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         write_seqlock(&state->seqlock);
1389         if (deleg_stateid != NULL) {
1390                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1391                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1392         }
1393         if (open_stateid != NULL)
1394                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1395         write_sequnlock(&state->seqlock);
1396         spin_lock(&state->owner->so_lock);
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                 trace_nfs4_cached_open(data->state);
1602                 goto out;
1603         }
1604
1605         ret = -EAGAIN;
1606         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1607                 goto err;
1608         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1609         ret = PTR_ERR(inode);
1610         if (IS_ERR(inode))
1611                 goto err;
1612         ret = -ENOMEM;
1613         state = nfs4_get_open_state(inode, data->owner);
1614         if (state == NULL)
1615                 goto err_put_inode;
1616         if (data->o_res.delegation_type != 0)
1617                 nfs4_opendata_check_deleg(data, state);
1618         update_open_stateid(state, &data->o_res.stateid, NULL,
1619                         data->o_arg.fmode);
1620         iput(inode);
1621 out:
1622         nfs_release_seqid(data->o_arg.seqid);
1623         return state;
1624 err_put_inode:
1625         iput(inode);
1626 err:
1627         return ERR_PTR(ret);
1628 }
1629
1630 static struct nfs4_state *
1631 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1632 {
1633         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1634                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1635         return _nfs4_opendata_to_nfs4_state(data);
1636 }
1637
1638 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1639 {
1640         struct nfs_inode *nfsi = NFS_I(state->inode);
1641         struct nfs_open_context *ctx;
1642
1643         spin_lock(&state->inode->i_lock);
1644         list_for_each_entry(ctx, &nfsi->open_files, list) {
1645                 if (ctx->state != state)
1646                         continue;
1647                 get_nfs_open_context(ctx);
1648                 spin_unlock(&state->inode->i_lock);
1649                 return ctx;
1650         }
1651         spin_unlock(&state->inode->i_lock);
1652         return ERR_PTR(-ENOENT);
1653 }
1654
1655 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1656                 struct nfs4_state *state, enum open_claim_type4 claim)
1657 {
1658         struct nfs4_opendata *opendata;
1659
1660         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1661                         NULL, NULL, claim, GFP_NOFS);
1662         if (opendata == NULL)
1663                 return ERR_PTR(-ENOMEM);
1664         opendata->state = state;
1665         atomic_inc(&state->count);
1666         return opendata;
1667 }
1668
1669 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1670                 fmode_t fmode)
1671 {
1672         struct nfs4_state *newstate;
1673         int ret;
1674
1675         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1676                 return 0;
1677         opendata->o_arg.open_flags = 0;
1678         opendata->o_arg.fmode = fmode;
1679         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1680                         NFS_SB(opendata->dentry->d_sb),
1681                         fmode, 0);
1682         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1683         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1684         nfs4_init_opendata_res(opendata);
1685         ret = _nfs4_recover_proc_open(opendata);
1686         if (ret != 0)
1687                 return ret;
1688         newstate = nfs4_opendata_to_nfs4_state(opendata);
1689         if (IS_ERR(newstate))
1690                 return PTR_ERR(newstate);
1691         if (newstate != opendata->state)
1692                 ret = -ESTALE;
1693         nfs4_close_state(newstate, fmode);
1694         return ret;
1695 }
1696
1697 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1698 {
1699         int ret;
1700
1701         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1702         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1703         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1704         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1705         /* memory barrier prior to reading state->n_* */
1706         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1707         clear_bit(NFS_OPEN_STATE, &state->flags);
1708         smp_rmb();
1709         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1710         if (ret != 0)
1711                 return ret;
1712         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1713         if (ret != 0)
1714                 return ret;
1715         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1716         if (ret != 0)
1717                 return ret;
1718         /*
1719          * We may have performed cached opens for all three recoveries.
1720          * Check if we need to update the current stateid.
1721          */
1722         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1723             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1724                 write_seqlock(&state->seqlock);
1725                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1726                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1727                 write_sequnlock(&state->seqlock);
1728         }
1729         return 0;
1730 }
1731
1732 /*
1733  * OPEN_RECLAIM:
1734  *      reclaim state on the server after a reboot.
1735  */
1736 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1737 {
1738         struct nfs_delegation *delegation;
1739         struct nfs4_opendata *opendata;
1740         fmode_t delegation_type = 0;
1741         int status;
1742
1743         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1744                         NFS4_OPEN_CLAIM_PREVIOUS);
1745         if (IS_ERR(opendata))
1746                 return PTR_ERR(opendata);
1747         rcu_read_lock();
1748         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1749         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1750                 delegation_type = delegation->type;
1751         rcu_read_unlock();
1752         opendata->o_arg.u.delegation_type = delegation_type;
1753         status = nfs4_open_recover(opendata, state);
1754         nfs4_opendata_put(opendata);
1755         return status;
1756 }
1757
1758 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1759 {
1760         struct nfs_server *server = NFS_SERVER(state->inode);
1761         struct nfs4_exception exception = { };
1762         int err;
1763         do {
1764                 err = _nfs4_do_open_reclaim(ctx, state);
1765                 trace_nfs4_open_reclaim(ctx, 0, err);
1766                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1767                         continue;
1768                 if (err != -NFS4ERR_DELAY)
1769                         break;
1770                 nfs4_handle_exception(server, err, &exception);
1771         } while (exception.retry);
1772         return err;
1773 }
1774
1775 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1776 {
1777         struct nfs_open_context *ctx;
1778         int ret;
1779
1780         ctx = nfs4_state_find_open_context(state);
1781         if (IS_ERR(ctx))
1782                 return -EAGAIN;
1783         ret = nfs4_do_open_reclaim(ctx, state);
1784         put_nfs_open_context(ctx);
1785         return ret;
1786 }
1787
1788 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1789 {
1790         switch (err) {
1791                 default:
1792                         printk(KERN_ERR "NFS: %s: unhandled error "
1793                                         "%d.\n", __func__, err);
1794                 case 0:
1795                 case -ENOENT:
1796                 case -EAGAIN:
1797                 case -ESTALE:
1798                         break;
1799                 case -NFS4ERR_BADSESSION:
1800                 case -NFS4ERR_BADSLOT:
1801                 case -NFS4ERR_BAD_HIGH_SLOT:
1802                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1803                 case -NFS4ERR_DEADSESSION:
1804                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1805                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1806                         return -EAGAIN;
1807                 case -NFS4ERR_STALE_CLIENTID:
1808                 case -NFS4ERR_STALE_STATEID:
1809                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1810                 case -NFS4ERR_EXPIRED:
1811                         /* Don't recall a delegation if it was lost */
1812                         nfs4_schedule_lease_recovery(server->nfs_client);
1813                         return -EAGAIN;
1814                 case -NFS4ERR_MOVED:
1815                         nfs4_schedule_migration_recovery(server);
1816                         return -EAGAIN;
1817                 case -NFS4ERR_LEASE_MOVED:
1818                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1819                         return -EAGAIN;
1820                 case -NFS4ERR_DELEG_REVOKED:
1821                 case -NFS4ERR_ADMIN_REVOKED:
1822                 case -NFS4ERR_BAD_STATEID:
1823                 case -NFS4ERR_OPENMODE:
1824                         nfs_inode_find_state_and_recover(state->inode,
1825                                         stateid);
1826                         nfs4_schedule_stateid_recovery(server, state);
1827                         return -EAGAIN;
1828                 case -NFS4ERR_DELAY:
1829                 case -NFS4ERR_GRACE:
1830                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1831                         ssleep(1);
1832                         return -EAGAIN;
1833                 case -ENOMEM:
1834                 case -NFS4ERR_DENIED:
1835                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1836                         return 0;
1837         }
1838         return err;
1839 }
1840
1841 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1842                 struct nfs4_state *state, const nfs4_stateid *stateid,
1843                 fmode_t type)
1844 {
1845         struct nfs_server *server = NFS_SERVER(state->inode);
1846         struct nfs4_opendata *opendata;
1847         int err = 0;
1848
1849         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1850                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1851         if (IS_ERR(opendata))
1852                 return PTR_ERR(opendata);
1853         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1854         write_seqlock(&state->seqlock);
1855         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1856         write_sequnlock(&state->seqlock);
1857         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1858         switch (type & (FMODE_READ|FMODE_WRITE)) {
1859         case FMODE_READ|FMODE_WRITE:
1860         case FMODE_WRITE:
1861                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1862                 if (err)
1863                         break;
1864                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1865                 if (err)
1866                         break;
1867         case FMODE_READ:
1868                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1869         }
1870         nfs4_opendata_put(opendata);
1871         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1872 }
1873
1874 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1875 {
1876         struct nfs4_opendata *data = calldata;
1877
1878         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1879                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1880 }
1881
1882 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1883 {
1884         struct nfs4_opendata *data = calldata;
1885
1886         nfs40_sequence_done(task, &data->c_res.seq_res);
1887
1888         data->rpc_status = task->tk_status;
1889         if (data->rpc_status == 0) {
1890                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1891                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1892                 renew_lease(data->o_res.server, data->timestamp);
1893                 data->rpc_done = 1;
1894         }
1895 }
1896
1897 static void nfs4_open_confirm_release(void *calldata)
1898 {
1899         struct nfs4_opendata *data = calldata;
1900         struct nfs4_state *state = NULL;
1901
1902         /* If this request hasn't been cancelled, do nothing */
1903         if (data->cancelled == 0)
1904                 goto out_free;
1905         /* In case of error, no cleanup! */
1906         if (!data->rpc_done)
1907                 goto out_free;
1908         state = nfs4_opendata_to_nfs4_state(data);
1909         if (!IS_ERR(state))
1910                 nfs4_close_state(state, data->o_arg.fmode);
1911 out_free:
1912         nfs4_opendata_put(data);
1913 }
1914
1915 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1916         .rpc_call_prepare = nfs4_open_confirm_prepare,
1917         .rpc_call_done = nfs4_open_confirm_done,
1918         .rpc_release = nfs4_open_confirm_release,
1919 };
1920
1921 /*
1922  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1923  */
1924 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1925 {
1926         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1927         struct rpc_task *task;
1928         struct  rpc_message msg = {
1929                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1930                 .rpc_argp = &data->c_arg,
1931                 .rpc_resp = &data->c_res,
1932                 .rpc_cred = data->owner->so_cred,
1933         };
1934         struct rpc_task_setup task_setup_data = {
1935                 .rpc_client = server->client,
1936                 .rpc_message = &msg,
1937                 .callback_ops = &nfs4_open_confirm_ops,
1938                 .callback_data = data,
1939                 .workqueue = nfsiod_workqueue,
1940                 .flags = RPC_TASK_ASYNC,
1941         };
1942         int status;
1943
1944         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1945         kref_get(&data->kref);
1946         data->rpc_done = 0;
1947         data->rpc_status = 0;
1948         data->timestamp = jiffies;
1949         if (data->is_recover)
1950                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1951         task = rpc_run_task(&task_setup_data);
1952         if (IS_ERR(task))
1953                 return PTR_ERR(task);
1954         status = nfs4_wait_for_completion_rpc_task(task);
1955         if (status != 0) {
1956                 data->cancelled = 1;
1957                 smp_wmb();
1958         } else
1959                 status = data->rpc_status;
1960         rpc_put_task(task);
1961         return status;
1962 }
1963
1964 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1965 {
1966         struct nfs4_opendata *data = calldata;
1967         struct nfs4_state_owner *sp = data->owner;
1968         struct nfs_client *clp = sp->so_server->nfs_client;
1969         enum open_claim_type4 claim = data->o_arg.claim;
1970
1971         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1972                 goto out_wait;
1973         /*
1974          * Check if we still need to send an OPEN call, or if we can use
1975          * a delegation instead.
1976          */
1977         if (data->state != NULL) {
1978                 struct nfs_delegation *delegation;
1979
1980                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1981                         goto out_no_action;
1982                 rcu_read_lock();
1983                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1984                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1985                         goto unlock_no_action;
1986                 rcu_read_unlock();
1987         }
1988         /* Update client id. */
1989         data->o_arg.clientid = clp->cl_clientid;
1990         switch (claim) {
1991         default:
1992                 break;
1993         case NFS4_OPEN_CLAIM_PREVIOUS:
1994         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1995         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1996                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1997         case NFS4_OPEN_CLAIM_FH:
1998                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1999                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2000         }
2001         data->timestamp = jiffies;
2002         if (nfs4_setup_sequence(data->o_arg.server,
2003                                 &data->o_arg.seq_args,
2004                                 &data->o_res.seq_res,
2005                                 task) != 0)
2006                 nfs_release_seqid(data->o_arg.seqid);
2007
2008         /* Set the create mode (note dependency on the session type) */
2009         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2010         if (data->o_arg.open_flags & O_EXCL) {
2011                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2012                 if (nfs4_has_persistent_session(clp))
2013                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2014                 else if (clp->cl_mvops->minor_version > 0)
2015                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2016         }
2017         return;
2018 unlock_no_action:
2019         trace_nfs4_cached_open(data->state);
2020         rcu_read_unlock();
2021 out_no_action:
2022         task->tk_action = NULL;
2023 out_wait:
2024         nfs4_sequence_done(task, &data->o_res.seq_res);
2025 }
2026
2027 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2028 {
2029         struct nfs4_opendata *data = calldata;
2030
2031         data->rpc_status = task->tk_status;
2032
2033         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2034                 return;
2035
2036         if (task->tk_status == 0) {
2037                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2038                         switch (data->o_res.f_attr->mode & S_IFMT) {
2039                         case S_IFREG:
2040                                 break;
2041                         case S_IFLNK:
2042                                 data->rpc_status = -ELOOP;
2043                                 break;
2044                         case S_IFDIR:
2045                                 data->rpc_status = -EISDIR;
2046                                 break;
2047                         default:
2048                                 data->rpc_status = -ENOTDIR;
2049                         }
2050                 }
2051                 renew_lease(data->o_res.server, data->timestamp);
2052                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2053                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2054         }
2055         data->rpc_done = 1;
2056 }
2057
2058 static void nfs4_open_release(void *calldata)
2059 {
2060         struct nfs4_opendata *data = calldata;
2061         struct nfs4_state *state = NULL;
2062
2063         /* If this request hasn't been cancelled, do nothing */
2064         if (data->cancelled == 0)
2065                 goto out_free;
2066         /* In case of error, no cleanup! */
2067         if (data->rpc_status != 0 || !data->rpc_done)
2068                 goto out_free;
2069         /* In case we need an open_confirm, no cleanup! */
2070         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2071                 goto out_free;
2072         state = nfs4_opendata_to_nfs4_state(data);
2073         if (!IS_ERR(state))
2074                 nfs4_close_state(state, data->o_arg.fmode);
2075 out_free:
2076         nfs4_opendata_put(data);
2077 }
2078
2079 static const struct rpc_call_ops nfs4_open_ops = {
2080         .rpc_call_prepare = nfs4_open_prepare,
2081         .rpc_call_done = nfs4_open_done,
2082         .rpc_release = nfs4_open_release,
2083 };
2084
2085 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2086 {
2087         struct inode *dir = d_inode(data->dir);
2088         struct nfs_server *server = NFS_SERVER(dir);
2089         struct nfs_openargs *o_arg = &data->o_arg;
2090         struct nfs_openres *o_res = &data->o_res;
2091         struct rpc_task *task;
2092         struct rpc_message msg = {
2093                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2094                 .rpc_argp = o_arg,
2095                 .rpc_resp = o_res,
2096                 .rpc_cred = data->owner->so_cred,
2097         };
2098         struct rpc_task_setup task_setup_data = {
2099                 .rpc_client = server->client,
2100                 .rpc_message = &msg,
2101                 .callback_ops = &nfs4_open_ops,
2102                 .callback_data = data,
2103                 .workqueue = nfsiod_workqueue,
2104                 .flags = RPC_TASK_ASYNC,
2105         };
2106         int status;
2107
2108         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2109         kref_get(&data->kref);
2110         data->rpc_done = 0;
2111         data->rpc_status = 0;
2112         data->cancelled = 0;
2113         data->is_recover = 0;
2114         if (isrecover) {
2115                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2116                 data->is_recover = 1;
2117         }
2118         task = rpc_run_task(&task_setup_data);
2119         if (IS_ERR(task))
2120                 return PTR_ERR(task);
2121         status = nfs4_wait_for_completion_rpc_task(task);
2122         if (status != 0) {
2123                 data->cancelled = 1;
2124                 smp_wmb();
2125         } else
2126                 status = data->rpc_status;
2127         rpc_put_task(task);
2128
2129         return status;
2130 }
2131
2132 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2133 {
2134         struct inode *dir = d_inode(data->dir);
2135         struct nfs_openres *o_res = &data->o_res;
2136         int status;
2137
2138         status = nfs4_run_open_task(data, 1);
2139         if (status != 0 || !data->rpc_done)
2140                 return status;
2141
2142         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2143
2144         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2145                 status = _nfs4_proc_open_confirm(data);
2146                 if (status != 0)
2147                         return status;
2148         }
2149
2150         return status;
2151 }
2152
2153 /*
2154  * Additional permission checks in order to distinguish between an
2155  * open for read, and an open for execute. This works around the
2156  * fact that NFSv4 OPEN treats read and execute permissions as being
2157  * the same.
2158  * Note that in the non-execute case, we want to turn off permission
2159  * checking if we just created a new file (POSIX open() semantics).
2160  */
2161 static int nfs4_opendata_access(struct rpc_cred *cred,
2162                                 struct nfs4_opendata *opendata,
2163                                 struct nfs4_state *state, fmode_t fmode,
2164                                 int openflags)
2165 {
2166         struct nfs_access_entry cache;
2167         u32 mask;
2168
2169         /* access call failed or for some reason the server doesn't
2170          * support any access modes -- defer access call until later */
2171         if (opendata->o_res.access_supported == 0)
2172                 return 0;
2173
2174         mask = 0;
2175         /*
2176          * Use openflags to check for exec, because fmode won't
2177          * always have FMODE_EXEC set when file open for exec.
2178          */
2179         if (openflags & __FMODE_EXEC) {
2180                 /* ONLY check for exec rights */
2181                 mask = MAY_EXEC;
2182         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2183                 mask = MAY_READ;
2184
2185         cache.cred = cred;
2186         cache.jiffies = jiffies;
2187         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2188         nfs_access_add_cache(state->inode, &cache);
2189
2190         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2191                 return 0;
2192
2193         /* even though OPEN succeeded, access is denied. Close the file */
2194         nfs4_close_state(state, fmode);
2195         return -EACCES;
2196 }
2197
2198 /*
2199  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2200  */
2201 static int _nfs4_proc_open(struct nfs4_opendata *data)
2202 {
2203         struct inode *dir = d_inode(data->dir);
2204         struct nfs_server *server = NFS_SERVER(dir);
2205         struct nfs_openargs *o_arg = &data->o_arg;
2206         struct nfs_openres *o_res = &data->o_res;
2207         int status;
2208
2209         status = nfs4_run_open_task(data, 0);
2210         if (!data->rpc_done)
2211                 return status;
2212         if (status != 0) {
2213                 if (status == -NFS4ERR_BADNAME &&
2214                                 !(o_arg->open_flags & O_CREAT))
2215                         return -ENOENT;
2216                 return status;
2217         }
2218
2219         nfs_fattr_map_and_free_names(server, &data->f_attr);
2220
2221         if (o_arg->open_flags & O_CREAT) {
2222                 update_changeattr(dir, &o_res->cinfo);
2223                 if (o_arg->open_flags & O_EXCL)
2224                         data->file_created = 1;
2225                 else if (o_res->cinfo.before != o_res->cinfo.after)
2226                         data->file_created = 1;
2227         }
2228         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2229                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2230         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2231                 status = _nfs4_proc_open_confirm(data);
2232                 if (status != 0)
2233                         return status;
2234         }
2235         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2236                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2237         return 0;
2238 }
2239
2240 static int nfs4_recover_expired_lease(struct nfs_server *server)
2241 {
2242         return nfs4_client_recover_expired_lease(server->nfs_client);
2243 }
2244
2245 /*
2246  * OPEN_EXPIRED:
2247  *      reclaim state on the server after a network partition.
2248  *      Assumes caller holds the appropriate lock
2249  */
2250 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2251 {
2252         struct nfs4_opendata *opendata;
2253         int ret;
2254
2255         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2256                         NFS4_OPEN_CLAIM_FH);
2257         if (IS_ERR(opendata))
2258                 return PTR_ERR(opendata);
2259         ret = nfs4_open_recover(opendata, state);
2260         if (ret == -ESTALE)
2261                 d_drop(ctx->dentry);
2262         nfs4_opendata_put(opendata);
2263         return ret;
2264 }
2265
2266 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2267 {
2268         struct nfs_server *server = NFS_SERVER(state->inode);
2269         struct nfs4_exception exception = { };
2270         int err;
2271
2272         do {
2273                 err = _nfs4_open_expired(ctx, state);
2274                 trace_nfs4_open_expired(ctx, 0, err);
2275                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2276                         continue;
2277                 switch (err) {
2278                 default:
2279                         goto out;
2280                 case -NFS4ERR_GRACE:
2281                 case -NFS4ERR_DELAY:
2282                         nfs4_handle_exception(server, err, &exception);
2283                         err = 0;
2284                 }
2285         } while (exception.retry);
2286 out:
2287         return err;
2288 }
2289
2290 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2291 {
2292         struct nfs_open_context *ctx;
2293         int ret;
2294
2295         ctx = nfs4_state_find_open_context(state);
2296         if (IS_ERR(ctx))
2297                 return -EAGAIN;
2298         ret = nfs4_do_open_expired(ctx, state);
2299         put_nfs_open_context(ctx);
2300         return ret;
2301 }
2302
2303 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2304 {
2305         nfs_remove_bad_delegation(state->inode);
2306         write_seqlock(&state->seqlock);
2307         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2308         write_sequnlock(&state->seqlock);
2309         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2310 }
2311
2312 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2313 {
2314         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2315                 nfs_finish_clear_delegation_stateid(state);
2316 }
2317
2318 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2319 {
2320         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2321         nfs40_clear_delegation_stateid(state);
2322         return nfs4_open_expired(sp, state);
2323 }
2324
2325 #if defined(CONFIG_NFS_V4_1)
2326 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2327 {
2328         struct nfs_server *server = NFS_SERVER(state->inode);
2329         nfs4_stateid stateid;
2330         struct nfs_delegation *delegation;
2331         struct rpc_cred *cred;
2332         int status;
2333
2334         /* Get the delegation credential for use by test/free_stateid */
2335         rcu_read_lock();
2336         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2337         if (delegation == NULL) {
2338                 rcu_read_unlock();
2339                 return;
2340         }
2341
2342         nfs4_stateid_copy(&stateid, &delegation->stateid);
2343         cred = get_rpccred(delegation->cred);
2344         rcu_read_unlock();
2345         status = nfs41_test_stateid(server, &stateid, cred);
2346         trace_nfs4_test_delegation_stateid(state, NULL, status);
2347
2348         if (status != NFS_OK) {
2349                 /* Free the stateid unless the server explicitly
2350                  * informs us the stateid is unrecognized. */
2351                 if (status != -NFS4ERR_BAD_STATEID)
2352                         nfs41_free_stateid(server, &stateid, cred);
2353                 nfs_finish_clear_delegation_stateid(state);
2354         }
2355
2356         put_rpccred(cred);
2357 }
2358
2359 /**
2360  * nfs41_check_open_stateid - possibly free an open stateid
2361  *
2362  * @state: NFSv4 state for an inode
2363  *
2364  * Returns NFS_OK if recovery for this stateid is now finished.
2365  * Otherwise a negative NFS4ERR value is returned.
2366  */
2367 static int nfs41_check_open_stateid(struct nfs4_state *state)
2368 {
2369         struct nfs_server *server = NFS_SERVER(state->inode);
2370         nfs4_stateid *stateid = &state->open_stateid;
2371         struct rpc_cred *cred = state->owner->so_cred;
2372         int status;
2373
2374         /* If a state reset has been done, test_stateid is unneeded */
2375         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2376             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2377             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2378                 return -NFS4ERR_BAD_STATEID;
2379
2380         status = nfs41_test_stateid(server, stateid, cred);
2381         trace_nfs4_test_open_stateid(state, NULL, status);
2382         if (status != NFS_OK) {
2383                 /* Free the stateid unless the server explicitly
2384                  * informs us the stateid is unrecognized. */
2385                 if (status != -NFS4ERR_BAD_STATEID)
2386                         nfs41_free_stateid(server, stateid, cred);
2387
2388                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2389                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2390                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2391                 clear_bit(NFS_OPEN_STATE, &state->flags);
2392         }
2393         return status;
2394 }
2395
2396 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2397 {
2398         int status;
2399
2400         nfs41_check_delegation_stateid(state);
2401         status = nfs41_check_open_stateid(state);
2402         if (status != NFS_OK)
2403                 status = nfs4_open_expired(sp, state);
2404         return status;
2405 }
2406 #endif
2407
2408 /*
2409  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2410  * fields corresponding to attributes that were used to store the verifier.
