = kmalloc(len, GFP_KERNEL);
5071         if (!str)
5072                 return -ENOMEM;
5073
5074         rcu_read_lock();
5075         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5076                         clp->cl_ipaddr,
5077                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5078                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5079         rcu_read_unlock();
5080
5081         clp->cl_owner_id = str;
5082         return 0;
5083 }
5084
5085 static int
5086 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5087 {
5088         int result;
5089         size_t len;
5090         char *str;
5091
5092         len = 10 + 10 + 1 + 10 + 1 +
5093                 strlen(nfs4_client_id_uniquifier) + 1 +
5094                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5095
5096         if (len > NFS4_OPAQUE_LIMIT + 1)
5097                 return -EINVAL;
5098
5099         /*
5100          * Since this string is allocated at mount time, and held until the
5101          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5102          * about a memory-reclaim deadlock.
5103          */
5104         str = kmalloc(len, GFP_KERNEL);
5105         if (!str)
5106                 return -ENOMEM;
5107
5108         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5109                         clp->rpc_ops->version, clp->cl_minorversion,
5110                         nfs4_client_id_uniquifier,
5111                         clp->cl_rpcclient->cl_nodename);
5112         clp->cl_owner_id = str;
5113         return 0;
5114 }
5115
5116 static int
5117 nfs4_init_uniform_client_string(struct nfs_client *clp)
5118 {
5119         int result;
5120         size_t len;
5121         char *str;
5122
5123         if (clp->cl_owner_id != NULL)
5124                 return 0;
5125
5126         if (nfs4_client_id_uniquifier[0] != '\0')
5127                 return nfs4_init_uniquifier_client_string(clp);
5128
5129         len = 10 + 10 + 1 + 10 + 1 +
5130                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5131
5132         if (len > NFS4_OPAQUE_LIMIT + 1)
5133                 return -EINVAL;
5134
5135         /*
5136          * Since this string is allocated at mount time, and held until the
5137          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5138          * about a memory-reclaim deadlock.
5139          */
5140         str = kmalloc(len, GFP_KERNEL);
5141         if (!str)
5142                 return -ENOMEM;
5143
5144         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5145                         clp->rpc_ops->version, clp->cl_minorversion,
5146                         clp->cl_rpcclient->cl_nodename);
5147         clp->cl_owner_id = str;
5148         return 0;
5149 }
5150
5151 /*
5152  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5153  * services.  Advertise one based on the address family of the
5154  * clientaddr.
5155  */
5156 static unsigned int
5157 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5158 {
5159         if (strchr(clp->cl_ipaddr, ':') != NULL)
5160                 return scnprintf(buf, len, "tcp6");
5161         else
5162                 return scnprintf(buf, len, "tcp");
5163 }
5164
5165 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5166 {
5167         struct nfs4_setclientid *sc = calldata;
5168
5169         if (task->tk_status == 0)
5170                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5171 }
5172
5173 static const struct rpc_call_ops nfs4_setclientid_ops = {
5174         .rpc_call_done = nfs4_setclientid_done,
5175 };
5176
5177 /**
5178  * nfs4_proc_setclientid - Negotiate client ID
5179  * @clp: state data structure
5180  * @program: RPC program for NFSv4 callback service
5181  * @port: IP port number for NFS4 callback service
5182  * @cred: RPC credential to use for this call
5183  * @res: where to place the result
5184  *
5185  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5186  */
5187 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5188                 unsigned short port, struct rpc_cred *cred,
5189                 struct nfs4_setclientid_res *res)
5190 {
5191         nfs4_verifier sc_verifier;
5192         struct nfs4_setclientid setclientid = {
5193                 .sc_verifier = &sc_verifier,
5194                 .sc_prog = program,
5195                 .sc_clnt = clp,
5196         };
5197         struct rpc_message msg = {
5198                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5199                 .rpc_argp = &setclientid,
5200                 .rpc_resp = res,
5201                 .rpc_cred = cred,
5202         };
5203         struct rpc_task *task;
5204         struct rpc_task_setup task_setup_data = {
5205                 .rpc_client = clp->cl_rpcclient,
5206                 .rpc_message = &msg,
5207                 .callback_ops = &nfs4_setclientid_ops,
5208                 .callback_data = &setclientid,
5209                 .flags = RPC_TASK_TIMEOUT,
5210         };
5211         int status;
5212
5213         /* nfs_client_id4 */
5214         nfs4_init_boot_verifier(clp, &sc_verifier);
5215
5216         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5217                 status = nfs4_init_uniform_client_string(clp);
5218         else
5219                 status = nfs4_init_nonuniform_client_string(clp);
5220
5221         if (status)
5222                 goto out;
5223
5224         /* cb_client4 */
5225         setclientid.sc_netid_len =
5226                                 nfs4_init_callback_netid(clp,
5227                                                 setclientid.sc_netid,
5228                                                 sizeof(setclientid.sc_netid));
5229         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5230                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5231                                 clp->cl_ipaddr, port >> 8, port & 255);
5232
5233         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5234                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5235                 clp->cl_owner_id);
5236         task = rpc_run_task(&task_setup_data);
5237         if (IS_ERR(task)) {
5238                 status = PTR_ERR(task);
5239                 goto out;
5240         }
5241         status = task->tk_status;
5242         if (setclientid.sc_cred) {
5243                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5244                 put_rpccred(setclientid.sc_cred);
5245         }
5246         rpc_put_task(task);
5247 out:
5248         trace_nfs4_setclientid(clp, status);
5249         dprintk("NFS reply setclientid: %d\n", status);
5250         return status;
5251 }
5252
5253 /**
5254  * nfs4_proc_setclientid_confirm - Confirm client ID
5255  * @clp: state data structure
5256  * @res: result of a previous SETCLIENTID
5257  * @cred: RPC credential to use for this call
5258  *
5259  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5260  */
5261 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5262                 struct nfs4_setclientid_res *arg,
5263                 struct rpc_cred *cred)
5264 {
5265         struct rpc_message msg = {
5266                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5267                 .rpc_argp = arg,
5268                 .rpc_cred = cred,
5269         };
5270         int status;
5271
5272         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5273                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5274                 clp->cl_clientid);
5275         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5276         trace_nfs4_setclientid_confirm(clp, status);
5277         dprintk("NFS reply setclientid_confirm: %d\n", status);
5278         return status;
5279 }
5280
5281 struct nfs4_delegreturndata {
5282         struct nfs4_delegreturnargs args;
5283         struct nfs4_delegreturnres res;
5284         struct nfs_fh fh;
5285         nfs4_stateid stateid;
5286         unsigned long timestamp;
5287         struct nfs_fattr fattr;
5288         int rpc_status;
5289         struct inode *inode;
5290         bool roc;
5291         u32 roc_barrier;
5292 };
5293
5294 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5295 {
5296         struct nfs4_delegreturndata *data = calldata;
5297
5298         if (!nfs4_sequence_done(task, &data->res.seq_res))
5299                 return;
5300
5301         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5302         switch (task->tk_status) {
5303         case 0:
5304                 renew_lease(data->res.server, data->timestamp);
5305         case -NFS4ERR_ADMIN_REVOKED:
5306         case -NFS4ERR_DELEG_REVOKED:
5307         case -NFS4ERR_BAD_STATEID:
5308         case -NFS4ERR_OLD_STATEID:
5309         case -NFS4ERR_STALE_STATEID:
5310         case -NFS4ERR_EXPIRED:
5311                 task->tk_status = 0;
5312                 if (data->roc)
5313                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5314                 break;
5315         default:
5316                 if (nfs4_async_handle_error(task, data->res.server,
5317                                             NULL, NULL) == -EAGAIN) {
5318                         rpc_restart_call_prepare(task);
5319                         return;
5320                 }
5321         }
5322         data->rpc_status = task->tk_status;
5323 }
5324
5325 static void nfs4_delegreturn_release(void *calldata)
5326 {
5327         struct nfs4_delegreturndata *data = calldata;
5328         struct inode *inode = data->inode;
5329
5330         if (inode) {
5331                 if (data->roc)
5332                         pnfs_roc_release(inode);
5333                 nfs_iput_and_deactive(inode);
5334         }
5335         kfree(calldata);
5336 }
5337
5338 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5339 {
5340         struct nfs4_delegreturndata *d_data;
5341
5342         d_data = (struct nfs4_delegreturndata *)data;
5343
5344         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5345                 return;
5346
5347         if (d_data->roc)
5348                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5349
5350         nfs4_setup_sequence(d_data->res.server,
5351                         &d_data->args.seq_args,
5352                         &d_data->res.seq_res,
5353                         task);
5354 }
5355
5356 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5357         .rpc_call_prepare = nfs4_delegreturn_prepare,
5358         .rpc_call_done = nfs4_delegreturn_done,
5359         .rpc_release = nfs4_delegreturn_release,
5360 };
5361
5362 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5363 {
5364         struct nfs4_delegreturndata *data;
5365         struct nfs_server *server = NFS_SERVER(inode);
5366         struct rpc_task *task;
5367         struct rpc_message msg = {
5368                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5369                 .rpc_cred = cred,
5370         };
5371         struct rpc_task_setup task_setup_data = {
5372                 .rpc_client = server->client,
5373                 .rpc_message = &msg,
5374                 .callback_ops = &nfs4_delegreturn_ops,
5375                 .flags = RPC_TASK_ASYNC,
5376         };
5377         int status = 0;
5378
5379         data = kzalloc(sizeof(*data), GFP_NOFS);
5380         if (data == NULL)
5381                 return -ENOMEM;
5382         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5383         data->args.fhandle = &data->fh;
5384         data->args.stateid = &data->stateid;
5385         data->args.bitmask = server->cache_consistency_bitmask;
5386         nfs_copy_fh(&data->fh, NFS_FH(inode));
5387         nfs4_stateid_copy(&data->stateid, stateid);
5388         data->res.fattr = &data->fattr;
5389         data->res.server = server;
5390         nfs_fattr_init(data->res.fattr);
5391         data->timestamp = jiffies;
5392         data->rpc_status = 0;
5393         data->inode = nfs_igrab_and_active(inode);
5394         if (data->inode)
5395                 data->roc = nfs4_roc(inode);
5396
5397         task_setup_data.callback_data = data;
5398         msg.rpc_argp = &data->args;
5399         msg.rpc_resp = &data->res;
5400         task = rpc_run_task(&task_setup_data);
5401         if (IS_ERR(task))
5402                 return PTR_ERR(task);
5403         if (!issync)
5404                 goto out;
5405         status = nfs4_wait_for_completion_rpc_task(task);
5406         if (status != 0)
5407                 goto out;
5408         status = data->rpc_status;
5409         if (status == 0)
5410                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5411         else
5412                 nfs_refresh_inode(inode, &data->fattr);
5413 out:
5414         rpc_put_task(task);
5415         return status;
5416 }
5417
5418 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5419 {
5420         struct nfs_server *server = NFS_SERVER(inode);
5421         struct nfs4_exception exception = { };
5422         int err;
5423         do {
5424                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5425                 trace_nfs4_delegreturn(inode, err);
5426                 switch (err) {
5427                         case -NFS4ERR_STALE_STATEID:
5428                         case -NFS4ERR_EXPIRED:
5429                         case 0:
5430                                 return 0;
5431                 }
5432                 err = nfs4_handle_exception(server, err, &exception);
5433         } while (exception.retry);
5434         return err;
5435 }
5436
5437 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5438 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5439
5440 /*
5441  * sleep, with exponential backoff, and retry the LOCK operation.
