fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 /* file attributes which can be mapped to nfs attributes */
  78 #define NFS4_VALID_ATTRS (ATTR_MODE \
  79         | ATTR_UID \
  80         | ATTR_GID \
  81         | ATTR_SIZE \
  82         | ATTR_ATIME \
  83         | ATTR_MTIME \
  84         | ATTR_CTIME \
  85         | ATTR_ATIME_SET \
  86         | ATTR_MTIME_SET)
  87
  88 struct nfs4_opendata;
  89 static int _nfs4_proc_open(struct nfs4_opendata *data);
  90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 101                 struct rpc_cred *);
 102 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 103                 struct rpc_cred *, bool);
 104 #endif
 105
 106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 107 static inline struct nfs4_label *
 108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 109         struct iattr *sattr, struct nfs4_label *label)
 110 {
 111         int err;
 112
 113         if (label == NULL)
 114                 return NULL;
 115
 116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 117                 return NULL;
 118
 119         err = security_dentry_init_security(dentry, sattr->ia_mode,
 120                                 &dentry->d_name, (void **)&label->label, &label->len);
 121         if (err == 0)
 122                 return label;
 123
 124         return NULL;
 125 }
 126 static inline void
 127 nfs4_label_release_security(struct nfs4_label *label)
 128 {
 129         if (label)
 130                 security_release_secctx(label->label, label->len);
 131 }
 132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 133 {
 134         if (label)
 135                 return server->attr_bitmask;
 136
 137         return server->attr_bitmask_nl;
 138 }
 139 #else
 140 static inline struct nfs4_label *
 141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 142         struct iattr *sattr, struct nfs4_label *l)
 143 { return NULL; }
 144 static inline void
 145 nfs4_label_release_security(struct nfs4_label *label)
 146 { return; }
 147 static inline u32 *
 148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 149 { return server->attr_bitmask; }
 150 #endif
 151
 152 /* Prevent leaks of NFSv4 errors into userland */
 153 static int nfs4_map_errors(int err)
 154 {
 155         if (err >= -1000)
 156                 return err;
 157         switch (err) {
 158         case -NFS4ERR_RESOURCE:
 159         case -NFS4ERR_LAYOUTTRYLATER:
 160         case -NFS4ERR_RECALLCONFLICT:
 161                 return -EREMOTEIO;
 162         case -NFS4ERR_WRONGSEC:
 163         case -NFS4ERR_WRONG_CRED:
 164                 return -EPERM;
 165         case -NFS4ERR_BADOWNER:
 166         case -NFS4ERR_BADNAME:
 167                 return -EINVAL;
 168         case -NFS4ERR_SHARE_DENIED:
 169                 return -EACCES;
 170         case -NFS4ERR_MINOR_VERS_MISMATCH:
 171                 return -EPROTONOSUPPORT;
 172         case -NFS4ERR_FILE_OPEN:
 173                 return -EBUSY;
 174         default:
 175                 dprintk("%s could not handle NFSv4 error %d\n",
 176                                 __func__, -err);
 177                 break;
 178         }
 179         return -EIO;
 180 }
 181
 182 /*
 183  * This is our standard bitmap for GETATTR requests.
 184  */
 185 const u32 nfs4_fattr_bitmap[3] = {
 186         FATTR4_WORD0_TYPE
 187         | FATTR4_WORD0_CHANGE
 188         | FATTR4_WORD0_SIZE
 189         | FATTR4_WORD0_FSID
 190         | FATTR4_WORD0_FILEID,
 191         FATTR4_WORD1_MODE
 192         | FATTR4_WORD1_NUMLINKS
 193         | FATTR4_WORD1_OWNER
 194         | FATTR4_WORD1_OWNER_GROUP
 195         | FATTR4_WORD1_RAWDEV
 196         | FATTR4_WORD1_SPACE_USED
 197         | FATTR4_WORD1_TIME_ACCESS
 198         | FATTR4_WORD1_TIME_METADATA
 199         | FATTR4_WORD1_TIME_MODIFY
 200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 202         FATTR4_WORD2_SECURITY_LABEL
 203 #endif
 204 };
 205
 206 static const u32 nfs4_pnfs_open_bitmap[3] = {
 207         FATTR4_WORD0_TYPE
 208         | FATTR4_WORD0_CHANGE
 209         | FATTR4_WORD0_SIZE
 210         | FATTR4_WORD0_FSID
 211         | FATTR4_WORD0_FILEID,
 212         FATTR4_WORD1_MODE
 213         | FATTR4_WORD1_NUMLINKS
 214         | FATTR4_WORD1_OWNER
 215         | FATTR4_WORD1_OWNER_GROUP
 216         | FATTR4_WORD1_RAWDEV
 217         | FATTR4_WORD1_SPACE_USED
 218         | FATTR4_WORD1_TIME_ACCESS
 219         | FATTR4_WORD1_TIME_METADATA
 220         | FATTR4_WORD1_TIME_MODIFY,
 221         FATTR4_WORD2_MDSTHRESHOLD
 222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 223         | FATTR4_WORD2_SECURITY_LABEL
 224 #endif
 225 };
 226
 227 static const u32 nfs4_open_noattr_bitmap[3] = {
 228         FATTR4_WORD0_TYPE
 229         | FATTR4_WORD0_FILEID,
 230 };
 231
 232 const u32 nfs4_statfs_bitmap[3] = {
 233         FATTR4_WORD0_FILES_AVAIL
 234         | FATTR4_WORD0_FILES_FREE
 235         | FATTR4_WORD0_FILES_TOTAL,
 236         FATTR4_WORD1_SPACE_AVAIL
 237         | FATTR4_WORD1_SPACE_FREE
 238         | FATTR4_WORD1_SPACE_TOTAL
 239 };
 240
 241 const u32 nfs4_pathconf_bitmap[3] = {
 242         FATTR4_WORD0_MAXLINK
 243         | FATTR4_WORD0_MAXNAME,
 244         0
 245 };
 246
 247 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 248                         | FATTR4_WORD0_MAXREAD
 249                         | FATTR4_WORD0_MAXWRITE
 250                         | FATTR4_WORD0_LEASE_TIME,
 251                         FATTR4_WORD1_TIME_DELTA
 252                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 253                         FATTR4_WORD2_LAYOUT_BLKSIZE
 254                         | FATTR4_WORD2_CLONE_BLKSIZE
 255 };
 256
 257 const u32 nfs4_fs_locations_bitmap[3] = {
 258         FATTR4_WORD0_TYPE
 259         | FATTR4_WORD0_CHANGE
 260         | FATTR4_WORD0_SIZE
 261         | FATTR4_WORD0_FSID
 262         | FATTR4_WORD0_FILEID
 263         | FATTR4_WORD0_FS_LOCATIONS,
 264         FATTR4_WORD1_MODE
 265         | FATTR4_WORD1_NUMLINKS
 266         | FATTR4_WORD1_OWNER
 267         | FATTR4_WORD1_OWNER_GROUP
 268         | FATTR4_WORD1_RAWDEV
 269         | FATTR4_WORD1_SPACE_USED
 270         | FATTR4_WORD1_TIME_ACCESS
 271         | FATTR4_WORD1_TIME_METADATA
 272         | FATTR4_WORD1_TIME_MODIFY
 273         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 274 };
 275
 276 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 277                 struct nfs4_readdir_arg *readdir)
 278 {
 279         __be32 *start, *p;
 280
 281         if (cookie > 2) {
 282                 readdir->cookie = cookie;
 283                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 284                 return;
 285         }
 286
 287         readdir->cookie = 0;
 288         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 289         if (cookie == 2)
 290                 return;
 291
 292         /*
 293          * NFSv4 servers do not return entries for '.' and '..'