2411  * Make sure we clobber those fields in the later setattr call
2412  */
2413 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2414                                 struct iattr *sattr, struct nfs4_label **label)
2415 {
2416         const u32 *attrset = opendata->o_res.attrset;
2417
2418         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2419             !(sattr->ia_valid & ATTR_ATIME_SET))
2420                 sattr->ia_valid |= ATTR_ATIME;
2421
2422         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2423             !(sattr->ia_valid & ATTR_MTIME_SET))
2424                 sattr->ia_valid |= ATTR_MTIME;
2425
2426         /* Except MODE, it seems harmless of setting twice. */
2427         if ((attrset[1] & FATTR4_WORD1_MODE))
2428                 sattr->ia_valid &= ~ATTR_MODE;
2429
2430         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2431                 *label = NULL;
2432 }
2433
2434 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2435                 fmode_t fmode,
2436                 int flags,
2437                 struct nfs_open_context *ctx)
2438 {
2439         struct nfs4_state_owner *sp = opendata->owner;
2440         struct nfs_server *server = sp->so_server;
2441         struct dentry *dentry;
2442         struct nfs4_state *state;
2443         unsigned int seq;
2444         int ret;
2445
2446         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2447
2448         ret = _nfs4_proc_open(opendata);
2449         if (ret != 0)
2450                 goto out;
2451
2452         state = nfs4_opendata_to_nfs4_state(opendata);
2453         ret = PTR_ERR(state);
2454         if (IS_ERR(state))
2455                 goto out;
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 if (dentry != ctx->dentry) {
2467                         dput(ctx->dentry);
2468                         ctx->dentry = dget(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         ctx->state = state;
2479         if (d_inode(dentry) == state->inode) {
2480                 nfs_inode_attach_open_context(ctx);
2481                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2482                         nfs4_schedule_stateid_recovery(server, state);
2483         }
2484 out:
2485         return ret;
2486 }
2487
2488 /*
2489  * Returns a referenced nfs4_state
2490  */
2491 static int _nfs4_do_open(struct inode *dir,
2492                         struct nfs_open_context *ctx,
2493                         int flags,
2494                         struct iattr *sattr,
2495                         struct nfs4_label *label,
2496                         int *opened)
2497 {
2498         struct nfs4_state_owner  *sp;
2499         struct nfs4_state     *state = NULL;
2500         struct nfs_server       *server = NFS_SERVER(dir);
2501         struct nfs4_opendata *opendata;
2502         struct dentry *dentry = ctx->dentry;
2503         struct rpc_cred *cred = ctx->cred;
2504         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2505         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2506         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2507         struct nfs4_label *olabel = NULL;
2508         int status;
2509
2510         /* Protect against reboot recovery conflicts */
2511         status = -ENOMEM;
2512         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2513         if (sp == NULL) {
2514                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2515                 goto out_err;
2516         }
2517         status = nfs4_recover_expired_lease(server);
2518         if (status != 0)
2519                 goto err_put_state_owner;
2520         if (d_really_is_positive(dentry))
2521                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2522         status = -ENOMEM;
2523         if (d_really_is_positive(dentry))
2524                 claim = NFS4_OPEN_CLAIM_FH;
2525         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2526                         label, claim, GFP_KERNEL);
2527         if (opendata == NULL)
2528                 goto err_put_state_owner;
2529
2530         if (label) {
2531                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2532                 if (IS_ERR(olabel)) {
2533                         status = PTR_ERR(olabel);
2534                         goto err_opendata_put;
2535                 }
2536         }
2537
2538         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2539                 if (!opendata->f_attr.mdsthreshold) {
2540                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2541                         if (!opendata->f_attr.mdsthreshold)
2542                                 goto err_free_label;
2543                 }
2544                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2545         }
2546         if (d_really_is_positive(dentry))
2547                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2548
2549         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2550         if (status != 0)
2551                 goto err_free_label;
2552         state = ctx->state;
2553
2554         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2555             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2556                 nfs4_exclusive_attrset(opendata, sattr, &label);
2557
2558                 nfs_fattr_init(opendata->o_res.f_attr);
2559                 status = nfs4_do_setattr(state->inode, cred,
2560                                 opendata->o_res.f_attr, sattr,
2561                                 state, label, olabel);
2562                 if (status == 0) {
2563                         nfs_setattr_update_inode(state->inode, sattr,
2564                                         opendata->o_res.f_attr);
2565                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2566                 }
2567         }
2568         if (opened && opendata->file_created)
2569                 *opened |= FILE_CREATED;
2570
2571         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2572                 *ctx_th = opendata->f_attr.mdsthreshold;
2573                 opendata->f_attr.mdsthreshold = NULL;
2574         }
2575
2576         nfs4_label_free(olabel);
2577
2578         nfs4_opendata_put(opendata);
2579         nfs4_put_state_owner(sp);
2580         return 0;
2581 err_free_label:
2582         nfs4_label_free(olabel);
2583 err_opendata_put:
2584         nfs4_opendata_put(opendata);
2585 err_put_state_owner:
2586         nfs4_put_state_owner(sp);
2587 out_err:
2588         return status;
2589 }
2590
2591
2592 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2593                                         struct nfs_open_context *ctx,
2594                                         int flags,
2595                                         struct iattr *sattr,
2596                                         struct nfs4_label *label,
2597                                         int *opened)
2598 {
2599         struct nfs_server *server = NFS_SERVER(dir);
2600         struct nfs4_exception exception = { };
2601         struct nfs4_state *res;
2602         int status;
2603
2604         do {
2605                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2606                 res = ctx->state;
2607                 trace_nfs4_open_file(ctx, flags, status);
2608                 if (status == 0)
2609                         break;
2610                 /* NOTE: BAD_SEQID means the server and client disagree about the
2611                  * book-keeping w.r.t. state-changing operations
2612                  * (OPEN/CLOSE/LOCK/LOCKU...)
2613                  * It is actually a sign of a bug on the client or on the server.
2614                  *
2615                  * If we receive a BAD_SEQID error in the particular case of
2616                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2617                  * have unhashed the old state_owner for us, and that we can
2618                  * therefore safely retry using a new one. We should still warn
2619                  * the user though...
2620                  */
2621                 if (status == -NFS4ERR_BAD_SEQID) {
2622                         pr_warn_ratelimited("NFS: v4 server %s "
2623                                         " returned a bad sequence-id error!\n",
2624                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2625                         exception.retry = 1;
2626                         continue;
2627                 }
2628                 /*
2629                  * BAD_STATEID on OPEN means that the server cancelled our
2630                  * state before it received the OPEN_CONFIRM.
2631                  * Recover by retrying the request as per the discussion
2632                  * on Page 181 of RFC3530.
2633                  */
2634                 if (status == -NFS4ERR_BAD_STATEID) {
2635                         exception.retry = 1;
2636                         continue;
2637                 }
2638                 if (status == -EAGAIN) {
2639                         /* We must have found a delegation */
2640                         exception.retry = 1;
2641                         continue;
2642                 }
2643                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2644                         continue;
2645                 res = ERR_PTR(nfs4_handle_exception(server,
2646                                         status, &exception));
2647         } while (exception.retry);
2648         return res;
2649 }
2650
2651 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2652                             struct nfs_fattr *fattr, struct iattr *sattr,
2653                             struct nfs4_state *state, struct nfs4_label *ilabel,
2654                             struct nfs4_label *olabel)
2655 {
2656         struct nfs_server *server = NFS_SERVER(inode);
2657         struct nfs_setattrargs  arg = {
2658                 .fh             = NFS_FH(inode),
2659                 .iap            = sattr,
2660                 .server         = server,
2661                 .bitmask = server->attr_bitmask,
2662                 .label          = ilabel,
2663         };
2664         struct nfs_setattrres  res = {
2665                 .fattr          = fattr,
2666                 .label          = olabel,
2667                 .server         = server,
2668         };
2669         struct rpc_message msg = {
2670                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2671                 .rpc_argp       = &arg,
2672                 .rpc_resp       = &res,
2673                 .rpc_cred       = cred,
2674         };
2675         unsigned long timestamp = jiffies;
2676         fmode_t fmode;
2677         bool truncate;
2678         int status;
2679
2680         arg.bitmask = nfs4_bitmask(server, ilabel);
2681         if (ilabel)
2682                 arg.bitmask = nfs4_bitmask(server, olabel);
2683
2684         nfs_fattr_init(fattr);
2685
2686         /* Servers should only apply open mode checks for file size changes */
2687         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2688         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2689
2690         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2691                 /* Use that stateid */
2692         } else if (truncate && state != NULL) {
2693                 struct nfs_lockowner lockowner = {
2694                         .l_owner = current->files,
2695                         .l_pid = current->tgid,
2696                 };
2697                 if (!nfs4_valid_open_stateid(state))
2698                         return -EBADF;
2699                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2700                                 &lockowner) == -EIO)
2701                         return -EBADF;
2702         } else
2703                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2704
2705         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2706         if (status == 0 && state != NULL)
2707                 renew_lease(server, timestamp);
2708         trace_nfs4_setattr(inode, &arg.stateid, status);
2709         return status;
2710 }
2711
2712 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2713                            struct nfs_fattr *fattr, struct iattr *sattr,
2714                            struct nfs4_state *state, struct nfs4_label *ilabel,
2715                            struct nfs4_label *olabel)
2716 {
2717         struct nfs_server *server = NFS_SERVER(inode);
2718         struct nfs4_exception exception = {
2719                 .state = state,
2720                 .inode = inode,
2721         };
2722         int err;
2723         do {
2724                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2725                 switch (err) {
2726                 case -NFS4ERR_OPENMODE:
2727                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2728                                 pr_warn_once("NFSv4: server %s is incorrectly "
2729                                                 "applying open mode checks to "
2730                                                 "a SETATTR that is not "
2731                                                 "changing file size.\n",
2732                                                 server->nfs_client->cl_hostname);
2733                         }
2734                         if (state && !(state->state & FMODE_WRITE)) {
2735                                 err = -EBADF;
2736                                 if (sattr->ia_valid & ATTR_OPEN)
2737                                         err = -EACCES;
2738                                 goto out;
2739                         }
2740                 }
2741                 err = nfs4_handle_exception(server, err, &exception);
2742         } while (exception.retry);
2743 out:
2744         return err;
2745 }
2746
2747 static bool
2748 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2749 {
2750         if (inode == NULL || !nfs_have_layout(inode))
2751                 return false;
2752
2753         return pnfs_wait_on_layoutreturn(inode, task);
2754 }
2755
2756 struct nfs4_closedata {
2757         struct inode *inode;
2758         struct nfs4_state *state;
2759         struct nfs_closeargs arg;
2760         struct nfs_closeres res;
2761         struct nfs_fattr fattr;
2762         unsigned long timestamp;
2763         bool roc;
2764         u32 roc_barrier;
2765 };
2766
2767 static void nfs4_free_closedata(void *data)
2768 {
2769         struct nfs4_closedata *calldata = data;
2770         struct nfs4_state_owner *sp = calldata->state->owner;
2771         struct super_block *sb = calldata->state->inode->i_sb;
2772
2773         if (calldata->roc)
2774                 pnfs_roc_release(calldata->state->inode);
2775         nfs4_put_open_state(calldata->state);
2776         nfs_free_seqid(calldata->arg.seqid);
2777         nfs4_put_state_owner(sp);
2778         nfs_sb_deactive(sb);
2779         kfree(calldata);
2780 }
2781
2782 static void nfs4_close_done(struct rpc_task *task, void *data)
2783 {
2784         struct nfs4_closedata *calldata = data;
2785         struct nfs4_state *state = calldata->state;
2786         struct nfs_server *server = NFS_SERVER(calldata->inode);
2787         nfs4_stateid *res_stateid = NULL;
2788
2789         dprintk("%s: begin!\n", __func__);
2790         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2791                 return;
2792         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2793         /* hmm. we are done with the inode, and in the process of freeing
2794          * the state_owner. we keep this around to process errors
2795          */
2796         switch (task->tk_status) {
2797                 case 0:
2798                         res_stateid = &calldata->res.stateid;
2799                         if (calldata->roc)
2800                                 pnfs_roc_set_barrier(state->inode,
2801                                                      calldata->roc_barrier);
2802                         renew_lease(server, calldata->timestamp);
2803                         break;
2804                 case -NFS4ERR_ADMIN_REVOKED:
2805                 case -NFS4ERR_STALE_STATEID:
2806                 case -NFS4ERR_OLD_STATEID:
2807                 case -NFS4ERR_BAD_STATEID:
2808                 case -NFS4ERR_EXPIRED:
2809                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2810                                                 &state->open_stateid)) {
2811                                 rpc_restart_call_prepare(task);
2812                                 goto out_release;
2813                         }
2814                         if (calldata->arg.fmode == 0)
2815                                 break;
2816                 default:
2817                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2818                                 rpc_restart_call_prepare(task);
2819                                 goto out_release;
2820                         }
2821         }
2822         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2823                         res_stateid, calldata->arg.fmode);
2824 out_release:
2825         nfs_release_seqid(calldata->arg.seqid);
2826         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2827         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2828 }
2829
2830 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2831 {
2832         struct nfs4_closedata *calldata = data;
2833         struct nfs4_state *state = calldata->state;
2834         struct inode *inode = calldata->inode;
2835         bool is_rdonly, is_wronly, is_rdwr;
2836         int call_close = 0;
2837
2838         dprintk("%s: begin!\n", __func__);
2839         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2840                 goto out_wait;
2841
2842         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2843         spin_lock(&state->owner->so_lock);
2844         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2845         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2846         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2847         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2848         /* Calculate the change in open mode */
2849         calldata->arg.fmode = 0;
2850         if (state->n_rdwr == 0) {
2851                 if (state->n_rdonly == 0)
2852                         call_close |= is_rdonly;
2853                 else if (is_rdonly)
2854                         calldata->arg.fmode |= FMODE_READ;
2855                 if (state->n_wronly == 0)
2856                         call_close |= is_wronly;
2857                 else if (is_wronly)
2858                         calldata->arg.fmode |= FMODE_WRITE;
2859         } else if (is_rdwr)
2860                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2861
2862         if (calldata->arg.fmode == 0)
2863                 call_close |= is_rdwr;
2864
2865         if (!nfs4_valid_open_stateid(state))
2866                 call_close = 0;
2867         spin_unlock(&state->owner->so_lock);
2868
2869         if (!call_close) {
2870                 /* Note: exit _without_ calling nfs4_close_done */
2871                 goto out_no_action;
2872         }
2873
2874         if (nfs4_wait_on_layoutreturn(inode, task)) {
2875                 nfs_release_seqid(calldata->arg.seqid);
2876                 goto out_wait;
2877         }
2878
2879         if (calldata->arg.fmode == 0)
2880                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2881         if (calldata->roc)
2882                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2883
2884         calldata->arg.share_access =
2885                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2886                                 calldata->arg.fmode, 0);
2887
2888         nfs_fattr_init(calldata->res.fattr);
2889         calldata->timestamp = jiffies;
2890         if (nfs4_setup_sequence(NFS_SERVER(inode),
2891                                 &calldata->arg.seq_args,
2892                                 &calldata->res.seq_res,
2893                                 task) != 0)
2894                 nfs_release_seqid(calldata->arg.seqid);
2895         dprintk("%s: done!\n", __func__);
2896         return;
2897 out_no_action:
2898         task->tk_action = NULL;
2899 out_wait:
2900         nfs4_sequence_done(task, &calldata->res.seq_res);
2901 }
2902
2903 static const struct rpc_call_ops nfs4_close_ops = {
2904         .rpc_call_prepare = nfs4_close_prepare,
2905         .rpc_call_done = nfs4_close_done,
2906         .rpc_release = nfs4_free_closedata,
2907 };
2908
2909 static bool nfs4_roc(struct inode *inode)
2910 {
2911         if (!nfs_have_layout(inode))
2912                 return false;
2913         return pnfs_roc(inode);
2914 }
2915
2916 /*
2917  * It is possible for data to be read/written from a mem-mapped file
2918  * after the sys_close call (which hits the vfs layer as a flush).
2919  * This means that we can't safely call nfsv4 close on a file until
2920  * the inode is cleared. This in turn means that we are not good
2921  * NFSv4 citizens - we do not indicate to the server to update the file's
2922  * share state even when we are done with one of the three share
2923  * stateid's in the inode.
2924  *
2925  * NOTE: Caller must be holding the sp->so_owner semaphore!
2926  */
2927 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2928 {
2929         struct nfs_server *server = NFS_SERVER(state->inode);
2930         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2931         struct nfs4_closedata *calldata;
2932         struct nfs4_state_owner *sp = state->owner;
2933         struct rpc_task *task;
2934         struct rpc_message msg = {
2935                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2936                 .rpc_cred = state->owner->so_cred,
2937         };
2938         struct rpc_task_setup task_setup_data = {
2939                 .rpc_client = server->client,
2940                 .rpc_message = &msg,
2941                 .callback_ops = &nfs4_close_ops,
2942                 .workqueue = nfsiod_workqueue,
2943                 .flags = RPC_TASK_ASYNC,
2944         };
2945         int status = -ENOMEM;
2946
2947         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2948                 &task_setup_data.rpc_client, &msg);
2949
2950         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2951         if (calldata == NULL)
2952                 goto out;
2953         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2954         calldata->inode = state->inode;
2955         calldata->state = state;
2956         calldata->arg.fh = NFS_FH(state->inode);
2957         /* Serialization for the sequence id */
2958         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2959         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2960         if (IS_ERR(calldata->arg.seqid))
2961                 goto out_free_calldata;
2962         calldata->arg.fmode = 0;
2963         calldata->arg.bitmask = server->cache_consistency_bitmask;
2964         calldata->res.fattr = &calldata->fattr;
2965         calldata->res.seqid = calldata->arg.seqid;
2966         calldata->res.server = server;
2967         calldata->roc = nfs4_roc(state->inode);
2968         nfs_sb_active(calldata->inode->i_sb);
2969
2970         msg.rpc_argp = &calldata->arg;
2971         msg.rpc_resp = &calldata->res;
2972         task_setup_data.callback_data = calldata;
2973         task = rpc_run_task(&task_setup_data);
2974         if (IS_ERR(task))
2975                 return PTR_ERR(task);
2976         status = 0;
2977         if (wait)
2978                 status = rpc_wait_for_completion_task(task);
2979         rpc_put_task(task);
2980         return status;
2981 out_free_calldata:
2982         kfree(calldata);
2983 out:
2984         nfs4_put_open_state(state);
2985         nfs4_put_state_owner(sp);
2986         return status;
2987 }
2988
2989 static struct inode *
2990 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2991                 int open_flags, struct iattr *attr, int *opened)
2992 {
2993         struct nfs4_state *state;
2994         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2995
2996         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2997
2998         /* Protect against concurrent sillydeletes */
2999         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3000
3001         nfs4_label_release_security(label);
3002
3003         if (IS_ERR(state))
3004                 return ERR_CAST(state);
3005         return state->inode;
3006 }
3007
3008 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3009 {
3010         if (ctx->state == NULL)
3011                 return;
3012         if (is_sync)
3013                 nfs4_close_sync(ctx->state, ctx->mode);
3014         else
3015                 nfs4_close_state(ctx->state, ctx->mode);
3016 }
3017
3018 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3019 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3020 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3021
3022 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3023 {
3024         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3025         struct nfs4_server_caps_arg args = {
3026                 .fhandle = fhandle,
3027                 .bitmask = bitmask,
3028         };
3029         struct nfs4_server_caps_res res = {};
3030         struct rpc_message msg = {
3031                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3032                 .rpc_argp = &args,
3033                 .rpc_resp = &res,
3034         };
3035         int status;
3036
3037         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3038                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3039                      FATTR4_WORD0_LINK_SUPPORT |
3040                      FATTR4_WORD0_SYMLINK_SUPPORT |
3041                      FATTR4_WORD0_ACLSUPPORT;
3042         if (minorversion)
3043                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3044
3045         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3046         if (status == 0) {
3047                 /* Sanity check the server answers */
3048                 switch (minorversion) {
3049                 case 0:
3050                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3051                         res.attr_bitmask[2] = 0;
3052                         break;
3053                 case 1:
3054                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3055                         break;
3056                 case 2:
3057                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3058                 }
3059                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3060                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3061                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3062                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3063                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3064                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3065                                 NFS_CAP_SECURITY_LABEL);
3066                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3067                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3068                         server->caps |= NFS_CAP_ACLS;
3069                 if (res.has_links != 0)
3070                         server->caps |= NFS_CAP_HARDLINKS;
3071                 if (res.has_symlinks != 0)
3072                         server->caps |= NFS_CAP_SYMLINKS;
3073                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3074                         server->caps |= NFS_CAP_FILEID;
3075                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3076                         server->caps |= NFS_CAP_MODE;
3077                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3078                         server->caps |= NFS_CAP_NLINK;
3079                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3080                         server->caps |= NFS_CAP_OWNER;
3081                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3082                         server->caps |= NFS_CAP_OWNER_GROUP;
3083                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3084                         server->caps |= NFS_CAP_ATIME;
3085                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3086                         server->caps |= NFS_CAP_CTIME;
3087                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3088                         server->caps |= NFS_CAP_MTIME;
3089 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3090                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3091                         server->caps |= NFS_CAP_SECURITY_LABEL;
3092 #endif
3093                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3094                                 sizeof(server->attr_bitmask));
3095                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3096
3097                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3098                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3099                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3100                 server->cache_consistency_bitmask[2] = 0;
3101                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3102                         sizeof(server->exclcreat_bitmask));
3103                 server->acl_bitmask = res.acl_bitmask;
3104                 server->fh_expire_type = res.fh_expire_type;
3105         }
3106
3107         return status;
3108 }
3109
3110 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3111 {
3112         struct nfs4_exception exception = { };
3113         int err;
3114         do {
3115                 err = nfs4_handle_exception(server,
3116                                 _nfs4_server_capabilities(server, fhandle),
3117                                 &exception);
3118         } while (exception.retry);
3119         return err;
3120 }
3121
3122 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3123                 struct nfs_fsinfo *info)
3124 {
3125         u32 bitmask[3];
3126         struct nfs4_lookup_root_arg args = {
3127                 .bitmask = bitmask,
3128         };
3129         struct nfs4_lookup_res res = {
3130                 .server = server,
3131                 .fattr = info->fattr,
3132                 .fh = fhandle,
3133         };
3134         struct rpc_message msg = {
3135                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3136                 .rpc_argp = &args,
3137                 .rpc_resp = &res,
3138         };
3139
3140         bitmask[0] = nfs4_fattr_bitmap[0];
3141         bitmask[1] = nfs4_fattr_bitmap[1];
3142         /*
3143          * Process the label in the upcoming getfattr
3144          */
3145         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3146
3147         nfs_fattr_init(info->fattr);
3148         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3149 }
3150
3151 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3152                 struct nfs_fsinfo *info)
3153 {
3154         struct nfs4_exception exception = { };
3155         int err;
3156         do {
3157                 err = _nfs4_lookup_root(server, fhandle, info);
3158                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3159                 switch (err) {
3160                 case 0:
3161                 case -NFS4ERR_WRONGSEC:
3162                         goto out;
3163                 default:
3164                         err = nfs4_handle_exception(server, err, &exception);
3165                 }
3166         } while (exception.retry);
3167 out:
3168         return err;
3169 }
3170
3171 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3172                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3173 {
3174         struct rpc_auth_create_args auth_args = {
3175                 .pseudoflavor = flavor,
3176         };
3177         struct rpc_auth *auth;
3178         int ret;
3179
3180         auth = rpcauth_create(&auth_args, server->client);
3181         if (IS_ERR(auth)) {
3182                 ret = -EACCES;
3183                 goto out;
3184         }
3185         ret = nfs4_lookup_root(server, fhandle, info);
3186 out:
3187         return ret;
3188 }
3189
3190 /*
3191  * Retry pseudoroot lookup with various security flavors.  We do this when:
3192  *
3193  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3194  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3195  *
3196  * Returns zero on success, or a negative NFS4ERR value, or a
3197  * negative errno value.
3198  */
3199 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3200                               struct nfs_fsinfo *info)
3201 {
3202         /* Per 3530bis 15.33.5 */
3203         static const rpc_authflavor_t flav_array[] = {
3204                 RPC_AUTH_GSS_KRB5P,
3205                 RPC_AUTH_GSS_KRB5I,
3206                 RPC_AUTH_GSS_KRB5,
3207                 RPC_AUTH_UNIX,                  /* courtesy */
3208                 RPC_AUTH_NULL,
3209         };
3210         int status = -EPERM;
3211         size_t i;
3212
3213         if (server->auth_info.flavor_len > 0) {
3214                 /* try each flavor specified by user */
3215                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3216                         status = nfs4_lookup_root_sec(server, fhandle, info,
3217                                                 server->auth_info.flavors[i]);
3218                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3219                                 continue;
3220                         break;
3221                 }
3222         } else {
3223                 /* no flavors specified by user, try default list */
3224                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3225                         status = nfs4_lookup_root_sec(server, fhandle, info,
3226                                                       flav_array[i]);
3227                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3228                                 continue;
3229                         break;
3230                 }
3231         }
3232
3233         /*
3234          * -EACCESS could mean that the user doesn't have correct permissions
3235          * to access the mount.  It could also mean that we tried to mount
3236          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3237          * existing mount programs don't handle -EACCES very well so it should
3238          * be mapped to -EPERM instead.
3239          */
3240         if (status == -EACCES)
3241                 status = -EPERM;
3242         return status;
3243 }
3244
3245 static int nfs4_do_find_root_sec(struct nfs_server *server,
3246                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3247 {
3248         int mv = server->nfs_client->cl_minorversion;
3249         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3250 }
3251
3252 /**
3253  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3254  * @server: initialized nfs_server handle
3255  * @fhandle: we fill in the pseudo-fs root file handle
3256  * @info: we fill in an FSINFO struct
3257  * @auth_probe: probe the auth flavours
3258  *
3259  * Returns zero on success, or a negative errno.
3260  */
3261 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3262                          struct nfs_fsinfo *info,
3263                          bool auth_probe)
3264 {
3265         int status = 0;
3266
3267         if (!auth_probe)
3268                 status = nfs4_lookup_root(server, fhandle, info);
3269
3270         if (auth_probe || status == NFS4ERR_WRONGSEC)
3271                 status = nfs4_do_find_root_sec(server, fhandle, info);
3272
3273         if (status == 0)
3274                 status = nfs4_server_capabilities(server, fhandle);
3275         if (status == 0)
3276                 status = nfs4_do_fsinfo(server, fhandle, info);
3277
3278         return nfs4_map_errors(status);
3279 }
3280
3281 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3282                               struct nfs_fsinfo *info)
3283 {
3284         int error;
3285         struct nfs_fattr *fattr = info->fattr;
3286         struct nfs4_label *label = NULL;
3287
3288         error = nfs4_server_capabilities(server, mntfh);
3289         if (error < 0) {
3290                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3291                 return error;
3292         }
3293
3294         label = nfs4_label_alloc(server, GFP_KERNEL);
3295         if (IS_ERR(label))
3296                 return PTR_ERR(label);
3297
3298         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3299         if (error < 0) {
3300                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3301                 goto err_free_label;
3302         }
3303
3304         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3305             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3306                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3307
3308 err_free_label:
3309         nfs4_label_free(label);
3310
3311         return error;
3312 }
3313
3314 /*
3315  * Get locations and (maybe) other attributes of a referral.
3316  * Note that we'll actually follow the referral later when
3317  * we detect fsid mismatch in inode revalidation
3318  */
3319 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3320                              const struct qstr *name, struct nfs_fattr *fattr,
3321                              struct nfs_fh *fhandle)
3322 {
3323         int status = -ENOMEM;
3324         struct page *page = NULL;
3325         struct nfs4_fs_locations *locations = NULL;
3326
3327         page = alloc_page(GFP_KERNEL);
3328         if (page == NULL)
3329                 goto out;
3330         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3331         if (locations == NULL)
3332                 goto out;
3333
3334         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3335         if (status != 0)
3336                 goto out;
3337
3338         /*
3339          * If the fsid didn't change, this is a migration event, not a
3340          * referral.  Cause us to drop into the exception handler, which
3341          * will kick off migration recovery.
3342          */
3343         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3344                 dprintk("%s: server did not return a different fsid for"
3345                         " a referral at %s\n", __func__, name->name);
3346                 status = -NFS4ERR_MOVED;
3347                 goto out;
3348         }
3349         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3350         nfs_fixup_referral_attributes(&locations->fattr);
3351
3352         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3353         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3354         memset(fhandle, 0, sizeof(struct nfs_fh));
3355 out:
3356         if (page)
3357                 __free_page(page);
3358         kfree(locations);
3359         return status;
3360 }
3361
3362 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3363                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3364 {
3365         struct nfs4_getattr_arg args = {
3366                 .fh = fhandle,
3367                 .bitmask = server->attr_bitmask,
3368         };
3369         struct nfs4_getattr_res res = {
3370                 .fattr = fattr,
3371                 .label = label,
3372                 .server = server,
3373         };
3374         struct rpc_message msg = {
3375                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3376                 .rpc_argp = &args,
3377                 .rpc_resp = &res,
3378         };
3379
3380         args.bitmask = nfs4_bitmask(server, label);
3381
3382         nfs_fattr_init(fattr);
3383         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3384 }
3385
3386 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3387                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3388 {
3389         struct nfs4_exception exception = { };
3390         int err;
3391         do {
3392                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3393                 trace_nfs4_getattr(server, fhandle, fattr, err);
3394                 err = nfs4_handle_exception(server, err,
3395                                 &exception);
3396         } while (exception.retry);
3397         return err;
3398 }
3399
3400 /*
3401  * The file is not closed if it is opened due to the a request to change
3402  * the size of the file. The open call will not be needed once the
3403  * VFS layer lookup-intents are implemented.