5442  */
5443 static unsigned long
5444 nfs4_set_lock_task_retry(unsigned long timeout)
5445 {
5446         freezable_schedule_timeout_killable_unsafe(timeout);
5447         timeout <<= 1;
5448         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5449                 return NFS4_LOCK_MAXTIMEOUT;
5450         return timeout;
5451 }
5452
5453 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5454 {
5455         struct inode *inode = state->inode;
5456         struct nfs_server *server = NFS_SERVER(inode);
5457         struct nfs_client *clp = server->nfs_client;
5458         struct nfs_lockt_args arg = {
5459                 .fh = NFS_FH(inode),
5460                 .fl = request,
5461         };
5462         struct nfs_lockt_res res = {
5463                 .denied = request,
5464         };
5465         struct rpc_message msg = {
5466                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5467                 .rpc_argp       = &arg,
5468                 .rpc_resp       = &res,
5469                 .rpc_cred       = state->owner->so_cred,
5470         };
5471         struct nfs4_lock_state *lsp;
5472         int status;
5473
5474         arg.lock_owner.clientid = clp->cl_clientid;
5475         status = nfs4_set_lock_state(state, request);
5476         if (status != 0)
5477                 goto out;
5478         lsp = request->fl_u.nfs4_fl.owner;
5479         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5480         arg.lock_owner.s_dev = server->s_dev;
5481         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5482         switch (status) {
5483                 case 0:
5484                         request->fl_type = F_UNLCK;
5485                         break;
5486                 case -NFS4ERR_DENIED:
5487                         status = 0;
5488         }
5489         request->fl_ops->fl_release_private(request);
5490         request->fl_ops = NULL;
5491 out:
5492         return status;
5493 }
5494
5495 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5496 {
5497         struct nfs4_exception exception = { };
5498         int err;
5499
5500         do {
5501                 err = _nfs4_proc_getlk(state, cmd, request);
5502                 trace_nfs4_get_lock(request, state, cmd, err);
5503                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5504                                 &exception);
5505         } while (exception.retry);
5506         return err;
5507 }
5508
5509 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5510 {
5511         return locks_lock_inode_wait(inode, fl);
5512 }
5513
5514 struct nfs4_unlockdata {
5515         struct nfs_locku_args arg;
5516         struct nfs_locku_res res;
5517         struct nfs4_lock_state *lsp;
5518         struct nfs_open_context *ctx;
5519         struct file_lock fl;
5520         struct nfs_server *server;
5521         unsigned long timestamp;
5522 };
5523
5524 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5525                 struct nfs_open_context *ctx,
5526                 struct nfs4_lock_state *lsp,
5527                 struct nfs_seqid *seqid)
5528 {
5529         struct nfs4_unlockdata *p;
5530         struct inode *inode = lsp->ls_state->inode;
5531
5532         p = kzalloc(sizeof(*p), GFP_NOFS);
5533         if (p == NULL)
5534                 return NULL;
5535         p->arg.fh = NFS_FH(inode);
5536         p->arg.fl = &p->fl;
5537         p->arg.seqid = seqid;
5538         p->res.seqid = seqid;
5539         p->lsp = lsp;
5540         atomic_inc(&lsp->ls_count);
5541         /* Ensure we don't close file until we're done freeing locks! */
5542         p->ctx = get_nfs_open_context(ctx);
5543         memcpy(&p->fl, fl, sizeof(p->fl));
5544         p->server = NFS_SERVER(inode);
5545         return p;
5546 }
5547
5548 static void nfs4_locku_release_calldata(void *data)
5549 {
5550         struct nfs4_unlockdata *calldata = data;
5551         nfs_free_seqid(calldata->arg.seqid);
5552         nfs4_put_lock_state(calldata->lsp);
5553         put_nfs_open_context(calldata->ctx);
5554         kfree(calldata);
5555 }
5556
5557 static void nfs4_locku_done(struct rpc_task *task, void *data)
5558 {
5559         struct nfs4_unlockdata *calldata = data;
5560
5561         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5562                 return;
5563         switch (task->tk_status) {
5564                 case 0:
5565                         renew_lease(calldata->server, calldata->timestamp);
5566                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5567                         if (nfs4_update_lock_stateid(calldata->lsp,
5568                                         &calldata->res.stateid))
5569                                 break;
5570                 case -NFS4ERR_BAD_STATEID:
5571                 case -NFS4ERR_OLD_STATEID:
5572                 case -NFS4ERR_STALE_STATEID:
5573                 case -NFS4ERR_EXPIRED:
5574                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5575                                                 &calldata->lsp->ls_stateid))
5576                                 rpc_restart_call_prepare(task);
5577                         break;
5578                 default:
5579                         if (nfs4_async_handle_error(task, calldata->server,
5580                                                     NULL, NULL) == -EAGAIN)
5581                                 rpc_restart_call_prepare(task);
5582         }
5583         nfs_release_seqid(calldata->arg.seqid);
5584 }
5585
5586 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5587 {
5588         struct nfs4_unlockdata *calldata = data;
5589
5590         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5591                 goto out_wait;
5592         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5593         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5594                 /* Note: exit _without_ running nfs4_locku_done */
5595                 goto out_no_action;
5596         }
5597         calldata->timestamp = jiffies;
5598         if (nfs4_setup_sequence(calldata->server,
5599                                 &calldata->arg.seq_args,
5600                                 &calldata->res.seq_res,
5601                                 task) != 0)
5602                 nfs_release_seqid(calldata->arg.seqid);
5603         return;
5604 out_no_action:
5605         task->tk_action = NULL;
5606 out_wait:
5607         nfs4_sequence_done(task, &calldata->res.seq_res);
5608 }
5609
5610 static const struct rpc_call_ops nfs4_locku_ops = {
5611         .rpc_call_prepare = nfs4_locku_prepare,
5612         .rpc_call_done = nfs4_locku_done,
5613         .rpc_release = nfs4_locku_release_calldata,
5614 };
5615
5616 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5617                 struct nfs_open_context *ctx,
5618                 struct nfs4_lock_state *lsp,
5619                 struct nfs_seqid *seqid)
5620 {
5621         struct nfs4_unlockdata *data;
5622         struct rpc_message msg = {
5623                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5624                 .rpc_cred = ctx->cred,
5625         };
5626         struct rpc_task_setup task_setup_data = {
5627                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5628                 .rpc_message = &msg,
5629                 .callback_ops = &nfs4_locku_ops,
5630                 .workqueue = nfsiod_workqueue,
5631                 .flags = RPC_TASK_ASYNC,
5632         };
5633
5634         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5635                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5636
5637         /* Ensure this is an unlock - when canceling a lock, the
5638          * canceled lock is passed in, and it won't be an unlock.
5639          */
5640         fl->fl_type = F_UNLCK;
5641
5642         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5643         if (data == NULL) {
5644                 nfs_free_seqid(seqid);
5645                 return ERR_PTR(-ENOMEM);
5646         }
5647
5648         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5649         msg.rpc_argp = &data->arg;
5650         msg.rpc_resp = &data->res;
5651         task_setup_data.callback_data = data;
5652         return rpc_run_task(&task_setup_data);
5653 }
5654
5655 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5656 {
5657         struct inode *inode = state->inode;
5658         struct nfs4_state_owner *sp = state->owner;
5659         struct nfs_inode *nfsi = NFS_I(inode);
5660         struct nfs_seqid *seqid;
5661         struct nfs4_lock_state *lsp;
5662         struct rpc_task *task;
5663         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5664         int status = 0;
5665         unsigned char fl_flags = request->fl_flags;
5666
5667         status = nfs4_set_lock_state(state, request);
5668         /* Unlock _before_ we do the RPC call */
5669         request->fl_flags |= FL_EXISTS;
5670         /* Exclude nfs_delegation_claim_locks() */
5671         mutex_lock(&sp->so_delegreturn_mutex);
5672         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5673         down_read(&nfsi->rwsem);
5674         if (do_vfs_lock(inode, request) == -ENOENT) {
5675                 up_read(&nfsi->rwsem);
5676                 mutex_unlock(&sp->so_delegreturn_mutex);
5677                 goto out;
5678         }
5679         up_read(&nfsi->rwsem);
5680         mutex_unlock(&sp->so_delegreturn_mutex);
5681         if (status != 0)
5682                 goto out;
5683         /* Is this a delegated lock? */
5684         lsp = request->fl_u.nfs4_fl.owner;
5685         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5686                 goto out;
5687         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5688         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5689         status = -ENOMEM;
5690         if (IS_ERR(seqid))
5691                 goto out;
5692         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5693         status = PTR_ERR(task);
5694         if (IS_ERR(task))
5695                 goto out;
5696         status = nfs4_wait_for_completion_rpc_task(task);
5697         rpc_put_task(task);
5698 out:
5699         request->fl_flags = fl_flags;
5700         trace_nfs4_unlock(request, state, F_SETLK, status);
5701         return status;
5702 }
5703
5704 struct nfs4_lockdata {
5705         struct nfs_lock_args arg;
5706         struct nfs_lock_res res;
5707         struct nfs4_lock_state *lsp;
5708         struct nfs_open_context *ctx;
5709         struct file_lock fl;
5710         unsigned long timestamp;
5711         int rpc_status;
5712         int cancelled;
5713         struct nfs_server *server;
5714 };
5715
5716 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5717                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5718                 gfp_t gfp_mask)
5719 {
5720         struct nfs4_lockdata *p;
5721         struct inode *inode = lsp->ls_state->inode;
5722         struct nfs_server *server = NFS_SERVER(inode);
5723         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5724
5725         p = kzalloc(sizeof(*p), gfp_mask);
5726         if (p == NULL)
5727                 return NULL;
5728
5729         p->arg.fh = NFS_FH(inode);
5730         p->arg.fl = &p->fl;
5731         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5732         if (IS_ERR(p->arg.open_seqid))
5733                 goto out_free;
5734         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5735         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5736         if (IS_ERR(p->arg.lock_seqid))
5737                 goto out_free_seqid;
5738         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5739         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5740         p->arg.lock_owner.s_dev = server->s_dev;
5741         p->res.lock_seqid = p->arg.lock_seqid;
5742         p->lsp = lsp;
5743         p->server = server;
5744         atomic_inc(&lsp->ls_count);
5745         p->ctx = get_nfs_open_context(ctx);
5746         get_file(fl->fl_file);
5747         memcpy(&p->fl, fl, sizeof(p->fl));
5748         return p;
5749 out_free_seqid:
5750         nfs_free_seqid(p->arg.open_seqid);
5751 out_free:
5752         kfree(p);
5753         return NULL;
5754 }
5755
5756 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5757 {
5758         struct nfs4_lockdata *data = calldata;
5759         struct nfs4_state *state = data->lsp->ls_state;
5760
5761         dprintk("%s: begin!\n", __func__);
5762         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5763                 goto out_wait;
5764         /* Do we need to do an open_to_lock_owner? */
5765         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5766                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5767                         goto out_release_lock_seqid;
5768                 }
5769                 nfs4_stateid_copy(&data->arg.open_stateid,
5770                                 &state->open_stateid);
5771                 data->arg.new_lock_owner = 1;
5772                 data->res.open_seqid = data->arg.open_seqid;
5773         } else {
5774                 data->arg.new_lock_owner = 0;
5775                 nfs4_stateid_copy(&data->arg.lock_stateid,
5776                                 &data->lsp->ls_stateid);
5777         }
5778         if (!nfs4_valid_open_stateid(state)) {
5779                 data->rpc_status = -EBADF;
5780                 task->tk_action = NULL;
5781                 goto out_release_open_seqid;
5782         }
5783         data->timestamp = jiffies;
5784         if (nfs4_setup_sequence(data->server,
5785                                 &data->arg.seq_args,
5786                                 &data->res.seq_res,
5787                                 task) == 0)
5788                 return;
5789 out_release_open_seqid:
5790         nfs_release_seqid(data->arg.open_seqid);
5791 out_release_lock_seqid:
5792         nfs_release_seqid(data->arg.lock_seqid);
5793 out_wait:
5794         nfs4_sequence_done(task, &data->res.seq_res);
5795         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5796 }
5797
5798 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5799 {
5800         struct nfs4_lockdata *data = calldata;
5801         struct nfs4_lock_state *lsp = data->lsp;
5802
5803         dprintk("%s: begin!\n", __func__);
5804
5805         if (!nfs4_sequence_done(task, &data->res.seq_res))
5806                 return;
5807
5808         data->rpc_status = task->tk_status;
5809         switch (task->tk_status) {
5810         case 0:
5811                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5812                                 data->timestamp);
5813                 if (data->arg.new_lock) {
5814                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5815                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5816                                 rpc_restart_call_prepare(task);
5817                                 break;
5818                         }
5819                 }
5820                 if (data->arg.new_lock_owner != 0) {
5821                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5822                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5823                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5824                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5825                         rpc_restart_call_prepare(task);
5826                 break;
5827         case -NFS4ERR_BAD_STATEID:
5828         case -NFS4ERR_OLD_STATEID:
5829         case -NFS4ERR_STALE_STATEID:
5830         case -NFS4ERR_EXPIRED:
5831                 if (data->arg.new_lock_owner != 0) {
5832                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5833                                                 &lsp->ls_state->open_stateid))
5834                                 rpc_restart_call_prepare(task);
5835                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5836                                                 &lsp->ls_stateid))
5837                                 rpc_restart_call_prepare(task);
5838         }
5839         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5840 }
5841
5842 static void nfs4_lock_release(void *calldata)
5843 {
5844         struct nfs4_lockdata *data = calldata;
5845
5846         dprintk("%s: begin!\n", __func__);
5847         nfs_free_seqid(data->arg.open_seqid);
5848         if (data->cancelled != 0) {
5849                 struct rpc_task *task;
5850                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5851                                 data->arg.lock_seqid);
5852                 if (!IS_ERR(task))
5853                         rpc_put_task_async(task);
5854                 dprintk("%s: cancelling lock!\n", __func__);
5855         } else
5856                 nfs_free_seqid(data->arg.lock_seqid);
5857         nfs4_put_lock_state(data->lsp);
5858         put_nfs_open_context(data->ctx);
5859         fput(data->fl.fl_file);
5860         kfree(data);
5861         dprintk("%s: done!\n", __func__);
5862 }
5863
5864 static const struct rpc_call_ops nfs4_lock_ops = {
5865         .rpc_call_prepare = nfs4_lock_prepare,
5866         .rpc_call_done = nfs4_lock_done,
5867         .rpc_release = nfs4_lock_release,
5868 };
5869
5870 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5871 {
5872         switch (error) {
5873         case -NFS4ERR_ADMIN_REVOKED:
5874         case -NFS4ERR_BAD_STATEID:
5875                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5876                 if (new_lock_owner != 0 ||
5877                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5878                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5879                 break;
5880         case -NFS4ERR_STALE_STATEID:
5881                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5882         case -NFS4ERR_EXPIRED:
5883                 nfs4_schedule_lease_recovery(server->nfs_client);
5884         };
5885 }
5886
5887 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5888 {
5889         struct nfs4_lockdata *data;
5890         struct rpc_task *task;
5891         struct rpc_message msg = {
5892                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5893                 .rpc_cred = state->owner->so_cred,
5894         };
5895         struct rpc_task_setup task_setup_data = {
5896                 .rpc_client = NFS_CLIENT(state->inode),
5897                 .rpc_message = &msg,
5898                 .callback_ops = &nfs4_lock_ops,
5899                 .workqueue = nfsiod_workqueue,
5900                 .flags = RPC_TASK_ASYNC,
5901         };
5902         int ret;
5903
5904         dprintk("%s: begin!\n", __func__);
5905         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5906                         fl->fl_u.nfs4_fl.owner,
5907                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5908         if (data == NULL)
5909                 return -ENOMEM;
5910         if (IS_SETLKW(cmd))
5911                 data->arg.block = 1;
5912         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5913         msg.rpc_argp = &data->arg;
5914         msg.rpc_resp = &data->res;
5915         task_setup_data.callback_data = data;
5916         if (recovery_type > NFS_LOCK_NEW) {
5917                 if (recovery_type == NFS_LOCK_RECLAIM)
5918                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5919                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5920         } else
5921                 data->arg.new_lock = 1;
5922         task = rpc_run_task(&task_setup_data);
5923         if (IS_ERR(task))
5924                 return PTR_ERR(task);
5925         ret = nfs4_wait_for_completion_rpc_task(task);
5926         if (ret == 0) {
5927                 ret = data->rpc_status;
5928                 if (ret)
5929                         nfs4_handle_setlk_error(data->server, data->lsp,
5930                                         data->arg.new_lock_owner, ret);
5931         } else
5932                 data->cancelled = 1;
5933         rpc_put_task(task);
5934         dprintk("%s: done, ret = %d!\n", __func__, ret);
5935         return ret;
5936 }
5937
5938 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5939 {
5940         struct nfs_server *server = NFS_SERVER(state->inode);
5941         struct nfs4_exception exception = {
5942                 .inode = state->inode,
5943         };
5944         int err;
5945
5946         do {
5947                 /* Cache the lock if possible... */
5948                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5949                         return 0;
5950                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5951                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5952                 if (err != -NFS4ERR_DELAY)
5953                         break;
5954                 nfs4_handle_exception(server, err, &exception);
5955         } while (exception.retry);
5956         return err;
5957 }
5958
5959 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5960 {
5961         struct nfs_server *server = NFS_SERVER(state->inode);
5962         struct nfs4_exception exception = {
5963                 .inode = state->inode,
5964         };
5965         int err;
5966
5967         err = nfs4_set_lock_state(state, request);
5968         if (err != 0)
5969                 return err;
5970         if (!recover_lost_locks) {
5971                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5972                 return 0;
5973         }
5974         do {
5975                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5976                         return 0;
5977                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5978                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5979                 switch (err) {
5980                 default:
5981                         goto out;
5982                 case -NFS4ERR_GRACE:
5983                 case -NFS4ERR_DELAY:
5984                         nfs4_handle_exception(server, err, &exception);
5985                         err = 0;
5986                 }
5987         } while (exception.retry);
5988 out:
5989         return err;
5990 }
5991
5992 #if defined(CONFIG_NFS_V4_1)
5993 /**
5994  * nfs41_check_expired_locks - possibly free a lock stateid
5995  *
5996  * @state: NFSv4 state for an inode
5997  *
5998  * Returns NFS_OK if recovery for this stateid is now finished.