 294          * Therefore, we fake these entries here.  We let '.'
 295          * have cookie 0 and '..' have cookie 1.  Note that
 296          * when talking to the server, we always send cookie 0
 297          * instead of 1 or 2.
 298          */
 299         start = p = kmap_atomic(*readdir->pages);
 300
 301         if (cookie == 0) {
 302                 *p++ = xdr_one;                                  /* next */
 303                 *p++ = xdr_zero;                   /* cookie, first word */
 304                 *p++ = xdr_one;                   /* cookie, second word */
 305                 *p++ = xdr_one;                             /* entry len */
 306                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 307                 p++;
 308                 *p++ = xdr_one;                         /* bitmap length */
 309                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 310                 *p++ = htonl(8);              /* attribute buffer length */
 311                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 312         }
 313
 314         *p++ = xdr_one;                                  /* next */
 315         *p++ = xdr_zero;                   /* cookie, first word */
 316         *p++ = xdr_two;                   /* cookie, second word */
 317         *p++ = xdr_two;                             /* entry len */
 318         memcpy(p, "..\0\0", 4);                         /* entry */
 319         p++;
 320         *p++ = xdr_one;                         /* bitmap length */
 321         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 322         *p++ = htonl(8);              /* attribute buffer length */
 323         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 324
 325         readdir->pgbase = (char *)p - (char *)start;
 326         readdir->count -= readdir->pgbase;
 327         kunmap_atomic(start);
 328 }
 329
 330 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 331                 nfs4_stateid *stateid,
 332                 struct rpc_cred *cred)
 333 {
 334         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 335
 336         ops->test_and_free_expired(server, stateid, cred);
 337 }
 338
 339 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 340                 nfs4_stateid *stateid,
 341                 struct rpc_cred *cred)
 342 {
 343         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 344         nfs4_test_and_free_stateid(server, stateid, cred);
 345 }
 346
 347 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 348                 const nfs4_stateid *stateid,
 349                 struct rpc_cred *cred)
 350 {
 351         nfs4_stateid tmp;
 352
 353         nfs4_stateid_copy(&tmp, stateid);
 354         __nfs4_free_revoked_stateid(server, &tmp, cred);
 355 }
 356
 357 static long nfs4_update_delay(long *timeout)
 358 {
 359         long ret;
 360         if (!timeout)
 361                 return NFS4_POLL_RETRY_MAX;
 362         if (*timeout <= 0)
 363                 *timeout = NFS4_POLL_RETRY_MIN;
 364         if (*timeout > NFS4_POLL_RETRY_MAX)
 365                 *timeout = NFS4_POLL_RETRY_MAX;
 366         ret = *timeout;
 367         *timeout <<= 1;
 368         return ret;
 369 }
 370
 371 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 372 {
 373         int res = 0;
 374
 375         might_sleep();
 376
 377         freezable_schedule_timeout_killable_unsafe(
 378                 nfs4_update_delay(timeout));
 379         if (fatal_signal_pending(current))
 380                 res = -ERESTARTSYS;
 381         return res;
 382 }
 383
 384 /* This is the error handling routine for processes that are allowed
 385  * to sleep.
 386  */
 387 static int nfs4_do_handle_exception(struct nfs_server *server,
 388                 int errorcode, struct nfs4_exception *exception)
 389 {
 390         struct nfs_client *clp = server->nfs_client;
 391         struct nfs4_state *state = exception->state;
 392         const nfs4_stateid *stateid = exception->stateid;
 393         struct inode *inode = exception->inode;
 394         int ret = errorcode;
 395
 396         exception->delay = 0;
 397         exception->recovering = 0;
 398         exception->retry = 0;
 399
 400         if (stateid == NULL && state != NULL)
 401                 stateid = &state->stateid;
 402
 403         switch(errorcode) {
 404                 case 0:
 405                         return 0;
 406                 case -NFS4ERR_DELEG_REVOKED:
 407                 case -NFS4ERR_ADMIN_REVOKED:
 408                 case -NFS4ERR_EXPIRED:
 409                 case -NFS4ERR_BAD_STATEID:
 410                         if (inode != NULL && stateid != NULL) {
 411                                 nfs_inode_find_state_and_recover(inode,
 412                                                 stateid);
 413                                 goto wait_on_recovery;
 414                         }
 415                 case -NFS4ERR_OPENMODE:
 416                         if (inode) {
 417                                 int err;
 418
 419                                 err = nfs_async_inode_return_delegation(inode,
 420                                                 stateid);
 421                                 if (err == 0)
 422                                         goto wait_on_recovery;
 423                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 424                                         exception->retry = 1;
 425                                         break;
 426                                 }
 427                         }
 428                         if (state == NULL)
 429                                 break;
 430                         ret = nfs4_schedule_stateid_recovery(server, state);
 431                         if (ret < 0)
 432                                 break;
 433                         goto wait_on_recovery;
 434                 case -NFS4ERR_STALE_STATEID:
 435                 case -NFS4ERR_STALE_CLIENTID:
 436                         nfs4_schedule_lease_recovery(clp);
 437                         goto wait_on_recovery;
 438                 case -NFS4ERR_MOVED:
 439                         ret = nfs4_schedule_migration_recovery(server);
 440                         if (ret < 0)
 441                                 break;
 442                         goto wait_on_recovery;
 443                 case -NFS4ERR_LEASE_MOVED:
 444                         nfs4_schedule_lease_moved_recovery(clp);
 445                         goto wait_on_recovery;
 446 #if defined(CONFIG_NFS_V4_1)
 447                 case -NFS4ERR_BADSESSION:
 448                 case -NFS4ERR_BADSLOT:
 449                 case -NFS4ERR_BAD_HIGH_SLOT:
 450                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 451                 case -NFS4ERR_DEADSESSION:
 452                 case -NFS4ERR_SEQ_FALSE_RETRY:
 453                 case -NFS4ERR_SEQ_MISORDERED:
 454                         dprintk("%s ERROR: %d Reset session\n", __func__,
 455                                 errorcode);
 456                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 457                         goto wait_on_recovery;
 458 #endif /* defined(CONFIG_NFS_V4_1) */
 459                 case -NFS4ERR_FILE_OPEN:
 460                         if (exception->timeout > HZ) {
 461                                 /* We have retried a decent amount, time to
 462                                  * fail
 463                                  */
 464                                 ret = -EBUSY;
 465                                 break;
 466                         }
 467                 case -NFS4ERR_DELAY:
 468                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 469                 case -NFS4ERR_GRACE:
 470                 case -NFS4ERR_LAYOUTTRYLATER:
 471                 case -NFS4ERR_RECALLCONFLICT:
 472                         exception->delay = 1;
 473                         return 0;
 474
 475                 case -NFS4ERR_RETRY_UNCACHED_REP:
 476                 case -NFS4ERR_OLD_STATEID:
 477                         exception->retry = 1;
 478                         break;
 479                 case -NFS4ERR_BADOWNER:
 480                         /* The following works around a Linux server bug! */
 481                 case -NFS4ERR_BADNAME:
 482                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 483                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 484                                 exception->retry = 1;
 485                                 printk(KERN_WARNING "NFS: v4 server %s "
 486                                                 "does not accept raw "
 487                                                 "uid/gids. "
 488                                                 "Reenabling the idmapper.\n",
 489                                                 server->nfs_client->cl_hostname);
 490                         }
 491         }
 492         /* We failed to handle the error */
 493         return nfs4_map_errors(ret);
 494 wait_on_recovery:
 495         exception->recovering = 1;
 496         return 0;
 497 }
 498
 499 /* This is the error handling routine for processes that are allowed
 500  * to sleep.