3404  *
3405  * Close is called when the inode is destroyed.
3406  * If we haven't opened the file for O_WRONLY, we
3407  * need to in the size_change case to obtain a stateid.
3408  *
3409  * Got race?
3410  * Because OPEN is always done by name in nfsv4, it is
3411  * possible that we opened a different file by the same
3412  * name.  We can recognize this race condition, but we
3413  * can't do anything about it besides returning an error.
3414  *
3415  * This will be fixed with VFS changes (lookup-intent).
3416  */
3417 static int
3418 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3419                   struct iattr *sattr)
3420 {
3421         struct inode *inode = d_inode(dentry);
3422         struct rpc_cred *cred = NULL;
3423         struct nfs4_state *state = NULL;
3424         struct nfs4_label *label = NULL;
3425         int status;
3426
3427         if (pnfs_ld_layoutret_on_setattr(inode) &&
3428             sattr->ia_valid & ATTR_SIZE &&
3429             sattr->ia_size < i_size_read(inode))
3430                 pnfs_commit_and_return_layout(inode);
3431
3432         nfs_fattr_init(fattr);
3433
3434         /* Deal with open(O_TRUNC) */
3435         if (sattr->ia_valid & ATTR_OPEN)
3436                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3437
3438         /* Optimization: if the end result is no change, don't RPC */
3439         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3440                 return 0;
3441
3442         /* Search for an existing open(O_WRITE) file */
3443         if (sattr->ia_valid & ATTR_FILE) {
3444                 struct nfs_open_context *ctx;
3445
3446                 ctx = nfs_file_open_context(sattr->ia_file);
3447                 if (ctx) {
3448                         cred = ctx->cred;
3449                         state = ctx->state;
3450                 }
3451         }
3452
3453         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3454         if (IS_ERR(label))
3455                 return PTR_ERR(label);
3456
3457         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3458         if (status == 0) {
3459                 nfs_setattr_update_inode(inode, sattr, fattr);
3460                 nfs_setsecurity(inode, fattr, label);
3461         }
3462         nfs4_label_free(label);
3463         return status;
3464 }
3465
3466 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3467                 const struct qstr *name, struct nfs_fh *fhandle,
3468                 struct nfs_fattr *fattr, struct nfs4_label *label)
3469 {
3470         struct nfs_server *server = NFS_SERVER(dir);
3471         int                    status;
3472         struct nfs4_lookup_arg args = {
3473                 .bitmask = server->attr_bitmask,
3474                 .dir_fh = NFS_FH(dir),
3475                 .name = name,
3476         };
3477         struct nfs4_lookup_res res = {
3478                 .server = server,
3479                 .fattr = fattr,
3480                 .label = label,
3481                 .fh = fhandle,
3482         };
3483         struct rpc_message msg = {
3484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3485                 .rpc_argp = &args,
3486                 .rpc_resp = &res,
3487         };
3488
3489         args.bitmask = nfs4_bitmask(server, label);
3490
3491         nfs_fattr_init(fattr);
3492
3493         dprintk("NFS call  lookup %s\n", name->name);
3494         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3495         dprintk("NFS reply lookup: %d\n", status);
3496         return status;
3497 }
3498
3499 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3500 {
3501         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3502                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3503         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3504         fattr->nlink = 2;
3505 }
3506
3507 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3508                                    struct qstr *name, struct nfs_fh *fhandle,
3509                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3510 {
3511         struct nfs4_exception exception = { };
3512         struct rpc_clnt *client = *clnt;
3513         int err;
3514         do {
3515                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3516                 trace_nfs4_lookup(dir, name, err);
3517                 switch (err) {
3518                 case -NFS4ERR_BADNAME:
3519                         err = -ENOENT;
3520                         goto out;
3521                 case -NFS4ERR_MOVED:
3522                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3523                         if (err == -NFS4ERR_MOVED)
3524                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3525                         goto out;
3526                 case -NFS4ERR_WRONGSEC:
3527                         err = -EPERM;
3528                         if (client != *clnt)
3529                                 goto out;
3530                         client = nfs4_negotiate_security(client, dir, name);
3531                         if (IS_ERR(client))
3532                                 return PTR_ERR(client);
3533
3534                         exception.retry = 1;
3535                         break;
3536                 default:
3537                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3538                 }
3539         } while (exception.retry);
3540
3541 out:
3542         if (err == 0)
3543                 *clnt = client;
3544         else if (client != *clnt)
3545                 rpc_shutdown_client(client);
3546
3547         return err;
3548 }
3549
3550 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3551                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3552                             struct nfs4_label *label)
3553 {
3554         int status;
3555         struct rpc_clnt *client = NFS_CLIENT(dir);
3556
3557         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3558         if (client != NFS_CLIENT(dir)) {
3559                 rpc_shutdown_client(client);
3560                 nfs_fixup_secinfo_attributes(fattr);
3561         }
3562         return status;
3563 }
3564
3565 struct rpc_clnt *
3566 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3567                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3568 {
3569         struct rpc_clnt *client = NFS_CLIENT(dir);
3570         int status;
3571
3572         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3573         if (status < 0)
3574                 return ERR_PTR(status);
3575         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3576 }
3577
3578 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3579 {
3580         struct nfs_server *server = NFS_SERVER(inode);
3581         struct nfs4_accessargs args = {
3582                 .fh = NFS_FH(inode),
3583                 .bitmask = server->cache_consistency_bitmask,
3584         };
3585         struct nfs4_accessres res = {
3586                 .server = server,
3587         };
3588         struct rpc_message msg = {
3589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3590                 .rpc_argp = &args,
3591                 .rpc_resp = &res,
3592                 .rpc_cred = entry->cred,
3593         };
3594         int mode = entry->mask;
3595         int status = 0;
3596
3597         /*
3598          * Determine which access bits we want to ask for...
3599          */
3600         if (mode & MAY_READ)
3601                 args.access |= NFS4_ACCESS_READ;
3602         if (S_ISDIR(inode->i_mode)) {
3603                 if (mode & MAY_WRITE)
3604                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3605                 if (mode & MAY_EXEC)
3606                         args.access |= NFS4_ACCESS_LOOKUP;
3607         } else {
3608                 if (mode & MAY_WRITE)
3609                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3610                 if (mode & MAY_EXEC)
3611                         args.access |= NFS4_ACCESS_EXECUTE;
3612         }
3613
3614         res.fattr = nfs_alloc_fattr();
3615         if (res.fattr == NULL)
3616                 return -ENOMEM;
3617
3618         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3619         if (!status) {
3620                 nfs_access_set_mask(entry, res.access);
3621                 nfs_refresh_inode(inode, res.fattr);
3622         }
3623         nfs_free_fattr(res.fattr);
3624         return status;
3625 }
3626
3627 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3628 {
3629         struct nfs4_exception exception = { };
3630         int err;
3631         do {
3632                 err = _nfs4_proc_access(inode, entry);
3633                 trace_nfs4_access(inode, err);
3634                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3635                                 &exception);
3636         } while (exception.retry);
3637         return err;
3638 }
3639
3640 /*
3641  * TODO: For the time being, we don't try to get any attributes
3642  * along with any of the zero-copy operations READ, READDIR,
3643  * READLINK, WRITE.
3644  *
3645  * In the case of the first three, we want to put the GETATTR
3646  * after the read-type operation -- this is because it is hard
3647  * to predict the length of a GETATTR response in v4, and thus
3648  * align the READ data correctly.  This means that the GETATTR
3649  * may end up partially falling into the page cache, and we should
3650  * shift it into the 'tail' of the xdr_buf before processing.
3651  * To do this efficiently, we need to know the total length
3652  * of data received, which doesn't seem to be available outside
3653  * of the RPC layer.
3654  *
3655  * In the case of WRITE, we also want to put the GETATTR after
3656  * the operation -- in this case because we want to make sure
3657  * we get the post-operation mtime and size.
3658  *
3659  * Both of these changes to the XDR layer would in fact be quite
3660  * minor, but I decided to leave them for a subsequent patch.
3661  */
3662 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3663                 unsigned int pgbase, unsigned int pglen)
3664 {
3665         struct nfs4_readlink args = {
3666                 .fh       = NFS_FH(inode),
3667                 .pgbase   = pgbase,
3668                 .pglen    = pglen,
3669                 .pages    = &page,
3670         };
3671         struct nfs4_readlink_res res;
3672         struct rpc_message msg = {
3673                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3674                 .rpc_argp = &args,
3675                 .rpc_resp = &res,
3676         };
3677
3678         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3679 }
3680
3681 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3682                 unsigned int pgbase, unsigned int pglen)
3683 {
3684         struct nfs4_exception exception = { };
3685         int err;
3686         do {
3687                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3688                 trace_nfs4_readlink(inode, err);
3689                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3690                                 &exception);
3691         } while (exception.retry);
3692         return err;
3693 }
3694
3695 /*
3696  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3697  */
3698 static int
3699 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3700                  int flags)
3701 {
3702         struct nfs4_label l, *ilabel = NULL;
3703         struct nfs_open_context *ctx;
3704         struct nfs4_state *state;
3705         int status = 0;
3706
3707         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3708         if (IS_ERR(ctx))
3709                 return PTR_ERR(ctx);
3710
3711         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3712
3713         sattr->ia_mode &= ~current_umask();
3714         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3715         if (IS_ERR(state)) {
3716                 status = PTR_ERR(state);
3717                 goto out;
3718         }
3719 out:
3720         nfs4_label_release_security(ilabel);
3721         put_nfs_open_context(ctx);
3722         return status;
3723 }
3724
3725 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3726 {
3727         struct nfs_server *server = NFS_SERVER(dir);
3728         struct nfs_removeargs args = {
3729                 .fh = NFS_FH(dir),
3730                 .name = *name,
3731         };
3732         struct nfs_removeres res = {
3733                 .server = server,
3734         };
3735         struct rpc_message msg = {
3736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3737                 .rpc_argp = &args,
3738                 .rpc_resp = &res,
3739         };
3740         int status;
3741
3742         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3743         if (status == 0)
3744                 update_changeattr(dir, &res.cinfo);
3745         return status;
3746 }
3747
3748 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3749 {
3750         struct nfs4_exception exception = { };
3751         int err;
3752         do {
3753                 err = _nfs4_proc_remove(dir, name);
3754                 trace_nfs4_remove(dir, name, err);
3755                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3756                                 &exception);
3757         } while (exception.retry);
3758         return err;
3759 }
3760
3761 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3762 {
3763         struct nfs_server *server = NFS_SERVER(dir);
3764         struct nfs_removeargs *args = msg->rpc_argp;
3765         struct nfs_removeres *res = msg->rpc_resp;
3766
3767         res->server = server;
3768         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3769         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3770
3771         nfs_fattr_init(res->dir_attr);
3772 }
3773
3774 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3775 {
3776         nfs4_setup_sequence(NFS_SERVER(data->dir),
3777                         &data->args.seq_args,
3778                         &data->res.seq_res,
3779                         task);
3780 }
3781
3782 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3783 {
3784         struct nfs_unlinkdata *data = task->tk_calldata;
3785         struct nfs_removeres *res = &data->res;
3786
3787         if (!nfs4_sequence_done(task, &res->seq_res))
3788                 return 0;
3789         if (nfs4_async_handle_error(task, res->server, NULL,
3790                                     &data->timeout) == -EAGAIN)
3791                 return 0;
3792         update_changeattr(dir, &res->cinfo);
3793         return 1;
3794 }
3795
3796 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3797 {
3798         struct nfs_server *server = NFS_SERVER(dir);
3799         struct nfs_renameargs *arg = msg->rpc_argp;
3800         struct nfs_renameres *res = msg->rpc_resp;
3801
3802         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3803         res->server = server;
3804         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3805 }
3806
3807 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3808 {
3809         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3810                         &data->args.seq_args,
3811                         &data->res.seq_res,
3812                         task);
3813 }
3814
3815 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3816                                  struct inode *new_dir)
3817 {
3818         struct nfs_renamedata *data = task->tk_calldata;
3819         struct nfs_renameres *res = &data->res;
3820
3821         if (!nfs4_sequence_done(task, &res->seq_res))
3822                 return 0;
3823         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3824                 return 0;
3825
3826         update_changeattr(old_dir, &res->old_cinfo);
3827         update_changeattr(new_dir, &res->new_cinfo);
3828         return 1;
3829 }
3830
3831 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3832 {
3833         struct nfs_server *server = NFS_SERVER(inode);
3834         struct nfs4_link_arg arg = {
3835                 .fh     = NFS_FH(inode),
3836                 .dir_fh = NFS_FH(dir),
3837                 .name   = name,
3838                 .bitmask = server->attr_bitmask,
3839         };
3840         struct nfs4_link_res res = {
3841                 .server = server,
3842                 .label = NULL,
3843         };
3844         struct rpc_message msg = {
3845                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3846                 .rpc_argp = &arg,
3847                 .rpc_resp = &res,
3848         };
3849         int status = -ENOMEM;
3850
3851         res.fattr = nfs_alloc_fattr();
3852         if (res.fattr == NULL)
3853                 goto out;
3854
3855         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3856         if (IS_ERR(res.label)) {
3857                 status = PTR_ERR(res.label);
3858                 goto out;
3859         }
3860         arg.bitmask = nfs4_bitmask(server, res.label);
3861
3862         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3863         if (!status) {
3864                 update_changeattr(dir, &res.cinfo);
3865                 status = nfs_post_op_update_inode(inode, res.fattr);
3866                 if (!status)
3867                         nfs_setsecurity(inode, res.fattr, res.label);
3868         }
3869
3870
3871         nfs4_label_free(res.label);
3872
3873 out:
3874         nfs_free_fattr(res.fattr);
3875         return status;
3876 }
3877
3878 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3879 {
3880         struct nfs4_exception exception = { };
3881         int err;
3882         do {
3883                 err = nfs4_handle_exception(NFS_SERVER(inode),
3884                                 _nfs4_proc_link(inode, dir, name),
3885                                 &exception);
3886         } while (exception.retry);
3887         return err;
3888 }
3889
3890 struct nfs4_createdata {
3891         struct rpc_message msg;
3892         struct nfs4_create_arg arg;
3893         struct nfs4_create_res res;
3894         struct nfs_fh fh;
3895         struct nfs_fattr fattr;
3896         struct nfs4_label *label;
3897 };
3898
3899 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3900                 struct qstr *name, struct iattr *sattr, u32 ftype)
3901 {
3902         struct nfs4_createdata *data;
3903
3904         data = kzalloc(sizeof(*data), GFP_KERNEL);
3905         if (data != NULL) {
3906                 struct nfs_server *server = NFS_SERVER(dir);
3907
3908                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3909                 if (IS_ERR(data->label))
3910                         goto out_free;
3911
3912                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3913                 data->msg.rpc_argp = &data->arg;
3914                 data->msg.rpc_resp = &data->res;
3915                 data->arg.dir_fh = NFS_FH(dir);
3916                 data->arg.server = server;
3917                 data->arg.name = name;
3918                 data->arg.attrs = sattr;
3919                 data->arg.ftype = ftype;
3920                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3921                 data->res.server = server;
3922                 data->res.fh = &data->fh;
3923                 data->res.fattr = &data->fattr;
3924                 data->res.label = data->label;
3925                 nfs_fattr_init(data->res.fattr);
3926         }
3927         return data;
3928 out_free:
3929         kfree(data);
3930         return NULL;
3931 }
3932
3933 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3934 {
3935         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3936                                     &data->arg.seq_args, &data->res.seq_res, 1);
3937         if (status == 0) {
3938                 update_changeattr(dir, &data->res.dir_cinfo);
3939                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3940         }
3941         return status;
3942 }
3943
3944 static void nfs4_free_createdata(struct nfs4_createdata *data)
3945 {
3946         nfs4_label_free(data->label);
3947         kfree(data);
3948 }
3949
3950 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3951                 struct page *page, unsigned int len, struct iattr *sattr,
3952                 struct nfs4_label *label)
3953 {
3954         struct nfs4_createdata *data;
3955         int status = -ENAMETOOLONG;
3956
3957         if (len > NFS4_MAXPATHLEN)
3958                 goto out;
3959
3960         status = -ENOMEM;
3961         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3962         if (data == NULL)
3963                 goto out;
3964
3965         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3966         data->arg.u.symlink.pages = &page;
3967         data->arg.u.symlink.len = len;
3968         data->arg.label = label;
3969
3970         status = nfs4_do_create(dir, dentry, data);
3971
3972         nfs4_free_createdata(data);
3973 out:
3974         return status;
3975 }
3976
3977 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3978                 struct page *page, unsigned int len, struct iattr *sattr)
3979 {
3980         struct nfs4_exception exception = { };
3981         struct nfs4_label l, *label = NULL;
3982         int err;
3983
3984         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3985
3986         do {
3987                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3988                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3989                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3990                                 &exception);
3991         } while (exception.retry);
3992
3993         nfs4_label_release_security(label);
3994         return err;
3995 }
3996
3997 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3998                 struct iattr *sattr, struct nfs4_label *label)
3999 {
4000         struct nfs4_createdata *data;
4001         int status = -ENOMEM;
4002
4003         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4004         if (data == NULL)
4005                 goto out;
4006
4007         data->arg.label = label;
4008         status = nfs4_do_create(dir, dentry, data);
4009
4010         nfs4_free_createdata(data);
4011 out:
4012         return status;
4013 }
4014
4015 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4016                 struct iattr *sattr)
4017 {
4018         struct nfs4_exception exception = { };
4019         struct nfs4_label l, *label = NULL;
4020         int err;
4021
4022         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4023
4024         sattr->ia_mode &= ~current_umask();
4025         do {
4026                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4027                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4028                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4029                                 &exception);
4030         } while (exception.retry);
4031         nfs4_label_release_security(label);
4032
4033         return err;
4034 }
4035
4036 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4037                 u64 cookie, struct page **pages, unsigned int count, int plus)
4038 {
4039         struct inode            *dir = d_inode(dentry);
4040         struct nfs4_readdir_arg args = {
4041                 .fh = NFS_FH(dir),
4042                 .pages = pages,
4043                 .pgbase = 0,
4044                 .count = count,
4045                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4046                 .plus = plus,
4047         };
4048         struct nfs4_readdir_res res;
4049         struct rpc_message msg = {
4050                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4051                 .rpc_argp = &args,
4052                 .rpc_resp = &res,
4053                 .rpc_cred = cred,
4054         };
4055         int                     status;
4056
4057         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4058                         dentry,
4059                         (unsigned long long)cookie);
4060         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4061         res.pgbase = args.pgbase;
4062         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4063         if (status >= 0) {
4064                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4065                 status += args.pgbase;
4066         }
4067
4068         nfs_invalidate_atime(dir);
4069
4070         dprintk("%s: returns %d\n", __func__, status);
4071         return status;
4072 }
4073
4074 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4075                 u64 cookie, struct page **pages, unsigned int count, int plus)
4076 {
4077         struct nfs4_exception exception = { };
4078         int err;
4079         do {
4080                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4081                                 pages, count, plus);
4082                 trace_nfs4_readdir(d_inode(dentry), err);
4083                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4084                                 &exception);
4085         } while (exception.retry);
4086         return err;
4087 }
4088
4089 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4090                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4091 {
4092         struct nfs4_createdata *data;
4093         int mode = sattr->ia_mode;
4094         int status = -ENOMEM;
4095
4096         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4097         if (data == NULL)
4098                 goto out;
4099
4100         if (S_ISFIFO(mode))
4101                 data->arg.ftype = NF4FIFO;
4102         else if (S_ISBLK(mode)) {
4103                 data->arg.ftype = NF4BLK;
4104                 data->arg.u.device.specdata1 = MAJOR(rdev);
4105                 data->arg.u.device.specdata2 = MINOR(rdev);
4106         }
4107         else if (S_ISCHR(mode)) {
4108                 data->arg.ftype = NF4CHR;
4109                 data->arg.u.device.specdata1 = MAJOR(rdev);
4110                 data->arg.u.device.specdata2 = MINOR(rdev);
4111         } else if (!S_ISSOCK(mode)) {
4112                 status = -EINVAL;
4113                 goto out_free;
4114         }
4115
4116         data->arg.label = label;
4117         status = nfs4_do_create(dir, dentry, data);
4118 out_free:
4119         nfs4_free_createdata(data);
4120 out:
4121         return status;
4122 }
4123
4124 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4125                 struct iattr *sattr, dev_t rdev)
4126 {
4127         struct nfs4_exception exception = { };
4128         struct nfs4_label l, *label = NULL;
4129         int err;
4130
4131         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4132
4133         sattr->ia_mode &= ~current_umask();
4134         do {
4135                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4136                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4137                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4138                                 &exception);
4139         } while (exception.retry);
4140
4141         nfs4_label_release_security(label);
4142
4143         return err;
4144 }
4145
4146 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4147                  struct nfs_fsstat *fsstat)
4148 {
4149         struct nfs4_statfs_arg args = {
4150                 .fh = fhandle,
4151                 .bitmask = server->attr_bitmask,
4152         };
4153         struct nfs4_statfs_res res = {
4154                 .fsstat = fsstat,
4155         };
4156         struct rpc_message msg = {
4157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4158                 .rpc_argp = &args,
4159                 .rpc_resp = &res,
4160         };
4161
4162         nfs_fattr_init(fsstat->fattr);
4163         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4164 }
4165
4166 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4167 {
4168         struct nfs4_exception exception = { };
4169         int err;
4170         do {
4171                 err = nfs4_handle_exception(server,
4172                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4173                                 &exception);
4174         } while (exception.retry);
4175         return err;
4176 }
4177
4178 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4179                 struct nfs_fsinfo *fsinfo)
4180 {
4181         struct nfs4_fsinfo_arg args = {
4182                 .fh = fhandle,
4183                 .bitmask = server->attr_bitmask,
4184         };
4185         struct nfs4_fsinfo_res res = {
4186                 .fsinfo = fsinfo,
4187         };
4188         struct rpc_message msg = {
4189                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4190                 .rpc_argp = &args,
4191                 .rpc_resp = &res,
4192         };
4193
4194         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4195 }
4196
4197 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4198 {
4199         struct nfs4_exception exception = { };
4200         unsigned long now = jiffies;
4201         int err;
4202
4203         do {
4204                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4205                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4206                 if (err == 0) {
4207                         struct nfs_client *clp = server->nfs_client;
4208
4209                         spin_lock(&clp->cl_lock);
4210                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4211                         clp->cl_last_renewal = now;
4212                         spin_unlock(&clp->cl_lock);
4213                         break;
4214                 }
4215                 err = nfs4_handle_exception(server, err, &exception);
4216         } while (exception.retry);
4217         return err;
4218 }
4219
4220 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4221 {
4222         int error;
4223
4224         nfs_fattr_init(fsinfo->fattr);
4225         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4226         if (error == 0) {
4227                 /* block layout checks this! */
4228                 server->pnfs_blksize = fsinfo->blksize;
4229                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4230         }
4231
4232         return error;
4233 }
4234
4235 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4236                 struct nfs_pathconf *pathconf)
4237 {
4238         struct nfs4_pathconf_arg args = {
4239                 .fh = fhandle,
4240                 .bitmask = server->attr_bitmask,
4241         };
4242         struct nfs4_pathconf_res res = {
4243                 .pathconf = pathconf,
4244         };
4245         struct rpc_message msg = {
4246                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4247                 .rpc_argp = &args,
4248                 .rpc_resp = &res,
4249         };
4250
4251         /* None of the pathconf attributes are mandatory to implement */
4252         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4253                 memset(pathconf, 0, sizeof(*pathconf));
4254                 return 0;
4255         }
4256
4257         nfs_fattr_init(pathconf->fattr);
4258         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4259 }
4260
4261 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4262                 struct nfs_pathconf *pathconf)
4263 {
4264         struct nfs4_exception exception = { };
4265         int err;
4266
4267         do {
4268                 err = nfs4_handle_exception(server,
4269                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4270                                 &exception);
4271         } while (exception.retry);
4272         return err;
4273 }
4274
4275 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4276                 const struct nfs_open_context *ctx,
4277                 const struct nfs_lock_context *l_ctx,
4278                 fmode_t fmode)
4279 {
4280         const struct nfs_lockowner *lockowner = NULL;
4281
4282         if (l_ctx != NULL)
4283                 lockowner = &l_ctx->lockowner;
4284         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4285 }
4286 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4287
4288 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4289                 const struct nfs_open_context *ctx,
4290                 const struct nfs_lock_context *l_ctx,
4291                 fmode_t fmode)
4292 {
4293         nfs4_stateid current_stateid;
4294
4295         /* If the current stateid represents a lost lock, then exit */
4296         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4297                 return true;
4298         return nfs4_stateid_match(stateid, &current_stateid);
4299 }
4300
4301 static bool nfs4_error_stateid_expired(int err)
4302 {
4303         switch (err) {
4304         case -NFS4ERR_DELEG_REVOKED:
4305         case -NFS4ERR_ADMIN_REVOKED:
4306         case -NFS4ERR_BAD_STATEID:
4307         case -NFS4ERR_STALE_STATEID:
4308         case -NFS4ERR_OLD_STATEID:
4309         case -NFS4ERR_OPENMODE:
4310         case -NFS4ERR_EXPIRED:
4311                 return true;
4312         }
4313         return false;
4314 }
4315
4316 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4317 {
4318         nfs_invalidate_atime(hdr->inode);
4319 }
4320
4321 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4322 {
4323         struct nfs_server *server = NFS_SERVER(hdr->inode);
4324
4325         trace_nfs4_read(hdr, task->tk_status);
4326         if (nfs4_async_handle_error(task, server,
4327                                     hdr->args.context->state,
4328                                     NULL) == -EAGAIN) {
4329                 rpc_restart_call_prepare(task);
4330                 return -EAGAIN;
4331         }
4332
4333         __nfs4_read_done_cb(hdr);
4334         if (task->tk_status > 0)
4335                 renew_lease(server, hdr->timestamp);
4336         return 0;
4337 }
4338
4339 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4340                 struct nfs_pgio_args *args)
4341 {
4342
4343         if (!nfs4_error_stateid_expired(task->tk_status) ||
4344                 nfs4_stateid_is_current(&args->stateid,
4345                                 args->context,
4346                                 args->lock_context,
4347                                 FMODE_READ))
4348                 return false;
4349         rpc_restart_call_prepare(task);
4350         return true;
4351 }
4352
4353 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4354 {
4355
4356         dprintk("--> %s\n", __func__);
4357
4358         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4359                 return -EAGAIN;
4360         if (nfs4_read_stateid_changed(task, &hdr->args))
4361                 return -EAGAIN;
4362         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4363                                     nfs4_read_done_cb(task, hdr);
4364 }
4365
4366 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4367                                  struct rpc_message *msg)
4368 {
4369         hdr->timestamp   = jiffies;
4370         hdr->pgio_done_cb = nfs4_read_done_cb;
4371         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4372         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4373 }
4374
4375 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4376                                       struct nfs_pgio_header *hdr)
4377 {
4378         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4379                         &hdr->args.seq_args,
4380                         &hdr->res.seq_res,
4381                         task))
4382                 return 0;
4383         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4384                                 hdr->args.lock_context,
4385                                 hdr->rw_ops->rw_mode) == -EIO)
4386                 return -EIO;
4387         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4388                 return -EIO;
4389         return 0;
4390 }
4391
4392 static int nfs4_write_done_cb(struct rpc_task *task,
4393                               struct nfs_pgio_header *hdr)
4394 {
4395         struct inode *inode = hdr->inode;
4396
4397         trace_nfs4_write(hdr, task->tk_status);
4398         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4399                                     hdr->args.context->state,
4400                                     NULL) == -EAGAIN) {
4401                 rpc_restart_call_prepare(task);
4402                 return -EAGAIN;
4403         }
4404         if (task->tk_status >= 0) {
4405                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4406                 nfs_writeback_update_inode(hdr);
4407         }
4408         return 0;
4409 }
4410
4411 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4412                 struct nfs_pgio_args *args)
4413 {
4414
4415         if (!nfs4_error_stateid_expired(task->tk_status) ||
4416                 nfs4_stateid_is_current(&args->stateid,
4417                                 args->context,
4418                                 args->lock_context,
4419                                 FMODE_WRITE))
4420                 return false;
4421         rpc_restart_call_prepare(task);
4422         return true;
4423 }
4424
4425 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4426 {
4427         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4428                 return -EAGAIN;
4429         if (nfs4_write_stateid_changed(task, &hdr->args))
4430                 return -EAGAIN;
4431         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4432                 nfs4_write_done_cb(task, hdr);
4433 }
4434
4435 static
4436 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4437 {
4438         /* Don't request attributes for pNFS or O_DIRECT writes */
4439         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4440                 return false;
4441         /* Otherwise, request attributes if and only if we don't hold
4442          * a delegation
4443          */
4444         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4445 }
4446
4447 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4448                                   struct rpc_message *msg)
4449 {
4450         struct nfs_server *server = NFS_SERVER(hdr->inode);
4451
4452         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4453                 hdr->args.bitmask = NULL;
4454                 hdr->res.fattr = NULL;
4455         } else
4456                 hdr->args.bitmask = server->cache_consistency_bitmask;
4457
4458         if (!hdr->pgio_done_cb)
4459                 hdr->pgio_done_cb = nfs4_write_done_cb;
4460         hdr->res.server = server;
4461         hdr->timestamp   = jiffies;
4462
4463         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4464         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4465 }
4466
4467 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4468 {
4469         nfs4_setup_sequence(NFS_SERVER(data->inode),
4470                         &data->args.seq_args,
4471                         &data->res.seq_res,
4472                         task);
4473 }
4474
4475 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4476 {
4477         struct inode *inode = data->inode;
4478
4479         trace_nfs4_commit(data, task->tk_status);
4480         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4481                                     NULL, NULL) == -EAGAIN) {
4482                 rpc_restart_call_prepare(task);
4483                 return -EAGAIN;
4484         }
4485         return 0;
4486 }
4487
4488 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4489 {
4490         if (!nfs4_sequence_done(task, &data->res.seq_res))
4491                 return -EAGAIN;
4492         return data->commit_done_cb(task, data);
4493 }
4494
4495 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4496 {
4497         struct nfs_server *server = NFS_SERVER(data->inode);
4498
4499         if (data->commit_done_cb == NULL)
4500                 data->commit_done_cb = nfs4_commit_done_cb;
4501         data->res.server = server;
4502         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4503         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4504 }
4505
4506 struct nfs4_renewdata {
4507         struct nfs_client       *client;
4508         unsigned long           timestamp;
4509 };
4510
4511 /*
4512  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4513  * standalone procedure for queueing an asynchronous RENEW.