5999  * Otherwise a negative NFS4ERR value is returned.
6000  */
6001 static int nfs41_check_expired_locks(struct nfs4_state *state)
6002 {
6003         int status, ret = -NFS4ERR_BAD_STATEID;
6004         struct nfs4_lock_state *lsp;
6005         struct nfs_server *server = NFS_SERVER(state->inode);
6006
6007         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6008                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6009                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6010
6011                         status = nfs41_test_stateid(server,
6012                                         &lsp->ls_stateid,
6013                                         cred);
6014                         trace_nfs4_test_lock_stateid(state, lsp, status);
6015                         if (status != NFS_OK) {
6016                                 /* Free the stateid unless the server
6017                                  * informs us the stateid is unrecognized. */
6018                                 if (status != -NFS4ERR_BAD_STATEID)
6019                                         nfs41_free_stateid(server,
6020                                                         &lsp->ls_stateid,
6021                                                         cred);
6022                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6023                                 ret = status;
6024                         }
6025                 }
6026         };
6027
6028         return ret;
6029 }
6030
6031 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6032 {
6033         int status = NFS_OK;
6034
6035         if (test_bit(LK_STATE_IN_USE, &state->flags))
6036                 status = nfs41_check_expired_locks(state);
6037         if (status != NFS_OK)
6038                 status = nfs4_lock_expired(state, request);
6039         return status;
6040 }
6041 #endif
6042
6043 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6044 {
6045         struct nfs_inode *nfsi = NFS_I(state->inode);
6046         unsigned char fl_flags = request->fl_flags;
6047         int status = -ENOLCK;
6048
6049         if ((fl_flags & FL_POSIX) &&
6050                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6051                 goto out;
6052         /* Is this a delegated open? */
6053         status = nfs4_set_lock_state(state, request);
6054         if (status != 0)
6055                 goto out;
6056         request->fl_flags |= FL_ACCESS;
6057         status = do_vfs_lock(state->inode, request);
6058         if (status < 0)
6059                 goto out;
6060         down_read(&nfsi->rwsem);
6061         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6062                 /* Yes: cache locks! */
6063                 /* ...but avoid races with delegation recall... */
6064                 request->fl_flags = fl_flags & ~FL_SLEEP;
6065                 status = do_vfs_lock(state->inode, request);
6066                 up_read(&nfsi->rwsem);
6067                 goto out;
6068         }
6069         up_read(&nfsi->rwsem);
6070         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6071 out:
6072         request->fl_flags = fl_flags;
6073         return status;
6074 }
6075
6076 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6077 {
6078         struct nfs4_exception exception = {
6079                 .state = state,
6080                 .inode = state->inode,
6081         };
6082         int err;
6083
6084         do {
6085                 err = _nfs4_proc_setlk(state, cmd, request);
6086                 trace_nfs4_set_lock(request, state, cmd, err);
6087                 if (err == -NFS4ERR_DENIED)
6088                         err = -EAGAIN;
6089                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6090                                 err, &exception);
6091         } while (exception.retry);
6092         return err;
6093 }
6094
6095 static int
6096 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6097 {
6098         struct nfs_open_context *ctx;
6099         struct nfs4_state *state;
6100         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6101         int status;
6102
6103         /* verify open state */
6104         ctx = nfs_file_open_context(filp);
6105         state = ctx->state;
6106
6107         if (request->fl_start < 0 || request->fl_end < 0)
6108                 return -EINVAL;
6109
6110         if (IS_GETLK(cmd)) {
6111                 if (state != NULL)
6112                         return nfs4_proc_getlk(state, F_GETLK, request);
6113                 return 0;
6114         }
6115
6116         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6117                 return -EINVAL;
6118
6119         if (request->fl_type == F_UNLCK) {
6120                 if (state != NULL)
6121                         return nfs4_proc_unlck(state, cmd, request);
6122                 return 0;
6123         }
6124
6125         if (state == NULL)
6126                 return -ENOLCK;
6127         /*
6128          * Don't rely on the VFS having checked the file open mode,
6129          * since it won't do this for flock() locks.
6130          */
6131         switch (request->fl_type) {
6132         case F_RDLCK:
6133                 if (!(filp->f_mode & FMODE_READ))
6134                         return -EBADF;
6135                 break;
6136         case F_WRLCK:
6137                 if (!(filp->f_mode & FMODE_WRITE))
6138                         return -EBADF;
6139         }
6140
6141         do {
6142                 status = nfs4_proc_setlk(state, cmd, request);
6143                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6144                         break;
6145                 timeout = nfs4_set_lock_task_retry(timeout);
6146                 status = -ERESTARTSYS;
6147                 if (signalled())
6148                         break;
6149         } while(status < 0);
6150         return status;
6151 }
6152
6153 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6154 {
6155         struct nfs_server *server = NFS_SERVER(state->inode);
6156         int err;
6157
6158         err = nfs4_set_lock_state(state, fl);
6159         if (err != 0)
6160                 return err;
6161         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6162         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6163 }
6164
6165 struct nfs_release_lockowner_data {
6166         struct nfs4_lock_state *lsp;
6167         struct nfs_server *server;
6168         struct nfs_release_lockowner_args args;
6169         struct nfs_release_lockowner_res res;
6170         unsigned long timestamp;
6171 };
6172
6173 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6174 {
6175         struct nfs_release_lockowner_data *data = calldata;
6176         struct nfs_server *server = data->server;
6177         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6178                              &data->args.seq_args, &data->res.seq_res, task);
6179         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6180         data->timestamp = jiffies;
6181 }
6182
6183 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6184 {
6185         struct nfs_release_lockowner_data *data = calldata;
6186         struct nfs_server *server = data->server;
6187
6188         nfs40_sequence_done(task, &data->res.seq_res);
6189
6190         switch (task->tk_status) {
6191         case 0:
6192                 renew_lease(server, data->timestamp);
6193                 break;
6194         case -NFS4ERR_STALE_CLIENTID:
6195         case -NFS4ERR_EXPIRED:
6196                 nfs4_schedule_lease_recovery(server->nfs_client);
6197                 break;
6198         case -NFS4ERR_LEASE_MOVED:
6199         case -NFS4ERR_DELAY:
6200                 if (nfs4_async_handle_error(task, server,
6201                                             NULL, NULL) == -EAGAIN)
6202                         rpc_restart_call_prepare(task);
6203         }
6204 }
6205
6206 static void nfs4_release_lockowner_release(void *calldata)
6207 {
6208         struct nfs_release_lockowner_data *data = calldata;
6209         nfs4_free_lock_state(data->server, data->lsp);
6210         kfree(calldata);
6211 }
6212
6213 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6214         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6215         .rpc_call_done = nfs4_release_lockowner_done,
6216         .rpc_release = nfs4_release_lockowner_release,
6217 };
6218
6219 static void
6220 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6221 {
6222         struct nfs_release_lockowner_data *data;
6223         struct rpc_message msg = {
6224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6225         };
6226
6227         if (server->nfs_client->cl_mvops->minor_version != 0)
6228                 return;
6229
6230         data = kmalloc(sizeof(*data), GFP_NOFS);
6231         if (!data)
6232                 return;
6233         data->lsp = lsp;
6234         data->server = server;
6235         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6236         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6237         data->args.lock_owner.s_dev = server->s_dev;
6238
6239         msg.rpc_argp = &data->args;
6240         msg.rpc_resp = &data->res;
6241         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6242         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6243 }
6244
6245 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6246
6247 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6248                                    struct dentry *dentry, const char *key,
6249                                    const void *buf, size_t buflen,
6250                                    int flags)
6251 {
6252         if (strcmp(key, "") != 0)
6253                 return -EINVAL;
6254
6255         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6256 }
6257
6258 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6259                                    struct dentry *dentry, const char *key,
6260                                    void *buf, size_t buflen)
6261 {
6262         if (strcmp(key, "") != 0)
6263                 return -EINVAL;
6264
6265         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6266 }
6267
6268 static size_t nfs4_xattr_list_nfs4_acl(const struct xattr_handler *handler,
6269                                        struct dentry *dentry, char *list,
6270                                        size_t list_len, const char *name,
6271                                        size_t name_len)
6272 {
6273         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6274
6275         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6276                 return 0;
6277
6278         if (list && len <= list_len)
6279                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6280         return len;
6281 }
6282
6283 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6284 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6285 {
6286         return server->caps & NFS_CAP_SECURITY_LABEL;
6287 }
6288
6289 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6290                                      struct dentry *dentry, const char *key,
6291                                      const void *buf, size_t buflen,
6292                                      int flags)
6293 {
6294         if (security_ismaclabel(key))
6295                 return nfs4_set_security_label(dentry, buf, buflen);
6296
6297         return -EOPNOTSUPP;
6298 }
6299
6300 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6301                                      struct dentry *dentry, const char *key,
6302                                      void *buf, size_t buflen)
6303 {
6304         if (security_ismaclabel(key))
6305                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6306         return -EOPNOTSUPP;
6307 }
6308
6309 static size_t nfs4_xattr_list_nfs4_label(const struct xattr_handler *handler,
6310                                          struct dentry *dentry, char *list,
6311                                          size_t list_len, const char *name,
6312                                          size_t name_len)
6313 {
6314         size_t len = 0;
6315
6316         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6317                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6318                 if (list && len <= list_len)
6319                         security_inode_listsecurity(d_inode(dentry), list, len);
6320         }
6321         return len;
6322 }
6323
6324 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6325         .prefix = XATTR_SECURITY_PREFIX,
6326         .list   = nfs4_xattr_list_nfs4_label,
6327         .get    = nfs4_xattr_get_nfs4_label,
6328         .set    = nfs4_xattr_set_nfs4_label,
6329 };
6330 #endif
6331
6332
6333 /*
6334  * nfs_fhget will use either the mounted_on_fileid or the fileid
6335  */
6336 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6337 {
6338         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6339                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6340               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6341               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6342                 return;
6343
6344         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6345                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6346         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6347         fattr->nlink = 2;
6348 }
6349
6350 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6351                                    const struct qstr *name,
6352                                    struct nfs4_fs_locations *fs_locations,
6353                                    struct page *page)
6354 {
6355         struct nfs_server *server = NFS_SERVER(dir);
6356         u32 bitmask[3] = {
6357                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6358         };
6359         struct nfs4_fs_locations_arg args = {
6360                 .dir_fh = NFS_FH(dir),
6361                 .name = name,
6362                 .page = page,
6363                 .bitmask = bitmask,
6364         };
6365         struct nfs4_fs_locations_res res = {
6366                 .fs_locations = fs_locations,
6367         };
6368         struct rpc_message msg = {
6369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6370                 .rpc_argp = &args,
6371                 .rpc_resp = &res,
6372         };
6373         int status;
6374
6375         dprintk("%s: start\n", __func__);
6376
6377         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6378          * is not supported */
6379         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6380                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6381         else
6382                 bitmask[0] |= FATTR4_WORD0_FILEID;
6383
6384         nfs_fattr_init(&fs_locations->fattr);
6385         fs_locations->server = server;
6386         fs_locations->nlocations = 0;
6387         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6388         dprintk("%s: returned status = %d\n", __func__, status);
6389         return status;
6390 }
6391
6392 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6393                            const struct qstr *name,
6394                            struct nfs4_fs_locations *fs_locations,
6395                            struct page *page)
6396 {
6397         struct nfs4_exception exception = { };
6398         int err;
6399         do {
6400                 err = _nfs4_proc_fs_locations(client, dir, name,
6401                                 fs_locations, page);
6402                 trace_nfs4_get_fs_locations(dir, name, err);
6403                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6404                                 &exception);
6405         } while (exception.retry);
6406         return err;
6407 }
6408
6409 /*
6410  * This operation also signals the server that this client is
6411  * performing migration recovery.  The server can stop returning
6412  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6413  * appended to this compound to identify the client ID which is
6414  * performing recovery.
6415  */
6416 static int _nfs40_proc_get_locations(struct inode *inode,
6417                                      struct nfs4_fs_locations *locations,
6418                                      struct page *page, struct rpc_cred *cred)
6419 {
6420         struct nfs_server *server = NFS_SERVER(inode);
6421         struct rpc_clnt *clnt = server->client;
6422         u32 bitmask[2] = {
6423                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6424         };
6425         struct nfs4_fs_locations_arg args = {
6426                 .clientid       = server->nfs_client->cl_clientid,
6427                 .fh             = NFS_FH(inode),
6428                 .page           = page,
6429                 .bitmask        = bitmask,
6430                 .migration      = 1,            /* skip LOOKUP */
6431                 .renew          = 1,            /* append RENEW */
6432         };
6433         struct nfs4_fs_locations_res res = {
6434                 .fs_locations   = locations,
6435                 .migration      = 1,
6436                 .renew          = 1,
6437         };
6438         struct rpc_message msg = {
6439                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6440                 .rpc_argp       = &args,
6441                 .rpc_resp       = &res,
6442                 .rpc_cred       = cred,
6443         };
6444         unsigned long now = jiffies;
6445         int status;
6446
6447         nfs_fattr_init(&locations->fattr);
6448         locations->server = server;
6449         locations->nlocations = 0;
6450
6451         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6452         nfs4_set_sequence_privileged(&args.seq_args);
6453         status = nfs4_call_sync_sequence(clnt, server, &msg,
6454                                         &args.seq_args, &res.seq_res);
6455         if (status)
6456                 return status;
6457
6458         renew_lease(server, now);
6459         return 0;
6460 }
6461
6462 #ifdef CONFIG_NFS_V4_1
6463
6464 /*
6465  * This operation also signals the server that this client is
6466  * performing migration recovery.  The server can stop asserting
6467  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6468  * performing this operation is identified in the SEQUENCE
6469  * operation in this compound.
6470  *
6471  * When the client supports GETATTR(fs_locations_info), it can
6472  * be plumbed in here.