 501  */
 502 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 503 {
 504         struct nfs_client *clp = server->nfs_client;
 505         int ret;
 506
 507         ret = nfs4_do_handle_exception(server, errorcode, exception);
 508         if (exception->delay) {
 509                 ret = nfs4_delay(server->client, &exception->timeout);
 510                 goto out_retry;
 511         }
 512         if (exception->recovering) {
 513                 ret = nfs4_wait_clnt_recover(clp);
 514                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 515                         return -EIO;
 516                 goto out_retry;
 517         }
 518         return ret;
 519 out_retry:
 520         if (ret == 0)
 521                 exception->retry = 1;
 522         return ret;
 523 }
 524
 525 static int
 526 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 527                 int errorcode, struct nfs4_exception *exception)
 528 {
 529         struct nfs_client *clp = server->nfs_client;
 530         int ret;
 531
 532         ret = nfs4_do_handle_exception(server, errorcode, exception);
 533         if (exception->delay) {
 534                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 535                 goto out_retry;
 536         }
 537         if (exception->recovering) {
 538                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 539                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 540                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 541                 goto out_retry;
 542         }
 543         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 544                 ret = -EIO;
 545         return ret;
 546 out_retry:
 547         if (ret == 0)
 548                 exception->retry = 1;
 549         return ret;
 550 }
 551
 552 static int
 553 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 554                         struct nfs4_state *state, long *timeout)
 555 {
 556         struct nfs4_exception exception = {
 557                 .state = state,
 558         };
 559
 560         if (task->tk_status >= 0)
 561                 return 0;
 562         if (timeout)
 563                 exception.timeout = *timeout;
 564         task->tk_status = nfs4_async_handle_exception(task, server,
 565                         task->tk_status,
 566                         &exception);
 567         if (exception.delay && timeout)
 568                 *timeout = exception.timeout;
 569         if (exception.retry)
 570                 return -EAGAIN;
 571         return 0;
 572 }
 573
 574 /*
 575  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 576  * or 'false' otherwise.
 577  */
 578 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 579 {
 580         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 581
 582         if (flavor == RPC_AUTH_GSS_KRB5I ||
 583             flavor == RPC_AUTH_GSS_KRB5P)
 584                 return true;
 585
 586         return false;
 587 }
 588
 589 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 590 {
 591         spin_lock(&clp->cl_lock);
 592         if (time_before(clp->cl_last_renewal,timestamp))
 593                 clp->cl_last_renewal = timestamp;
 594         spin_unlock(&clp->cl_lock);
 595 }
 596
 597 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 598 {
 599         struct nfs_client *clp = server->nfs_client;
 600
 601         if (!nfs4_has_session(clp))
 602                 do_renew_lease(clp, timestamp);
 603 }
 604
 605 struct nfs4_call_sync_data {
 606         const struct nfs_server *seq_server;
 607         struct nfs4_sequence_args *seq_args;
 608         struct nfs4_sequence_res *seq_res;
 609 };
 610
 611 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 612                         struct nfs4_sequence_res *res, int cache_reply)
 613 {
 614         args->sa_slot = NULL;
 615         args->sa_cache_this = cache_reply;
 616         args->sa_privileged = 0;
 617
 618         res->sr_slot = NULL;
 619 }
 620
 621 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 622 {
 623         args->sa_privileged = 1;
 624 }
 625
 626 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 627                          struct nfs4_sequence_args *args,
 628                          struct nfs4_sequence_res *res,
 629                          struct rpc_task *task)
 630 {
 631         struct nfs4_slot *slot;
 632
 633         /* slot already allocated? */
 634         if (res->sr_slot != NULL)
 635                 goto out_start;
 636
 637         spin_lock(&tbl->slot_tbl_lock);
 638         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 639                 goto out_sleep;
 640
 641         slot = nfs4_alloc_slot(tbl);
 642         if (IS_ERR(slot)) {
 643                 if (slot == ERR_PTR(-ENOMEM))
 644                         task->tk_timeout = HZ >> 2;
 645                 goto out_sleep;
 646         }
 647         spin_unlock(&tbl->slot_tbl_lock);
 648
 649         slot->privileged = args->sa_privileged ? 1 : 0;
 650         args->sa_slot = slot;
 651         res->sr_slot = slot;
 652
 653 out_start:
 654         rpc_call_start(task);
 655         return 0;
 656
 657 out_sleep:
 658         if (args->sa_privileged)
 659                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 660                                 NULL, RPC_PRIORITY_PRIVILEGED);
 661         else
 662                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 663         spin_unlock(&tbl->slot_tbl_lock);
 664         return -EAGAIN;
 665 }
 666 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 667
 668 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 669 {
 670         struct nfs4_slot *slot = res->sr_slot;
 671         struct nfs4_slot_table *tbl;
 672
 673         tbl = slot->table;
 674         spin_lock(&tbl->slot_tbl_lock);
 675         if (!nfs41_wake_and_assign_slot(tbl, slot))
 676                 nfs4_free_slot(tbl, slot);
 677         spin_unlock(&tbl->slot_tbl_lock);
 678
 679         res->sr_slot = NULL;
 680 }
 681
 682 static int nfs40_sequence_done(struct rpc_task *task,
 683                                struct nfs4_sequence_res *res)
 684 {
 685         if (res->sr_slot != NULL)
 686                 nfs40_sequence_free_slot(res);
 687         return 1;
 688 }
 689
 690 #if defined(CONFIG_NFS_V4_1)
 691
 692 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 693 {
 694         struct nfs4_session *session;
 695         struct nfs4_slot_table *tbl;
 696         struct nfs4_slot *slot = res->sr_slot;
 697         bool send_new_highest_used_slotid = false;
 698
 699         tbl = slot->table;
 700         session = tbl->session;
 701
 702         /* Bump the slot sequence number */
 703         if (slot->seq_done)
 704                 slot->seq_nr++;
 705         slot->seq_done = 0;
 706
 707         spin_lock(&tbl->slot_tbl_lock);
 708         /* Be nice to the server: try to ensure that the last transmitted
 709          * value for highest_user_slotid <= target_highest_slotid
 710          */
 711         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 712                 send_new_highest_used_slotid = true;
 713
 714         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 715                 send_new_highest_used_slotid = false;
 716                 goto out_unlock;
 717         }
 718         nfs4_free_slot(tbl, slot);
 719
 720         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 721                 send_new_highest_used_slotid = false;
 722 out_unlock:
 723         spin_unlock(&tbl->slot_tbl_lock);
 724         res->sr_slot = NULL;
 725         if (send_new_highest_used_slotid)
 726                 nfs41_notify_server(session->clp);
 727         if (waitqueue_active(&tbl->slot_waitq))
 728                 wake_up_all(&tbl->slot_waitq);
 729 }
 730
 731 static int nfs41_sequence_process(struct rpc_task *task,
 732                 struct nfs4_sequence_res *res)
 733 {
 734         struct nfs4_session *session;
 735         struct nfs4_slot *slot = res->sr_slot;
 736         struct nfs_client *clp;
 737         bool interrupted = false;
 738         int ret = 1;
 739
 740         if (slot == NULL)
 741                 goto out_noaction;
 742         /* don't increment the sequence number if the task wasn't sent */
 743         if (!RPC_WAS_SENT(task))
 744                 goto out;
 745
 746         session = slot->table->session;
 747
 748         if (slot->interrupted) {
 749                 slot->interrupted = 0;
 750                 interrupted = true;
 751         }
 752
 753         trace_nfs4_sequence_done(session, res);
 754         /* Check the SEQUENCE operation status */
 755         switch (res->sr_status) {
 756         case 0:
 757                 /* If previous op on slot was interrupted and we reused
 758                  * the seq# and got a reply from the cache, then retry
 759                  */
 760                 if (task->tk_status == -EREMOTEIO && interrupted) {
 761                         ++slot->seq_nr;
 762                         goto retry_nowait;
 763                 }
 764                 /* Update the slot's sequence and clientid lease timer */
 765                 slot->seq_done = 1;
 766                 clp = session->clp;
 767                 do_renew_lease(clp, res->sr_timestamp);
 768                 /* Check sequence flags */
 769                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 770                                 !!slot->privileged);
 771                 nfs41_update_target_slotid(slot->table, slot, res);
 772                 break;
 773         case 1:
 774                 /*
 775                  * sr_status remains 1 if an RPC level error occurred.