4514  */
4515 static void nfs4_renew_release(void *calldata)
4516 {
4517         struct nfs4_renewdata *data = calldata;
4518         struct nfs_client *clp = data->client;
4519
4520         if (atomic_read(&clp->cl_count) > 1)
4521                 nfs4_schedule_state_renewal(clp);
4522         nfs_put_client(clp);
4523         kfree(data);
4524 }
4525
4526 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4527 {
4528         struct nfs4_renewdata *data = calldata;
4529         struct nfs_client *clp = data->client;
4530         unsigned long timestamp = data->timestamp;
4531
4532         trace_nfs4_renew_async(clp, task->tk_status);
4533         switch (task->tk_status) {
4534         case 0:
4535                 break;
4536         case -NFS4ERR_LEASE_MOVED:
4537                 nfs4_schedule_lease_moved_recovery(clp);
4538                 break;
4539         default:
4540                 /* Unless we're shutting down, schedule state recovery! */
4541                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4542                         return;
4543                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4544                         nfs4_schedule_lease_recovery(clp);
4545                         return;
4546                 }
4547                 nfs4_schedule_path_down_recovery(clp);
4548         }
4549         do_renew_lease(clp, timestamp);
4550 }
4551
4552 static const struct rpc_call_ops nfs4_renew_ops = {
4553         .rpc_call_done = nfs4_renew_done,
4554         .rpc_release = nfs4_renew_release,
4555 };
4556
4557 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4558 {
4559         struct rpc_message msg = {
4560                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4561                 .rpc_argp       = clp,
4562                 .rpc_cred       = cred,
4563         };
4564         struct nfs4_renewdata *data;
4565
4566         if (renew_flags == 0)
4567                 return 0;
4568         if (!atomic_inc_not_zero(&clp->cl_count))
4569                 return -EIO;
4570         data = kmalloc(sizeof(*data), GFP_NOFS);
4571         if (data == NULL)
4572                 return -ENOMEM;
4573         data->client = clp;
4574         data->timestamp = jiffies;
4575         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4576                         &nfs4_renew_ops, data);
4577 }
4578
4579 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4580 {
4581         struct rpc_message msg = {
4582                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4583                 .rpc_argp       = clp,
4584                 .rpc_cred       = cred,
4585         };
4586         unsigned long now = jiffies;
4587         int status;
4588
4589         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4590         if (status < 0)
4591                 return status;
4592         do_renew_lease(clp, now);
4593         return 0;
4594 }
4595
4596 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4597 {
4598         return server->caps & NFS_CAP_ACLS;
4599 }
4600
4601 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4602  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4603  * the stack.
4604  */
4605 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4606
4607 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4608                 struct page **pages)
4609 {
4610         struct page *newpage, **spages;
4611         int rc = 0;
4612         size_t len;
4613         spages = pages;
4614
4615         do {
4616                 len = min_t(size_t, PAGE_SIZE, buflen);
4617                 newpage = alloc_page(GFP_KERNEL);
4618
4619                 if (newpage == NULL)
4620                         goto unwind;
4621                 memcpy(page_address(newpage), buf, len);
4622                 buf += len;
4623                 buflen -= len;
4624                 *pages++ = newpage;
4625                 rc++;
4626         } while (buflen != 0);
4627
4628         return rc;
4629
4630 unwind:
4631         for(; rc > 0; rc--)
4632                 __free_page(spages[rc-1]);
4633         return -ENOMEM;
4634 }
4635
4636 struct nfs4_cached_acl {
4637         int cached;
4638         size_t len;
4639         char data[0];
4640 };
4641
4642 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4643 {
4644         struct nfs_inode *nfsi = NFS_I(inode);
4645
4646         spin_lock(&inode->i_lock);
4647         kfree(nfsi->nfs4_acl);
4648         nfsi->nfs4_acl = acl;
4649         spin_unlock(&inode->i_lock);
4650 }
4651
4652 static void nfs4_zap_acl_attr(struct inode *inode)
4653 {
4654         nfs4_set_cached_acl(inode, NULL);
4655 }
4656
4657 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4658 {
4659         struct nfs_inode *nfsi = NFS_I(inode);
4660         struct nfs4_cached_acl *acl;
4661         int ret = -ENOENT;
4662
4663         spin_lock(&inode->i_lock);
4664         acl = nfsi->nfs4_acl;
4665         if (acl == NULL)
4666                 goto out;
4667         if (buf == NULL) /* user is just asking for length */
4668                 goto out_len;
4669         if (acl->cached == 0)
4670                 goto out;
4671         ret = -ERANGE; /* see getxattr(2) man page */
4672         if (acl->len > buflen)
4673                 goto out;
4674         memcpy(buf, acl->data, acl->len);
4675 out_len:
4676         ret = acl->len;
4677 out:
4678         spin_unlock(&inode->i_lock);
4679         return ret;
4680 }
4681
4682 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4683 {
4684         struct nfs4_cached_acl *acl;
4685         size_t buflen = sizeof(*acl) + acl_len;
4686
4687         if (buflen <= PAGE_SIZE) {
4688                 acl = kmalloc(buflen, GFP_KERNEL);
4689                 if (acl == NULL)
4690                         goto out;
4691                 acl->cached = 1;
4692                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4693         } else {
4694                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4695                 if (acl == NULL)
4696                         goto out;
4697                 acl->cached = 0;
4698         }
4699         acl->len = acl_len;
4700 out:
4701         nfs4_set_cached_acl(inode, acl);
4702 }
4703
4704 /*
4705  * The getxattr API returns the required buffer length when called with a
4706  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4707  * the required buf.  On a NULL buf, we send a page of data to the server
4708  * guessing that the ACL request can be serviced by a page. If so, we cache
4709  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4710  * the cache. If not so, we throw away the page, and cache the required
4711  * length. The next getxattr call will then produce another round trip to
4712  * the server, this time with the input buf of the required size.
4713  */
4714 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4715 {
4716         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4717         struct nfs_getaclargs args = {
4718                 .fh = NFS_FH(inode),
4719                 .acl_pages = pages,
4720                 .acl_len = buflen,
4721         };
4722         struct nfs_getaclres res = {
4723                 .acl_len = buflen,
4724         };
4725         struct rpc_message msg = {
4726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4727                 .rpc_argp = &args,
4728                 .rpc_resp = &res,
4729         };
4730         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4731         int ret = -ENOMEM, i;
4732
4733         /* As long as we're doing a round trip to the server anyway,
4734          * let's be prepared for a page of acl data. */
4735         if (npages == 0)
4736                 npages = 1;
4737         if (npages > ARRAY_SIZE(pages))
4738                 return -ERANGE;
4739
4740         for (i = 0; i < npages; i++) {
4741                 pages[i] = alloc_page(GFP_KERNEL);
4742                 if (!pages[i])
4743                         goto out_free;
4744         }
4745
4746         /* for decoding across pages */
4747         res.acl_scratch = alloc_page(GFP_KERNEL);
4748         if (!res.acl_scratch)
4749                 goto out_free;
4750
4751         args.acl_len = npages * PAGE_SIZE;
4752
4753         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4754                 __func__, buf, buflen, npages, args.acl_len);
4755         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4756                              &msg, &args.seq_args, &res.seq_res, 0);
4757         if (ret)
4758                 goto out_free;
4759
4760         /* Handle the case where the passed-in buffer is too short */
4761         if (res.acl_flags & NFS4_ACL_TRUNC) {
4762                 /* Did the user only issue a request for the acl length? */
4763                 if (buf == NULL)
4764                         goto out_ok;
4765                 ret = -ERANGE;
4766                 goto out_free;
4767         }
4768         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4769         if (buf) {
4770                 if (res.acl_len > buflen) {
4771                         ret = -ERANGE;
4772                         goto out_free;
4773                 }
4774                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4775         }
4776 out_ok:
4777         ret = res.acl_len;
4778 out_free:
4779         for (i = 0; i < npages; i++)
4780                 if (pages[i])
4781                         __free_page(pages[i]);
4782         if (res.acl_scratch)
4783                 __free_page(res.acl_scratch);
4784         return ret;
4785 }
4786
4787 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4788 {
4789         struct nfs4_exception exception = { };
4790         ssize_t ret;
4791         do {
4792                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4793                 trace_nfs4_get_acl(inode, ret);
4794                 if (ret >= 0)
4795                         break;
4796                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4797         } while (exception.retry);
4798         return ret;
4799 }
4800
4801 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4802 {
4803         struct nfs_server *server = NFS_SERVER(inode);
4804         int ret;
4805
4806         if (!nfs4_server_supports_acls(server))
4807                 return -EOPNOTSUPP;
4808         ret = nfs_revalidate_inode(server, inode);
4809         if (ret < 0)
4810                 return ret;
4811         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4812                 nfs_zap_acl_cache(inode);
4813         ret = nfs4_read_cached_acl(inode, buf, buflen);
4814         if (ret != -ENOENT)
4815                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4816                  * but no cached acl data, just the acl length */
4817                 return ret;
4818         return nfs4_get_acl_uncached(inode, buf, buflen);
4819 }
4820
4821 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4822 {
4823         struct nfs_server *server = NFS_SERVER(inode);
4824         struct page *pages[NFS4ACL_MAXPAGES];
4825         struct nfs_setaclargs arg = {
4826                 .fh             = NFS_FH(inode),
4827                 .acl_pages      = pages,
4828                 .acl_len        = buflen,
4829         };
4830         struct nfs_setaclres res;
4831         struct rpc_message msg = {
4832                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4833                 .rpc_argp       = &arg,
4834                 .rpc_resp       = &res,
4835         };
4836         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4837         int ret, i;
4838
4839         if (!nfs4_server_supports_acls(server))
4840                 return -EOPNOTSUPP;
4841         if (npages > ARRAY_SIZE(pages))
4842                 return -ERANGE;
4843         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4844         if (i < 0)
4845                 return i;
4846         nfs4_inode_return_delegation(inode);
4847         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4848
4849         /*
4850          * Free each page after tx, so the only ref left is
4851          * held by the network stack
4852          */
4853         for (; i > 0; i--)
4854                 put_page(pages[i-1]);
4855
4856         /*
4857          * Acl update can result in inode attribute update.
4858          * so mark the attribute cache invalid.
4859          */
4860         spin_lock(&inode->i_lock);
4861         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4862         spin_unlock(&inode->i_lock);
4863         nfs_access_zap_cache(inode);
4864         nfs_zap_acl_cache(inode);
4865         return ret;
4866 }
4867
4868 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4869 {
4870         struct nfs4_exception exception = { };
4871         int err;
4872         do {
4873                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4874                 trace_nfs4_set_acl(inode, err);
4875                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4876                                 &exception);
4877         } while (exception.retry);
4878         return err;
4879 }
4880
4881 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4882 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4883                                         size_t buflen)
4884 {
4885         struct nfs_server *server = NFS_SERVER(inode);
4886         struct nfs_fattr fattr;
4887         struct nfs4_label label = {0, 0, buflen, buf};
4888
4889         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4890         struct nfs4_getattr_arg arg = {
4891                 .fh             = NFS_FH(inode),
4892                 .bitmask        = bitmask,
4893         };
4894         struct nfs4_getattr_res res = {
4895                 .fattr          = &fattr,
4896                 .label          = &label,
4897                 .server         = server,
4898         };
4899         struct rpc_message msg = {
4900                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4901                 .rpc_argp       = &arg,
4902                 .rpc_resp       = &res,
4903         };
4904         int ret;
4905
4906         nfs_fattr_init(&fattr);
4907
4908         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4909         if (ret)
4910                 return ret;
4911         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4912                 return -ENOENT;
4913         if (buflen < label.len)
4914                 return -ERANGE;
4915         return 0;
4916 }
4917
4918 static int nfs4_get_security_label(struct inode *inode, void *buf,
4919                                         size_t buflen)
4920 {
4921         struct nfs4_exception exception = { };
4922         int err;
4923
4924         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4925                 return -EOPNOTSUPP;
4926
4927         do {
4928                 err = _nfs4_get_security_label(inode, buf, buflen);
4929                 trace_nfs4_get_security_label(inode, err);
4930                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4931                                 &exception);
4932         } while (exception.retry);
4933         return err;
4934 }
4935
4936 static int _nfs4_do_set_security_label(struct inode *inode,
4937                 struct nfs4_label *ilabel,
4938                 struct nfs_fattr *fattr,
4939                 struct nfs4_label *olabel)
4940 {
4941
4942         struct iattr sattr = {0};
4943         struct nfs_server *server = NFS_SERVER(inode);
4944         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4945         struct nfs_setattrargs arg = {
4946                 .fh             = NFS_FH(inode),
4947                 .iap            = &sattr,
4948                 .server         = server,
4949                 .bitmask        = bitmask,
4950                 .label          = ilabel,
4951         };
4952         struct nfs_setattrres res = {
4953                 .fattr          = fattr,
4954                 .label          = olabel,
4955                 .server         = server,
4956         };
4957         struct rpc_message msg = {
4958                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4959                 .rpc_argp       = &arg,
4960                 .rpc_resp       = &res,
4961         };
4962         int status;
4963
4964         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4965
4966         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4967         if (status)
4968                 dprintk("%s failed: %d\n", __func__, status);
4969
4970         return status;
4971 }
4972
4973 static int nfs4_do_set_security_label(struct inode *inode,
4974                 struct nfs4_label *ilabel,
4975                 struct nfs_fattr *fattr,
4976                 struct nfs4_label *olabel)
4977 {
4978         struct nfs4_exception exception = { };
4979         int err;
4980
4981         do {
4982                 err = _nfs4_do_set_security_label(inode, ilabel,
4983                                 fattr, olabel);
4984                 trace_nfs4_set_security_label(inode, err);
4985                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4986                                 &exception);
4987         } while (exception.retry);
4988         return err;
4989 }
4990
4991 static int
4992 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4993 {
4994         struct nfs4_label ilabel, *olabel = NULL;
4995         struct nfs_fattr fattr;
4996         struct rpc_cred *cred;
4997         struct inode *inode = d_inode(dentry);
4998         int status;
4999
5000         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5001                 return -EOPNOTSUPP;
5002
5003         nfs_fattr_init(&fattr);
5004
5005         ilabel.pi = 0;
5006         ilabel.lfs = 0;
5007         ilabel.label = (char *)buf;
5008         ilabel.len = buflen;
5009
5010         cred = rpc_lookup_cred();
5011         if (IS_ERR(cred))
5012                 return PTR_ERR(cred);
5013
5014         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5015         if (IS_ERR(olabel)) {
5016                 status = -PTR_ERR(olabel);
5017                 goto out;
5018         }
5019
5020         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5021         if (status == 0)
5022                 nfs_setsecurity(inode, &fattr, olabel);
5023
5024         nfs4_label_free(olabel);
5025 out:
5026         put_rpccred(cred);
5027         return status;
5028 }
5029 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5030
5031
5032 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5033                                     nfs4_verifier *bootverf)
5034 {
5035         __be32 verf[2];
5036
5037         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5038                 /* An impossible timestamp guarantees this value
5039                  * will never match a generated boot time. */
5040                 verf[0] = 0;
5041                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5042         } else {
5043                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5044                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5045                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5046         }
5047         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5048 }
5049
5050 static int
5051 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5052 {
5053         int result;
5054         size_t len;
5055         char *str;
5056
5057         if (clp->cl_owner_id != NULL)
5058                 return 0;
5059
5060         rcu_read_lock();
5061         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5062                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5063                 1 +
5064                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5065                 1;
5066         rcu_read_unlock();
5067
5068         if (len > NFS4_OPAQUE_LIMIT + 1)
5069                 return -EINVAL;
5070
5071         /*
5072          * Since this string is allocated at mount time, and held until the
5073          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5074          * about a memory-reclaim deadlock.
5075          */
5076         str = kmalloc(len, GFP_KERNEL);
5077         if (!str)
5078                 return -ENOMEM;
5079
5080         rcu_read_lock();
5081         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5082                         clp->cl_ipaddr,
5083                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5084                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5085         rcu_read_unlock();
5086
5087         clp->cl_owner_id = str;
5088         return 0;
5089 }
5090
5091 static int
5092 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5093 {
5094         int result;
5095         size_t len;
5096         char *str;
5097
5098         len = 10 + 10 + 1 + 10 + 1 +
5099                 strlen(nfs4_client_id_uniquifier) + 1 +
5100                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5101
5102         if (len > NFS4_OPAQUE_LIMIT + 1)
5103                 return -EINVAL;
5104
5105         /*
5106          * Since this string is allocated at mount time, and held until the
5107          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5108          * about a memory-reclaim deadlock.
5109          */
5110         str = kmalloc(len, GFP_KERNEL);
5111         if (!str)
5112                 return -ENOMEM;
5113
5114         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5115                         clp->rpc_ops->version, clp->cl_minorversion,
5116                         nfs4_client_id_uniquifier,
5117                         clp->cl_rpcclient->cl_nodename);
5118         clp->cl_owner_id = str;
5119         return 0;
5120 }
5121
5122 static int
5123 nfs4_init_uniform_client_string(struct nfs_client *clp)
5124 {
5125         int result;
5126         size_t len;
5127         char *str;
5128
5129         if (clp->cl_owner_id != NULL)
5130                 return 0;
5131
5132         if (nfs4_client_id_uniquifier[0] != '\0')
5133                 return nfs4_init_uniquifier_client_string(clp);
5134
5135         len = 10 + 10 + 1 + 10 + 1 +
5136                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5137
5138         if (len > NFS4_OPAQUE_LIMIT + 1)
5139                 return -EINVAL;
5140
5141         /*
5142          * Since this string is allocated at mount time, and held until the
5143          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5144          * about a memory-reclaim deadlock.
5145          */
5146         str = kmalloc(len, GFP_KERNEL);
5147         if (!str)
5148                 return -ENOMEM;
5149
5150         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5151                         clp->rpc_ops->version, clp->cl_minorversion,
5152                         clp->cl_rpcclient->cl_nodename);
5153         clp->cl_owner_id = str;
5154         return 0;
5155 }
5156
5157 /*
5158  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5159  * services.  Advertise one based on the address family of the
5160  * clientaddr.
5161  */
5162 static unsigned int
5163 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5164 {
5165         if (strchr(clp->cl_ipaddr, ':') != NULL)
5166                 return scnprintf(buf, len, "tcp6");
5167         else
5168                 return scnprintf(buf, len, "tcp");
5169 }
5170
5171 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5172 {
5173         struct nfs4_setclientid *sc = calldata;
5174
5175         if (task->tk_status == 0)
5176                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5177 }
5178
5179 static const struct rpc_call_ops nfs4_setclientid_ops = {
5180         .rpc_call_done = nfs4_setclientid_done,
5181 };
5182
5183 /**
5184  * nfs4_proc_setclientid - Negotiate client ID
5185  * @clp: state data structure
5186  * @program: RPC program for NFSv4 callback service
5187  * @port: IP port number for NFS4 callback service
5188  * @cred: RPC credential to use for this call
5189  * @res: where to place the result
5190  *
5191  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5192  */
5193 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5194                 unsigned short port, struct rpc_cred *cred,
5195                 struct nfs4_setclientid_res *res)
5196 {
5197         nfs4_verifier sc_verifier;
5198         struct nfs4_setclientid setclientid = {
5199                 .sc_verifier = &sc_verifier,
5200                 .sc_prog = program,
5201                 .sc_clnt = clp,
5202         };
5203         struct rpc_message msg = {
5204                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5205                 .rpc_argp = &setclientid,
5206                 .rpc_resp = res,
5207                 .rpc_cred = cred,
5208         };
5209         struct rpc_task *task;
5210         struct rpc_task_setup task_setup_data = {
5211                 .rpc_client = clp->cl_rpcclient,
5212                 .rpc_message = &msg,
5213                 .callback_ops = &nfs4_setclientid_ops,
5214                 .callback_data = &setclientid,
5215                 .flags = RPC_TASK_TIMEOUT,
5216         };
5217         int status;
5218
5219         /* nfs_client_id4 */
5220         nfs4_init_boot_verifier(clp, &sc_verifier);
5221
5222         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5223                 status = nfs4_init_uniform_client_string(clp);
5224         else
5225                 status = nfs4_init_nonuniform_client_string(clp);
5226
5227         if (status)
5228                 goto out;
5229
5230         /* cb_client4 */
5231         setclientid.sc_netid_len =
5232                                 nfs4_init_callback_netid(clp,
5233                                                 setclientid.sc_netid,
5234                                                 sizeof(setclientid.sc_netid));
5235         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5236                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5237                                 clp->cl_ipaddr, port >> 8, port & 255);
5238
5239         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5240                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5241                 clp->cl_owner_id);
5242         task = rpc_run_task(&task_setup_data);
5243         if (IS_ERR(task)) {
5244                 status = PTR_ERR(task);
5245                 goto out;
5246         }
5247         status = task->tk_status;
5248         if (setclientid.sc_cred) {
5249                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5250                 put_rpccred(setclientid.sc_cred);
5251         }
5252         rpc_put_task(task);
5253 out:
5254         trace_nfs4_setclientid(clp, status);
5255         dprintk("NFS reply setclientid: %d\n", status);
5256         return status;
5257 }
5258
5259 /**
5260  * nfs4_proc_setclientid_confirm - Confirm client ID
5261  * @clp: state data structure
5262  * @res: result of a previous SETCLIENTID
5263  * @cred: RPC credential to use for this call
5264  *
5265  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5266  */
5267 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5268                 struct nfs4_setclientid_res *arg,
5269                 struct rpc_cred *cred)
5270 {
5271         struct rpc_message msg = {
5272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5273                 .rpc_argp = arg,
5274                 .rpc_cred = cred,
5275         };
5276         int status;
5277
5278         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5279                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5280                 clp->cl_clientid);
5281         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5282         trace_nfs4_setclientid_confirm(clp, status);
5283         dprintk("NFS reply setclientid_confirm: %d\n", status);
5284         return status;
5285 }
5286
5287 struct nfs4_delegreturndata {
5288         struct nfs4_delegreturnargs args;
5289         struct nfs4_delegreturnres res;
5290         struct nfs_fh fh;
5291         nfs4_stateid stateid;
5292         unsigned long timestamp;
5293         struct nfs_fattr fattr;
5294         int rpc_status;
5295         struct inode *inode;
5296         bool roc;
5297         u32 roc_barrier;
5298 };
5299
5300 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5301 {
5302         struct nfs4_delegreturndata *data = calldata;
5303
5304         if (!nfs4_sequence_done(task, &data->res.seq_res))
5305                 return;
5306
5307         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5308         switch (task->tk_status) {
5309         case 0:
5310                 renew_lease(data->res.server, data->timestamp);
5311         case -NFS4ERR_ADMIN_REVOKED:
5312         case -NFS4ERR_DELEG_REVOKED:
5313         case -NFS4ERR_BAD_STATEID:
5314         case -NFS4ERR_OLD_STATEID:
5315         case -NFS4ERR_STALE_STATEID:
5316         case -NFS4ERR_EXPIRED:
5317                 task->tk_status = 0;
5318                 if (data->roc)
5319                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5320                 break;
5321         default:
5322                 if (nfs4_async_handle_error(task, data->res.server,
5323                                             NULL, NULL) == -EAGAIN) {
5324                         rpc_restart_call_prepare(task);
5325                         return;
5326                 }
5327         }
5328         data->rpc_status = task->tk_status;
5329 }
5330
5331 static void nfs4_delegreturn_release(void *calldata)
5332 {
5333         struct nfs4_delegreturndata *data = calldata;
5334         struct inode *inode = data->inode;
5335
5336         if (inode) {
5337                 if (data->roc)
5338                         pnfs_roc_release(inode);
5339                 nfs_iput_and_deactive(inode);
5340         }
5341         kfree(calldata);
5342 }
5343
5344 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5345 {
5346         struct nfs4_delegreturndata *d_data;
5347
5348         d_data = (struct nfs4_delegreturndata *)data;
5349
5350         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5351                 return;
5352
5353         if (d_data->roc)
5354                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5355
5356         nfs4_setup_sequence(d_data->res.server,
5357                         &d_data->args.seq_args,
5358                         &d_data->res.seq_res,
5359                         task);
5360 }
5361
5362 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5363         .rpc_call_prepare = nfs4_delegreturn_prepare,
5364         .rpc_call_done = nfs4_delegreturn_done,
5365         .rpc_release = nfs4_delegreturn_release,
5366 };
5367
5368 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5369 {
5370         struct nfs4_delegreturndata *data;
5371         struct nfs_server *server = NFS_SERVER(inode);
5372         struct rpc_task *task;
5373         struct rpc_message msg = {
5374                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5375                 .rpc_cred = cred,
5376         };
5377         struct rpc_task_setup task_setup_data = {
5378                 .rpc_client = server->client,
5379                 .rpc_message = &msg,
5380                 .callback_ops = &nfs4_delegreturn_ops,
5381                 .flags = RPC_TASK_ASYNC,
5382         };
5383         int status = 0;
5384
5385         data = kzalloc(sizeof(*data), GFP_NOFS);
5386         if (data == NULL)
5387                 return -ENOMEM;
5388         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5389         data->args.fhandle = &data->fh;
5390         data->args.stateid = &data->stateid;
5391         data->args.bitmask = server->cache_consistency_bitmask;
5392         nfs_copy_fh(&data->fh, NFS_FH(inode));
5393         nfs4_stateid_copy(&data->stateid, stateid);
5394         data->res.fattr = &data->fattr;
5395         data->res.server = server;
5396         nfs_fattr_init(data->res.fattr);
5397         data->timestamp = jiffies;
5398         data->rpc_status = 0;
5399         data->inode = nfs_igrab_and_active(inode);
5400         if (data->inode)
5401                 data->roc = nfs4_roc(inode);
5402
5403         task_setup_data.callback_data = data;
5404         msg.rpc_argp = &data->args;
5405         msg.rpc_resp = &data->res;
5406         task = rpc_run_task(&task_setup_data);
5407         if (IS_ERR(task))
5408                 return PTR_ERR(task);
5409         if (!issync)
5410                 goto out;
5411         status = nfs4_wait_for_completion_rpc_task(task);
5412         if (status != 0)
5413                 goto out;
5414         status = data->rpc_status;
5415         if (status == 0)
5416                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5417         else
5418                 nfs_refresh_inode(inode, &data->fattr);
5419 out:
5420         rpc_put_task(task);
5421         return status;
5422 }
5423
5424 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5425 {
5426         struct nfs_server *server = NFS_SERVER(inode);
5427         struct nfs4_exception exception = { };
5428         int err;
5429         do {
5430                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5431                 trace_nfs4_delegreturn(inode, stateid, err);
5432                 switch (err) {
5433                         case -NFS4ERR_STALE_STATEID:
5434                         case -NFS4ERR_EXPIRED:
5435                         case 0:
5436                                 return 0;
5437                 }
5438                 err = nfs4_handle_exception(server, err, &exception);
5439         } while (exception.retry);
5440         return err;
5441 }
5442
5443 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5444 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5445
5446 /*
5447  * sleep, with exponential backoff, and retry the LOCK operation.