6473  */
6474 static int _nfs41_proc_get_locations(struct inode *inode,
6475                                      struct nfs4_fs_locations *locations,
6476                                      struct page *page, struct rpc_cred *cred)
6477 {
6478         struct nfs_server *server = NFS_SERVER(inode);
6479         struct rpc_clnt *clnt = server->client;
6480         u32 bitmask[2] = {
6481                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6482         };
6483         struct nfs4_fs_locations_arg args = {
6484                 .fh             = NFS_FH(inode),
6485                 .page           = page,
6486                 .bitmask        = bitmask,
6487                 .migration      = 1,            /* skip LOOKUP */
6488         };
6489         struct nfs4_fs_locations_res res = {
6490                 .fs_locations   = locations,
6491                 .migration      = 1,
6492         };
6493         struct rpc_message msg = {
6494                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6495                 .rpc_argp       = &args,
6496                 .rpc_resp       = &res,
6497                 .rpc_cred       = cred,
6498         };
6499         int status;
6500
6501         nfs_fattr_init(&locations->fattr);
6502         locations->server = server;
6503         locations->nlocations = 0;
6504
6505         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6506         nfs4_set_sequence_privileged(&args.seq_args);
6507         status = nfs4_call_sync_sequence(clnt, server, &msg,
6508                                         &args.seq_args, &res.seq_res);
6509         if (status == NFS4_OK &&
6510             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6511                 status = -NFS4ERR_LEASE_MOVED;
6512         return status;
6513 }
6514
6515 #endif  /* CONFIG_NFS_V4_1 */
6516
6517 /**
6518  * nfs4_proc_get_locations - discover locations for a migrated FSID
6519  * @inode: inode on FSID that is migrating
6520  * @locations: result of query
6521  * @page: buffer
6522  * @cred: credential to use for this operation
6523  *
6524  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6525  * operation failed, or a negative errno if a local error occurred.
6526  *
6527  * On success, "locations" is filled in, but if the server has
6528  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6529  * asserted.
6530  *
6531  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6532  * from this client that require migration recovery.
6533  */
6534 int nfs4_proc_get_locations(struct inode *inode,
6535                             struct nfs4_fs_locations *locations,
6536                             struct page *page, struct rpc_cred *cred)
6537 {
6538         struct nfs_server *server = NFS_SERVER(inode);
6539         struct nfs_client *clp = server->nfs_client;
6540         const struct nfs4_mig_recovery_ops *ops =
6541                                         clp->cl_mvops->mig_recovery_ops;
6542         struct nfs4_exception exception = { };
6543         int status;
6544
6545         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6546                 (unsigned long long)server->fsid.major,
6547                 (unsigned long long)server->fsid.minor,
6548                 clp->cl_hostname);
6549         nfs_display_fhandle(NFS_FH(inode), __func__);
6550
6551         do {
6552                 status = ops->get_locations(inode, locations, page, cred);
6553                 if (status != -NFS4ERR_DELAY)
6554                         break;
6555                 nfs4_handle_exception(server, status, &exception);
6556         } while (exception.retry);
6557         return status;
6558 }
6559
6560 /*
6561  * This operation also signals the server that this client is
6562  * performing "lease moved" recovery.  The server can stop
6563  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6564  * is appended to this compound to identify the client ID which is
6565  * performing recovery.
6566  */
6567 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6568 {
6569         struct nfs_server *server = NFS_SERVER(inode);
6570         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6571         struct rpc_clnt *clnt = server->client;
6572         struct nfs4_fsid_present_arg args = {
6573                 .fh             = NFS_FH(inode),
6574                 .clientid       = clp->cl_clientid,
6575                 .renew          = 1,            /* append RENEW */
6576         };
6577         struct nfs4_fsid_present_res res = {
6578                 .renew          = 1,
6579         };
6580         struct rpc_message msg = {
6581                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6582                 .rpc_argp       = &args,
6583                 .rpc_resp       = &res,
6584                 .rpc_cred       = cred,
6585         };
6586         unsigned long now = jiffies;
6587         int status;
6588
6589         res.fh = nfs_alloc_fhandle();
6590         if (res.fh == NULL)
6591                 return -ENOMEM;
6592
6593         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6594         nfs4_set_sequence_privileged(&args.seq_args);
6595         status = nfs4_call_sync_sequence(clnt, server, &msg,
6596                                                 &args.seq_args, &res.seq_res);
6597         nfs_free_fhandle(res.fh);
6598         if (status)
6599                 return status;
6600
6601         do_renew_lease(clp, now);
6602         return 0;
6603 }
6604
6605 #ifdef CONFIG_NFS_V4_1
6606
6607 /*
6608  * This operation also signals the server that this client is
6609  * performing "lease moved" recovery.  The server can stop asserting
6610  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6611  * this operation is identified in the SEQUENCE operation in this
6612  * compound.
6613  */
6614 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6615 {
6616         struct nfs_server *server = NFS_SERVER(inode);
6617         struct rpc_clnt *clnt = server->client;
6618         struct nfs4_fsid_present_arg args = {
6619                 .fh             = NFS_FH(inode),
6620         };
6621         struct nfs4_fsid_present_res res = {
6622         };
6623         struct rpc_message msg = {
6624                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6625                 .rpc_argp       = &args,
6626                 .rpc_resp       = &res,
6627                 .rpc_cred       = cred,
6628         };
6629         int status;
6630
6631         res.fh = nfs_alloc_fhandle();
6632         if (res.fh == NULL)
6633                 return -ENOMEM;
6634
6635         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6636         nfs4_set_sequence_privileged(&args.seq_args);
6637         status = nfs4_call_sync_sequence(clnt, server, &msg,
6638                                                 &args.seq_args, &res.seq_res);
6639         nfs_free_fhandle(res.fh);
6640         if (status == NFS4_OK &&
6641             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6642                 status = -NFS4ERR_LEASE_MOVED;
6643         return status;
6644 }
6645
6646 #endif  /* CONFIG_NFS_V4_1 */
6647
6648 /**
6649  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6650  * @inode: inode on FSID to check
6651  * @cred: credential to use for this operation
6652  *
6653  * Server indicates whether the FSID is present, moved, or not
6654  * recognized.  This operation is necessary to clear a LEASE_MOVED
6655  * condition for this client ID.
6656  *
6657  * Returns NFS4_OK if the FSID is present on this server,
6658  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6659  *  NFS4ERR code if some error occurred on the server, or a
6660  *  negative errno if a local failure occurred.
6661  */
6662 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6663 {
6664         struct nfs_server *server = NFS_SERVER(inode);
6665         struct nfs_client *clp = server->nfs_client;
6666         const struct nfs4_mig_recovery_ops *ops =
6667                                         clp->cl_mvops->mig_recovery_ops;
6668         struct nfs4_exception exception = { };
6669         int status;
6670
6671         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6672                 (unsigned long long)server->fsid.major,
6673                 (unsigned long long)server->fsid.minor,
6674                 clp->cl_hostname);
6675         nfs_display_fhandle(NFS_FH(inode), __func__);
6676
6677         do {
6678                 status = ops->fsid_present(inode, cred);
6679                 if (status != -NFS4ERR_DELAY)
6680                         break;
6681                 nfs4_handle_exception(server, status, &exception);
6682         } while (exception.retry);
6683         return status;
6684 }
6685
6686 /**
6687  * If 'use_integrity' is true and the state managment nfs_client
6688  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6689  * and the machine credential as per RFC3530bis and RFC5661 Security
6690  * Considerations sections. Otherwise, just use the user cred with the
6691  * filesystem's rpc_client.
6692  */
6693 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6694 {
6695         int status;
6696         struct nfs4_secinfo_arg args = {
6697                 .dir_fh = NFS_FH(dir),
6698                 .name   = name,
6699         };
6700         struct nfs4_secinfo_res res = {
6701                 .flavors     = flavors,
6702         };
6703         struct rpc_message msg = {
6704                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6705                 .rpc_argp = &args,
6706                 .rpc_resp = &res,
6707         };
6708         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6709         struct rpc_cred *cred = NULL;
6710
6711         if (use_integrity) {
6712                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6713                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6714                 msg.rpc_cred = cred;
6715         }
6716
6717         dprintk("NFS call  secinfo %s\n", name->name);
6718
6719         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6720                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6721
6722         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6723                                 &res.seq_res, 0);
6724         dprintk("NFS reply  secinfo: %d\n", status);
6725
6726         if (cred)
6727                 put_rpccred(cred);
6728
6729         return status;
6730 }
6731
6732 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6733                       struct nfs4_secinfo_flavors *flavors)
6734 {
6735         struct nfs4_exception exception = { };
6736         int err;
6737         do {
6738                 err = -NFS4ERR_WRONGSEC;
6739
6740                 /* try to use integrity protection with machine cred */
6741                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6742                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6743
6744                 /*
6745                  * if unable to use integrity protection, or SECINFO with
6746                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6747                  * disallowed by spec, but exists in deployed servers) use
6748                  * the current filesystem's rpc_client and the user cred.
6749                  */
6750                 if (err == -NFS4ERR_WRONGSEC)
6751                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6752
6753                 trace_nfs4_secinfo(dir, name, err);
6754                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6755                                 &exception);
6756         } while (exception.retry);
6757         return err;
6758 }
6759
6760 #ifdef CONFIG_NFS_V4_1
6761 /*
6762  * Check the exchange flags returned by the server for invalid flags, having
6763  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6764  * DS flags set.
6765  */
6766 static int nfs4_check_cl_exchange_flags(u32 flags)
6767 {
6768         if (flags & ~EXCHGID4_FLAG_MASK_R)
6769                 goto out_inval;
6770         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6771             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6772                 goto out_inval;
6773         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6774                 goto out_inval;
6775         return NFS_OK;
6776 out_inval:
6777         return -NFS4ERR_INVAL;
6778 }
6779
6780 static bool
6781 nfs41_same_server_scope(struct nfs41_server_scope *a,
6782                         struct nfs41_server_scope *b)
6783 {
6784         if (a->server_scope_sz == b->server_scope_sz &&
6785             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6786                 return true;
6787
6788         return false;
6789 }
6790
6791 /*
6792  * nfs4_proc_bind_conn_to_session()
6793  *
6794  * The 4.1 client currently uses the same TCP connection for the
6795  * fore and backchannel.
6796  */
6797 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6798 {
6799         int status;
6800         struct nfs41_bind_conn_to_session_args args = {
6801                 .client = clp,
6802                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6803         };
6804         struct nfs41_bind_conn_to_session_res res;
6805         struct rpc_message msg = {
6806                 .rpc_proc =
6807                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6808                 .rpc_argp = &args,
6809                 .rpc_resp = &res,
6810                 .rpc_cred = cred,
6811         };
6812
6813         dprintk("--> %s\n", __func__);
6814
6815         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6816         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6817                 args.dir = NFS4_CDFC4_FORE;
6818
6819         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6820         trace_nfs4_bind_conn_to_session(clp, status);
6821         if (status == 0) {
6822                 if (memcmp(res.sessionid.data,
6823                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6824                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6825                         status = -EIO;
6826                         goto out;
6827                 }
6828                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6829                         dprintk("NFS: %s: Unexpected direction from server\n",
6830                                 __func__);
6831                         status = -EIO;
6832                         goto out;
6833                 }
6834                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6835                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6836                                 __func__);
6837                         status = -EIO;
6838                         goto out;
6839                 }
6840         }
6841 out:
6842         dprintk("<-- %s status= %d\n", __func__, status);
6843         return status;
6844 }
6845
6846 /*
6847  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6848  * and operations we'd like to see to enable certain features in the allow map
6849  */
6850 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6851         .how = SP4_MACH_CRED,
6852         .enforce.u.words = {
6853                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6854                       1 << (OP_EXCHANGE_ID - 32) |
6855                       1 << (OP_CREATE_SESSION - 32) |
6856                       1 << (OP_DESTROY_SESSION - 32) |
6857                       1 << (OP_DESTROY_CLIENTID - 32)
6858         },
6859         .allow.u.words = {
6860                 [0] = 1 << (OP_CLOSE) |
6861                       1 << (OP_LOCKU) |
6862                       1 << (OP_COMMIT),
6863                 [1] = 1 << (OP_SECINFO - 32) |
6864                       1 << (OP_SECINFO_NO_NAME - 32) |
6865                       1 << (OP_TEST_STATEID - 32) |
6866                       1 << (OP_FREE_STATEID - 32) |
6867                       1 << (OP_WRITE - 32)
6868         }
6869 };
6870
6871 /*
6872  * Select the state protection mode for client `clp' given the server results
6873  * from exchange_id in `sp'.
6874  *
6875  * Returns 0 on success, negative errno otherwise.
6876  */
6877 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6878                                  struct nfs41_state_protection *sp)
6879 {
6880         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6881                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6882                       1 << (OP_EXCHANGE_ID - 32) |
6883                       1 << (OP_CREATE_SESSION - 32) |
6884                       1 << (OP_DESTROY_SESSION - 32) |
6885                       1 << (OP_DESTROY_CLIENTID - 32)
6886         };
6887         unsigned int i;
6888
6889         if (sp->how == SP4_MACH_CRED) {
6890                 /* Print state protect result */
6891                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6892                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6893                         if (test_bit(i, sp->enforce.u.longs))
6894                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6895                         if (test_bit(i, sp->allow.u.longs))
6896                                 dfprintk(MOUNT, "  allow op %d\n", i);
6897                 }
6898
6899                 /* make sure nothing is on enforce list that isn't supported */
6900                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6901                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6902                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6903                                 return -EINVAL;
6904                         }
6905                 }
6906
6907                 /*
6908                  * Minimal mode - state operations are allowed to use machine
6909                  * credential.  Note this already happens by default, so the
6910                  * client doesn't have to do anything more than the negotiation.
6911                  *
6912                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6913                  *       we're already using the machine cred for exchange_id
6914                  *       and will never use a different cred.
6915                  */
6916                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6917                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6918                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6919                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6920                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6921                         dfprintk(MOUNT, "  minimal mode enabled\n");
6922                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6923                 } else {
6924                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6925                         return -EINVAL;
6926                 }
6927
6928                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6929                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6930                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6931                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6932                 }
6933
6934                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6935                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6936                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6937                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6938                 }
6939
6940                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6941                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6942                         dfprintk(MOUNT, "  stateid mode enabled\n");
6943                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6944                 }
6945
6946                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6947                         dfprintk(MOUNT, "  write mode enabled\n");
6948                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6949                 }
6950
6951                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6952                         dfprintk(MOUNT, "  commit mode enabled\n");
6953                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6954                 }
6955         }
6956
6957         return 0;
6958 }
6959
6960 /*
6961  * _nfs4_proc_exchange_id()
6962  *
6963  * Wrapper for EXCHANGE_ID operation.