 776                  * The server may or may not have processed the sequence
 777                  * operation..
 778                  * Mark the slot as having hosted an interrupted RPC call.
 779                  */
 780                 slot->interrupted = 1;
 781                 goto out;
 782         case -NFS4ERR_DELAY:
 783                 /* The server detected a resend of the RPC call and
 784                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 785                  * of RFC5661.
 786                  */
 787                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 788                         __func__,
 789                         slot->slot_nr,
 790                         slot->seq_nr);
 791                 goto out_retry;
 792         case -NFS4ERR_BADSLOT:
 793                 /*
 794                  * The slot id we used was probably retired. Try again
 795                  * using a different slot id.
 796                  */
 797                 goto retry_nowait;
 798         case -NFS4ERR_SEQ_MISORDERED:
 799                 /*
 800                  * Was the last operation on this sequence interrupted?
 801                  * If so, retry after bumping the sequence number.
 802                  */
 803                 if (interrupted) {
 804                         ++slot->seq_nr;
 805                         goto retry_nowait;
 806                 }
 807                 /*
 808                  * Could this slot have been previously retired?
 809                  * If so, then the server may be expecting seq_nr = 1!
 810                  */
 811                 if (slot->seq_nr != 1) {
 812                         slot->seq_nr = 1;
 813                         goto retry_nowait;
 814                 }
 815                 break;
 816         case -NFS4ERR_SEQ_FALSE_RETRY:
 817                 ++slot->seq_nr;
 818                 goto retry_nowait;
 819         case -NFS4ERR_DEADSESSION:
 820         case -NFS4ERR_BADSESSION:
 821                 nfs4_schedule_session_recovery(session, res->sr_status);
 822                 goto retry_nowait;
 823         default:
 824                 /* Just update the slot sequence no. */
 825                 slot->seq_done = 1;
 826         }
 827 out:
 828         /* The session may be reset by one of the error handlers. */
 829         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 830 out_noaction:
 831         return ret;
 832 retry_nowait:
 833         if (rpc_restart_call_prepare(task)) {
 834                 nfs41_sequence_free_slot(res);
 835                 task->tk_status = 0;
 836                 ret = 0;
 837         }
 838         goto out;
 839 out_retry:
 840         if (!rpc_restart_call(task))
 841                 goto out;
 842         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 843         return 0;
 844 }
 845
 846 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 847 {
 848         if (!nfs41_sequence_process(task, res))
 849                 return 0;
 850         if (res->sr_slot != NULL)
 851                 nfs41_sequence_free_slot(res);
 852         return 1;
 853
 854 }
 855 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 856
 857 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 858 {
 859         if (res->sr_slot == NULL)
 860                 return 1;
 861         if (res->sr_slot->table->session != NULL)
 862                 return nfs41_sequence_process(task, res);
 863         return nfs40_sequence_done(task, res);
 864 }
 865
 866 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 867 {
 868         if (res->sr_slot != NULL) {
 869                 if (res->sr_slot->table->session != NULL)
 870                         nfs41_sequence_free_slot(res);
 871                 else
 872                         nfs40_sequence_free_slot(res);
 873         }
 874 }
 875
 876 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 877 {
 878         if (res->sr_slot == NULL)
 879                 return 1;
 880         if (!res->sr_slot->table->session)
 881                 return nfs40_sequence_done(task, res);
 882         return nfs41_sequence_done(task, res);
 883 }
 884 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 885
 886 int nfs41_setup_sequence(struct nfs4_session *session,
 887                                 struct nfs4_sequence_args *args,
 888                                 struct nfs4_sequence_res *res,
 889                                 struct rpc_task *task)
 890 {
 891         struct nfs4_slot *slot;
 892         struct nfs4_slot_table *tbl;
 893
 894         dprintk("--> %s\n", __func__);
 895         /* slot already allocated? */
 896         if (res->sr_slot != NULL)
 897                 goto out_success;
 898
 899         tbl = &session->fc_slot_table;
 900
 901         task->tk_timeout = 0;
 902
 903         spin_lock(&tbl->slot_tbl_lock);
 904         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 905             !args->sa_privileged) {
 906                 /* The state manager will wait until the slot table is empty */
 907                 dprintk("%s session is draining\n", __func__);
 908                 goto out_sleep;
 909         }
 910
 911         slot = nfs4_alloc_slot(tbl);
 912         if (IS_ERR(slot)) {
 913                 /* If out of memory, try again in 1/4 second */
 914                 if (slot == ERR_PTR(-ENOMEM))
 915                         task->tk_timeout = HZ >> 2;
 916                 dprintk("<-- %s: no free slots\n", __func__);
 917                 goto out_sleep;
 918         }
 919         spin_unlock(&tbl->slot_tbl_lock);
 920
 921         slot->privileged = args->sa_privileged ? 1 : 0;
 922         args->sa_slot = slot;
 923
 924         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 925                         slot->slot_nr, slot->seq_nr);
 926
 927         res->sr_slot = slot;
 928         res->sr_timestamp = jiffies;
 929         res->sr_status_flags = 0;
 930         /*
 931          * sr_status is only set in decode_sequence, and so will remain
 932          * set to 1 if an rpc level failure occurs.
 933          */
 934         res->sr_status = 1;
 935         trace_nfs4_setup_sequence(session, args);
 936 out_success:
 937         rpc_call_start(task);
 938         return 0;
 939 out_sleep:
 940         /* Privileged tasks are queued with top priority */
 941         if (args->sa_privileged)
 942                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 943                                 NULL, RPC_PRIORITY_PRIVILEGED);
 944         else
 945                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 946         spin_unlock(&tbl->slot_tbl_lock);
 947         return -EAGAIN;
 948 }
 949 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 950
 951 static int nfs4_setup_sequence(const struct nfs_server *server,
 952                                struct nfs4_sequence_args *args,
 953                                struct nfs4_sequence_res *res,
 954                                struct rpc_task *task)
 955 {
 956         struct nfs4_session *session = nfs4_get_session(server);
 957         int ret = 0;
 958
 959         if (!session)
 960                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 961                                             args, res, task);
 962
 963         dprintk("--> %s clp %p session %p sr_slot %u\n",
 964                 __func__, session->clp, session, res->sr_slot ?