5448  */
5449 static unsigned long
5450 nfs4_set_lock_task_retry(unsigned long timeout)
5451 {
5452         freezable_schedule_timeout_killable_unsafe(timeout);
5453         timeout <<= 1;
5454         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5455                 return NFS4_LOCK_MAXTIMEOUT;
5456         return timeout;
5457 }
5458
5459 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5460 {
5461         struct inode *inode = state->inode;
5462         struct nfs_server *server = NFS_SERVER(inode);
5463         struct nfs_client *clp = server->nfs_client;
5464         struct nfs_lockt_args arg = {
5465                 .fh = NFS_FH(inode),
5466                 .fl = request,
5467         };
5468         struct nfs_lockt_res res = {
5469                 .denied = request,
5470         };
5471         struct rpc_message msg = {
5472                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5473                 .rpc_argp       = &arg,
5474                 .rpc_resp       = &res,
5475                 .rpc_cred       = state->owner->so_cred,
5476         };
5477         struct nfs4_lock_state *lsp;
5478         int status;
5479
5480         arg.lock_owner.clientid = clp->cl_clientid;
5481         status = nfs4_set_lock_state(state, request);
5482         if (status != 0)
5483                 goto out;
5484         lsp = request->fl_u.nfs4_fl.owner;
5485         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5486         arg.lock_owner.s_dev = server->s_dev;
5487         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5488         switch (status) {
5489                 case 0:
5490                         request->fl_type = F_UNLCK;
5491                         break;
5492                 case -NFS4ERR_DENIED:
5493                         status = 0;
5494         }
5495         request->fl_ops->fl_release_private(request);
5496         request->fl_ops = NULL;
5497 out:
5498         return status;
5499 }
5500
5501 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5502 {
5503         struct nfs4_exception exception = { };
5504         int err;
5505
5506         do {
5507                 err = _nfs4_proc_getlk(state, cmd, request);
5508                 trace_nfs4_get_lock(request, state, cmd, err);
5509                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5510                                 &exception);
5511         } while (exception.retry);
5512         return err;
5513 }
5514
5515 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5516 {
5517         return locks_lock_inode_wait(inode, fl);
5518 }
5519
5520 struct nfs4_unlockdata {
5521         struct nfs_locku_args arg;
5522         struct nfs_locku_res res;
5523         struct nfs4_lock_state *lsp;
5524         struct nfs_open_context *ctx;
5525         struct file_lock fl;
5526         struct nfs_server *server;
5527         unsigned long timestamp;
5528 };
5529
5530 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5531                 struct nfs_open_context *ctx,
5532                 struct nfs4_lock_state *lsp,
5533                 struct nfs_seqid *seqid)
5534 {
5535         struct nfs4_unlockdata *p;
5536         struct inode *inode = lsp->ls_state->inode;
5537
5538         p = kzalloc(sizeof(*p), GFP_NOFS);
5539         if (p == NULL)
5540                 return NULL;
5541         p->arg.fh = NFS_FH(inode);
5542         p->arg.fl = &p->fl;
5543         p->arg.seqid = seqid;
5544         p->res.seqid = seqid;
5545         p->lsp = lsp;
5546         atomic_inc(&lsp->ls_count);
5547         /* Ensure we don't close file until we're done freeing locks! */
5548         p->ctx = get_nfs_open_context(ctx);
5549         memcpy(&p->fl, fl, sizeof(p->fl));
5550         p->server = NFS_SERVER(inode);
5551         return p;
5552 }
5553
5554 static void nfs4_locku_release_calldata(void *data)
5555 {
5556         struct nfs4_unlockdata *calldata = data;
5557         nfs_free_seqid(calldata->arg.seqid);
5558         nfs4_put_lock_state(calldata->lsp);
5559         put_nfs_open_context(calldata->ctx);
5560         kfree(calldata);
5561 }
5562
5563 static void nfs4_locku_done(struct rpc_task *task, void *data)
5564 {
5565         struct nfs4_unlockdata *calldata = data;
5566
5567         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5568                 return;
5569         switch (task->tk_status) {
5570                 case 0:
5571                         renew_lease(calldata->server, calldata->timestamp);
5572                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5573                         if (nfs4_update_lock_stateid(calldata->lsp,
5574                                         &calldata->res.stateid))
5575                                 break;
5576                 case -NFS4ERR_BAD_STATEID:
5577                 case -NFS4ERR_OLD_STATEID:
5578                 case -NFS4ERR_STALE_STATEID:
5579                 case -NFS4ERR_EXPIRED:
5580                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5581                                                 &calldata->lsp->ls_stateid))
5582                                 rpc_restart_call_prepare(task);
5583                         break;
5584                 default:
5585                         if (nfs4_async_handle_error(task, calldata->server,
5586                                                     NULL, NULL) == -EAGAIN)
5587                                 rpc_restart_call_prepare(task);
5588         }
5589         nfs_release_seqid(calldata->arg.seqid);
5590 }
5591
5592 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5593 {
5594         struct nfs4_unlockdata *calldata = data;
5595
5596         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5597                 goto out_wait;
5598         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5599         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5600                 /* Note: exit _without_ running nfs4_locku_done */
5601                 goto out_no_action;
5602         }
5603         calldata->timestamp = jiffies;
5604         if (nfs4_setup_sequence(calldata->server,
5605                                 &calldata->arg.seq_args,
5606                                 &calldata->res.seq_res,
5607                                 task) != 0)
5608                 nfs_release_seqid(calldata->arg.seqid);
5609         return;
5610 out_no_action:
5611         task->tk_action = NULL;
5612 out_wait:
5613         nfs4_sequence_done(task, &calldata->res.seq_res);
5614 }
5615
5616 static const struct rpc_call_ops nfs4_locku_ops = {
5617         .rpc_call_prepare = nfs4_locku_prepare,
5618         .rpc_call_done = nfs4_locku_done,
5619         .rpc_release = nfs4_locku_release_calldata,
5620 };
5621
5622 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5623                 struct nfs_open_context *ctx,
5624                 struct nfs4_lock_state *lsp,
5625                 struct nfs_seqid *seqid)
5626 {
5627         struct nfs4_unlockdata *data;
5628         struct rpc_message msg = {
5629                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5630                 .rpc_cred = ctx->cred,
5631         };
5632         struct rpc_task_setup task_setup_data = {
5633                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5634                 .rpc_message = &msg,
5635                 .callback_ops = &nfs4_locku_ops,
5636                 .workqueue = nfsiod_workqueue,
5637                 .flags = RPC_TASK_ASYNC,
5638         };
5639
5640         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5641                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5642
5643         /* Ensure this is an unlock - when canceling a lock, the
5644          * canceled lock is passed in, and it won't be an unlock.
5645          */
5646         fl->fl_type = F_UNLCK;
5647
5648         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5649         if (data == NULL) {
5650                 nfs_free_seqid(seqid);
5651                 return ERR_PTR(-ENOMEM);
5652         }
5653
5654         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5655         msg.rpc_argp = &data->arg;
5656         msg.rpc_resp = &data->res;
5657         task_setup_data.callback_data = data;
5658         return rpc_run_task(&task_setup_data);
5659 }
5660
5661 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5662 {
5663         struct inode *inode = state->inode;
5664         struct nfs4_state_owner *sp = state->owner;
5665         struct nfs_inode *nfsi = NFS_I(inode);
5666         struct nfs_seqid *seqid;
5667         struct nfs4_lock_state *lsp;
5668         struct rpc_task *task;
5669         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5670         int status = 0;
5671         unsigned char fl_flags = request->fl_flags;
5672
5673         status = nfs4_set_lock_state(state, request);
5674         /* Unlock _before_ we do the RPC call */
5675         request->fl_flags |= FL_EXISTS;
5676         /* Exclude nfs_delegation_claim_locks() */
5677         mutex_lock(&sp->so_delegreturn_mutex);
5678         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5679         down_read(&nfsi->rwsem);
5680         if (do_vfs_lock(inode, request) == -ENOENT) {
5681                 up_read(&nfsi->rwsem);
5682                 mutex_unlock(&sp->so_delegreturn_mutex);
5683                 goto out;
5684         }
5685         up_read(&nfsi->rwsem);
5686         mutex_unlock(&sp->so_delegreturn_mutex);
5687         if (status != 0)
5688                 goto out;
5689         /* Is this a delegated lock? */
5690         lsp = request->fl_u.nfs4_fl.owner;
5691         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5692                 goto out;
5693         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5694         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5695         status = -ENOMEM;
5696         if (IS_ERR(seqid))
5697                 goto out;
5698         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5699         status = PTR_ERR(task);
5700         if (IS_ERR(task))
5701                 goto out;
5702         status = nfs4_wait_for_completion_rpc_task(task);
5703         rpc_put_task(task);
5704 out:
5705         request->fl_flags = fl_flags;
5706         trace_nfs4_unlock(request, state, F_SETLK, status);
5707         return status;
5708 }
5709
5710 struct nfs4_lockdata {
5711         struct nfs_lock_args arg;
5712         struct nfs_lock_res res;
5713         struct nfs4_lock_state *lsp;
5714         struct nfs_open_context *ctx;
5715         struct file_lock fl;
5716         unsigned long timestamp;
5717         int rpc_status;
5718         int cancelled;
5719         struct nfs_server *server;
5720 };
5721
5722 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5723                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5724                 gfp_t gfp_mask)
5725 {
5726         struct nfs4_lockdata *p;
5727         struct inode *inode = lsp->ls_state->inode;
5728         struct nfs_server *server = NFS_SERVER(inode);
5729         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5730
5731         p = kzalloc(sizeof(*p), gfp_mask);
5732         if (p == NULL)
5733                 return NULL;
5734
5735         p->arg.fh = NFS_FH(inode);
5736         p->arg.fl = &p->fl;
5737         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5738         if (IS_ERR(p->arg.open_seqid))
5739                 goto out_free;
5740         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5741         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5742         if (IS_ERR(p->arg.lock_seqid))
5743                 goto out_free_seqid;
5744         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5745         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5746         p->arg.lock_owner.s_dev = server->s_dev;
5747         p->res.lock_seqid = p->arg.lock_seqid;
5748         p->lsp = lsp;
5749         p->server = server;
5750         atomic_inc(&lsp->ls_count);
5751         p->ctx = get_nfs_open_context(ctx);
5752         get_file(fl->fl_file);
5753         memcpy(&p->fl, fl, sizeof(p->fl));
5754         return p;
5755 out_free_seqid:
5756         nfs_free_seqid(p->arg.open_seqid);
5757 out_free:
5758         kfree(p);
5759         return NULL;
5760 }
5761
5762 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5763 {
5764         struct nfs4_lockdata *data = calldata;
5765         struct nfs4_state *state = data->lsp->ls_state;
5766
5767         dprintk("%s: begin!\n", __func__);
5768         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5769                 goto out_wait;
5770         /* Do we need to do an open_to_lock_owner? */
5771         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5772                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5773                         goto out_release_lock_seqid;
5774                 }
5775                 nfs4_stateid_copy(&data->arg.open_stateid,
5776                                 &state->open_stateid);
5777                 data->arg.new_lock_owner = 1;
5778                 data->res.open_seqid = data->arg.open_seqid;
5779         } else {
5780                 data->arg.new_lock_owner = 0;
5781                 nfs4_stateid_copy(&data->arg.lock_stateid,
5782                                 &data->lsp->ls_stateid);
5783         }
5784         if (!nfs4_valid_open_stateid(state)) {
5785                 data->rpc_status = -EBADF;
5786                 task->tk_action = NULL;
5787                 goto out_release_open_seqid;
5788         }
5789         data->timestamp = jiffies;
5790         if (nfs4_setup_sequence(data->server,
5791                                 &data->arg.seq_args,
5792                                 &data->res.seq_res,
5793                                 task) == 0)
5794                 return;
5795 out_release_open_seqid:
5796         nfs_release_seqid(data->arg.open_seqid);
5797 out_release_lock_seqid:
5798         nfs_release_seqid(data->arg.lock_seqid);
5799 out_wait:
5800         nfs4_sequence_done(task, &data->res.seq_res);
5801         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5802 }
5803
5804 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5805 {
5806         struct nfs4_lockdata *data = calldata;
5807         struct nfs4_lock_state *lsp = data->lsp;
5808
5809         dprintk("%s: begin!\n", __func__);
5810
5811         if (!nfs4_sequence_done(task, &data->res.seq_res))
5812                 return;
5813
5814         data->rpc_status = task->tk_status;
5815         switch (task->tk_status) {
5816         case 0:
5817                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5818                                 data->timestamp);
5819                 if (data->arg.new_lock) {
5820                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5821                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5822                                 rpc_restart_call_prepare(task);
5823                                 break;
5824                         }
5825                 }
5826                 if (data->arg.new_lock_owner != 0) {
5827                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5828                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5829                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5830                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5831                         rpc_restart_call_prepare(task);
5832                 break;
5833         case -NFS4ERR_BAD_STATEID:
5834         case -NFS4ERR_OLD_STATEID:
5835         case -NFS4ERR_STALE_STATEID:
5836         case -NFS4ERR_EXPIRED:
5837                 if (data->arg.new_lock_owner != 0) {
5838                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5839                                                 &lsp->ls_state->open_stateid))
5840                                 rpc_restart_call_prepare(task);
5841                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5842                                                 &lsp->ls_stateid))
5843                                 rpc_restart_call_prepare(task);
5844         }
5845         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5846 }
5847
5848 static void nfs4_lock_release(void *calldata)
5849 {
5850         struct nfs4_lockdata *data = calldata;
5851
5852         dprintk("%s: begin!\n", __func__);
5853         nfs_free_seqid(data->arg.open_seqid);
5854         if (data->cancelled != 0) {
5855                 struct rpc_task *task;
5856                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5857                                 data->arg.lock_seqid);
5858                 if (!IS_ERR(task))
5859                         rpc_put_task_async(task);
5860                 dprintk("%s: cancelling lock!\n", __func__);
5861         } else
5862                 nfs_free_seqid(data->arg.lock_seqid);
5863         nfs4_put_lock_state(data->lsp);
5864         put_nfs_open_context(data->ctx);
5865         fput(data->fl.fl_file);
5866         kfree(data);
5867         dprintk("%s: done!\n", __func__);
5868 }
5869
5870 static const struct rpc_call_ops nfs4_lock_ops = {
5871         .rpc_call_prepare = nfs4_lock_prepare,
5872         .rpc_call_done = nfs4_lock_done,
5873         .rpc_release = nfs4_lock_release,
5874 };
5875
5876 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5877 {
5878         switch (error) {
5879         case -NFS4ERR_ADMIN_REVOKED:
5880         case -NFS4ERR_BAD_STATEID:
5881                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5882                 if (new_lock_owner != 0 ||
5883                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5884                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5885                 break;
5886         case -NFS4ERR_STALE_STATEID:
5887                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5888         case -NFS4ERR_EXPIRED:
5889                 nfs4_schedule_lease_recovery(server->nfs_client);
5890         };
5891 }
5892
5893 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5894 {
5895         struct nfs4_lockdata *data;
5896         struct rpc_task *task;
5897         struct rpc_message msg = {
5898                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5899                 .rpc_cred = state->owner->so_cred,
5900         };
5901         struct rpc_task_setup task_setup_data = {
5902                 .rpc_client = NFS_CLIENT(state->inode),
5903                 .rpc_message = &msg,
5904                 .callback_ops = &nfs4_lock_ops,
5905                 .workqueue = nfsiod_workqueue,
5906                 .flags = RPC_TASK_ASYNC,
5907         };
5908         int ret;
5909
5910         dprintk("%s: begin!\n", __func__);
5911         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5912                         fl->fl_u.nfs4_fl.owner,
5913                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5914         if (data == NULL)
5915                 return -ENOMEM;
5916         if (IS_SETLKW(cmd))
5917                 data->arg.block = 1;
5918         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5919         msg.rpc_argp = &data->arg;
5920         msg.rpc_resp = &data->res;
5921         task_setup_data.callback_data = data;
5922         if (recovery_type > NFS_LOCK_NEW) {
5923                 if (recovery_type == NFS_LOCK_RECLAIM)
5924                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5925                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5926         } else
5927                 data->arg.new_lock = 1;
5928         task = rpc_run_task(&task_setup_data);
5929         if (IS_ERR(task))
5930                 return PTR_ERR(task);
5931         ret = nfs4_wait_for_completion_rpc_task(task);
5932         if (ret == 0) {
5933                 ret = data->rpc_status;
5934                 if (ret)
5935                         nfs4_handle_setlk_error(data->server, data->lsp,
5936                                         data->arg.new_lock_owner, ret);
5937         } else
5938                 data->cancelled = 1;
5939         rpc_put_task(task);
5940         dprintk("%s: done, ret = %d!\n", __func__, ret);
5941         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5942         return ret;
5943 }
5944
5945 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5946 {
5947         struct nfs_server *server = NFS_SERVER(state->inode);
5948         struct nfs4_exception exception = {
5949                 .inode = state->inode,
5950         };
5951         int err;
5952
5953         do {
5954                 /* Cache the lock if possible... */
5955                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5956                         return 0;
5957                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5958                 if (err != -NFS4ERR_DELAY)
5959                         break;
5960                 nfs4_handle_exception(server, err, &exception);
5961         } while (exception.retry);
5962         return err;
5963 }
5964
5965 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5966 {
5967         struct nfs_server *server = NFS_SERVER(state->inode);
5968         struct nfs4_exception exception = {
5969                 .inode = state->inode,
5970         };
5971         int err;
5972
5973         err = nfs4_set_lock_state(state, request);
5974         if (err != 0)
5975                 return err;
5976         if (!recover_lost_locks) {
5977                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5978                 return 0;
5979         }
5980         do {
5981                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5982                         return 0;
5983                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5984                 switch (err) {
5985                 default:
5986                         goto out;
5987                 case -NFS4ERR_GRACE:
5988                 case -NFS4ERR_DELAY:
5989                         nfs4_handle_exception(server, err, &exception);
5990                         err = 0;
5991                 }
5992         } while (exception.retry);
5993 out:
5994         return err;
5995 }
5996
5997 #if defined(CONFIG_NFS_V4_1)
5998 /**
5999  * nfs41_check_expired_locks - possibly free a lock stateid
6000  *
6001  * @state: NFSv4 state for an inode
6002  *
6003  * Returns NFS_OK if recovery for this stateid is now finished.
6004  * Otherwise a negative NFS4ERR value is returned.
6005  */
6006 static int nfs41_check_expired_locks(struct nfs4_state *state)
6007 {
6008         int status, ret = -NFS4ERR_BAD_STATEID;
6009         struct nfs4_lock_state *lsp;
6010         struct nfs_server *server = NFS_SERVER(state->inode);
6011
6012         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6013                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6014                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6015
6016                         status = nfs41_test_stateid(server,
6017                                         &lsp->ls_stateid,
6018                                         cred);
6019                         trace_nfs4_test_lock_stateid(state, lsp, status);
6020                         if (status != NFS_OK) {
6021                                 /* Free the stateid unless the server
6022                                  * informs us the stateid is unrecognized. */
6023                                 if (status != -NFS4ERR_BAD_STATEID)
6024                                         nfs41_free_stateid(server,
6025                                                         &lsp->ls_stateid,
6026                                                         cred);
6027                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6028                                 ret = status;
6029                         }
6030                 }
6031         };
6032
6033         return ret;
6034 }
6035
6036 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6037 {
6038         int status = NFS_OK;
6039
6040         if (test_bit(LK_STATE_IN_USE, &state->flags))
6041                 status = nfs41_check_expired_locks(state);
6042         if (status != NFS_OK)
6043                 status = nfs4_lock_expired(state, request);
6044         return status;
6045 }
6046 #endif
6047
6048 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6049 {
6050         struct nfs_inode *nfsi = NFS_I(state->inode);
6051         unsigned char fl_flags = request->fl_flags;
6052         int status = -ENOLCK;
6053
6054         if ((fl_flags & FL_POSIX) &&
6055                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6056                 goto out;
6057         /* Is this a delegated open? */
6058         status = nfs4_set_lock_state(state, request);
6059         if (status != 0)
6060                 goto out;
6061         request->fl_flags |= FL_ACCESS;
6062         status = do_vfs_lock(state->inode, request);
6063         if (status < 0)
6064                 goto out;
6065         down_read(&nfsi->rwsem);
6066         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6067                 /* Yes: cache locks! */
6068                 /* ...but avoid races with delegation recall... */
6069                 request->fl_flags = fl_flags & ~FL_SLEEP;
6070                 status = do_vfs_lock(state->inode, request);
6071                 up_read(&nfsi->rwsem);
6072                 goto out;
6073         }
6074         up_read(&nfsi->rwsem);
6075         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6076 out:
6077         request->fl_flags = fl_flags;
6078         return status;
6079 }
6080
6081 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6082 {
6083         struct nfs4_exception exception = {
6084                 .state = state,
6085                 .inode = state->inode,
6086         };
6087         int err;
6088
6089         do {
6090                 err = _nfs4_proc_setlk(state, cmd, request);
6091                 if (err == -NFS4ERR_DENIED)
6092                         err = -EAGAIN;
6093                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6094                                 err, &exception);
6095         } while (exception.retry);
6096         return err;
6097 }
6098
6099 static int
6100 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6101 {
6102         struct nfs_open_context *ctx;
6103         struct nfs4_state *state;
6104         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6105         int status;
6106
6107         /* verify open state */
6108         ctx = nfs_file_open_context(filp);
6109         state = ctx->state;
6110
6111         if (request->fl_start < 0 || request->fl_end < 0)
6112                 return -EINVAL;
6113
6114         if (IS_GETLK(cmd)) {
6115                 if (state != NULL)
6116                         return nfs4_proc_getlk(state, F_GETLK, request);
6117                 return 0;
6118         }
6119
6120         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6121                 return -EINVAL;
6122
6123         if (request->fl_type == F_UNLCK) {
6124                 if (state != NULL)
6125                         return nfs4_proc_unlck(state, cmd, request);
6126                 return 0;
6127         }
6128
6129         if (state == NULL)
6130                 return -ENOLCK;
6131         /*
6132          * Don't rely on the VFS having checked the file open mode,
6133          * since it won't do this for flock() locks.