6964  */
6965 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6966         u32 sp4_how)
6967 {
6968         nfs4_verifier verifier;
6969         struct nfs41_exchange_id_args args = {
6970                 .verifier = &verifier,
6971                 .client = clp,
6972 #ifdef CONFIG_NFS_V4_1_MIGRATION
6973                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6974                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6975                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6976 #else
6977                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6978                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6979 #endif
6980         };
6981         struct nfs41_exchange_id_res res = {
6982                 0
6983         };
6984         int status;
6985         struct rpc_message msg = {
6986                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6987                 .rpc_argp = &args,
6988                 .rpc_resp = &res,
6989                 .rpc_cred = cred,
6990         };
6991
6992         nfs4_init_boot_verifier(clp, &verifier);
6993
6994         status = nfs4_init_uniform_client_string(clp);
6995         if (status)
6996                 goto out;
6997
6998         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6999                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7000                 clp->cl_owner_id);
7001
7002         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7003                                         GFP_NOFS);
7004         if (unlikely(res.server_owner == NULL)) {
7005                 status = -ENOMEM;
7006                 goto out;
7007         }
7008
7009         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7010                                         GFP_NOFS);
7011         if (unlikely(res.server_scope == NULL)) {
7012                 status = -ENOMEM;
7013                 goto out_server_owner;
7014         }
7015
7016         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7017         if (unlikely(res.impl_id == NULL)) {
7018                 status = -ENOMEM;
7019                 goto out_server_scope;
7020         }
7021
7022         switch (sp4_how) {
7023         case SP4_NONE:
7024                 args.state_protect.how = SP4_NONE;
7025                 break;
7026
7027         case SP4_MACH_CRED:
7028                 args.state_protect = nfs4_sp4_mach_cred_request;
7029                 break;
7030
7031         default:
7032                 /* unsupported! */
7033                 WARN_ON_ONCE(1);
7034                 status = -EINVAL;
7035                 goto out_impl_id;
7036         }
7037
7038         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7039         trace_nfs4_exchange_id(clp, status);
7040         if (status == 0)
7041                 status = nfs4_check_cl_exchange_flags(res.flags);
7042
7043         if (status == 0)
7044                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7045
7046         if (status == 0) {
7047                 clp->cl_clientid = res.clientid;
7048                 clp->cl_exchange_flags = res.flags;
7049                 /* Client ID is not confirmed */
7050                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7051                         clear_bit(NFS4_SESSION_ESTABLISHED,
7052                                         &clp->cl_session->session_state);
7053                         clp->cl_seqid = res.seqid;
7054                 }
7055
7056                 kfree(clp->cl_serverowner);
7057                 clp->cl_serverowner = res.server_owner;
7058                 res.server_owner = NULL;
7059
7060                 /* use the most recent implementation id */
7061                 kfree(clp->cl_implid);
7062                 clp->cl_implid = res.impl_id;
7063                 res.impl_id = NULL;
7064
7065                 if (clp->cl_serverscope != NULL &&
7066                     !nfs41_same_server_scope(clp->cl_serverscope,
7067                                              res.server_scope)) {
7068                         dprintk("%s: server_scope mismatch detected\n",
7069                                 __func__);
7070                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7071                         kfree(clp->cl_serverscope);
7072                         clp->cl_serverscope = NULL;
7073                 }
7074
7075                 if (clp->cl_serverscope == NULL) {
7076                         clp->cl_serverscope = res.server_scope;
7077                         res.server_scope = NULL;
7078                 }
7079         }
7080
7081 out_impl_id:
7082         kfree(res.impl_id);
7083 out_server_scope:
7084         kfree(res.server_scope);
7085 out_server_owner:
7086         kfree(res.server_owner);
7087 out:
7088         if (clp->cl_implid != NULL)
7089                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7090                         "domain: %s, name: %s, date: %llu,%u\n",
7091                         clp->cl_implid->domain, clp->cl_implid->name,
7092                         clp->cl_implid->date.seconds,
7093                         clp->cl_implid->date.nseconds);
7094         dprintk("NFS reply exchange_id: %d\n", status);
7095         return status;
7096 }
7097
7098 /*
7099  * nfs4_proc_exchange_id()
7100  *
7101  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7102  *
7103  * Since the clientid has expired, all compounds using sessions
7104  * associated with the stale clientid will be returning
7105  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7106  * be in some phase of session reset.
7107  *
7108  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7109  */
7110 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7111 {
7112         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7113         int status;
7114
7115         /* try SP4_MACH_CRED if krb5i/p */
7116         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7117             authflavor == RPC_AUTH_GSS_KRB5P) {
7118                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7119                 if (!status)
7120                         return 0;
7121         }
7122
7123         /* try SP4_NONE */
7124         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7125 }
7126
7127 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7128                 struct rpc_cred *cred)
7129 {
7130         struct rpc_message msg = {
7131                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7132                 .rpc_argp = clp,
7133                 .rpc_cred = cred,
7134         };
7135         int status;
7136
7137         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7138         trace_nfs4_destroy_clientid(clp, status);
7139         if (status)
7140                 dprintk("NFS: Got error %d from the server %s on "
7141                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7142         return status;
7143 }
7144
7145 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7146                 struct rpc_cred *cred)
7147 {
7148         unsigned int loop;
7149         int ret;
7150
7151         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7152                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7153                 switch (ret) {
7154                 case -NFS4ERR_DELAY:
7155                 case -NFS4ERR_CLIENTID_BUSY:
7156                         ssleep(1);
7157                         break;
7158                 default:
7159                         return ret;
7160                 }
7161         }
7162         return 0;
7163 }
7164
7165 int nfs4_destroy_clientid(struct nfs_client *clp)
7166 {
7167         struct rpc_cred *cred;
7168         int ret = 0;
7169
7170         if (clp->cl_mvops->minor_version < 1)
7171                 goto out;
7172         if (clp->cl_exchange_flags == 0)
7173                 goto out;
7174         if (clp->cl_preserve_clid)
7175                 goto out;
7176         cred = nfs4_get_clid_cred(clp);
7177         ret = nfs4_proc_destroy_clientid(clp, cred);
7178         if (cred)
7179                 put_rpccred(cred);
7180         switch (ret) {
7181         case 0:
7182         case -NFS4ERR_STALE_CLIENTID:
7183                 clp->cl_exchange_flags = 0;
7184         }
7185 out:
7186         return ret;
7187 }
7188
7189 struct nfs4_get_lease_time_data {
7190         struct nfs4_get_lease_time_args *args;
7191         struct nfs4_get_lease_time_res *res;
7192         struct nfs_client *clp;
7193 };
7194
7195 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7196                                         void *calldata)
7197 {
7198         struct nfs4_get_lease_time_data *data =
7199                         (struct nfs4_get_lease_time_data *)calldata;
7200
7201         dprintk("--> %s\n", __func__);
7202         /* just setup sequence, do not trigger session recovery
7203            since we're invoked within one */
7204         nfs41_setup_sequence(data->clp->cl_session,
7205                         &data->args->la_seq_args,
7206                         &data->res->lr_seq_res,
7207                         task);
7208         dprintk("<-- %s\n", __func__);
7209 }
7210
7211 /*
7212  * Called from nfs4_state_manager thread for session setup, so don't recover
7213  * from sequence operation or clientid errors.
7214  */
7215 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7216 {
7217         struct nfs4_get_lease_time_data *data =
7218                         (struct nfs4_get_lease_time_data *)calldata;
7219
7220         dprintk("--> %s\n", __func__);
7221         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7222                 return;
7223         switch (task->tk_status) {
7224         case -NFS4ERR_DELAY:
7225         case -NFS4ERR_GRACE:
7226                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7227                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7228                 task->tk_status = 0;
7229                 /* fall through */
7230         case -NFS4ERR_RETRY_UNCACHED_REP:
7231                 rpc_restart_call_prepare(task);
7232                 return;
7233         }
7234         dprintk("<-- %s\n", __func__);
7235 }
7236
7237 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7238         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7239         .rpc_call_done = nfs4_get_lease_time_done,
7240 };
7241
7242 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7243 {
7244         struct rpc_task *task;
7245         struct nfs4_get_lease_time_args args;
7246         struct nfs4_get_lease_time_res res = {
7247                 .lr_fsinfo = fsinfo,
7248         };
7249         struct nfs4_get_lease_time_data data = {
7250                 .args = &args,
7251                 .res = &res,
7252                 .clp = clp,
7253         };
7254         struct rpc_message msg = {
7255                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7256                 .rpc_argp = &args,
7257                 .rpc_resp = &res,
7258         };
7259         struct rpc_task_setup task_setup = {
7260                 .rpc_client = clp->cl_rpcclient,
7261                 .rpc_message = &msg,
7262                 .callback_ops = &nfs4_get_lease_time_ops,
7263                 .callback_data = &data,
7264                 .flags = RPC_TASK_TIMEOUT,
7265         };
7266         int status;
7267
7268         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7269         nfs4_set_sequence_privileged(&args.la_seq_args);
7270         dprintk("--> %s\n", __func__);
7271         task = rpc_run_task(&task_setup);
7272
7273         if (IS_ERR(task))
7274                 status = PTR_ERR(task);
7275         else {
7276                 status = task->tk_status;
7277                 rpc_put_task(task);
7278         }
7279         dprintk("<-- %s return %d\n", __func__, status);
7280
7281         return status;
7282 }
7283
7284 /*
7285  * Initialize the values to be used by the client in CREATE_SESSION
7286  * If nfs4_init_session set the fore channel request and response sizes,
7287  * use them.
7288  *
7289  * Set the back channel max_resp_sz_cached to zero to force the client to
7290  * always set csa_cachethis to FALSE because the current implementation
7291  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7292  */
7293 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7294 {
7295         unsigned int max_rqst_sz, max_resp_sz;
7296
7297         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7298         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7299
7300         /* Fore channel attributes */
7301         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7302         args->fc_attrs.max_resp_sz = max_resp_sz;
7303         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7304         args->fc_attrs.max_reqs = max_session_slots;
7305
7306         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7307                 "max_ops=%u max_reqs=%u\n",
7308                 __func__,
7309                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7310                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7311
7312         /* Back channel attributes */
7313         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7314         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7315         args->bc_attrs.max_resp_sz_cached = 0;
7316         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7317         args->bc_attrs.max_reqs = 1;
7318
7319         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7320                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7321                 __func__,
7322                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7323                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7324                 args->bc_attrs.max_reqs);
7325 }
7326
7327 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7328                 struct nfs41_create_session_res *res)
7329 {
7330         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7331         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7332
7333         if (rcvd->max_resp_sz > sent->max_resp_sz)
7334                 return -EINVAL;
7335         /*
7336          * Our requested max_ops is the minimum we need; we're not
7337          * prepared to break up compounds into smaller pieces than that.
7338          * So, no point even trying to continue if the server won't
7339          * cooperate:
7340          */
7341         if (rcvd->max_ops < sent->max_ops)
7342                 return -EINVAL;
7343         if (rcvd->max_reqs == 0)
7344                 return -EINVAL;
7345         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7346                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7347         return 0;
7348 }
7349
7350 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7351                 struct nfs41_create_session_res *res)
7352 {
7353         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7354         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7355
7356         if (!(res->flags & SESSION4_BACK_CHAN))
7357                 goto out;
7358         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7359                 return -EINVAL;
7360         if (rcvd->max_resp_sz < sent->max_resp_sz)
7361                 return -EINVAL;
7362         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7363                 return -EINVAL;
7364         /* These would render the backchannel useless: */
7365         if (rcvd->max_ops != sent->max_ops)
7366                 return -EINVAL;
7367         if (rcvd->max_reqs != sent->max_reqs)
7368                 return -EINVAL;
7369 out:
7370         return 0;
7371 }
7372
7373 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7374                                      struct nfs41_create_session_res *res)
7375 {
7376         int ret;
7377
7378         ret = nfs4_verify_fore_channel_attrs(args, res);
7379         if (ret)
7380                 return ret;
7381         return nfs4_verify_back_channel_attrs(args, res);
7382 }
7383
7384 static void nfs4_update_session(struct nfs4_session *session,
7385                 struct nfs41_create_session_res *res)
7386 {
7387         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7388         /* Mark client id and session as being confirmed */
7389         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7390         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7391         session->flags = res->flags;
7392         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7393         if (res->flags & SESSION4_BACK_CHAN)
7394                 memcpy(&session->bc_attrs, &res->bc_attrs,
7395                                 sizeof(session->bc_attrs));
7396 }
7397
7398 static int _nfs4_proc_create_session(struct nfs_client *clp,
7399                 struct rpc_cred *cred)
7400 {
7401         struct nfs4_session *session = clp->cl_session;
7402         struct nfs41_create_session_args args = {
7403                 .client = clp,
7404                 .clientid = clp->cl_clientid,
7405                 .seqid = clp->cl_seqid,
7406                 .cb_program = NFS4_CALLBACK,
7407         };
7408         struct nfs41_create_session_res res;
7409
7410         struct rpc_message msg = {
7411                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7412                 .rpc_argp = &args,
7413                 .rpc_resp = &res,
7414                 .rpc_cred = cred,
7415         };
7416         int status;
7417
7418         nfs4_init_channel_attrs(&args);
7419         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7420
7421         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7422         trace_nfs4_create_session(clp, status);
7423
7424         switch (status) {
7425         case -NFS4ERR_STALE_CLIENTID:
7426         case -NFS4ERR_DELAY:
7427         case -ETIMEDOUT:
7428         case -EACCES:
7429         case -EAGAIN:
7430                 goto out;
7431         };
7432
7433         clp->cl_seqid++;
7434         if (!status) {
7435                 /* Verify the session's negotiated channel_attrs values */
7436                 status = nfs4_verify_channel_attrs(&args, &res);
7437                 /* Increment the clientid slot sequence id */
7438                 if (status)
7439                         goto out;
7440                 nfs4_update_session(session, &res);
7441         }
7442 out:
7443         return status;
7444 }
7445
7446 /*
7447  * Issues a CREATE_SESSION operation to the server.
7448  * It is the responsibility of the caller to verify the session is
7449  * expired before calling this routine.
7450  */
7451 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7452 {
7453         int status;
7454         unsigned *ptr;
7455         struct nfs4_session *session = clp->cl_session;
7456
7457         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7458
7459         status = _nfs4_proc_create_session(clp, cred);
7460         if (status)
7461                 goto out;
7462
7463         /* Init or reset the session slot tables */
7464         status = nfs4_setup_session_slot_tables(session);
7465         dprintk("slot table setup returned %d\n", status);
7466         if (status)
7467                 goto out;
7468
7469         ptr = (unsigned *)&session->sess_id.data[0];
7470         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7471                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7472 out:
7473         dprintk("<-- %s\n", __func__);
7474         return status;
7475 }
7476
7477 /*
7478  * Issue the over-the-wire RPC DESTROY_SESSION.