 965                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 966
 967         ret = nfs41_setup_sequence(session, args, res, task);
 968
 969         dprintk("<-- %s status=%d\n", __func__, ret);
 970         return ret;
 971 }
 972
 973 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 974 {
 975         struct nfs4_call_sync_data *data = calldata;
 976         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 977
 978         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 979
 980         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 981 }
 982
 983 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 984 {
 985         struct nfs4_call_sync_data *data = calldata;
 986
 987         nfs41_sequence_done(task, data->seq_res);
 988 }
 989
 990 static const struct rpc_call_ops nfs41_call_sync_ops = {
 991         .rpc_call_prepare = nfs41_call_sync_prepare,
 992         .rpc_call_done = nfs41_call_sync_done,
 993 };
 994
 995 #else   /* !CONFIG_NFS_V4_1 */
 996
 997 static int nfs4_setup_sequence(const struct nfs_server *server,
 998                                struct nfs4_sequence_args *args,
 999                                struct nfs4_sequence_res *res,
1000                                struct rpc_task *task)
1001 {
1002         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
1003                                     args, res, task);
1004 }
1005
1006 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1007 {
1008         return nfs40_sequence_done(task, res);
1009 }
1010
1011 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1012 {
1013         if (res->sr_slot != NULL)
1014                 nfs40_sequence_free_slot(res);
1015 }
1016
1017 int nfs4_sequence_done(struct rpc_task *task,
1018                        struct nfs4_sequence_res *res)
1019 {
1020         return nfs40_sequence_done(task, res);
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1023
1024 #endif  /* !CONFIG_NFS_V4_1 */
1025
1026 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1027 {
1028         struct nfs4_call_sync_data *data = calldata;
1029         nfs4_setup_sequence(data->seq_server,
1030                                 data->seq_args, data->seq_res, task);
1031 }
1032
1033 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1034 {
1035         struct nfs4_call_sync_data *data = calldata;
1036         nfs4_sequence_done(task, data->seq_res);
1037 }
1038
1039 static const struct rpc_call_ops nfs40_call_sync_ops = {
1040         .rpc_call_prepare = nfs40_call_sync_prepare,
1041         .rpc_call_done = nfs40_call_sync_done,
1042 };
1043
1044 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1045                                    struct nfs_server *server,
1046                                    struct rpc_message *msg,
1047                                    struct nfs4_sequence_args *args,
1048                                    struct nfs4_sequence_res *res)
1049 {
1050         int ret;
1051         struct rpc_task *task;
1052         struct nfs_client *clp = server->nfs_client;
1053         struct nfs4_call_sync_data data = {
1054                 .seq_server = server,
1055                 .seq_args = args,
1056                 .seq_res = res,
1057         };
1058         struct rpc_task_setup task_setup = {
1059                 .rpc_client = clnt,
1060                 .rpc_message = msg,
1061                 .callback_ops = clp->cl_mvops->call_sync_ops,
1062                 .callback_data = &data
1063         };
1064
1065         task = rpc_run_task(&task_setup);
1066         if (IS_ERR(task))
1067                 ret = PTR_ERR(task);
1068         else {
1069                 ret = task->tk_status;
1070                 rpc_put_task(task);
1071         }
1072         return ret;
1073 }
1074
1075 int nfs4_call_sync(struct rpc_clnt *clnt,
1076                    struct nfs_server *server,
1077                    struct rpc_message *msg,
1078                    struct nfs4_sequence_args *args,
1079                    struct nfs4_sequence_res *res,
1080                    int cache_reply)
1081 {
1082         nfs4_init_sequence(args, res, cache_reply);
1083         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1084 }
1085
1086 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1087 {
1088         struct nfs_inode *nfsi = NFS_I(dir);
1089
1090         spin_lock(&dir->i_lock);
1091         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1092         if (!cinfo->atomic || cinfo->before != dir->i_version)
1093                 nfs_force_lookup_revalidate(dir);
1094         dir->i_version = cinfo->after;
1095         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1096         nfs_fscache_invalidate(dir);
1097         spin_unlock(&dir->i_lock);
1098 }
1099
1100 struct nfs4_opendata {
1101         struct kref kref;
1102         struct nfs_openargs o_arg;
1103         struct nfs_openres o_res;
1104         struct nfs_open_confirmargs c_arg;
1105         struct nfs_open_confirmres c_res;
1106         struct nfs4_string owner_name;
1107         struct nfs4_string group_name;
1108         struct nfs4_label *a_label;
1109         struct nfs_fattr f_attr;
1110         struct nfs4_label *f_label;
1111         struct dentry *dir;
1112         struct dentry *dentry;
1113         struct nfs4_state_owner *owner;
1114         struct nfs4_state *state;
1115         struct iattr attrs;
1116         unsigned long timestamp;
1117         unsigned int rpc_done : 1;
1118         unsigned int file_created : 1;
1119         unsigned int is_recover : 1;
1120         int rpc_status;
1121         int cancelled;
1122 };
1123
1124 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1125                 int err, struct nfs4_exception *exception)
1126 {
1127         if (err != -EINVAL)
1128                 return false;
1129         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1130                 return false;
1131         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1132         exception->retry = 1;
1133         return true;
1134 }
1135
1136 static u32
1137 nfs4_map_atomic_open_share(struct nfs_server *server,
1138                 fmode_t fmode, int openflags)
1139 {
1140         u32 res = 0;
1141
1142         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1143         case FMODE_READ:
1144                 res = NFS4_SHARE_ACCESS_READ;
1145                 break;
1146         case FMODE_WRITE:
1147                 res = NFS4_SHARE_ACCESS_WRITE;
1148                 break;
1149         case FMODE_READ|FMODE_WRITE:
1150                 res = NFS4_SHARE_ACCESS_BOTH;
1151         }
1152         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1153                 goto out;
1154         /* Want no delegation if we're using O_DIRECT */
1155         if (openflags & O_DIRECT)
1156                 res |= NFS4_SHARE_WANT_NO_DELEG;
1157 out:
1158         return res;
1159 }
1160
1161 static enum open_claim_type4
1162 nfs4_map_atomic_open_claim(struct nfs_server *server,
1163                 enum open_claim_type4 claim)
1164 {
1165         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1166                 return claim;
1167         switch (claim) {
1168         default:
1169                 return claim;
1170         case NFS4_OPEN_CLAIM_FH:
1171                 return NFS4_OPEN_CLAIM_NULL;
1172         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1173                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1174         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1175                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1176         }
1177 }
1178
1179 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1180 {
1181         p->o_res.f_attr = &p->f_attr;
1182         p->o_res.f_label = p->f_label;
1183         p->o_res.seqid = p->o_arg.seqid;
1184         p->c_res.seqid = p->c_arg.seqid;
1185         p->o_res.server = p->o_arg.server;
1186         p->o_res.access_request = p->o_arg.access;
1187         nfs_fattr_init(&p->f_attr);
1188         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1189 }
1190
1191 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1192                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1193                 const struct iattr *attrs,
1194                 struct nfs4_label *label,
1195                 enum open_claim_type4 claim,
1196                 gfp_t gfp_mask)
1197 {
1198         struct dentry *parent = dget_parent(dentry);
1199         struct inode *dir = d_inode(parent);
1200         struct nfs_server *server = NFS_SERVER(dir);
1201         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1202         struct nfs4_opendata *p;
1203
1204         p = kzalloc(sizeof(*p), gfp_mask);
1205         if (p == NULL)
1206                 goto err;
1207
1208         p->f_label = nfs4_label_alloc(server, gfp_mask);
1209         if (IS_ERR(p->f_label))
1210                 goto err_free_p;
1211
1212         p->a_label = nfs4_label_alloc(server, gfp_mask);
1213         if (IS_ERR(p->a_label))
1214                 goto err_free_f;
1215
1216         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1217         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1218         if (IS_ERR(p->o_arg.seqid))
1219                 goto err_free_label;
1220         nfs_sb_active(dentry->d_sb);
1221         p->dentry = dget(dentry);
1222         p->dir = parent;
1223         p->owner = sp;
1224         atomic_inc(&sp->so_count);
1225         p->o_arg.open_flags = flags;
1226         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1227         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1228         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1229                         fmode, flags);