6134          */
6135         switch (request->fl_type) {
6136         case F_RDLCK:
6137                 if (!(filp->f_mode & FMODE_READ))
6138                         return -EBADF;
6139                 break;
6140         case F_WRLCK:
6141                 if (!(filp->f_mode & FMODE_WRITE))
6142                         return -EBADF;
6143         }
6144
6145         do {
6146                 status = nfs4_proc_setlk(state, cmd, request);
6147                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6148                         break;
6149                 timeout = nfs4_set_lock_task_retry(timeout);
6150                 status = -ERESTARTSYS;
6151                 if (signalled())
6152                         break;
6153         } while(status < 0);
6154         return status;
6155 }
6156
6157 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6158 {
6159         struct nfs_server *server = NFS_SERVER(state->inode);
6160         int err;
6161
6162         err = nfs4_set_lock_state(state, fl);
6163         if (err != 0)
6164                 return err;
6165         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6166         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6167 }
6168
6169 struct nfs_release_lockowner_data {
6170         struct nfs4_lock_state *lsp;
6171         struct nfs_server *server;
6172         struct nfs_release_lockowner_args args;
6173         struct nfs_release_lockowner_res res;
6174         unsigned long timestamp;
6175 };
6176
6177 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6178 {
6179         struct nfs_release_lockowner_data *data = calldata;
6180         struct nfs_server *server = data->server;
6181         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6182                              &data->args.seq_args, &data->res.seq_res, task);
6183         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6184         data->timestamp = jiffies;
6185 }
6186
6187 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6188 {
6189         struct nfs_release_lockowner_data *data = calldata;
6190         struct nfs_server *server = data->server;
6191
6192         nfs40_sequence_done(task, &data->res.seq_res);
6193
6194         switch (task->tk_status) {
6195         case 0:
6196                 renew_lease(server, data->timestamp);
6197                 break;
6198         case -NFS4ERR_STALE_CLIENTID:
6199         case -NFS4ERR_EXPIRED:
6200                 nfs4_schedule_lease_recovery(server->nfs_client);
6201                 break;
6202         case -NFS4ERR_LEASE_MOVED:
6203         case -NFS4ERR_DELAY:
6204                 if (nfs4_async_handle_error(task, server,
6205                                             NULL, NULL) == -EAGAIN)
6206                         rpc_restart_call_prepare(task);
6207         }
6208 }
6209
6210 static void nfs4_release_lockowner_release(void *calldata)
6211 {
6212         struct nfs_release_lockowner_data *data = calldata;
6213         nfs4_free_lock_state(data->server, data->lsp);
6214         kfree(calldata);
6215 }
6216
6217 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6218         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6219         .rpc_call_done = nfs4_release_lockowner_done,
6220         .rpc_release = nfs4_release_lockowner_release,
6221 };
6222
6223 static void
6224 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6225 {
6226         struct nfs_release_lockowner_data *data;
6227         struct rpc_message msg = {
6228                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6229         };
6230
6231         if (server->nfs_client->cl_mvops->minor_version != 0)
6232                 return;
6233
6234         data = kmalloc(sizeof(*data), GFP_NOFS);
6235         if (!data)
6236                 return;
6237         data->lsp = lsp;
6238         data->server = server;
6239         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6240         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6241         data->args.lock_owner.s_dev = server->s_dev;
6242
6243         msg.rpc_argp = &data->args;
6244         msg.rpc_resp = &data->res;
6245         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6246         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6247 }
6248
6249 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6250
6251 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6252                                    struct dentry *dentry, const char *key,
6253                                    const void *buf, size_t buflen,
6254                                    int flags)
6255 {
6256         if (strcmp(key, "") != 0)
6257                 return -EINVAL;
6258
6259         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6260 }
6261
6262 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6263                                    struct dentry *dentry, const char *key,
6264                                    void *buf, size_t buflen)
6265 {
6266         if (strcmp(key, "") != 0)
6267                 return -EINVAL;
6268
6269         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6270 }
6271
6272 static size_t nfs4_xattr_list_nfs4_acl(const struct xattr_handler *handler,
6273                                        struct dentry *dentry, char *list,
6274                                        size_t list_len, const char *name,
6275                                        size_t name_len)
6276 {
6277         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6278
6279         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6280                 return 0;
6281
6282         if (list && len <= list_len)
6283                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6284         return len;
6285 }
6286
6287 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6288 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6289 {
6290         return server->caps & NFS_CAP_SECURITY_LABEL;
6291 }
6292
6293 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6294                                      struct dentry *dentry, const char *key,
6295                                      const void *buf, size_t buflen,
6296                                      int flags)
6297 {
6298         if (security_ismaclabel(key))
6299                 return nfs4_set_security_label(dentry, buf, buflen);
6300
6301         return -EOPNOTSUPP;
6302 }
6303
6304 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6305                                      struct dentry *dentry, const char *key,
6306                                      void *buf, size_t buflen)
6307 {
6308         if (security_ismaclabel(key))
6309                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6310         return -EOPNOTSUPP;
6311 }
6312
6313 static size_t nfs4_xattr_list_nfs4_label(const struct xattr_handler *handler,
6314                                          struct dentry *dentry, char *list,
6315                                          size_t list_len, const char *name,
6316                                          size_t name_len)
6317 {
6318         size_t len = 0;
6319
6320         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6321                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6322                 if (list && len <= list_len)
6323                         security_inode_listsecurity(d_inode(dentry), list, len);
6324         }
6325         return len;
6326 }
6327
6328 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6329         .prefix = XATTR_SECURITY_PREFIX,
6330         .list   = nfs4_xattr_list_nfs4_label,
6331         .get    = nfs4_xattr_get_nfs4_label,
6332         .set    = nfs4_xattr_set_nfs4_label,
6333 };
6334 #endif
6335
6336
6337 /*
6338  * nfs_fhget will use either the mounted_on_fileid or the fileid
6339  */
6340 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6341 {
6342         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6343                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6344               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6345               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6346                 return;
6347
6348         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6349                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6350         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6351         fattr->nlink = 2;
6352 }
6353
6354 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6355                                    const struct qstr *name,
6356                                    struct nfs4_fs_locations *fs_locations,
6357                                    struct page *page)
6358 {
6359         struct nfs_server *server = NFS_SERVER(dir);
6360         u32 bitmask[3] = {
6361                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6362         };
6363         struct nfs4_fs_locations_arg args = {
6364                 .dir_fh = NFS_FH(dir),
6365                 .name = name,
6366                 .page = page,
6367                 .bitmask = bitmask,
6368         };
6369         struct nfs4_fs_locations_res res = {
6370                 .fs_locations = fs_locations,
6371         };
6372         struct rpc_message msg = {
6373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6374                 .rpc_argp = &args,
6375                 .rpc_resp = &res,
6376         };
6377         int status;
6378
6379         dprintk("%s: start\n", __func__);
6380
6381         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6382          * is not supported */
6383         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6384                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6385         else
6386                 bitmask[0] |= FATTR4_WORD0_FILEID;
6387
6388         nfs_fattr_init(&fs_locations->fattr);
6389         fs_locations->server = server;
6390         fs_locations->nlocations = 0;
6391         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6392         dprintk("%s: returned status = %d\n", __func__, status);
6393         return status;
6394 }
6395
6396 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6397                            const struct qstr *name,
6398                            struct nfs4_fs_locations *fs_locations,
6399                            struct page *page)
6400 {
6401         struct nfs4_exception exception = { };
6402         int err;
6403         do {
6404                 err = _nfs4_proc_fs_locations(client, dir, name,
6405                                 fs_locations, page);
6406                 trace_nfs4_get_fs_locations(dir, name, err);
6407                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6408                                 &exception);
6409         } while (exception.retry);
6410         return err;
6411 }
6412
6413 /*
6414  * This operation also signals the server that this client is
6415  * performing migration recovery.  The server can stop returning
6416  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6417  * appended to this compound to identify the client ID which is
6418  * performing recovery.
6419  */
6420 static int _nfs40_proc_get_locations(struct inode *inode,
6421                                      struct nfs4_fs_locations *locations,
6422                                      struct page *page, struct rpc_cred *cred)
6423 {
6424         struct nfs_server *server = NFS_SERVER(inode);
6425         struct rpc_clnt *clnt = server->client;
6426         u32 bitmask[2] = {
6427                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6428         };
6429         struct nfs4_fs_locations_arg args = {
6430                 .clientid       = server->nfs_client->cl_clientid,
6431                 .fh             = NFS_FH(inode),
6432                 .page           = page,
6433                 .bitmask        = bitmask,
6434                 .migration      = 1,            /* skip LOOKUP */
6435                 .renew          = 1,            /* append RENEW */
6436         };
6437         struct nfs4_fs_locations_res res = {
6438                 .fs_locations   = locations,
6439                 .migration      = 1,
6440                 .renew          = 1,
6441         };
6442         struct rpc_message msg = {
6443                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6444                 .rpc_argp       = &args,
6445                 .rpc_resp       = &res,
6446                 .rpc_cred       = cred,
6447         };
6448         unsigned long now = jiffies;
6449         int status;
6450
6451         nfs_fattr_init(&locations->fattr);
6452         locations->server = server;
6453         locations->nlocations = 0;
6454
6455         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6456         nfs4_set_sequence_privileged(&args.seq_args);
6457         status = nfs4_call_sync_sequence(clnt, server, &msg,
6458                                         &args.seq_args, &res.seq_res);
6459         if (status)
6460                 return status;
6461
6462         renew_lease(server, now);
6463         return 0;
6464 }
6465
6466 #ifdef CONFIG_NFS_V4_1
6467
6468 /*
6469  * This operation also signals the server that this client is
6470  * performing migration recovery.  The server can stop asserting
6471  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6472  * performing this operation is identified in the SEQUENCE
6473  * operation in this compound.
6474  *
6475  * When the client supports GETATTR(fs_locations_info), it can
6476  * be plumbed in here.
6477  */
6478 static int _nfs41_proc_get_locations(struct inode *inode,
6479                                      struct nfs4_fs_locations *locations,
6480                                      struct page *page, struct rpc_cred *cred)
6481 {
6482         struct nfs_server *server = NFS_SERVER(inode);
6483         struct rpc_clnt *clnt = server->client;
6484         u32 bitmask[2] = {
6485                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6486         };
6487         struct nfs4_fs_locations_arg args = {
6488                 .fh             = NFS_FH(inode),
6489                 .page           = page,
6490                 .bitmask        = bitmask,
6491                 .migration      = 1,            /* skip LOOKUP */
6492         };
6493         struct nfs4_fs_locations_res res = {
6494                 .fs_locations   = locations,
6495                 .migration      = 1,
6496         };
6497         struct rpc_message msg = {
6498                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6499                 .rpc_argp       = &args,
6500                 .rpc_resp       = &res,
6501                 .rpc_cred       = cred,
6502         };
6503         int status;
6504
6505         nfs_fattr_init(&locations->fattr);
6506         locations->server = server;
6507         locations->nlocations = 0;
6508
6509         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6510         nfs4_set_sequence_privileged(&args.seq_args);
6511         status = nfs4_call_sync_sequence(clnt, server, &msg,
6512                                         &args.seq_args, &res.seq_res);
6513         if (status == NFS4_OK &&
6514             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6515                 status = -NFS4ERR_LEASE_MOVED;
6516         return status;
6517 }
6518
6519 #endif  /* CONFIG_NFS_V4_1 */
6520
6521 /**
6522  * nfs4_proc_get_locations - discover locations for a migrated FSID
6523  * @inode: inode on FSID that is migrating
6524  * @locations: result of query
6525  * @page: buffer
6526  * @cred: credential to use for this operation
6527  *
6528  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6529  * operation failed, or a negative errno if a local error occurred.
6530  *
6531  * On success, "locations" is filled in, but if the server has
6532  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6533  * asserted.
6534  *
6535  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6536  * from this client that require migration recovery.
6537  */
6538 int nfs4_proc_get_locations(struct inode *inode,
6539                             struct nfs4_fs_locations *locations,
6540                             struct page *page, struct rpc_cred *cred)
6541 {
6542         struct nfs_server *server = NFS_SERVER(inode);
6543         struct nfs_client *clp = server->nfs_client;
6544         const struct nfs4_mig_recovery_ops *ops =
6545                                         clp->cl_mvops->mig_recovery_ops;
6546         struct nfs4_exception exception = { };
6547         int status;
6548
6549         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6550                 (unsigned long long)server->fsid.major,
6551                 (unsigned long long)server->fsid.minor,
6552                 clp->cl_hostname);
6553         nfs_display_fhandle(NFS_FH(inode), __func__);
6554
6555         do {
6556                 status = ops->get_locations(inode, locations, page, cred);
6557                 if (status != -NFS4ERR_DELAY)
6558                         break;
6559                 nfs4_handle_exception(server, status, &exception);
6560         } while (exception.retry);
6561         return status;
6562 }
6563
6564 /*
6565  * This operation also signals the server that this client is
6566  * performing "lease moved" recovery.  The server can stop
6567  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6568  * is appended to this compound to identify the client ID which is
6569  * performing recovery.
6570  */
6571 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6572 {
6573         struct nfs_server *server = NFS_SERVER(inode);
6574         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6575         struct rpc_clnt *clnt = server->client;
6576         struct nfs4_fsid_present_arg args = {
6577                 .fh             = NFS_FH(inode),
6578                 .clientid       = clp->cl_clientid,
6579                 .renew          = 1,            /* append RENEW */
6580         };
6581         struct nfs4_fsid_present_res res = {
6582                 .renew          = 1,
6583         };
6584         struct rpc_message msg = {
6585                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6586                 .rpc_argp       = &args,
6587                 .rpc_resp       = &res,
6588                 .rpc_cred       = cred,
6589         };
6590         unsigned long now = jiffies;
6591         int status;
6592
6593         res.fh = nfs_alloc_fhandle();
6594         if (res.fh == NULL)
6595                 return -ENOMEM;
6596
6597         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6598         nfs4_set_sequence_privileged(&args.seq_args);
6599         status = nfs4_call_sync_sequence(clnt, server, &msg,
6600                                                 &args.seq_args, &res.seq_res);
6601         nfs_free_fhandle(res.fh);
6602         if (status)
6603                 return status;
6604
6605         do_renew_lease(clp, now);
6606         return 0;
6607 }
6608
6609 #ifdef CONFIG_NFS_V4_1
6610
6611 /*
6612  * This operation also signals the server that this client is
6613  * performing "lease moved" recovery.  The server can stop asserting
6614  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6615  * this operation is identified in the SEQUENCE operation in this
6616  * compound.
6617  */
6618 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6619 {
6620         struct nfs_server *server = NFS_SERVER(inode);
6621         struct rpc_clnt *clnt = server->client;
6622         struct nfs4_fsid_present_arg args = {
6623                 .fh             = NFS_FH(inode),
6624         };
6625         struct nfs4_fsid_present_res res = {
6626         };
6627         struct rpc_message msg = {
6628                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6629                 .rpc_argp       = &args,
6630                 .rpc_resp       = &res,
6631                 .rpc_cred       = cred,
6632         };
6633         int status;
6634
6635         res.fh = nfs_alloc_fhandle();
6636         if (res.fh == NULL)
6637                 return -ENOMEM;
6638
6639         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6640         nfs4_set_sequence_privileged(&args.seq_args);
6641         status = nfs4_call_sync_sequence(clnt, server, &msg,
6642                                                 &args.seq_args, &res.seq_res);
6643         nfs_free_fhandle(res.fh);
6644         if (status == NFS4_OK &&
6645             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6646                 status = -NFS4ERR_LEASE_MOVED;
6647         return status;
6648 }
6649
6650 #endif  /* CONFIG_NFS_V4_1 */
6651
6652 /**
6653  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6654  * @inode: inode on FSID to check
6655  * @cred: credential to use for this operation
6656  *
6657  * Server indicates whether the FSID is present, moved, or not
6658  * recognized.  This operation is necessary to clear a LEASE_MOVED
6659  * condition for this client ID.
6660  *
6661  * Returns NFS4_OK if the FSID is present on this server,
6662  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6663  *  NFS4ERR code if some error occurred on the server, or a
6664  *  negative errno if a local failure occurred.
6665  */
6666 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6667 {
6668         struct nfs_server *server = NFS_SERVER(inode);
6669         struct nfs_client *clp = server->nfs_client;
6670         const struct nfs4_mig_recovery_ops *ops =
6671                                         clp->cl_mvops->mig_recovery_ops;
6672         struct nfs4_exception exception = { };
6673         int status;
6674
6675         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6676                 (unsigned long long)server->fsid.major,
6677                 (unsigned long long)server->fsid.minor,
6678                 clp->cl_hostname);
6679         nfs_display_fhandle(NFS_FH(inode), __func__);
6680
6681         do {
6682                 status = ops->fsid_present(inode, cred);
6683                 if (status != -NFS4ERR_DELAY)
6684                         break;
6685                 nfs4_handle_exception(server, status, &exception);
6686         } while (exception.retry);
6687         return status;
6688 }
6689
6690 /**
6691  * If 'use_integrity' is true and the state managment nfs_client
6692  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6693  * and the machine credential as per RFC3530bis and RFC5661 Security
6694  * Considerations sections. Otherwise, just use the user cred with the
6695  * filesystem's rpc_client.
6696  */
6697 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6698 {
6699         int status;
6700         struct nfs4_secinfo_arg args = {
6701                 .dir_fh = NFS_FH(dir),
6702                 .name   = name,
6703         };
6704         struct nfs4_secinfo_res res = {
6705                 .flavors     = flavors,
6706         };
6707         struct rpc_message msg = {
6708                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6709                 .rpc_argp = &args,
6710                 .rpc_resp = &res,
6711         };
6712         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6713         struct rpc_cred *cred = NULL;
6714
6715         if (use_integrity) {
6716                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6717                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6718                 msg.rpc_cred = cred;
6719         }
6720
6721         dprintk("NFS call  secinfo %s\n", name->name);
6722
6723         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6724                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6725
6726         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6727                                 &res.seq_res, 0);
6728         dprintk("NFS reply  secinfo: %d\n", status);
6729
6730         if (cred)
6731                 put_rpccred(cred);
6732
6733         return status;
6734 }
6735
6736 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6737                       struct nfs4_secinfo_flavors *flavors)
6738 {
6739         struct nfs4_exception exception = { };
6740         int err;
6741         do {
6742                 err = -NFS4ERR_WRONGSEC;
6743
6744                 /* try to use integrity protection with machine cred */
6745                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6746                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6747
6748                 /*
6749                  * if unable to use integrity protection, or SECINFO with
6750                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6751                  * disallowed by spec, but exists in deployed servers) use
6752                  * the current filesystem's rpc_client and the user cred.
6753                  */
6754                 if (err == -NFS4ERR_WRONGSEC)
6755                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6756
6757                 trace_nfs4_secinfo(dir, name, err);
6758                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6759                                 &exception);
6760         } while (exception.retry);
6761         return err;
6762 }
6763
6764 #ifdef CONFIG_NFS_V4_1
6765 /*
6766  * Check the exchange flags returned by the server for invalid flags, having
6767  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6768  * DS flags set.
6769  */
6770 static int nfs4_check_cl_exchange_flags(u32 flags)
6771 {
6772         if (flags & ~EXCHGID4_FLAG_MASK_R)
6773                 goto out_inval;
6774         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6775             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6776                 goto out_inval;
6777         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6778                 goto out_inval;
6779         return NFS_OK;
6780 out_inval:
6781         return -NFS4ERR_INVAL;
6782 }
6783
6784 static bool
6785 nfs41_same_server_scope(struct nfs41_server_scope *a,
6786                         struct nfs41_server_scope *b)
6787 {
6788         if (a->server_scope_sz == b->server_scope_sz &&
6789             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6790                 return true;
6791
6792         return false;
6793 }
6794
6795 /*
6796  * nfs4_proc_bind_conn_to_session()
6797  *
6798  * The 4.1 client currently uses the same TCP connection for the
6799  * fore and backchannel.
6800  */
6801 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6802 {
6803         int status;
6804         struct nfs41_bind_conn_to_session_args args = {
6805                 .client = clp,
6806                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6807         };
6808         struct nfs41_bind_conn_to_session_res res;
6809         struct rpc_message msg = {
6810                 .rpc_proc =
6811                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6812                 .rpc_argp = &args,
6813                 .rpc_resp = &res,
6814                 .rpc_cred = cred,
6815         };
6816
6817         dprintk("--> %s\n", __func__);
6818
6819         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6820         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6821                 args.dir = NFS4_CDFC4_FORE;
6822
6823         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6824         trace_nfs4_bind_conn_to_session(clp, status);
6825         if (status == 0) {
6826                 if (memcmp(res.sessionid.data,
6827                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6828                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6829                         status = -EIO;
6830                         goto out;
6831                 }
6832                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6833                         dprintk("NFS: %s: Unexpected direction from server\n",
6834                                 __func__);
6835                         status = -EIO;
6836                         goto out;
6837                 }
6838                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6839                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6840                                 __func__);
6841                         status = -EIO;
6842                         goto out;
6843                 }
6844         }
6845 out:
6846         dprintk("<-- %s status= %d\n", __func__, status);
6847         return status;
6848 }
6849
6850 /*
6851  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6852  * and operations we'd like to see to enable certain features in the allow map
6853  */
6854 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6855         .how = SP4_MACH_CRED,
6856         .enforce.u.words = {
6857                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6858                       1 << (OP_EXCHANGE_ID - 32) |
6859                       1 << (OP_CREATE_SESSION - 32) |
6860                       1 << (OP_DESTROY_SESSION - 32) |
6861                       1 << (OP_DESTROY_CLIENTID - 32)
6862         },
6863         .allow.u.words = {
6864                 [0] = 1 << (OP_CLOSE) |
6865                       1 << (OP_LOCKU) |
6866                       1 << (OP_COMMIT),
6867                 [1] = 1 << (OP_SECINFO - 32) |
6868                       1 << (OP_SECINFO_NO_NAME - 32) |
6869                       1 << (OP_TEST_STATEID - 32) |
6870                       1 << (OP_FREE_STATEID - 32) |
6871                       1 << (OP_WRITE - 32)
6872         }
6873 };
6874
6875 /*
6876  * Select the state protection mode for client `clp' given the server results
6877  * from exchange_id in `sp'.
6878  *
6879  * Returns 0 on success, negative errno otherwise.
6880  */
6881 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6882                                  struct nfs41_state_protection *sp)
6883 {
6884         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6885                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6886                       1 << (OP_EXCHANGE_ID - 32) |
6887                       1 << (OP_CREATE_SESSION - 32) |
6888                       1 << (OP_DESTROY_SESSION - 32) |
6889                       1 << (OP_DESTROY_CLIENTID - 32)
6890         };
6891         unsigned int i;
6892
6893         if (sp->how == SP4_MACH_CRED) {
6894                 /* Print state protect result */
6895                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6896                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6897                         if (test_bit(i, sp->enforce.u.longs))
6898                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6899                         if (test_bit(i, sp->allow.u.longs))
6900                                 dfprintk(MOUNT, "  allow op %d\n", i);
6901                 }
6902
6903                 /* make sure nothing is on enforce list that isn't supported */
6904                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6905                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6906                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6907                                 return -EINVAL;
6908                         }
6909                 }
6910
6911                 /*
6912                  * Minimal mode - state operations are allowed to use machine
6913                  * credential.  Note this already happens by default, so the
6914                  * client doesn't have to do anything more than the negotiation.
6915                  *
6916                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6917                  *       we're already using the machine cred for exchange_id
6918                  *       and will never use a different cred.
6919                  */
6920                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6921                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6922                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6923                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6924                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6925                         dfprintk(MOUNT, "  minimal mode enabled\n");
6926                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6927                 } else {
6928                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6929                         return -EINVAL;
6930                 }
6931
6932                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6933                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6934                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6935                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6936                 }
6937
6938                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6939                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6940                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6941                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6942                 }
6943
6944                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6945                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6946                         dfprintk(MOUNT, "  stateid mode enabled\n");
6947                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6948                 }
6949
6950                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6951                         dfprintk(MOUNT, "  write mode enabled\n");
6952                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6953                 }
6954
6955                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6956                         dfprintk(MOUNT, "  commit mode enabled\n");
6957                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6958                 }
6959         }
6960
6961         return 0;
6962 }
6963
6964 /*
6965  * _nfs4_proc_exchange_id()
6966  *
6967  * Wrapper for EXCHANGE_ID operation.
6968  */
6969 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6970         u32 sp4_how)
6971 {
6972         nfs4_verifier verifier;
6973         struct nfs41_exchange_id_args args = {
6974                 .verifier = &verifier,
6975                 .client = clp,
6976 #ifdef CONFIG_NFS_V4_1_MIGRATION
6977                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6978                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6979                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6980 #else
6981                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6982                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6983 #endif
6984         };
6985         struct nfs41_exchange_id_res res = {
6986                 0
6987         };
6988         int status;
6989         struct rpc_message msg = {
6990                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6991                 .rpc_argp = &args,
6992                 .rpc_resp = &res,
6993                 .rpc_cred = cred,
6994         };
6995
6996         nfs4_init_boot_verifier(clp, &verifier);
6997
6998         status = nfs4_init_uniform_client_string(clp);
6999         if (status)
7000                 goto out;
7001
7002         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7003                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7004                 clp->cl_owner_id);
7005
7006         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7007                                         GFP_NOFS);
7008         if (unlikely(res.server_owner == NULL)) {
7009                 status = -ENOMEM;
7010                 goto out;
7011         }
7012
7013         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7014                                         GFP_NOFS);
7015         if (unlikely(res.server_scope == NULL)) {
7016                 status = -ENOMEM;
7017                 goto out_server_owner;
7018         }
7019
7020         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7021         if (unlikely(res.impl_id == NULL)) {
7022                 status = -ENOMEM;
7023                 goto out_server_scope;
7024         }
7025
7026         switch (sp4_how) {
7027         case SP4_NONE:
7028                 args.state_protect.how = SP4_NONE;
7029                 break;
7030
7031         case SP4_MACH_CRED:
7032                 args.state_protect = nfs4_sp4_mach_cred_request;
7033                 break;
7034
7035         default:
7036                 /* unsupported! */
7037                 WARN_ON_ONCE(1);
7038                 status = -EINVAL;
7039                 goto out_impl_id;
7040         }
7041
7042         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7043         trace_nfs4_exchange_id(clp, status);
7044         if (status == 0)
7045                 status = nfs4_check_cl_exchange_flags(res.flags);
7046
7047         if (status == 0)
7048                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7049
7050         if (status == 0) {
7051                 clp->cl_clientid = res.clientid;
7052                 clp->cl_exchange_flags = res.flags;
7053                 /* Client ID is not confirmed */
7054                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7055                         clear_bit(NFS4_SESSION_ESTABLISHED,
7056                                         &clp->cl_session->session_state);
7057                         clp->cl_seqid = res.seqid;
7058                 }
7059
7060                 kfree(clp->cl_serverowner);
7061                 clp->cl_serverowner = res.server_owner;
7062                 res.server_owner = NULL;
7063
7064                 /* use the most recent implementation id */
7065                 kfree(clp->cl_implid);
7066                 clp->cl_implid = res.impl_id;
7067                 res.impl_id = NULL;
7068
7069                 if (clp->cl_serverscope != NULL &&
7070                     !nfs41_same_server_scope(clp->cl_serverscope,
7071                                              res.server_scope)) {
7072                         dprintk("%s: server_scope mismatch detected\n",
7073                                 __func__);
7074                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7075                         kfree(clp->cl_serverscope);
7076                         clp->cl_serverscope = NULL;
7077                 }
7078
7079                 if (clp->cl_serverscope == NULL) {
7080                         clp->cl_serverscope = res.server_scope;
7081                         res.server_scope = NULL;
7082                 }
7083         }
7084
7085 out_impl_id:
7086         kfree(res.impl_id);
7087 out_server_scope:
7088         kfree(res.server_scope);
7089 out_server_owner:
7090         kfree(res.server_owner);
7091 out:
7092         if (clp->cl_implid != NULL)
7093                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7094                         "domain: %s, name: %s, date: %llu,%u\n",
7095                         clp->cl_implid->domain, clp->cl_implid->name,
7096                         clp->cl_implid->date.seconds,
7097                         clp->cl_implid->date.nseconds);
7098         dprintk("NFS reply exchange_id: %d\n", status);
7099         return status;
7100 }
7101
7102 /*
7103  * nfs4_proc_exchange_id()
7104  *
7105  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7106  *
7107  * Since the clientid has expired, all compounds using sessions
7108  * associated with the stale clientid will be returning
7109  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7110  * be in some phase of session reset.