7479  * The caller must serialize access to this routine.
7480  */
7481 int nfs4_proc_destroy_session(struct nfs4_session *session,
7482                 struct rpc_cred *cred)
7483 {
7484         struct rpc_message msg = {
7485                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7486                 .rpc_argp = session,
7487                 .rpc_cred = cred,
7488         };
7489         int status = 0;
7490
7491         dprintk("--> nfs4_proc_destroy_session\n");
7492
7493         /* session is still being setup */
7494         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7495                 return 0;
7496
7497         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7498         trace_nfs4_destroy_session(session->clp, status);
7499
7500         if (status)
7501                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7502                         "Session has been destroyed regardless...\n", status);
7503
7504         dprintk("<-- nfs4_proc_destroy_session\n");
7505         return status;
7506 }
7507
7508 /*
7509  * Renew the cl_session lease.
7510  */
7511 struct nfs4_sequence_data {
7512         struct nfs_client *clp;
7513         struct nfs4_sequence_args args;
7514         struct nfs4_sequence_res res;
7515 };
7516
7517 static void nfs41_sequence_release(void *data)
7518 {
7519         struct nfs4_sequence_data *calldata = data;
7520         struct nfs_client *clp = calldata->clp;
7521
7522         if (atomic_read(&clp->cl_count) > 1)
7523                 nfs4_schedule_state_renewal(clp);
7524         nfs_put_client(clp);
7525         kfree(calldata);
7526 }
7527
7528 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7529 {
7530         switch(task->tk_status) {
7531         case -NFS4ERR_DELAY:
7532                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7533                 return -EAGAIN;
7534         default:
7535                 nfs4_schedule_lease_recovery(clp);
7536         }
7537         return 0;
7538 }
7539
7540 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7541 {
7542         struct nfs4_sequence_data *calldata = data;
7543         struct nfs_client *clp = calldata->clp;
7544
7545         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7546                 return;
7547
7548         trace_nfs4_sequence(clp, task->tk_status);
7549         if (task->tk_status < 0) {
7550                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7551                 if (atomic_read(&clp->cl_count) == 1)
7552                         goto out;
7553
7554                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7555                         rpc_restart_call_prepare(task);
7556                         return;
7557                 }
7558         }
7559         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7560 out:
7561         dprintk("<-- %s\n", __func__);
7562 }
7563
7564 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7565 {
7566         struct nfs4_sequence_data *calldata = data;
7567         struct nfs_client *clp = calldata->clp;
7568         struct nfs4_sequence_args *args;
7569         struct nfs4_sequence_res *res;
7570
7571         args = task->tk_msg.rpc_argp;
7572         res = task->tk_msg.rpc_resp;
7573
7574         nfs41_setup_sequence(clp->cl_session, args, res, task);
7575 }
7576
7577 static const struct rpc_call_ops nfs41_sequence_ops = {
7578         .rpc_call_done = nfs41_sequence_call_done,
7579         .rpc_call_prepare = nfs41_sequence_prepare,
7580         .rpc_release = nfs41_sequence_release,
7581 };
7582
7583 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7584                 struct rpc_cred *cred,
7585                 bool is_privileged)
7586 {
7587         struct nfs4_sequence_data *calldata;
7588         struct rpc_message msg = {
7589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7590                 .rpc_cred = cred,
7591         };
7592         struct rpc_task_setup task_setup_data = {
7593                 .rpc_client = clp->cl_rpcclient,
7594                 .rpc_message = &msg,
7595                 .callback_ops = &nfs41_sequence_ops,
7596                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7597         };
7598
7599         if (!atomic_inc_not_zero(&clp->cl_count))
7600                 return ERR_PTR(-EIO);
7601         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7602         if (calldata == NULL) {
7603                 nfs_put_client(clp);
7604                 return ERR_PTR(-ENOMEM);
7605         }
7606         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7607         if (is_privileged)
7608                 nfs4_set_sequence_privileged(&calldata->args);
7609         msg.rpc_argp = &calldata->args;
7610         msg.rpc_resp = &calldata->res;
7611         calldata->clp = clp;
7612         task_setup_data.callback_data = calldata;
7613
7614         return rpc_run_task(&task_setup_data);
7615 }
7616
7617 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7618 {
7619         struct rpc_task *task;
7620         int ret = 0;
7621
7622         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7623                 return -EAGAIN;
7624         task = _nfs41_proc_sequence(clp, cred, false);
7625         if (IS_ERR(task))
7626                 ret = PTR_ERR(task);
7627         else
7628                 rpc_put_task_async(task);
7629         dprintk("<-- %s status=%d\n", __func__, ret);
7630         return ret;
7631 }
7632
7633 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7634 {
7635         struct rpc_task *task;
7636         int ret;
7637
7638         task = _nfs41_proc_sequence(clp, cred, true);
7639         if (IS_ERR(task)) {
7640                 ret = PTR_ERR(task);
7641                 goto out;
7642         }
7643         ret = rpc_wait_for_completion_task(task);
7644         if (!ret)
7645                 ret = task->tk_status;
7646         rpc_put_task(task);
7647 out:
7648         dprintk("<-- %s status=%d\n", __func__, ret);
7649         return ret;
7650 }
7651
7652 struct nfs4_reclaim_complete_data {
7653         struct nfs_client *clp;
7654         struct nfs41_reclaim_complete_args arg;
7655         struct nfs41_reclaim_complete_res res;
7656 };
7657
7658 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7659 {
7660         struct nfs4_reclaim_complete_data *calldata = data;
7661
7662         nfs41_setup_sequence(calldata->clp->cl_session,
7663                         &calldata->arg.seq_args,
7664                         &calldata->res.seq_res,
7665                         task);
7666 }
7667
7668 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7669 {
7670         switch(task->tk_status) {
7671         case 0:
7672         case -NFS4ERR_COMPLETE_ALREADY:
7673         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7674                 break;
7675         case -NFS4ERR_DELAY:
7676                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7677                 /* fall through */
7678         case -NFS4ERR_RETRY_UNCACHED_REP:
7679                 return -EAGAIN;
7680         default:
7681                 nfs4_schedule_lease_recovery(clp);
7682         }
7683         return 0;
7684 }
7685
7686 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7687 {
7688         struct nfs4_reclaim_complete_data *calldata = data;
7689         struct nfs_client *clp = calldata->clp;
7690         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7691
7692         dprintk("--> %s\n", __func__);
7693         if (!nfs41_sequence_done(task, res))
7694                 return;
7695
7696         trace_nfs4_reclaim_complete(clp, task->tk_status);
7697         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7698                 rpc_restart_call_prepare(task);
7699                 return;
7700         }
7701         dprintk("<-- %s\n", __func__);
7702 }
7703
7704 static void nfs4_free_reclaim_complete_data(void *data)
7705 {
7706         struct nfs4_reclaim_complete_data *calldata = data;
7707
7708         kfree(calldata);
7709 }
7710
7711 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7712         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7713         .rpc_call_done = nfs4_reclaim_complete_done,
7714         .rpc_release = nfs4_free_reclaim_complete_data,
7715 };
7716
7717 /*
7718  * Issue a global reclaim complete.
7719  */
7720 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7721                 struct rpc_cred *cred)
7722 {
7723         struct nfs4_reclaim_complete_data *calldata;
7724         struct rpc_task *task;
7725         struct rpc_message msg = {
7726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7727                 .rpc_cred = cred,
7728         };
7729         struct rpc_task_setup task_setup_data = {
7730                 .rpc_client = clp->cl_rpcclient,
7731                 .rpc_message = &msg,
7732                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7733                 .flags = RPC_TASK_ASYNC,
7734         };
7735         int status = -ENOMEM;
7736
7737         dprintk("--> %s\n", __func__);
7738         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7739         if (calldata == NULL)
7740                 goto out;
7741         calldata->clp = clp;
7742         calldata->arg.one_fs = 0;
7743
7744         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7745         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7746         msg.rpc_argp = &calldata->arg;
7747         msg.rpc_resp = &calldata->res;
7748         task_setup_data.callback_data = calldata;
7749         task = rpc_run_task(&task_setup_data);
7750         if (IS_ERR(task)) {
7751                 status = PTR_ERR(task);
7752                 goto out;
7753         }
7754         status = nfs4_wait_for_completion_rpc_task(task);
7755         if (status == 0)
7756                 status = task->tk_status;
7757         rpc_put_task(task);
7758         return 0;
7759 out:
7760         dprintk("<-- %s status=%d\n", __func__, status);
7761         return status;
7762 }
7763
7764 static void
7765 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7766 {
7767         struct nfs4_layoutget *lgp = calldata;
7768         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7769         struct nfs4_session *session = nfs4_get_session(server);
7770
7771         dprintk("--> %s\n", __func__);
7772         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7773          * right now covering the LAYOUTGET we are about to send.
7774          * However, that is not so catastrophic, and there seems
7775          * to be no way to prevent it completely.
7776          */
7777         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7778                                 &lgp->res.seq_res, task))
7779                 return;
7780         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7781                                           NFS_I(lgp->args.inode)->layout,
7782                                           &lgp->args.range,
7783                                           lgp->args.ctx->state)) {
7784                 rpc_exit(task, NFS4_OK);
7785         }
7786 }
7787
7788 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7789 {
7790         struct nfs4_layoutget *lgp = calldata;
7791         struct inode *inode = lgp->args.inode;
7792         struct nfs_server *server = NFS_SERVER(inode);
7793         struct pnfs_layout_hdr *lo;
7794         struct nfs4_state *state = NULL;
7795         unsigned long timeo, now, giveup;
7796
7797         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7798
7799         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7800                 goto out;
7801
7802         switch (task->tk_status) {
7803         case 0:
7804                 goto out;
7805         /*
7806          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7807          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7808          */
7809         case -NFS4ERR_BADLAYOUT:
7810                 goto out_overflow;
7811         /*
7812          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7813          * (or clients) writing to the same RAID stripe except when
7814          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7815          */
7816         case -NFS4ERR_LAYOUTTRYLATER:
7817                 if (lgp->args.minlength == 0)
7818                         goto out_overflow;
7819         /*
7820          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7821          * existing layout before getting a new one).
7822          */
7823         case -NFS4ERR_RECALLCONFLICT:
7824                 timeo = rpc_get_timeout(task->tk_client);
7825                 giveup = lgp->args.timestamp + timeo;
7826                 now = jiffies;
7827                 if (time_after(giveup, now)) {
7828                         unsigned long delay;
7829
7830                         /* Delay for:
7831                          * - Not less then NFS4_POLL_RETRY_MIN.
7832                          * - One last time a jiffie before we give up
7833                          * - exponential backoff (time_now minus start_attempt)
7834                          */
7835                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7836                                     min((giveup - now - 1),
7837                                         now - lgp->args.timestamp));
7838
7839                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7840                                 __func__, delay);
7841                         rpc_delay(task, delay);
7842                         /* Do not call nfs4_async_handle_error() */
7843                         goto out_restart;
7844                 }
7845                 break;
7846         case -NFS4ERR_EXPIRED:
7847         case -NFS4ERR_BAD_STATEID:
7848                 spin_lock(&inode->i_lock);
7849                 if (nfs4_stateid_match(&lgp->args.stateid,
7850                                         &lgp->args.ctx->state->stateid)) {
7851                         spin_unlock(&inode->i_lock);
7852                         /* If the open stateid was bad, then recover it. */
7853                         state = lgp->args.ctx->state;
7854                         break;
7855                 }
7856                 lo = NFS_I(inode)->layout;
7857                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7858                                         &lo->plh_stateid)) {
7859                         LIST_HEAD(head);
7860
7861                         /*
7862                          * Mark the bad layout state as invalid, then retry
7863                          * with the current stateid.