1230         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1231          * will return permission denied for all bits until close */
1232         if (!(flags & O_EXCL)) {
1233                 /* ask server to check for all possible rights as results
1234                  * are cached */
1235                 switch (p->o_arg.claim) {
1236                 default:
1237                         break;
1238                 case NFS4_OPEN_CLAIM_NULL:
1239                 case NFS4_OPEN_CLAIM_FH:
1240                         p->o_arg.access = NFS4_ACCESS_READ |
1241                                 NFS4_ACCESS_MODIFY |
1242                                 NFS4_ACCESS_EXTEND |
1243                                 NFS4_ACCESS_EXECUTE;
1244                 }
1245         }
1246         p->o_arg.clientid = server->nfs_client->cl_clientid;
1247         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1248         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1249         p->o_arg.name = &dentry->d_name;
1250         p->o_arg.server = server;
1251         p->o_arg.bitmask = nfs4_bitmask(server, label);
1252         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1253         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1254         switch (p->o_arg.claim) {
1255         case NFS4_OPEN_CLAIM_NULL:
1256         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1257         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1258                 p->o_arg.fh = NFS_FH(dir);
1259                 break;
1260         case NFS4_OPEN_CLAIM_PREVIOUS:
1261         case NFS4_OPEN_CLAIM_FH:
1262         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1263         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1264                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1265         }
1266         if (attrs != NULL && attrs->ia_valid != 0) {
1267                 __u32 verf[2];
1268
1269                 p->o_arg.u.attrs = &p->attrs;
1270                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1271
1272                 verf[0] = jiffies;
1273                 verf[1] = current->pid;
1274                 memcpy(p->o_arg.u.verifier.data, verf,
1275                                 sizeof(p->o_arg.u.verifier.data));
1276         }
1277         p->c_arg.fh = &p->o_res.fh;
1278         p->c_arg.stateid = &p->o_res.stateid;
1279         p->c_arg.seqid = p->o_arg.seqid;
1280         nfs4_init_opendata_res(p);
1281         kref_init(&p->kref);
1282         return p;
1283
1284 err_free_label:
1285         nfs4_label_free(p->a_label);
1286 err_free_f:
1287         nfs4_label_free(p->f_label);
1288 err_free_p:
1289         kfree(p);
1290 err:
1291         dput(parent);
1292         return NULL;
1293 }
1294
1295 static void nfs4_opendata_free(struct kref *kref)
1296 {
1297         struct nfs4_opendata *p = container_of(kref,
1298                         struct nfs4_opendata, kref);
1299         struct super_block *sb = p->dentry->d_sb;
1300
1301         nfs_free_seqid(p->o_arg.seqid);
1302         nfs4_sequence_free_slot(&p->o_res.seq_res);
1303         if (p->state != NULL)
1304                 nfs4_put_open_state(p->state);
1305         nfs4_put_state_owner(p->owner);
1306
1307         nfs4_label_free(p->a_label);
1308         nfs4_label_free(p->f_label);
1309
1310         dput(p->dir);
1311         dput(p->dentry);
1312         nfs_sb_deactive(sb);
1313         nfs_fattr_free_names(&p->f_attr);
1314         kfree(p->f_attr.mdsthreshold);
1315         kfree(p);
1316 }
1317
1318 static void nfs4_opendata_put(struct nfs4_opendata *p)
1319 {
1320         if (p != NULL)
1321                 kref_put(&p->kref, nfs4_opendata_free);
1322 }
1323
1324 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1325 {
1326         int ret;
1327
1328         ret = rpc_wait_for_completion_task(task);
1329         return ret;
1330 }
1331
1332 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1333                 fmode_t fmode)
1334 {
1335         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1336         case FMODE_READ|FMODE_WRITE:
1337                 return state->n_rdwr != 0;
1338         case FMODE_WRITE:
1339                 return state->n_wronly != 0;
1340         case FMODE_READ:
1341                 return state->n_rdonly != 0;
1342         }
1343         WARN_ON_ONCE(1);
1344         return false;
1345 }
1346
1347 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1348 {
1349         int ret = 0;
1350
1351         if (open_mode & (O_EXCL|O_TRUNC))
1352                 goto out;
1353         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1354                 case FMODE_READ:
1355                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1356                                 && state->n_rdonly != 0;
1357                         break;
1358                 case FMODE_WRITE:
1359                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1360                                 && state->n_wronly != 0;
1361                         break;
1362                 case FMODE_READ|FMODE_WRITE:
1363                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1364                                 && state->n_rdwr != 0;
1365         }
1366 out:
1367         return ret;
1368 }
1369
1370 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1371                 enum open_claim_type4 claim)
1372 {
1373         if (delegation == NULL)
1374                 return 0;
1375         if ((delegation->type & fmode) != fmode)
1376                 return 0;
1377         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1378                 return 0;
1379         switch (claim) {
1380         case NFS4_OPEN_CLAIM_NULL:
1381         case NFS4_OPEN_CLAIM_FH:
1382                 break;
1383         case NFS4_OPEN_CLAIM_PREVIOUS:
1384                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1385                         break;
1386         default:
1387                 return 0;
1388         }
1389         nfs_mark_delegation_referenced(delegation);
1390         return 1;
1391 }
1392
1393 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1394 {
1395         switch (fmode) {
1396                 case FMODE_WRITE:
1397                         state->n_wronly++;
1398                         break;
1399                 case FMODE_READ:
1400                         state->n_rdonly++;
1401                         break;
1402                 case FMODE_READ|FMODE_WRITE:
1403                         state->n_rdwr++;
1404         }
1405         nfs4_state_set_mode_locked(state, state->state | fmode);
1406 }
1407
1408 #ifdef CONFIG_NFS_V4_1
1409 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1410 {
1411         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1412                 return true;
1413         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1414                 return true;
1415         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1416                 return true;
1417         return false;
1418 }
1419 #endif /* CONFIG_NFS_V4_1 */
1420
1421 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1422 {
1423         struct nfs_client *clp = state->owner->so_server->nfs_client;
1424         bool need_recover = false;
1425
1426         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1427                 need_recover = true;
1428         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1429                 need_recover = true;
1430         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1431                 need_recover = true;
1432         if (need_recover)
1433                 nfs4_state_mark_reclaim_nograce(clp, state);
1434 }
1435
1436 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1437                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1438 {
1439         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1440                 return true;
1441         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1442                 nfs4_stateid_copy(freeme, &state->open_stateid);
1443                 nfs_test_and_clear_all_open_stateid(state);
1444                 return true;
1445         }
1446         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1447                 return true;
1448         return false;
1449 }
1450
1451 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1452 {
1453         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1454                 return;
1455         if (state->n_wronly)
1456                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1457         if (state->n_rdonly)
1458                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1459         if (state->n_rdwr)
1460                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1461         set_bit(NFS_OPEN_STATE, &state->flags);
1462 }
1463
1464 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1465                 nfs4_stateid *stateid, fmode_t fmode)
1466 {
1467         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1468         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1469         case FMODE_WRITE:
1470                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1471                 break;
1472         case FMODE_READ:
1473                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1474                 break;
1475         case 0:
1476                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1477                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1478                 clear_bit(NFS_OPEN_STATE, &state->flags);
1479         }
1480         if (stateid == NULL)
1481                 return;