7111  *
7112  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7113  */
7114 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7115 {
7116         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7117         int status;
7118
7119         /* try SP4_MACH_CRED if krb5i/p */
7120         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7121             authflavor == RPC_AUTH_GSS_KRB5P) {
7122                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7123                 if (!status)
7124                         return 0;
7125         }
7126
7127         /* try SP4_NONE */
7128         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7129 }
7130
7131 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7132                 struct rpc_cred *cred)
7133 {
7134         struct rpc_message msg = {
7135                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7136                 .rpc_argp = clp,
7137                 .rpc_cred = cred,
7138         };
7139         int status;
7140
7141         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7142         trace_nfs4_destroy_clientid(clp, status);
7143         if (status)
7144                 dprintk("NFS: Got error %d from the server %s on "
7145                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7146         return status;
7147 }
7148
7149 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7150                 struct rpc_cred *cred)
7151 {
7152         unsigned int loop;
7153         int ret;
7154
7155         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7156                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7157                 switch (ret) {
7158                 case -NFS4ERR_DELAY:
7159                 case -NFS4ERR_CLIENTID_BUSY:
7160                         ssleep(1);
7161                         break;
7162                 default:
7163                         return ret;
7164                 }
7165         }
7166         return 0;
7167 }
7168
7169 int nfs4_destroy_clientid(struct nfs_client *clp)
7170 {
7171         struct rpc_cred *cred;
7172         int ret = 0;
7173
7174         if (clp->cl_mvops->minor_version < 1)
7175                 goto out;
7176         if (clp->cl_exchange_flags == 0)
7177                 goto out;
7178         if (clp->cl_preserve_clid)
7179                 goto out;
7180         cred = nfs4_get_clid_cred(clp);
7181         ret = nfs4_proc_destroy_clientid(clp, cred);
7182         if (cred)
7183                 put_rpccred(cred);
7184         switch (ret) {
7185         case 0:
7186         case -NFS4ERR_STALE_CLIENTID:
7187                 clp->cl_exchange_flags = 0;
7188         }
7189 out:
7190         return ret;
7191 }
7192
7193 struct nfs4_get_lease_time_data {
7194         struct nfs4_get_lease_time_args *args;
7195         struct nfs4_get_lease_time_res *res;
7196         struct nfs_client *clp;
7197 };
7198
7199 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7200                                         void *calldata)
7201 {
7202         struct nfs4_get_lease_time_data *data =
7203                         (struct nfs4_get_lease_time_data *)calldata;
7204
7205         dprintk("--> %s\n", __func__);
7206         /* just setup sequence, do not trigger session recovery
7207            since we're invoked within one */
7208         nfs41_setup_sequence(data->clp->cl_session,
7209                         &data->args->la_seq_args,
7210                         &data->res->lr_seq_res,
7211                         task);
7212         dprintk("<-- %s\n", __func__);
7213 }
7214
7215 /*
7216  * Called from nfs4_state_manager thread for session setup, so don't recover
7217  * from sequence operation or clientid errors.
7218  */
7219 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7220 {
7221         struct nfs4_get_lease_time_data *data =
7222                         (struct nfs4_get_lease_time_data *)calldata;
7223
7224         dprintk("--> %s\n", __func__);
7225         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7226                 return;
7227         switch (task->tk_status) {
7228         case -NFS4ERR_DELAY:
7229         case -NFS4ERR_GRACE:
7230                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7231                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7232                 task->tk_status = 0;
7233                 /* fall through */
7234         case -NFS4ERR_RETRY_UNCACHED_REP:
7235                 rpc_restart_call_prepare(task);
7236                 return;
7237         }
7238         dprintk("<-- %s\n", __func__);
7239 }
7240
7241 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7242         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7243         .rpc_call_done = nfs4_get_lease_time_done,
7244 };
7245
7246 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7247 {
7248         struct rpc_task *task;
7249         struct nfs4_get_lease_time_args args;
7250         struct nfs4_get_lease_time_res res = {
7251                 .lr_fsinfo = fsinfo,
7252         };
7253         struct nfs4_get_lease_time_data data = {
7254                 .args = &args,
7255                 .res = &res,
7256                 .clp = clp,
7257         };
7258         struct rpc_message msg = {
7259                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7260                 .rpc_argp = &args,
7261                 .rpc_resp = &res,
7262         };
7263         struct rpc_task_setup task_setup = {
7264                 .rpc_client = clp->cl_rpcclient,
7265                 .rpc_message = &msg,
7266                 .callback_ops = &nfs4_get_lease_time_ops,
7267                 .callback_data = &data,
7268                 .flags = RPC_TASK_TIMEOUT,
7269         };
7270         int status;
7271
7272         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7273         nfs4_set_sequence_privileged(&args.la_seq_args);
7274         dprintk("--> %s\n", __func__);
7275         task = rpc_run_task(&task_setup);
7276
7277         if (IS_ERR(task))
7278                 status = PTR_ERR(task);
7279         else {
7280                 status = task->tk_status;
7281                 rpc_put_task(task);
7282         }
7283         dprintk("<-- %s return %d\n", __func__, status);
7284
7285         return status;
7286 }
7287
7288 /*
7289  * Initialize the values to be used by the client in CREATE_SESSION
7290  * If nfs4_init_session set the fore channel request and response sizes,
7291  * use them.
7292  *
7293  * Set the back channel max_resp_sz_cached to zero to force the client to
7294  * always set csa_cachethis to FALSE because the current implementation
7295  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7296  */
7297 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7298 {
7299         unsigned int max_rqst_sz, max_resp_sz;
7300
7301         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7302         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7303
7304         /* Fore channel attributes */
7305         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7306         args->fc_attrs.max_resp_sz = max_resp_sz;
7307         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7308         args->fc_attrs.max_reqs = max_session_slots;
7309
7310         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7311                 "max_ops=%u max_reqs=%u\n",
7312                 __func__,
7313                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7314                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7315
7316         /* Back channel attributes */
7317         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7318         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7319         args->bc_attrs.max_resp_sz_cached = 0;
7320         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7321         args->bc_attrs.max_reqs = 1;
7322
7323         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7324                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7325                 __func__,
7326                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7327                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7328                 args->bc_attrs.max_reqs);
7329 }
7330
7331 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7332                 struct nfs41_create_session_res *res)
7333 {
7334         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7335         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7336
7337         if (rcvd->max_resp_sz > sent->max_resp_sz)
7338                 return -EINVAL;
7339         /*
7340          * Our requested max_ops is the minimum we need; we're not
7341          * prepared to break up compounds into smaller pieces than that.
7342          * So, no point even trying to continue if the server won't
7343          * cooperate:
7344          */
7345         if (rcvd->max_ops < sent->max_ops)
7346                 return -EINVAL;
7347         if (rcvd->max_reqs == 0)
7348                 return -EINVAL;
7349         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7350                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7351         return 0;
7352 }
7353
7354 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7355                 struct nfs41_create_session_res *res)
7356 {
7357         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7358         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7359
7360         if (!(res->flags & SESSION4_BACK_CHAN))
7361                 goto out;
7362         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7363                 return -EINVAL;
7364         if (rcvd->max_resp_sz < sent->max_resp_sz)
7365                 return -EINVAL;
7366         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7367                 return -EINVAL;
7368         /* These would render the backchannel useless: */
7369         if (rcvd->max_ops != sent->max_ops)
7370                 return -EINVAL;
7371         if (rcvd->max_reqs != sent->max_reqs)
7372                 return -EINVAL;
7373 out:
7374         return 0;
7375 }
7376
7377 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7378                                      struct nfs41_create_session_res *res)
7379 {
7380         int ret;
7381
7382         ret = nfs4_verify_fore_channel_attrs(args, res);
7383         if (ret)
7384                 return ret;
7385         return nfs4_verify_back_channel_attrs(args, res);
7386 }
7387
7388 static void nfs4_update_session(struct nfs4_session *session,
7389                 struct nfs41_create_session_res *res)
7390 {
7391         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7392         /* Mark client id and session as being confirmed */
7393         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7394         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7395         session->flags = res->flags;
7396         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7397         if (res->flags & SESSION4_BACK_CHAN)
7398                 memcpy(&session->bc_attrs, &res->bc_attrs,
7399                                 sizeof(session->bc_attrs));
7400 }
7401
7402 static int _nfs4_proc_create_session(struct nfs_client *clp,
7403                 struct rpc_cred *cred)
7404 {
7405         struct nfs4_session *session = clp->cl_session;
7406         struct nfs41_create_session_args args = {
7407                 .client = clp,
7408                 .clientid = clp->cl_clientid,
7409                 .seqid = clp->cl_seqid,
7410                 .cb_program = NFS4_CALLBACK,
7411         };
7412         struct nfs41_create_session_res res;
7413
7414         struct rpc_message msg = {
7415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7416                 .rpc_argp = &args,
7417                 .rpc_resp = &res,
7418                 .rpc_cred = cred,
7419         };
7420         int status;
7421
7422         nfs4_init_channel_attrs(&args);
7423         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7424
7425         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7426         trace_nfs4_create_session(clp, status);
7427
7428         if (!status) {
7429                 /* Verify the session's negotiated channel_attrs values */
7430                 status = nfs4_verify_channel_attrs(&args, &res);
7431                 /* Increment the clientid slot sequence id */
7432                 if (clp->cl_seqid == res.seqid)
7433                         clp->cl_seqid++;
7434                 if (status)
7435                         goto out;
7436                 nfs4_update_session(session, &res);
7437         }
7438 out:
7439         return status;
7440 }
7441
7442 /*
7443  * Issues a CREATE_SESSION operation to the server.
7444  * It is the responsibility of the caller to verify the session is
7445  * expired before calling this routine.
7446  */
7447 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7448 {
7449         int status;
7450         unsigned *ptr;
7451         struct nfs4_session *session = clp->cl_session;
7452
7453         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7454
7455         status = _nfs4_proc_create_session(clp, cred);
7456         if (status)
7457                 goto out;
7458
7459         /* Init or reset the session slot tables */
7460         status = nfs4_setup_session_slot_tables(session);
7461         dprintk("slot table setup returned %d\n", status);
7462         if (status)
7463                 goto out;
7464
7465         ptr = (unsigned *)&session->sess_id.data[0];
7466         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7467                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7468 out:
7469         dprintk("<-- %s\n", __func__);
7470         return status;
7471 }
7472
7473 /*
7474  * Issue the over-the-wire RPC DESTROY_SESSION.
7475  * The caller must serialize access to this routine.
7476  */
7477 int nfs4_proc_destroy_session(struct nfs4_session *session,
7478                 struct rpc_cred *cred)
7479 {
7480         struct rpc_message msg = {
7481                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7482                 .rpc_argp = session,
7483                 .rpc_cred = cred,
7484         };
7485         int status = 0;
7486
7487         dprintk("--> nfs4_proc_destroy_session\n");
7488
7489         /* session is still being setup */
7490         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7491                 return 0;
7492
7493         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7494         trace_nfs4_destroy_session(session->clp, status);
7495
7496         if (status)
7497                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7498                         "Session has been destroyed regardless...\n", status);
7499
7500         dprintk("<-- nfs4_proc_destroy_session\n");
7501         return status;
7502 }
7503
7504 /*
7505  * Renew the cl_session lease.
7506  */
7507 struct nfs4_sequence_data {
7508         struct nfs_client *clp;
7509         struct nfs4_sequence_args args;
7510         struct nfs4_sequence_res res;
7511 };
7512
7513 static void nfs41_sequence_release(void *data)
7514 {
7515         struct nfs4_sequence_data *calldata = data;
7516         struct nfs_client *clp = calldata->clp;
7517
7518         if (atomic_read(&clp->cl_count) > 1)
7519                 nfs4_schedule_state_renewal(clp);
7520         nfs_put_client(clp);
7521         kfree(calldata);
7522 }
7523
7524 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7525 {
7526         switch(task->tk_status) {
7527         case -NFS4ERR_DELAY:
7528                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7529                 return -EAGAIN;
7530         default:
7531                 nfs4_schedule_lease_recovery(clp);
7532         }
7533         return 0;
7534 }
7535
7536 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7537 {
7538         struct nfs4_sequence_data *calldata = data;
7539         struct nfs_client *clp = calldata->clp;
7540
7541         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7542                 return;
7543
7544         trace_nfs4_sequence(clp, task->tk_status);
7545         if (task->tk_status < 0) {
7546                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7547                 if (atomic_read(&clp->cl_count) == 1)
7548                         goto out;
7549
7550                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7551                         rpc_restart_call_prepare(task);
7552                         return;
7553                 }
7554         }
7555         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7556 out:
7557         dprintk("<-- %s\n", __func__);
7558 }
7559
7560 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7561 {
7562         struct nfs4_sequence_data *calldata = data;
7563         struct nfs_client *clp = calldata->clp;
7564         struct nfs4_sequence_args *args;
7565         struct nfs4_sequence_res *res;
7566
7567         args = task->tk_msg.rpc_argp;
7568         res = task->tk_msg.rpc_resp;
7569
7570         nfs41_setup_sequence(clp->cl_session, args, res, task);
7571 }
7572
7573 static const struct rpc_call_ops nfs41_sequence_ops = {
7574         .rpc_call_done = nfs41_sequence_call_done,
7575         .rpc_call_prepare = nfs41_sequence_prepare,
7576         .rpc_release = nfs41_sequence_release,
7577 };
7578
7579 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7580                 struct rpc_cred *cred,
7581                 bool is_privileged)
7582 {
7583         struct nfs4_sequence_data *calldata;
7584         struct rpc_message msg = {
7585                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7586                 .rpc_cred = cred,
7587         };
7588         struct rpc_task_setup task_setup_data = {
7589                 .rpc_client = clp->cl_rpcclient,
7590                 .rpc_message = &msg,
7591                 .callback_ops = &nfs41_sequence_ops,
7592                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7593         };
7594
7595         if (!atomic_inc_not_zero(&clp->cl_count))
7596                 return ERR_PTR(-EIO);
7597         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7598         if (calldata == NULL) {
7599                 nfs_put_client(clp);
7600                 return ERR_PTR(-ENOMEM);
7601         }
7602         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7603         if (is_privileged)
7604                 nfs4_set_sequence_privileged(&calldata->args);
7605         msg.rpc_argp = &calldata->args;
7606         msg.rpc_resp = &calldata->res;
7607         calldata->clp = clp;
7608         task_setup_data.callback_data = calldata;
7609
7610         return rpc_run_task(&task_setup_data);
7611 }
7612
7613 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7614 {
7615         struct rpc_task *task;
7616         int ret = 0;
7617
7618         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7619                 return -EAGAIN;
7620         task = _nfs41_proc_sequence(clp, cred, false);
7621         if (IS_ERR(task))
7622                 ret = PTR_ERR(task);
7623         else
7624                 rpc_put_task_async(task);
7625         dprintk("<-- %s status=%d\n", __func__, ret);
7626         return ret;
7627 }
7628
7629 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7630 {
7631         struct rpc_task *task;
7632         int ret;
7633
7634         task = _nfs41_proc_sequence(clp, cred, true);
7635         if (IS_ERR(task)) {
7636                 ret = PTR_ERR(task);
7637                 goto out;
7638         }
7639         ret = rpc_wait_for_completion_task(task);
7640         if (!ret)
7641                 ret = task->tk_status;
7642         rpc_put_task(task);
7643 out:
7644         dprintk("<-- %s status=%d\n", __func__, ret);
7645         return ret;
7646 }
7647
7648 struct nfs4_reclaim_complete_data {
7649         struct nfs_client *clp;
7650         struct nfs41_reclaim_complete_args arg;
7651         struct nfs41_reclaim_complete_res res;
7652 };
7653
7654 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7655 {
7656         struct nfs4_reclaim_complete_data *calldata = data;
7657
7658         nfs41_setup_sequence(calldata->clp->cl_session,
7659                         &calldata->arg.seq_args,
7660                         &calldata->res.seq_res,
7661                         task);
7662 }
7663
7664 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7665 {
7666         switch(task->tk_status) {
7667         case 0:
7668         case -NFS4ERR_COMPLETE_ALREADY:
7669         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7670                 break;
7671         case -NFS4ERR_DELAY:
7672                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7673                 /* fall through */
7674         case -NFS4ERR_RETRY_UNCACHED_REP:
7675                 return -EAGAIN;
7676         default:
7677                 nfs4_schedule_lease_recovery(clp);
7678         }
7679         return 0;
7680 }
7681
7682 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7683 {
7684         struct nfs4_reclaim_complete_data *calldata = data;
7685         struct nfs_client *clp = calldata->clp;
7686         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7687
7688         dprintk("--> %s\n", __func__);
7689         if (!nfs41_sequence_done(task, res))
7690                 return;
7691
7692         trace_nfs4_reclaim_complete(clp, task->tk_status);
7693         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7694                 rpc_restart_call_prepare(task);
7695                 return;
7696         }
7697         dprintk("<-- %s\n", __func__);
7698 }
7699
7700 static void nfs4_free_reclaim_complete_data(void *data)
7701 {
7702         struct nfs4_reclaim_complete_data *calldata = data;
7703
7704         kfree(calldata);
7705 }
7706
7707 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7708         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7709         .rpc_call_done = nfs4_reclaim_complete_done,
7710         .rpc_release = nfs4_free_reclaim_complete_data,
7711 };
7712
7713 /*
7714  * Issue a global reclaim complete.
7715  */
7716 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7717                 struct rpc_cred *cred)
7718 {
7719         struct nfs4_reclaim_complete_data *calldata;
7720         struct rpc_task *task;
7721         struct rpc_message msg = {
7722                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7723                 .rpc_cred = cred,
7724         };
7725         struct rpc_task_setup task_setup_data = {
7726                 .rpc_client = clp->cl_rpcclient,
7727                 .rpc_message = &msg,
7728                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7729                 .flags = RPC_TASK_ASYNC,
7730         };
7731         int status = -ENOMEM;
7732
7733         dprintk("--> %s\n", __func__);
7734         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7735         if (calldata == NULL)
7736                 goto out;
7737         calldata->clp = clp;
7738         calldata->arg.one_fs = 0;
7739
7740         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7741         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7742         msg.rpc_argp = &calldata->arg;
7743         msg.rpc_resp = &calldata->res;
7744         task_setup_data.callback_data = calldata;
7745         task = rpc_run_task(&task_setup_data);
7746         if (IS_ERR(task)) {
7747                 status = PTR_ERR(task);
7748                 goto out;
7749         }
7750         status = nfs4_wait_for_completion_rpc_task(task);
7751         if (status == 0)
7752                 status = task->tk_status;
7753         rpc_put_task(task);
7754         return 0;
7755 out:
7756         dprintk("<-- %s status=%d\n", __func__, status);
7757         return status;
7758 }
7759
7760 static void
7761 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7762 {
7763         struct nfs4_layoutget *lgp = calldata;
7764         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7765         struct nfs4_session *session = nfs4_get_session(server);
7766
7767         dprintk("--> %s\n", __func__);
7768         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7769          * right now covering the LAYOUTGET we are about to send.
7770          * However, that is not so catastrophic, and there seems
7771          * to be no way to prevent it completely.
7772          */
7773         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7774                                 &lgp->res.seq_res, task))
7775                 return;
7776         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7777                                           NFS_I(lgp->args.inode)->layout,
7778                                           &lgp->args.range,
7779                                           lgp->args.ctx->state)) {
7780                 rpc_exit(task, NFS4_OK);
7781         }
7782 }
7783
7784 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7785 {
7786         struct nfs4_layoutget *lgp = calldata;
7787         struct inode *inode = lgp->args.inode;
7788         struct nfs_server *server = NFS_SERVER(inode);
7789         struct pnfs_layout_hdr *lo;
7790         struct nfs4_state *state = NULL;
7791         unsigned long timeo, now, giveup;
7792
7793         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7794
7795         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7796                 goto out;
7797
7798         switch (task->tk_status) {
7799         case 0:
7800                 goto out;
7801         /*
7802          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7803          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7804          */
7805         case -NFS4ERR_BADLAYOUT:
7806                 goto out_overflow;
7807         /*
7808          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7809          * (or clients) writing to the same RAID stripe except when
7810          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7811          */
7812         case -NFS4ERR_LAYOUTTRYLATER:
7813                 if (lgp->args.minlength == 0)
7814                         goto out_overflow;
7815         /*
7816          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7817          * existing layout before getting a new one).
7818          */
7819         case -NFS4ERR_RECALLCONFLICT:
7820                 timeo = rpc_get_timeout(task->tk_client);
7821                 giveup = lgp->args.timestamp + timeo;
7822                 now = jiffies;
7823                 if (time_after(giveup, now)) {
7824                         unsigned long delay;
7825
7826                         /* Delay for:
7827                          * - Not less then NFS4_POLL_RETRY_MIN.
7828                          * - One last time a jiffie before we give up
7829                          * - exponential backoff (time_now minus start_attempt)
7830                          */
7831                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7832                                     min((giveup - now - 1),
7833                                         now - lgp->args.timestamp));
7834
7835                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7836                                 __func__, delay);
7837                         rpc_delay(task, delay);
7838                         /* Do not call nfs4_async_handle_error() */
7839                         goto out_restart;
7840                 }
7841                 break;
7842         case -NFS4ERR_EXPIRED:
7843         case -NFS4ERR_BAD_STATEID:
7844                 spin_lock(&inode->i_lock);
7845                 if (nfs4_stateid_match(&lgp->args.stateid,
7846                                         &lgp->args.ctx->state->stateid)) {
7847                         spin_unlock(&inode->i_lock);
7848                         /* If the open stateid was bad, then recover it. */
7849                         state = lgp->args.ctx->state;
7850                         break;
7851                 }
7852                 lo = NFS_I(inode)->layout;
7853                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7854                                         &lo->plh_stateid)) {
7855                         LIST_HEAD(head);
7856
7857                         /*
7858                          * Mark the bad layout state as invalid, then retry
7859                          * with the current stateid.