7864                          */
7865                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7866                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7867                         spin_unlock(&inode->i_lock);
7868                         pnfs_free_lseg_list(&head);
7869                 } else
7870                         spin_unlock(&inode->i_lock);
7871                 goto out_restart;
7872         }
7873         if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN)
7874                 goto out_restart;
7875 out:
7876         dprintk("<-- %s\n", __func__);
7877         return;
7878 out_restart:
7879         task->tk_status = 0;
7880         rpc_restart_call_prepare(task);
7881         return;
7882 out_overflow:
7883         task->tk_status = -EOVERFLOW;
7884         goto out;
7885 }
7886
7887 static size_t max_response_pages(struct nfs_server *server)
7888 {
7889         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7890         return nfs_page_array_len(0, max_resp_sz);
7891 }
7892
7893 static void nfs4_free_pages(struct page **pages, size_t size)
7894 {
7895         int i;
7896
7897         if (!pages)
7898                 return;
7899
7900         for (i = 0; i < size; i++) {
7901                 if (!pages[i])
7902                         break;
7903                 __free_page(pages[i]);
7904         }
7905         kfree(pages);
7906 }
7907
7908 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7909 {
7910         struct page **pages;
7911         int i;
7912
7913         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7914         if (!pages) {
7915                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7916                 return NULL;
7917         }
7918
7919         for (i = 0; i < size; i++) {
7920                 pages[i] = alloc_page(gfp_flags);
7921                 if (!pages[i]) {
7922                         dprintk("%s: failed to allocate page\n", __func__);
7923                         nfs4_free_pages(pages, size);
7924                         return NULL;
7925                 }
7926         }
7927
7928         return pages;
7929 }
7930
7931 static void nfs4_layoutget_release(void *calldata)
7932 {
7933         struct nfs4_layoutget *lgp = calldata;
7934         struct inode *inode = lgp->args.inode;
7935         struct nfs_server *server = NFS_SERVER(inode);
7936         size_t max_pages = max_response_pages(server);
7937
7938         dprintk("--> %s\n", __func__);
7939         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7940         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7941         put_nfs_open_context(lgp->args.ctx);
7942         kfree(calldata);
7943         dprintk("<-- %s\n", __func__);
7944 }
7945
7946 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7947         .rpc_call_prepare = nfs4_layoutget_prepare,
7948         .rpc_call_done = nfs4_layoutget_done,
7949         .rpc_release = nfs4_layoutget_release,
7950 };
7951
7952 struct pnfs_layout_segment *
7953 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7954 {
7955         struct inode *inode = lgp->args.inode;
7956         struct nfs_server *server = NFS_SERVER(inode);
7957         size_t max_pages = max_response_pages(server);
7958         struct rpc_task *task;
7959         struct rpc_message msg = {
7960                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7961                 .rpc_argp = &lgp->args,
7962                 .rpc_resp = &lgp->res,
7963                 .rpc_cred = lgp->cred,
7964         };
7965         struct rpc_task_setup task_setup_data = {
7966                 .rpc_client = server->client,
7967                 .rpc_message = &msg,
7968                 .callback_ops = &nfs4_layoutget_call_ops,
7969                 .callback_data = lgp,
7970                 .flags = RPC_TASK_ASYNC,
7971         };
7972         struct pnfs_layout_segment *lseg = NULL;
7973         int status = 0;
7974
7975         dprintk("--> %s\n", __func__);
7976
7977         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7978         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7979
7980         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7981         if (!lgp->args.layout.pages) {
7982                 nfs4_layoutget_release(lgp);
7983                 return ERR_PTR(-ENOMEM);
7984         }
7985         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7986         lgp->args.timestamp = jiffies;
7987
7988         lgp->res.layoutp = &lgp->args.layout;
7989         lgp->res.seq_res.sr_slot = NULL;
7990         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7991
7992         task = rpc_run_task(&task_setup_data);
7993         if (IS_ERR(task))
7994                 return ERR_CAST(task);
7995         status = nfs4_wait_for_completion_rpc_task(task);
7996         if (status == 0)
7997                 status = task->tk_status;
7998         trace_nfs4_layoutget(lgp->args.ctx,
7999                         &lgp->args.range,
8000                         &lgp->res.range,
8001                         status);
8002         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8003         if (status == 0 && lgp->res.layoutp->len)
8004                 lseg = pnfs_layout_process(lgp);
8005         rpc_put_task(task);
8006         dprintk("<-- %s status=%d\n", __func__, status);
8007         if (status)
8008                 return ERR_PTR(status);
8009         return lseg;
8010 }
8011
8012 static void
8013 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8014 {
8015         struct nfs4_layoutreturn *lrp = calldata;
8016
8017         dprintk("--> %s\n", __func__);
8018         nfs41_setup_sequence(lrp->clp->cl_session,
8019                         &lrp->args.seq_args,
8020                         &lrp->res.seq_res,
8021                         task);
8022 }
8023
8024 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8025 {
8026         struct nfs4_layoutreturn *lrp = calldata;
8027         struct nfs_server *server;
8028
8029         dprintk("--> %s\n", __func__);
8030
8031         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8032                 return;
8033
8034         server = NFS_SERVER(lrp->args.inode);
8035         switch (task->tk_status) {
8036         default:
8037                 task->tk_status = 0;
8038         case 0:
8039                 break;
8040         case -NFS4ERR_DELAY:
8041                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8042                         break;
8043                 rpc_restart_call_prepare(task);
8044                 return;
8045         }
8046         dprintk("<-- %s\n", __func__);
8047 }
8048
8049 static void nfs4_layoutreturn_release(void *calldata)
8050 {
8051         struct nfs4_layoutreturn *lrp = calldata;
8052         struct pnfs_layout_hdr *lo = lrp->args.layout;
8053         LIST_HEAD(freeme);
8054
8055         dprintk("--> %s\n", __func__);
8056         spin_lock(&lo->plh_inode->i_lock);
8057         if (lrp->res.lrs_present)
8058                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8059         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8060         pnfs_clear_layoutreturn_waitbit(lo);
8061         spin_unlock(&lo->plh_inode->i_lock);
8062         pnfs_free_lseg_list(&freeme);
8063         pnfs_put_layout_hdr(lrp->args.layout);
8064         nfs_iput_and_deactive(lrp->inode);
8065         kfree(calldata);
8066         dprintk("<-- %s\n", __func__);
8067 }
8068
8069 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8070         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8071         .rpc_call_done = nfs4_layoutreturn_done,
8072         .rpc_release = nfs4_layoutreturn_release,
8073 };
8074
8075 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8076 {
8077         struct rpc_task *task;
8078         struct rpc_message msg = {
8079                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8080                 .rpc_argp = &lrp->args,
8081                 .rpc_resp = &lrp->res,
8082                 .rpc_cred = lrp->cred,
8083         };
8084         struct rpc_task_setup task_setup_data = {
8085                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8086                 .rpc_message = &msg,
8087                 .callback_ops = &nfs4_layoutreturn_call_ops,
8088                 .callback_data = lrp,
8089         };
8090         int status = 0;
8091
8092         dprintk("--> %s\n", __func__);
8093         if (!sync) {
8094                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8095                 if (!lrp->inode) {
8096                         nfs4_layoutreturn_release(lrp);
8097                         return -EAGAIN;
8098                 }
8099                 task_setup_data.flags |= RPC_TASK_ASYNC;
8100         }
8101         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8102         task = rpc_run_task(&task_setup_data);
8103         if (IS_ERR(task))
8104                 return PTR_ERR(task);
8105         if (sync)
8106                 status = task->tk_status;
8107         trace_nfs4_layoutreturn(lrp->args.inode, status);
8108         dprintk("<-- %s status=%d\n", __func__, status);
8109         rpc_put_task(task);
8110         return status;
8111 }
8112
8113 static int
8114 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8115                 struct pnfs_device *pdev,
8116                 struct rpc_cred *cred)
8117 {
8118         struct nfs4_getdeviceinfo_args args = {
8119                 .pdev = pdev,
8120                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8121                         NOTIFY_DEVICEID4_DELETE,
8122         };
8123         struct nfs4_getdeviceinfo_res res = {
8124                 .pdev = pdev,
8125         };
8126         struct rpc_message msg = {
8127                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8128                 .rpc_argp = &args,
8129                 .rpc_resp = &res,
8130                 .rpc_cred = cred,
8131         };
8132         int status;
8133
8134         dprintk("--> %s\n", __func__);
8135         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8136         if (res.notification & ~args.notify_types)
8137                 dprintk("%s: unsupported notification\n", __func__);
8138         if (res.notification != args.notify_types)
8139                 pdev->nocache = 1;
8140
8141         dprintk("<-- %s status=%d\n", __func__, status);
8142
8143         return status;
8144 }
8145
8146 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8147                 struct pnfs_device *pdev,
8148                 struct rpc_cred *cred)
8149 {
8150         struct nfs4_exception exception = { };
8151         int err;
8152
8153         do {
8154                 err = nfs4_handle_exception(server,
8155                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8156                                         &exception);
8157         } while (exception.retry);
8158         return err;
8159 }
8160 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8161
8162 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8163 {
8164         struct nfs4_layoutcommit_data *data = calldata;
8165         struct nfs_server *server = NFS_SERVER(data->args.inode);
8166         struct nfs4_session *session = nfs4_get_session(server);
8167
8168         nfs41_setup_sequence(session,
8169                         &data->args.seq_args,
8170                         &data->res.seq_res,
8171                         task);
8172 }
8173
8174 static void
8175 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8176 {
8177         struct nfs4_layoutcommit_data *data = calldata;
8178         struct nfs_server *server = NFS_SERVER(data->args.inode);
8179
8180         if (!nfs41_sequence_done(task, &data->res.seq_res))
8181                 return;
8182
8183         switch (task->tk_status) { /* Just ignore these failures */
8184         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8185         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8186         case -NFS4ERR_BADLAYOUT:     /* no layout */
8187         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8188                 task->tk_status = 0;
8189         case 0:
8190                 break;
8191         default:
8192                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8193                         rpc_restart_call_prepare(task);
8194                         return;
8195                 }
8196         }
8197 }
8198
8199 static void nfs4_layoutcommit_release(void *calldata)
8200 {
8201         struct nfs4_layoutcommit_data *data = calldata;
8202
8203         pnfs_cleanup_layoutcommit(data);
8204         nfs_post_op_update_inode_force_wcc(data->args.inode,
8205                                            data->res.fattr);
8206         put_rpccred(data->cred);
8207         nfs_iput_and_deactive(data->inode);
8208         kfree(data);
8209 }
8210
8211 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8212         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8213         .rpc_call_done = nfs4_layoutcommit_done,
8214         .rpc_release = nfs4_layoutcommit_release,
8215 };
8216
8217 int
8218 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8219 {
8220         struct rpc_message msg = {
8221                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8222                 .rpc_argp = &data->args,
8223                 .rpc_resp = &data->res,
8224                 .rpc_cred = data->cred,
8225         };
8226         struct rpc_task_setup task_setup_data = {
8227                 .task = &data->task,
8228                 .rpc_client = NFS_CLIENT(data->args.inode),
8229                 .rpc_message = &msg,
8230                 .callback_ops = &nfs4_layoutcommit_ops,
8231                 .callback_data = data,
8232         };
8233         struct rpc_task *task;
8234         int status = 0;
8235
8236         dprintk("NFS: initiating layoutcommit call. sync %d "
8237                 "lbw: %llu inode %lu\n", sync,
8238                 data->args.lastbytewritten,
8239                 data->args.inode->i_ino);
8240
8241         if (!sync) {
8242                 data->inode = nfs_igrab_and_active(data->args.inode);
8243                 if (data->inode == NULL) {
8244                         nfs4_layoutcommit_release(data);
8245                         return -EAGAIN;
8246                 }
8247                 task_setup_data.flags = RPC_TASK_ASYNC;
8248         }
8249         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8250         task = rpc_run_task(&task_setup_data);
8251         if (IS_ERR(task))
8252                 return PTR_ERR(task);
8253         if (sync)
8254                 status = task->tk_status;
8255         trace_nfs4_layoutcommit(data->args.inode, status);
8256         dprintk("%s: status %d\n", __func__, status);
8257         rpc_put_task(task);
8258         return status;
8259 }
8260
8261 /**
8262  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8263  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8264  */
8265 static int
8266 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8267                     struct nfs_fsinfo *info,
8268                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8269 {
8270         struct nfs41_secinfo_no_name_args args = {
8271                 .style = SECINFO_STYLE_CURRENT_FH,
8272         };
8273         struct nfs4_secinfo_res res = {
8274                 .flavors = flavors,
8275         };
8276         struct rpc_message msg = {
8277                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8278                 .rpc_argp = &args,
8279                 .rpc_resp = &res,
8280         };
8281         struct rpc_clnt *clnt = server->client;
8282         struct rpc_cred *cred = NULL;
8283         int status;
8284
8285         if (use_integrity) {
8286                 clnt = server->nfs_client->cl_rpcclient;
8287                 cred = nfs4_get_clid_cred(server->nfs_client);
8288                 msg.rpc_cred = cred;
8289         }
8290
8291         dprintk("--> %s\n", __func__);
8292         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8293                                 &res.seq_res, 0);
8294         dprintk("<-- %s status=%d\n", __func__, status);
8295
8296         if (cred)
8297                 put_rpccred(cred);
8298
8299         return status;
8300 }
8301
8302 static int
8303 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8304                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8305 {
8306         struct nfs4_exception exception = { };
8307         int err;
8308         do {
8309                 /* first try using integrity protection */
8310                 err = -NFS4ERR_WRONGSEC;
8311
8312                 /* try to use integrity protection with machine cred */
8313                 if (_nfs4_is_integrity_protected(server->nfs_client))
8314                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8315                                                           flavors, true);
8316
8317                 /*
8318                  * if unable to use integrity protection, or SECINFO with
8319                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8320                  * disallowed by spec, but exists in deployed servers) use
8321                  * the current filesystem's rpc_client and the user cred.
8322                  */
8323                 if (err == -NFS4ERR_WRONGSEC)
8324                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8325                                                           flavors, false);
8326
8327                 switch (err) {
8328                 case 0:
8329                 case -NFS4ERR_WRONGSEC:
8330                 case -ENOTSUPP:
8331                         goto out;
8332                 default:
8333                         err = nfs4_handle_exception(server, err, &exception);
8334                 }
8335         } while (exception.retry);
8336 out:
8337         return err;
8338 }
8339
8340 static int
8341 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8342                     struct nfs_fsinfo *info)
8343 {
8344         int err;
8345         struct page *page;
8346         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8347         struct nfs4_secinfo_flavors *flavors;
8348         struct nfs4_secinfo4 *secinfo;
8349         int i;
8350
8351         page = alloc_page(GFP_KERNEL);
8352         if (!page) {
8353                 err = -ENOMEM;
8354                 goto out;
8355         }
8356
8357         flavors = page_address(page);
8358         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8359
8360         /*
8361          * Fall back on "guess and check" method if
8362          * the server doesn't support SECINFO_NO_NAME
8363          */
8364         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8365                 err = nfs4_find_root_sec(server, fhandle, info);
8366                 goto out_freepage;
8367         }
8368         if (err)
8369                 goto out_freepage;
8370
8371         for (i = 0; i < flavors->num_flavors; i++) {
8372                 secinfo = &flavors->flavors[i];
8373
8374                 switch (secinfo->flavor) {
8375                 case RPC_AUTH_NULL:
8376                 case RPC_AUTH_UNIX:
8377                 case RPC_AUTH_GSS:
8378                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8379                                         &secinfo->flavor_info);
8380                         break;
8381                 default:
8382                         flavor = RPC_AUTH_MAXFLAVOR;
8383                         break;
8384                 }
8385
8386                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8387                         flavor = RPC_AUTH_MAXFLAVOR;
8388
8389                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8390                         err = nfs4_lookup_root_sec(server, fhandle,
8391                                                    info, flavor);
8392                         if (!err)
8393                                 break;
8394                 }
8395         }
8396
8397         if (flavor == RPC_AUTH_MAXFLAVOR)
8398                 err = -EPERM;
8399
8400 out_freepage:
8401         put_page(page);
8402         if (err == -EACCES)
8403                 return -EPERM;
8404 out:
8405         return err;
8406 }
8407
8408 static int _nfs41_test_stateid(struct nfs_server *server,
8409                 nfs4_stateid *stateid,
8410                 struct rpc_cred *cred)
8411 {
8412         int status;
8413         struct nfs41_test_stateid_args args = {
8414                 .stateid = stateid,
8415         };
8416         struct nfs41_test_stateid_res res;
8417         struct rpc_message msg = {
8418                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8419                 .rpc_argp = &args,
8420                 .rpc_resp = &res,
8421                 .rpc_cred = cred,
8422         };
8423         struct rpc_clnt *rpc_client = server->client;
8424
8425         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8426                 &rpc_client, &msg);
8427
8428         dprintk("NFS call  test_stateid %p\n", stateid);
8429         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8430         nfs4_set_sequence_privileged(&args.seq_args);
8431         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8432                         &args.seq_args, &res.seq_res);
8433         if (status != NFS_OK) {
8434                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8435                 return status;
8436         }
8437         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8438         return -res.status;
8439 }
8440
8441 /**
8442  * nfs41_test_stateid - perform a TEST_STATEID operation
8443  *
8444  * @server: server / transport on which to perform the operation
8445  * @stateid: state ID to test
8446  * @cred: credential
8447  *
8448  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8449  * Otherwise a negative NFS4ERR value is returned if the operation
8450  * failed or the state ID is not currently valid.