1482         /* Handle OPEN+OPEN_DOWNGRADE races */
1483         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1484             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1485                 nfs_resync_open_stateid_locked(state);
1486                 return;
1487         }
1488         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1489                 nfs4_stateid_copy(&state->stateid, stateid);
1490         nfs4_stateid_copy(&state->open_stateid, stateid);
1491 }
1492
1493 static void nfs_clear_open_stateid(struct nfs4_state *state,
1494         nfs4_stateid *arg_stateid,
1495         nfs4_stateid *stateid, fmode_t fmode)
1496 {
1497         write_seqlock(&state->seqlock);
1498         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1499         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1500                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1501         write_sequnlock(&state->seqlock);
1502         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1503                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1504 }
1505
1506 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1507                 const nfs4_stateid *stateid, fmode_t fmode,
1508                 nfs4_stateid *freeme)
1509 {
1510         switch (fmode) {
1511                 case FMODE_READ:
1512                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1513                         break;
1514                 case FMODE_WRITE:
1515                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1516                         break;
1517                 case FMODE_READ|FMODE_WRITE:
1518                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1519         }
1520         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1521                 return;
1522         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1523                 nfs4_stateid_copy(&state->stateid, stateid);
1524         nfs4_stateid_copy(&state->open_stateid, stateid);
1525 }
1526
1527 static void __update_open_stateid(struct nfs4_state *state,
1528                 const nfs4_stateid *open_stateid,
1529                 const nfs4_stateid *deleg_stateid,
1530                 fmode_t fmode,
1531                 nfs4_stateid *freeme)
1532 {
1533         /*
1534          * Protect the call to nfs4_state_set_mode_locked and
1535          * serialise the stateid update
1536          */
1537         spin_lock(&state->owner->so_lock);
1538         write_seqlock(&state->seqlock);
1539         if (deleg_stateid != NULL) {
1540                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1541                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1542         }
1543         if (open_stateid != NULL)
1544                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1545         write_sequnlock(&state->seqlock);
1546         update_open_stateflags(state, fmode);
1547         spin_unlock(&state->owner->so_lock);
1548 }
1549
1550 static int update_open_stateid(struct nfs4_state *state,
1551                 const nfs4_stateid *open_stateid,
1552                 const nfs4_stateid *delegation,
1553                 fmode_t fmode)
1554 {
1555         struct nfs_server *server = NFS_SERVER(state->inode);
1556         struct nfs_client *clp = server->nfs_client;
1557         struct nfs_inode *nfsi = NFS_I(state->inode);
1558         struct nfs_delegation *deleg_cur;
1559         nfs4_stateid freeme = { };
1560         int ret = 0;
1561
1562         fmode &= (FMODE_READ|FMODE_WRITE);
1563
1564         rcu_read_lock();
1565         deleg_cur = rcu_dereference(nfsi->delegation);
1566         if (deleg_cur == NULL)
1567                 goto no_delegation;
1568
1569         spin_lock(&deleg_cur->lock);
1570         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1571            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1572             (deleg_cur->type & fmode) != fmode)
1573                 goto no_delegation_unlock;
1574
1575         if (delegation == NULL)
1576                 delegation = &deleg_cur->stateid;
1577         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1578                 goto no_delegation_unlock;
1579
1580         nfs_mark_delegation_referenced(deleg_cur);
1581         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1582                         fmode, &freeme);
1583         ret = 1;
1584 no_delegation_unlock:
1585         spin_unlock(&deleg_cur->lock);
1586 no_delegation:
1587         rcu_read_unlock();
1588
1589         if (!ret && open_stateid != NULL) {
1590                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1591                 ret = 1;
1592         }
1593         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1594                 nfs4_schedule_state_manager(clp);
1595         if (freeme.type != 0)
1596                 nfs4_test_and_free_stateid(server, &freeme,
1597                                 state->owner->so_cred);
1598
1599         return ret;
1600 }
1601
1602 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1603                 const nfs4_stateid *stateid)
1604 {
1605         struct nfs4_state *state = lsp->ls_state;
1606         bool ret = false;
1607
1608         spin_lock(&state->state_lock);
1609         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1610                 goto out_noupdate;
1611         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1612                 goto out_noupdate;
1613         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1614         ret = true;
1615 out_noupdate:
1616         spin_unlock(&state->state_lock);
1617         return ret;
1618 }
1619
1620 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1621 {
1622         struct nfs_delegation *delegation;
1623
1624         rcu_read_lock();
1625         delegation = rcu_dereference(NFS_I(inode)->delegation);
1626         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1627                 rcu_read_unlock();
1628                 return;
1629         }
1630         rcu_read_unlock();
1631         nfs4_inode_return_delegation(inode);
1632 }
1633
1634 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1635 {
1636         struct nfs4_state *state = opendata->state;
1637         struct nfs_inode *nfsi = NFS_I(state->inode);
1638         struct nfs_delegation *delegation;
1639         int open_mode = opendata->o_arg.open_flags;
1640         fmode_t fmode = opendata->o_arg.fmode;
1641         enum open_claim_type4 claim = opendata->o_arg.claim;
1642         nfs4_stateid stateid;
1643         int ret = -EAGAIN;
1644
1645         for (;;) {
1646                 spin_lock(&state->owner->so_lock);
1647                 if (can_open_cached(state, fmode, open_mode)) {
1648                         update_open_stateflags(state, fmode);
1649                         spin_unlock(&state->owner->so_lock);
1650                         goto out_return_state;
1651                 }
1652                 spin_unlock(&state->owner->so_lock);
1653                 rcu_read_lock();
1654                 delegation = rcu_dereference(nfsi->delegation);
1655                 if (!can_open_delegated(delegation, fmode, claim)) {
1656                         rcu_read_unlock();
1657                         break;
1658                 }
1659                 /* Save the delegation */
1660                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1661                 rcu_read_unlock();
1662                 nfs_release_seqid(opendata->o_arg.seqid);
1663                 if (!opendata->is_recover) {
1664                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1665                         if (ret != 0)
1666                                 goto out;
1667                 }
1668                 ret = -EAGAIN;
1669
1670                 /* Try to update the stateid using the delegation */
1671                 if (update_open_stateid(state, NULL, &stateid, fmode))
1672                         goto out_return_state;
1673         }
1674 out:
1675         return ERR_PTR(ret);
1676 out_return_state:
1677         atomic_inc(&state->count);
1678         return state;
1679 }
1680
1681 static void
1682 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1683 {
1684         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1685         struct nfs_delegation *delegation;
1686         int delegation_flags = 0;
1687
1688         rcu_read_lock();
1689         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1690         if (delegation)
1691                 delegation_flags = delegation->flags;
1692         rcu_read_unlock();
1693         switch (data->o_arg.claim) {
1694         default:
1695                 break;
1696         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1697         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1698                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1699                                    "returning a delegation for "
1700                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1701                                    clp->cl_hostname);
1702                 return;
1703         }
1704         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1705                 nfs_inode_set_delegation(state->inode,
1706                                          data->owner->so_cred,
1707                                          &data->o_res);
1708         else
1709                 nfs_inode_reclaim_delegation(state->inode,
1710                                              data->owner->so_cred,
1711                                              &data->o_res);
1712 }
1713
1714 /*
1715  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1716  * and update the nfs4_state.