7860                          */
7861                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7862                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7863                         spin_unlock(&inode->i_lock);
7864                         pnfs_free_lseg_list(&head);
7865                 } else
7866                         spin_unlock(&inode->i_lock);
7867                 goto out_restart;
7868         }
7869         if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN)
7870                 goto out_restart;
7871 out:
7872         dprintk("<-- %s\n", __func__);
7873         return;
7874 out_restart:
7875         task->tk_status = 0;
7876         rpc_restart_call_prepare(task);
7877         return;
7878 out_overflow:
7879         task->tk_status = -EOVERFLOW;
7880         goto out;
7881 }
7882
7883 static size_t max_response_pages(struct nfs_server *server)
7884 {
7885         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7886         return nfs_page_array_len(0, max_resp_sz);
7887 }
7888
7889 static void nfs4_free_pages(struct page **pages, size_t size)
7890 {
7891         int i;
7892
7893         if (!pages)
7894                 return;
7895
7896         for (i = 0; i < size; i++) {
7897                 if (!pages[i])
7898                         break;
7899                 __free_page(pages[i]);
7900         }
7901         kfree(pages);
7902 }
7903
7904 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7905 {
7906         struct page **pages;
7907         int i;
7908
7909         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7910         if (!pages) {
7911                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7912                 return NULL;
7913         }
7914
7915         for (i = 0; i < size; i++) {
7916                 pages[i] = alloc_page(gfp_flags);
7917                 if (!pages[i]) {
7918                         dprintk("%s: failed to allocate page\n", __func__);
7919                         nfs4_free_pages(pages, size);
7920                         return NULL;
7921                 }
7922         }
7923
7924         return pages;
7925 }
7926
7927 static void nfs4_layoutget_release(void *calldata)
7928 {
7929         struct nfs4_layoutget *lgp = calldata;
7930         struct inode *inode = lgp->args.inode;
7931         struct nfs_server *server = NFS_SERVER(inode);
7932         size_t max_pages = max_response_pages(server);
7933
7934         dprintk("--> %s\n", __func__);
7935         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7936         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7937         put_nfs_open_context(lgp->args.ctx);
7938         kfree(calldata);
7939         dprintk("<-- %s\n", __func__);
7940 }
7941
7942 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7943         .rpc_call_prepare = nfs4_layoutget_prepare,
7944         .rpc_call_done = nfs4_layoutget_done,
7945         .rpc_release = nfs4_layoutget_release,
7946 };
7947
7948 struct pnfs_layout_segment *
7949 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7950 {
7951         struct inode *inode = lgp->args.inode;
7952         struct nfs_server *server = NFS_SERVER(inode);
7953         size_t max_pages = max_response_pages(server);
7954         struct rpc_task *task;
7955         struct rpc_message msg = {
7956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7957                 .rpc_argp = &lgp->args,
7958                 .rpc_resp = &lgp->res,
7959                 .rpc_cred = lgp->cred,
7960         };
7961         struct rpc_task_setup task_setup_data = {
7962                 .rpc_client = server->client,
7963                 .rpc_message = &msg,
7964                 .callback_ops = &nfs4_layoutget_call_ops,
7965                 .callback_data = lgp,
7966                 .flags = RPC_TASK_ASYNC,
7967         };
7968         struct pnfs_layout_segment *lseg = NULL;
7969         int status = 0;
7970
7971         dprintk("--> %s\n", __func__);
7972
7973         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7974         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7975
7976         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7977         if (!lgp->args.layout.pages) {
7978                 nfs4_layoutget_release(lgp);
7979                 return ERR_PTR(-ENOMEM);
7980         }
7981         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7982         lgp->args.timestamp = jiffies;
7983
7984         lgp->res.layoutp = &lgp->args.layout;
7985         lgp->res.seq_res.sr_slot = NULL;
7986         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7987
7988         task = rpc_run_task(&task_setup_data);
7989         if (IS_ERR(task))
7990                 return ERR_CAST(task);
7991         status = nfs4_wait_for_completion_rpc_task(task);
7992         if (status == 0)
7993                 status = task->tk_status;
7994         trace_nfs4_layoutget(lgp->args.ctx,
7995                         &lgp->args.range,
7996                         &lgp->res.range,
7997                         &lgp->res.stateid,
7998                         status);
7999         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8000         if (status == 0 && lgp->res.layoutp->len)
8001                 lseg = pnfs_layout_process(lgp);
8002         rpc_put_task(task);
8003         dprintk("<-- %s status=%d\n", __func__, status);
8004         if (status)
8005                 return ERR_PTR(status);
8006         return lseg;
8007 }
8008
8009 static void
8010 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8011 {
8012         struct nfs4_layoutreturn *lrp = calldata;
8013
8014         dprintk("--> %s\n", __func__);
8015         nfs41_setup_sequence(lrp->clp->cl_session,
8016                         &lrp->args.seq_args,
8017                         &lrp->res.seq_res,
8018                         task);
8019 }
8020
8021 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8022 {
8023         struct nfs4_layoutreturn *lrp = calldata;
8024         struct nfs_server *server;
8025
8026         dprintk("--> %s\n", __func__);
8027
8028         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8029                 return;
8030
8031         server = NFS_SERVER(lrp->args.inode);
8032         switch (task->tk_status) {
8033         default:
8034                 task->tk_status = 0;
8035         case 0:
8036                 break;
8037         case -NFS4ERR_DELAY:
8038                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8039                         break;
8040                 rpc_restart_call_prepare(task);
8041                 return;
8042         }
8043         dprintk("<-- %s\n", __func__);
8044 }
8045
8046 static void nfs4_layoutreturn_release(void *calldata)
8047 {
8048         struct nfs4_layoutreturn *lrp = calldata;
8049         struct pnfs_layout_hdr *lo = lrp->args.layout;
8050         LIST_HEAD(freeme);
8051
8052         dprintk("--> %s\n", __func__);
8053         spin_lock(&lo->plh_inode->i_lock);
8054         if (lrp->res.lrs_present)
8055                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8056         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8057         pnfs_clear_layoutreturn_waitbit(lo);
8058         lo->plh_block_lgets--;
8059         spin_unlock(&lo->plh_inode->i_lock);
8060         pnfs_free_lseg_list(&freeme);
8061         pnfs_put_layout_hdr(lrp->args.layout);
8062         nfs_iput_and_deactive(lrp->inode);
8063         kfree(calldata);
8064         dprintk("<-- %s\n", __func__);
8065 }
8066
8067 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8068         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8069         .rpc_call_done = nfs4_layoutreturn_done,
8070         .rpc_release = nfs4_layoutreturn_release,
8071 };
8072
8073 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8074 {
8075         struct rpc_task *task;
8076         struct rpc_message msg = {
8077                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8078                 .rpc_argp = &lrp->args,
8079                 .rpc_resp = &lrp->res,
8080                 .rpc_cred = lrp->cred,
8081         };
8082         struct rpc_task_setup task_setup_data = {
8083                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8084                 .rpc_message = &msg,
8085                 .callback_ops = &nfs4_layoutreturn_call_ops,
8086                 .callback_data = lrp,
8087         };
8088         int status = 0;
8089
8090         dprintk("--> %s\n", __func__);
8091         if (!sync) {
8092                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8093                 if (!lrp->inode) {
8094                         nfs4_layoutreturn_release(lrp);
8095                         return -EAGAIN;
8096                 }
8097                 task_setup_data.flags |= RPC_TASK_ASYNC;
8098         }
8099         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8100         task = rpc_run_task(&task_setup_data);
8101         if (IS_ERR(task))
8102                 return PTR_ERR(task);
8103         if (sync)
8104                 status = task->tk_status;
8105         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8106         dprintk("<-- %s status=%d\n", __func__, status);
8107         rpc_put_task(task);
8108         return status;
8109 }
8110
8111 static int
8112 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8113                 struct pnfs_device *pdev,
8114                 struct rpc_cred *cred)
8115 {
8116         struct nfs4_getdeviceinfo_args args = {
8117                 .pdev = pdev,
8118                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8119                         NOTIFY_DEVICEID4_DELETE,
8120         };
8121         struct nfs4_getdeviceinfo_res res = {
8122                 .pdev = pdev,
8123         };
8124         struct rpc_message msg = {
8125                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8126                 .rpc_argp = &args,
8127                 .rpc_resp = &res,
8128                 .rpc_cred = cred,
8129         };
8130         int status;
8131
8132         dprintk("--> %s\n", __func__);
8133         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8134         if (res.notification & ~args.notify_types)
8135                 dprintk("%s: unsupported notification\n", __func__);
8136         if (res.notification != args.notify_types)
8137                 pdev->nocache = 1;
8138
8139         dprintk("<-- %s status=%d\n", __func__, status);
8140
8141         return status;
8142 }
8143
8144 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8145                 struct pnfs_device *pdev,
8146                 struct rpc_cred *cred)
8147 {
8148         struct nfs4_exception exception = { };
8149         int err;
8150
8151         do {
8152                 err = nfs4_handle_exception(server,
8153                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8154                                         &exception);
8155         } while (exception.retry);
8156         return err;
8157 }
8158 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8159
8160 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8161 {
8162         struct nfs4_layoutcommit_data *data = calldata;
8163         struct nfs_server *server = NFS_SERVER(data->args.inode);
8164         struct nfs4_session *session = nfs4_get_session(server);
8165
8166         nfs41_setup_sequence(session,
8167                         &data->args.seq_args,
8168                         &data->res.seq_res,
8169                         task);
8170 }
8171
8172 static void
8173 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8174 {
8175         struct nfs4_layoutcommit_data *data = calldata;
8176         struct nfs_server *server = NFS_SERVER(data->args.inode);
8177
8178         if (!nfs41_sequence_done(task, &data->res.seq_res))
8179                 return;
8180
8181         switch (task->tk_status) { /* Just ignore these failures */
8182         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8183         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8184         case -NFS4ERR_BADLAYOUT:     /* no layout */
8185         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8186                 task->tk_status = 0;
8187         case 0:
8188                 break;
8189         default:
8190                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8191                         rpc_restart_call_prepare(task);
8192                         return;
8193                 }
8194         }
8195 }
8196
8197 static void nfs4_layoutcommit_release(void *calldata)
8198 {
8199         struct nfs4_layoutcommit_data *data = calldata;
8200
8201         pnfs_cleanup_layoutcommit(data);
8202         nfs_post_op_update_inode_force_wcc(data->args.inode,
8203                                            data->res.fattr);
8204         put_rpccred(data->cred);
8205         nfs_iput_and_deactive(data->inode);
8206         kfree(data);
8207 }
8208
8209 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8210         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8211         .rpc_call_done = nfs4_layoutcommit_done,
8212         .rpc_release = nfs4_layoutcommit_release,
8213 };
8214
8215 int
8216 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8217 {
8218         struct rpc_message msg = {
8219                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8220                 .rpc_argp = &data->args,
8221                 .rpc_resp = &data->res,
8222                 .rpc_cred = data->cred,
8223         };
8224         struct rpc_task_setup task_setup_data = {
8225                 .task = &data->task,
8226                 .rpc_client = NFS_CLIENT(data->args.inode),
8227                 .rpc_message = &msg,
8228                 .callback_ops = &nfs4_layoutcommit_ops,
8229                 .callback_data = data,
8230         };
8231         struct rpc_task *task;
8232         int status = 0;
8233
8234         dprintk("NFS: initiating layoutcommit call. sync %d "
8235                 "lbw: %llu inode %lu\n", sync,
8236                 data->args.lastbytewritten,
8237                 data->args.inode->i_ino);
8238
8239         if (!sync) {
8240                 data->inode = nfs_igrab_and_active(data->args.inode);
8241                 if (data->inode == NULL) {
8242                         nfs4_layoutcommit_release(data);
8243                         return -EAGAIN;
8244                 }
8245                 task_setup_data.flags = RPC_TASK_ASYNC;
8246         }
8247         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8248         task = rpc_run_task(&task_setup_data);
8249         if (IS_ERR(task))
8250                 return PTR_ERR(task);
8251         if (sync)
8252                 status = task->tk_status;
8253         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8254         dprintk("%s: status %d\n", __func__, status);
8255         rpc_put_task(task);
8256         return status;
8257 }
8258
8259 /**
8260  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8261  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8262  */
8263 static int
8264 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8265                     struct nfs_fsinfo *info,
8266                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8267 {
8268         struct nfs41_secinfo_no_name_args args = {
8269                 .style = SECINFO_STYLE_CURRENT_FH,
8270         };
8271         struct nfs4_secinfo_res res = {
8272                 .flavors = flavors,
8273         };
8274         struct rpc_message msg = {
8275                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8276                 .rpc_argp = &args,
8277                 .rpc_resp = &res,
8278         };
8279         struct rpc_clnt *clnt = server->client;
8280         struct rpc_cred *cred = NULL;
8281         int status;
8282
8283         if (use_integrity) {
8284                 clnt = server->nfs_client->cl_rpcclient;
8285                 cred = nfs4_get_clid_cred(server->nfs_client);
8286                 msg.rpc_cred = cred;
8287         }
8288
8289         dprintk("--> %s\n", __func__);
8290         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8291                                 &res.seq_res, 0);
8292         dprintk("<-- %s status=%d\n", __func__, status);
8293
8294         if (cred)
8295                 put_rpccred(cred);
8296
8297         return status;
8298 }
8299
8300 static int
8301 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8302                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8303 {
8304         struct nfs4_exception exception = { };
8305         int err;
8306         do {
8307                 /* first try using integrity protection */
8308                 err = -NFS4ERR_WRONGSEC;
8309
8310                 /* try to use integrity protection with machine cred */
8311                 if (_nfs4_is_integrity_protected(server->nfs_client))
8312                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8313                                                           flavors, true);
8314
8315                 /*
8316                  * if unable to use integrity protection, or SECINFO with
8317                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8318                  * disallowed by spec, but exists in deployed servers) use
8319                  * the current filesystem's rpc_client and the user cred.
8320                  */
8321                 if (err == -NFS4ERR_WRONGSEC)
8322                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8323                                                           flavors, false);
8324
8325                 switch (err) {
8326                 case 0:
8327                 case -NFS4ERR_WRONGSEC:
8328                 case -ENOTSUPP:
8329                         goto out;
8330                 default:
8331                         err = nfs4_handle_exception(server, err, &exception);
8332                 }
8333         } while (exception.retry);
8334 out:
8335         return err;
8336 }
8337
8338 static int
8339 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8340                     struct nfs_fsinfo *info)
8341 {
8342         int err;
8343         struct page *page;
8344         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8345         struct nfs4_secinfo_flavors *flavors;
8346         struct nfs4_secinfo4 *secinfo;
8347         int i;
8348
8349         page = alloc_page(GFP_KERNEL);
8350         if (!page) {
8351                 err = -ENOMEM;
8352                 goto out;
8353         }
8354
8355         flavors = page_address(page);
8356         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8357
8358         /*
8359          * Fall back on "guess and check" method if
8360          * the server doesn't support SECINFO_NO_NAME
8361          */
8362         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8363                 err = nfs4_find_root_sec(server, fhandle, info);
8364                 goto out_freepage;
8365         }
8366         if (err)
8367                 goto out_freepage;
8368
8369         for (i = 0; i < flavors->num_flavors; i++) {
8370                 secinfo = &flavors->flavors[i];
8371
8372                 switch (secinfo->flavor) {
8373                 case RPC_AUTH_NULL:
8374                 case RPC_AUTH_UNIX:
8375                 case RPC_AUTH_GSS:
8376                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8377                                         &secinfo->flavor_info);
8378                         break;
8379                 default:
8380                         flavor = RPC_AUTH_MAXFLAVOR;
8381                         break;
8382                 }
8383
8384                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8385                         flavor = RPC_AUTH_MAXFLAVOR;
8386
8387                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8388                         err = nfs4_lookup_root_sec(server, fhandle,
8389                                                    info, flavor);
8390                         if (!err)
8391                                 break;
8392                 }
8393         }
8394
8395         if (flavor == RPC_AUTH_MAXFLAVOR)
8396                 err = -EPERM;
8397
8398 out_freepage:
8399         put_page(page);
8400         if (err == -EACCES)
8401                 return -EPERM;
8402 out:
8403         return err;
8404 }
8405
8406 static int _nfs41_test_stateid(struct nfs_server *server,
8407                 nfs4_stateid *stateid,
8408                 struct rpc_cred *cred)
8409 {
8410         int status;
8411         struct nfs41_test_stateid_args args = {
8412                 .stateid = stateid,
8413         };
8414         struct nfs41_test_stateid_res res;
8415         struct rpc_message msg = {
8416                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8417                 .rpc_argp = &args,
8418                 .rpc_resp = &res,
8419                 .rpc_cred = cred,
8420         };
8421         struct rpc_clnt *rpc_client = server->client;
8422
8423         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8424                 &rpc_client, &msg);
8425
8426         dprintk("NFS call  test_stateid %p\n", stateid);
8427         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8428         nfs4_set_sequence_privileged(&args.seq_args);
8429         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8430                         &args.seq_args, &res.seq_res);
8431         if (status != NFS_OK) {
8432                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8433                 return status;
8434         }
8435         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8436         return -res.status;
8437 }
8438
8439 /**
8440  * nfs41_test_stateid - perform a TEST_STATEID operation
8441  *
8442  * @server: server / transport on which to perform the operation
8443  * @stateid: state ID to test
8444  * @cred: credential
8445  *
8446  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8447  * Otherwise a negative NFS4ERR value is returned if the operation
8448  * failed or the state ID is not currently valid.
8449  */
8450 static int nfs41_test_stateid(struct nfs_server *server,
8451                 nfs4_stateid *stateid,
8452                 struct rpc_cred *cred)
8453 {
8454         struct nfs4_exception exception = { };
8455         int err;
8456         do {
8457                 err = _nfs41_test_stateid(server, stateid, cred);
8458                 if (err != -NFS4ERR_DELAY)
8459                         break;
8460                 nfs4_handle_exception(server, err, &exception);
8461         } while (exception.retry);
8462         return err;
8463 }
8464
8465 struct nfs_free_stateid_data {
8466         struct nfs_server *server;
8467         struct nfs41_free_stateid_args args;
8468         struct nfs41_free_stateid_res res;
8469 };
8470
8471 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8472 {
8473         struct nfs_free_stateid_data *data = calldata;
8474         nfs41_setup_sequence(nfs4_get_session(data->server),
8475                         &data->args.seq_args,
8476                         &data->res.seq_res,
8477                         task);
8478 }
8479
8480 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8481 {
8482         struct nfs_free_stateid_data *data = calldata;
8483
8484         nfs41_sequence_done(task, &data->res.seq_res);
8485
8486         switch (task->tk_status) {
8487         case -NFS4ERR_DELAY:
8488                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8489                         rpc_restart_call_prepare(task);
8490         }
8491 }
8492
8493 static void nfs41_free_stateid_release(void *calldata)
8494 {
8495         kfree(calldata);
8496 }
8497
8498 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8499         .rpc_call_prepare = nfs41_free_stateid_prepare,
8500         .rpc_call_done = nfs41_free_stateid_done,
8501         .rpc_release = nfs41_free_stateid_release,
8502 };
8503
8504 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8505                 nfs4_stateid *stateid,
8506                 struct rpc_cred *cred,
8507                 bool privileged)
8508 {
8509         struct rpc_message msg = {
8510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8511                 .rpc_cred = cred,
8512         };
8513         struct rpc_task_setup task_setup = {
8514                 .rpc_client = server->client,
8515                 .rpc_message = &msg,
8516                 .callback_ops = &nfs41_free_stateid_ops,
8517                 .flags = RPC_TASK_ASYNC,
8518         };
8519         struct nfs_free_stateid_data *data;
8520
8521         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8522                 &task_setup.rpc_client, &msg);
8523
8524         dprintk("NFS call  free_stateid %p\n", stateid);
8525         data = kmalloc(sizeof(*data), GFP_NOFS);
8526         if (!data)
8527                 return ERR_PTR(-ENOMEM);
8528         data->server = server;
8529         nfs4_stateid_copy(&data->args.stateid, stateid);
8530
8531         task_setup.callback_data = data;
8532
8533         msg.rpc_argp = &data->args;
8534         msg.rpc_resp = &data->res;
8535         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8536         if (privileged)
8537                 nfs4_set_sequence_privileged(&data->args.seq_args);
8538
8539         return rpc_run_task(&task_setup);
8540 }
8541
8542 /**
8543  * nfs41_free_stateid - perform a FREE_STATEID operation
8544  *
8545  * @server: server / transport on which to perform the operation
8546  * @stateid: state ID to release
8547  * @cred: credential
8548  *
8549  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8550  * negative NFS4ERR value is returned.
8551  */
8552 static int nfs41_free_stateid(struct nfs_server *server,
8553                 nfs4_stateid *stateid,
8554                 struct rpc_cred *cred)
8555 {
8556         struct rpc_task *task;
8557         int ret;
8558
8559         task = _nfs41_free_stateid(server, stateid, cred, true);
8560         if (IS_ERR(task))
8561                 return PTR_ERR(task);
8562         ret = rpc_wait_for_completion_task(task);
8563         if (!ret)
8564                 ret = task->tk_status;
8565         rpc_put_task(task);
8566         return ret;
8567 }
8568
8569 static void
8570 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8571 {
8572         struct rpc_task *task;
8573         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8574
8575         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8576         nfs4_free_lock_state(server, lsp);
8577         if (IS_ERR(task))
8578                 return;
8579         rpc_put_task(task);
8580 }
8581
8582 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8583                 const nfs4_stateid *s2)
8584 {
8585         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8586                 return false;
8587
8588         if (s1->seqid == s2->seqid)
8589                 return true;
8590         if (s1->seqid == 0 || s2->seqid == 0)
8591                 return true;
8592
8593         return false;
8594 }
8595
8596 #endif /* CONFIG_NFS_V4_1 */
8597
8598 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8599                 const nfs4_stateid *s2)
8600 {
8601         return nfs4_stateid_match(s1, s2);
8602 }
8603
8604
8605 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8606         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8607         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8608         .recover_open   = nfs4_open_reclaim,
8609         .recover_lock   = nfs4_lock_reclaim,
8610         .establish_clid = nfs4_init_clientid,
8611         .detect_trunking = nfs40_discover_server_trunking,
8612 };
8613
8614 #if defined(CONFIG_NFS_V4_1)
8615 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8616         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8617         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8618         .recover_open   = nfs4_open_reclaim,
8619         .recover_lock   = nfs4_lock_reclaim,
8620         .establish_clid = nfs41_init_clientid,
8621         .reclaim_complete = nfs41_proc_reclaim_complete,
8622         .detect_trunking = nfs41_discover_server_trunking,
8623 };
8624 #endif /* CONFIG_NFS_V4_1 */
8625
8626 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8627         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8628         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8629         .recover_open   = nfs40_open_expired,
8630         .recover_lock   = nfs4_lock_expired,
8631         .establish_clid = nfs4_init_clientid,
8632 };
8633
8634 #if defined(CONFIG_NFS_V4_1)
8635 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8636         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8637         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8638         .recover_open   = nfs41_open_expired,
8639         .recover_lock   = nfs41_lock_expired,
8640         .establish_clid = nfs41_init_clientid,
8641 };
8642 #endif /* CONFIG_NFS_V4_1 */
8643
8644 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8645         .sched_state_renewal = nfs4_proc_async_renew,
8646         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8647         .renew_lease = nfs4_proc_renew,
8648 };
8649
8650 #if defined(CONFIG_NFS_V4_1)
8651 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8652         .sched_state_renewal = nfs41_proc_async_sequence,
8653         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8654         .renew_lease = nfs4_proc_sequence,
8655 };
8656 #endif
8657
8658 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8659         .get_locations = _nfs40_proc_get_locations,
8660         .fsid_present = _nfs40_proc_fsid_present,
8661 };
8662
8663 #if defined(CONFIG_NFS_V4_1)
8664 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8665         .get_locations = _nfs41_proc_get_locations,
8666         .fsid_present = _nfs41_proc_fsid_present,
8667 };
8668 #endif  /* CONFIG_NFS_V4_1 */
8669
8670 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8671         .minor_version = 0,
8672         .init_caps = NFS_CAP_READDIRPLUS
8673                 | NFS_CAP_ATOMIC_OPEN
8674                 | NFS_CAP_POSIX_LOCK,
8675         .init_client = nfs40_init_client,
8676         .shutdown_client = nfs40_shutdown_client,
8677         .match_stateid = nfs4_match_stateid,
8678         .find_root_sec = nfs4_find_root_sec,
8679         .free_lock_state = nfs4_release_lockowner,
8680         .alloc_seqid = nfs_alloc_seqid,
8681         .call_sync_ops = &nfs40_call_sync_ops,
8682         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8683         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8684         .state_renewal_ops = &nfs40_state_renewal_ops,
8685         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8686 };
8687
8688 #if defined(CONFIG_NFS_V4_1)
8689 static struct nfs_seqid *
8690 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8691 {
8692         return NULL;
8693 }
8694
8695 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8696         .minor_version = 1,
8697         .init_caps = NFS_CAP_READDIRPLUS
8698                 | NFS_CAP_ATOMIC_OPEN
8699                 | NFS_CAP_POSIX_LOCK
8700                 | NFS_CAP_STATEID_NFSV41
8701                 | NFS_CAP_ATOMIC_OPEN_V1,
8702         .init_client = nfs41_init_client,
8703         .shutdown_client = nfs41_shutdown_client,
8704         .match_stateid = nfs41_match_stateid,
8705         .find_root_sec = nfs41_find_root_sec,
8706         .free_lock_state = nfs41_free_lock_state,
8707         .alloc_seqid = nfs_alloc_no_seqid,
8708         .call_sync_ops = &nfs41_call_sync_ops,
8709         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8710         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8711         .state_renewal_ops = &nfs41_state_renewal_ops,
8712         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8713 };
8714 #endif
8715
8716 #if defined(CONFIG_NFS_V4_2)
8717 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8718         .minor_version = 2,
8719         .init_caps = NFS_CAP_READDIRPLUS
8720                 | NFS_CAP_ATOMIC_OPEN
8721                 | NFS_CAP_POSIX_LOCK
8722                 | NFS_CAP_STATEID_NFSV41
8723                 | NFS_CAP_ATOMIC_OPEN_V1
8724                 | NFS_CAP_ALLOCATE
8725                 | NFS_CAP_DEALLOCATE
8726                 | NFS_CAP_SEEK
8727                 | NFS_CAP_LAYOUTSTATS
8728                 | NFS_CAP_CLONE,
8729         .init_client = nfs41_init_client,
8730         .shutdown_client = nfs41_shutdown_client,
8731         .match_stateid = nfs41_match_stateid,
8732         .find_root_sec = nfs41_find_root_sec,
8733         .free_lock_state = nfs41_free_lock_state,
8734         .call_sync_ops = &nfs41_call_sync_ops,
8735         .alloc_seqid = nfs_alloc_no_seqid,
8736         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8737         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8738         .state_renewal_ops = &nfs41_state_renewal_ops,
8739         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8740 };
8741 #endif
8742
8743 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8744         [0] = &nfs_v4_0_minor_ops,
8745 #if defined(CONFIG_NFS_V4_1)
8746         [1] = &nfs_v4_1_minor_ops,
8747 #endif
8748 #if defined(CONFIG_NFS_V4_2)
8749         [2] = &nfs_v4_2_minor_ops,
8750 #endif
8751 };
8752
8753 static const struct inode_operations nfs4_dir_inode_operations = {
8754         .create         = nfs_create,
8755         .lookup         = nfs_lookup,
8756         .atomic_open    = nfs_atomic_open,
8757         .link           = nfs_link,
8758         .unlink         = nfs_unlink,
8759         .symlink        = nfs_symlink,
8760         .mkdir          = nfs_mkdir,
8761         .rmdir          = nfs_rmdir,
8762         .mknod          = nfs_mknod,
8763         .rename         = nfs_rename,
8764         .permission     = nfs_permission,
8765         .getattr        = nfs_getattr,
8766         .setattr        = nfs_setattr,
8767         .getxattr       = generic_getxattr,
8768         .setxattr       = generic_setxattr,
8769         .listxattr      = generic_listxattr,
8770         .removexattr    = generic_removexattr,
8771 };
8772
8773 static const struct inode_operations nfs4_file_inode_operations = {
8774         .permission     = nfs_permission,
8775         .getattr        = nfs_getattr,
8776         .setattr        = nfs_setattr,
8777         .getxattr       = generic_getxattr,
8778         .setxattr       = generic_setxattr,
8779         .listxattr      = generic_listxattr,
8780         .removexattr    = generic_removexattr,
8781 };
8782
8783 const struct nfs_rpc_ops nfs_v4_clientops = {
8784         .version        = 4,                    /* protocol version */
8785         .dentry_ops     = &nfs4_dentry_operations,
8786         .dir_inode_ops  = &nfs4_dir_inode_operations,
8787         .file_inode_ops = &nfs4_file_inode_operations,
8788         .file_ops       = &nfs4_file_operations,
8789         .getroot        = nfs4_proc_get_root,
8790         .submount       = nfs4_submount,
8791         .try_mount      = nfs4_try_mount,
8792         .getattr        = nfs4_proc_getattr,
8793         .setattr        = nfs4_proc_setattr,
8794         .lookup         = nfs4_proc_lookup,
8795         .access         = nfs4_proc_access,
8796         .readlink       = nfs4_proc_readlink,
8797         .create         = nfs4_proc_create,
8798         .remove         = nfs4_proc_remove,
8799         .unlink_setup   = nfs4_proc_unlink_setup,
8800         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8801         .unlink_done    = nfs4_proc_unlink_done,
8802         .rename_setup   = nfs4_proc_rename_setup,
8803         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8804         .rename_done    = nfs4_proc_rename_done,
8805         .link           = nfs4_proc_link,
8806         .symlink        = nfs4_proc_symlink,
8807         .mkdir          = nfs4_proc_mkdir,
8808         .rmdir          = nfs4_proc_remove,
8809         .readdir        = nfs4_proc_readdir,
8810         .mknod          = nfs4_proc_mknod,
8811         .statfs         = nfs4_proc_statfs,
8812         .fsinfo         = nfs4_proc_fsinfo,
8813         .pathconf       = nfs4_proc_pathconf,
8814         .set_capabilities = nfs4_server_capabilities,
8815         .decode_dirent  = nfs4_decode_dirent,
8816         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8817         .read_setup     = nfs4_proc_read_setup,
8818         .read_done      = nfs4_read_done,
8819         .write_setup    = nfs4_proc_write_setup,
8820         .write_done     = nfs4_write_done,
8821         .commit_setup   = nfs4_proc_commit_setup,
8822         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8823         .commit_done    = nfs4_commit_done,
8824         .lock           = nfs4_proc_lock,
8825         .clear_acl_cache = nfs4_zap_acl_attr,
8826         .close_context  = nfs4_close_context,
8827         .open_context   = nfs4_atomic_open,
8828         .have_delegation = nfs4_have_delegation,
8829         .return_delegation = nfs4_inode_return_delegation,
8830         .alloc_client   = nfs4_alloc_client,
8831         .init_client    = nfs4_init_client,
8832         .free_client    = nfs4_free_client,
8833         .create_server  = nfs4_create_server,
8834         .clone_server   = nfs_clone_server,
8835 };
8836
8837 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8838         .prefix = XATTR_NAME_NFSV4_ACL,
8839         .list   = nfs4_xattr_list_nfs4_acl,
8840         .get    = nfs4_xattr_get_nfs4_acl,
8841         .set    = nfs4_xattr_set_nfs4_acl,
8842 };
8843
8844 const struct xattr_handler *nfs4_xattr_handlers[] = {
8845         &nfs4_xattr_nfs4_acl_handler,
8846 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8847         &nfs4_xattr_nfs4_label_handler,
8848 #endif
8849         NULL
8850 };
8851
8852 /*
8853  * Local variables:
8854  *  c-basic-offset: 8
8855  * End:
8856  */