8451  */
8452 static int nfs41_test_stateid(struct nfs_server *server,
8453                 nfs4_stateid *stateid,
8454                 struct rpc_cred *cred)
8455 {
8456         struct nfs4_exception exception = { };
8457         int err;
8458         do {
8459                 err = _nfs41_test_stateid(server, stateid, cred);
8460                 if (err != -NFS4ERR_DELAY)
8461                         break;
8462                 nfs4_handle_exception(server, err, &exception);
8463         } while (exception.retry);
8464         return err;
8465 }
8466
8467 struct nfs_free_stateid_data {
8468         struct nfs_server *server;
8469         struct nfs41_free_stateid_args args;
8470         struct nfs41_free_stateid_res res;
8471 };
8472
8473 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8474 {
8475         struct nfs_free_stateid_data *data = calldata;
8476         nfs41_setup_sequence(nfs4_get_session(data->server),
8477                         &data->args.seq_args,
8478                         &data->res.seq_res,
8479                         task);
8480 }
8481
8482 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8483 {
8484         struct nfs_free_stateid_data *data = calldata;
8485
8486         nfs41_sequence_done(task, &data->res.seq_res);
8487
8488         switch (task->tk_status) {
8489         case -NFS4ERR_DELAY:
8490                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8491                         rpc_restart_call_prepare(task);
8492         }
8493 }
8494
8495 static void nfs41_free_stateid_release(void *calldata)
8496 {
8497         kfree(calldata);
8498 }
8499
8500 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8501         .rpc_call_prepare = nfs41_free_stateid_prepare,
8502         .rpc_call_done = nfs41_free_stateid_done,
8503         .rpc_release = nfs41_free_stateid_release,
8504 };
8505
8506 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8507                 nfs4_stateid *stateid,
8508                 struct rpc_cred *cred,
8509                 bool privileged)
8510 {
8511         struct rpc_message msg = {
8512                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8513                 .rpc_cred = cred,
8514         };
8515         struct rpc_task_setup task_setup = {
8516                 .rpc_client = server->client,
8517                 .rpc_message = &msg,
8518                 .callback_ops = &nfs41_free_stateid_ops,
8519                 .flags = RPC_TASK_ASYNC,
8520         };
8521         struct nfs_free_stateid_data *data;
8522
8523         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8524                 &task_setup.rpc_client, &msg);
8525
8526         dprintk("NFS call  free_stateid %p\n", stateid);
8527         data = kmalloc(sizeof(*data), GFP_NOFS);
8528         if (!data)
8529                 return ERR_PTR(-ENOMEM);
8530         data->server = server;
8531         nfs4_stateid_copy(&data->args.stateid, stateid);
8532
8533         task_setup.callback_data = data;
8534
8535         msg.rpc_argp = &data->args;
8536         msg.rpc_resp = &data->res;
8537         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8538         if (privileged)
8539                 nfs4_set_sequence_privileged(&data->args.seq_args);
8540
8541         return rpc_run_task(&task_setup);
8542 }
8543
8544 /**
8545  * nfs41_free_stateid - perform a FREE_STATEID operation
8546  *
8547  * @server: server / transport on which to perform the operation
8548  * @stateid: state ID to release
8549  * @cred: credential
8550  *
8551  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8552  * negative NFS4ERR value is returned.
8553  */
8554 static int nfs41_free_stateid(struct nfs_server *server,
8555                 nfs4_stateid *stateid,
8556                 struct rpc_cred *cred)
8557 {
8558         struct rpc_task *task;
8559         int ret;
8560
8561         task = _nfs41_free_stateid(server, stateid, cred, true);
8562         if (IS_ERR(task))
8563                 return PTR_ERR(task);
8564         ret = rpc_wait_for_completion_task(task);
8565         if (!ret)
8566                 ret = task->tk_status;
8567         rpc_put_task(task);
8568         return ret;
8569 }
8570
8571 static void
8572 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8573 {
8574         struct rpc_task *task;
8575         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8576
8577         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8578         nfs4_free_lock_state(server, lsp);
8579         if (IS_ERR(task))
8580                 return;
8581         rpc_put_task(task);
8582 }
8583
8584 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8585                 const nfs4_stateid *s2)
8586 {
8587         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8588                 return false;
8589
8590         if (s1->seqid == s2->seqid)
8591                 return true;
8592         if (s1->seqid == 0 || s2->seqid == 0)
8593                 return true;
8594
8595         return false;
8596 }
8597
8598 #endif /* CONFIG_NFS_V4_1 */
8599
8600 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8601                 const nfs4_stateid *s2)
8602 {
8603         return nfs4_stateid_match(s1, s2);
8604 }
8605
8606
8607 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8608         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8609         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8610         .recover_open   = nfs4_open_reclaim,
8611         .recover_lock   = nfs4_lock_reclaim,
8612         .establish_clid = nfs4_init_clientid,
8613         .detect_trunking = nfs40_discover_server_trunking,
8614 };
8615
8616 #if defined(CONFIG_NFS_V4_1)
8617 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8618         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8619         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8620         .recover_open   = nfs4_open_reclaim,
8621         .recover_lock   = nfs4_lock_reclaim,
8622         .establish_clid = nfs41_init_clientid,
8623         .reclaim_complete = nfs41_proc_reclaim_complete,
8624         .detect_trunking = nfs41_discover_server_trunking,
8625 };
8626 #endif /* CONFIG_NFS_V4_1 */
8627
8628 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8629         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8630         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8631         .recover_open   = nfs40_open_expired,
8632         .recover_lock   = nfs4_lock_expired,
8633         .establish_clid = nfs4_init_clientid,
8634 };
8635
8636 #if defined(CONFIG_NFS_V4_1)
8637 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8638         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8639         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8640         .recover_open   = nfs41_open_expired,
8641         .recover_lock   = nfs41_lock_expired,
8642         .establish_clid = nfs41_init_clientid,
8643 };
8644 #endif /* CONFIG_NFS_V4_1 */
8645
8646 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8647         .sched_state_renewal = nfs4_proc_async_renew,
8648         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8649         .renew_lease = nfs4_proc_renew,
8650 };
8651
8652 #if defined(CONFIG_NFS_V4_1)
8653 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8654         .sched_state_renewal = nfs41_proc_async_sequence,
8655         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8656         .renew_lease = nfs4_proc_sequence,
8657 };
8658 #endif
8659
8660 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8661         .get_locations = _nfs40_proc_get_locations,
8662         .fsid_present = _nfs40_proc_fsid_present,
8663 };
8664
8665 #if defined(CONFIG_NFS_V4_1)
8666 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8667         .get_locations = _nfs41_proc_get_locations,
8668         .fsid_present = _nfs41_proc_fsid_present,
8669 };
8670 #endif  /* CONFIG_NFS_V4_1 */
8671
8672 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8673         .minor_version = 0,
8674         .init_caps = NFS_CAP_READDIRPLUS
8675                 | NFS_CAP_ATOMIC_OPEN
8676                 | NFS_CAP_POSIX_LOCK,
8677         .init_client = nfs40_init_client,
8678         .shutdown_client = nfs40_shutdown_client,
8679         .match_stateid = nfs4_match_stateid,
8680         .find_root_sec = nfs4_find_root_sec,
8681         .free_lock_state = nfs4_release_lockowner,
8682         .alloc_seqid = nfs_alloc_seqid,
8683         .call_sync_ops = &nfs40_call_sync_ops,
8684         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8685         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8686         .state_renewal_ops = &nfs40_state_renewal_ops,
8687         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8688 };
8689
8690 #if defined(CONFIG_NFS_V4_1)
8691 static struct nfs_seqid *
8692 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8693 {
8694         return NULL;
8695 }
8696
8697 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8698         .minor_version = 1,
8699         .init_caps = NFS_CAP_READDIRPLUS
8700                 | NFS_CAP_ATOMIC_OPEN
8701                 | NFS_CAP_POSIX_LOCK
8702                 | NFS_CAP_STATEID_NFSV41
8703                 | NFS_CAP_ATOMIC_OPEN_V1,
8704         .init_client = nfs41_init_client,
8705         .shutdown_client = nfs41_shutdown_client,
8706         .match_stateid = nfs41_match_stateid,
8707         .find_root_sec = nfs41_find_root_sec,
8708         .free_lock_state = nfs41_free_lock_state,
8709         .alloc_seqid = nfs_alloc_no_seqid,
8710         .call_sync_ops = &nfs41_call_sync_ops,
8711         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8712         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8713         .state_renewal_ops = &nfs41_state_renewal_ops,
8714         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8715 };
8716 #endif
8717
8718 #if defined(CONFIG_NFS_V4_2)
8719 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8720         .minor_version = 2,
8721         .init_caps = NFS_CAP_READDIRPLUS
8722                 | NFS_CAP_ATOMIC_OPEN
8723                 | NFS_CAP_POSIX_LOCK
8724                 | NFS_CAP_STATEID_NFSV41
8725                 | NFS_CAP_ATOMIC_OPEN_V1
8726                 | NFS_CAP_ALLOCATE
8727                 | NFS_CAP_DEALLOCATE
8728                 | NFS_CAP_SEEK
8729                 | NFS_CAP_LAYOUTSTATS
8730                 | NFS_CAP_CLONE,
8731         .init_client = nfs41_init_client,
8732         .shutdown_client = nfs41_shutdown_client,
8733         .match_stateid = nfs41_match_stateid,
8734         .find_root_sec = nfs41_find_root_sec,
8735         .free_lock_state = nfs41_free_lock_state,
8736         .call_sync_ops = &nfs41_call_sync_ops,
8737         .alloc_seqid = nfs_alloc_no_seqid,
8738         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8739         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8740         .state_renewal_ops = &nfs41_state_renewal_ops,
8741         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8742 };
8743 #endif
8744
8745 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8746         [0] = &nfs_v4_0_minor_ops,
8747 #if defined(CONFIG_NFS_V4_1)
8748         [1] = &nfs_v4_1_minor_ops,
8749 #endif
8750 #if defined(CONFIG_NFS_V4_2)
8751         [2] = &nfs_v4_2_minor_ops,
8752 #endif
8753 };
8754
8755 static const struct inode_operations nfs4_dir_inode_operations = {
8756         .create         = nfs_create,
8757         .lookup         = nfs_lookup,
8758         .atomic_open    = nfs_atomic_open,
8759         .link           = nfs_link,
8760         .unlink         = nfs_unlink,
8761         .symlink        = nfs_symlink,
8762         .mkdir          = nfs_mkdir,
8763         .rmdir          = nfs_rmdir,
8764         .mknod          = nfs_mknod,
8765         .rename         = nfs_rename,
8766         .permission     = nfs_permission,
8767         .getattr        = nfs_getattr,
8768         .setattr        = nfs_setattr,
8769         .getxattr       = generic_getxattr,
8770         .setxattr       = generic_setxattr,
8771         .listxattr      = generic_listxattr,
8772         .removexattr    = generic_removexattr,
8773 };
8774
8775 static const struct inode_operations nfs4_file_inode_operations = {
8776         .permission     = nfs_permission,
8777         .getattr        = nfs_getattr,
8778         .setattr        = nfs_setattr,
8779         .getxattr       = generic_getxattr,
8780         .setxattr       = generic_setxattr,
8781         .listxattr      = generic_listxattr,
8782         .removexattr    = generic_removexattr,
8783 };
8784
8785 const struct nfs_rpc_ops nfs_v4_clientops = {
8786         .version        = 4,                    /* protocol version */
8787         .dentry_ops     = &nfs4_dentry_operations,
8788         .dir_inode_ops  = &nfs4_dir_inode_operations,
8789         .file_inode_ops = &nfs4_file_inode_operations,
8790         .file_ops       = &nfs4_file_operations,
8791         .getroot        = nfs4_proc_get_root,
8792         .submount       = nfs4_submount,
8793         .try_mount      = nfs4_try_mount,
8794         .getattr        = nfs4_proc_getattr,
8795         .setattr        = nfs4_proc_setattr,
8796         .lookup         = nfs4_proc_lookup,
8797         .access         = nfs4_proc_access,
8798         .readlink       = nfs4_proc_readlink,
8799         .create         = nfs4_proc_create,
8800         .remove         = nfs4_proc_remove,
8801         .unlink_setup   = nfs4_proc_unlink_setup,
8802         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8803         .unlink_done    = nfs4_proc_unlink_done,
8804         .rename_setup   = nfs4_proc_rename_setup,
8805         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8806         .rename_done    = nfs4_proc_rename_done,
8807         .link           = nfs4_proc_link,
8808         .symlink        = nfs4_proc_symlink,
8809         .mkdir          = nfs4_proc_mkdir,
8810         .rmdir          = nfs4_proc_remove,
8811         .readdir        = nfs4_proc_readdir,
8812         .mknod          = nfs4_proc_mknod,
8813         .statfs         = nfs4_proc_statfs,
8814         .fsinfo         = nfs4_proc_fsinfo,
8815         .pathconf       = nfs4_proc_pathconf,
8816         .set_capabilities = nfs4_server_capabilities,
8817         .decode_dirent  = nfs4_decode_dirent,
8818         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8819         .read_setup     = nfs4_proc_read_setup,
8820         .read_done      = nfs4_read_done,
8821         .write_setup    = nfs4_proc_write_setup,
8822         .write_done     = nfs4_write_done,
8823         .commit_setup   = nfs4_proc_commit_setup,
8824         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8825         .commit_done    = nfs4_commit_done,
8826         .lock           = nfs4_proc_lock,
8827         .clear_acl_cache = nfs4_zap_acl_attr,
8828         .close_context  = nfs4_close_context,
8829         .open_context   = nfs4_atomic_open,
8830         .have_delegation = nfs4_have_delegation,
8831         .return_delegation = nfs4_inode_return_delegation,
8832         .alloc_client   = nfs4_alloc_client,
8833         .init_client    = nfs4_init_client,
8834         .free_client    = nfs4_free_client,
8835         .create_server  = nfs4_create_server,
8836         .clone_server   = nfs_clone_server,
8837 };
8838
8839 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8840         .prefix = XATTR_NAME_NFSV4_ACL,
8841         .list   = nfs4_xattr_list_nfs4_acl,
8842         .get    = nfs4_xattr_get_nfs4_acl,
8843         .set    = nfs4_xattr_set_nfs4_acl,
8844 };
8845
8846 const struct xattr_handler *nfs4_xattr_handlers[] = {
8847         &nfs4_xattr_nfs4_acl_handler,
8848 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8849         &nfs4_xattr_nfs4_label_handler,
8850 #endif
8851         NULL
8852 };
8853
8854 /*
8855  * Local variables:
8856  *  c-basic-offset: 8
8857  * End:
8858  */