1717  */
1718 static struct nfs4_state *
1719 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1720 {
1721         struct inode *inode = data->state->inode;
1722         struct nfs4_state *state = data->state;
1723         int ret;
1724
1725         if (!data->rpc_done) {
1726                 if (data->rpc_status) {
1727                         ret = data->rpc_status;
1728                         goto err;
1729                 }
1730                 /* cached opens have already been processed */
1731                 goto update;
1732         }
1733
1734         ret = nfs_refresh_inode(inode, &data->f_attr);
1735         if (ret)
1736                 goto err;
1737
1738         if (data->o_res.delegation_type != 0)
1739                 nfs4_opendata_check_deleg(data, state);
1740 update:
1741         update_open_stateid(state, &data->o_res.stateid, NULL,
1742                             data->o_arg.fmode);
1743         atomic_inc(&state->count);
1744
1745         return state;
1746 err:
1747         return ERR_PTR(ret);
1748
1749 }
1750
1751 static struct nfs4_state *
1752 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1753 {
1754         struct inode *inode;
1755         struct nfs4_state *state = NULL;
1756         int ret;
1757
1758         if (!data->rpc_done) {
1759                 state = nfs4_try_open_cached(data);
1760                 trace_nfs4_cached_open(data->state);
1761                 goto out;
1762         }
1763
1764         ret = -EAGAIN;
1765         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1766                 goto err;
1767         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1768         ret = PTR_ERR(inode);
1769         if (IS_ERR(inode))
1770                 goto err;
1771         ret = -ENOMEM;
1772         state = nfs4_get_open_state(inode, data->owner);
1773         if (state == NULL)
1774                 goto err_put_inode;
1775         if (data->o_res.delegation_type != 0)
1776                 nfs4_opendata_check_deleg(data, state);
1777         update_open_stateid(state, &data->o_res.stateid, NULL,
1778                         data->o_arg.fmode);
1779         iput(inode);
1780 out:
1781         nfs_release_seqid(data->o_arg.seqid);
1782         return state;
1783 err_put_inode:
1784         iput(inode);
1785 err:
1786         return ERR_PTR(ret);
1787 }
1788
1789 static struct nfs4_state *
1790 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1791 {
1792         struct nfs4_state *ret;
1793
1794         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1795                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1796         else
1797                 ret = _nfs4_opendata_to_nfs4_state(data);
1798         nfs4_sequence_free_slot(&data->o_res.seq_res);
1799         return ret;
1800 }
1801
1802 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1803 {
1804         struct nfs_inode *nfsi = NFS_I(state->inode);
1805         struct nfs_open_context *ctx;
1806
1807         spin_lock(&state->inode->i_lock);
1808         list_for_each_entry(ctx, &nfsi->open_files, list) {
1809                 if (ctx->state != state)
1810                         continue;
1811                 get_nfs_open_context(ctx);
1812                 spin_unlock(&state->inode->i_lock);
1813                 return ctx;
1814         }
1815         spin_unlock(&state->inode->i_lock);
1816         return ERR_PTR(-ENOENT);
1817 }
1818
1819 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1820                 struct nfs4_state *state, enum open_claim_type4 claim)
1821 {
1822         struct nfs4_opendata *opendata;
1823
1824         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1825                         NULL, NULL, claim, GFP_NOFS);
1826         if (opendata == NULL)
1827                 return ERR_PTR(-ENOMEM);
1828         opendata->state = state;
1829         atomic_inc(&state->count);
1830         return opendata;
1831 }
1832
1833 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1834                 fmode_t fmode)
1835 {
1836         struct nfs4_state *newstate;
1837         int ret;
1838
1839         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1840                 return 0;
1841         opendata->o_arg.open_flags = 0;
1842         opendata->o_arg.fmode = fmode;
1843         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1844                         NFS_SB(opendata->dentry->d_sb),
1845                         fmode, 0);
1846         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1847         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1848         nfs4_init_opendata_res(opendata);
1849         ret = _nfs4_recover_proc_open(opendata);
1850         if (ret != 0)
1851                 return ret;
1852         newstate = nfs4_opendata_to_nfs4_state(opendata);
1853         if (IS_ERR(newstate))
1854                 return PTR_ERR(newstate);
1855         if (newstate != opendata->state)
1856                 ret = -ESTALE;
1857         nfs4_close_state(newstate, fmode);
1858         return ret;
1859 }
1860
1861 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1862 {
1863         int ret;
1864
1865         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1866         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1867         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1868         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1869         /* memory barrier prior to reading state->n_* */
1870         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1871         clear_bit(NFS_OPEN_STATE, &state->flags);
1872         smp_rmb();
1873         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1874         if (ret != 0)
1875                 return ret;
1876         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1877         if (ret != 0)
1878                 return ret;
1879         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1880         if (ret != 0)
1881                 return ret;
1882         /*
1883          * We may have performed cached opens for all three recoveries.
1884          * Check if we need to update the current stateid.
1885          */
1886         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1887             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1888                 write_seqlock(&state->seqlock);
1889                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1890                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1891                 write_sequnlock(&state->seqlock);
1892         }
1893         return 0;
1894 }
1895
1896 /*
1897  * OPEN_RECLAIM:
1898  *      reclaim state on the server after a reboot.
1899  */
1900 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1901 {
1902         struct nfs_delegation *delegation;
1903         struct nfs4_opendata *opendata;
1904         fmode_t delegation_type = 0;
1905         int status;
1906
1907         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1908                         NFS4_OPEN_CLAIM_PREVIOUS);
1909         if (IS_ERR(opendata))
1910                 return PTR_ERR(opendata);
1911         rcu_read_lock();
1912         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1913         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1914                 delegation_type = delegation->type;
1915         rcu_read_unlock();
1916         opendata->o_arg.u.delegation_type = delegation_type;
1917         status = nfs4_open_recover(opendata, state);
1918         nfs4_opendata_put(opendata);
1919         return status;
1920 }
1921
1922 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1923 {
1924         struct nfs_server *server = NFS_SERVER(state->inode);
1925         struct nfs4_exception exception = { };
1926         int err;
1927         do {
1928                 err = _nfs4_do_open_reclaim(ctx, state);
1929                 trace_nfs4_open_reclaim(ctx, 0, err);
1930                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1931                         continue;
1932                 if (err != -NFS4ERR_DELAY)
1933                         break;
1934                 nfs4_handle_exception(server, err, &exception);
1935         } while (exception.retry);
1936         return err;
1937 }
1938
1939 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1940 {
1941         struct nfs_open_context *ctx;
1942         int ret;
1943
1944         ctx = nfs4_state_find_open_context(state);
1945         if (IS_ERR(ctx))
1946                 return -EAGAIN;
1947         ret = nfs4_do_open_reclaim(ctx, state);
1948         put_nfs_open_context(ctx);
1949         return ret;
1950 }
1951
1952 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1953 {
1954         switch (err) {
1955                 default:
1956                         printk(KERN_ERR "NFS: %s: unhandled error "
1957                                         "%d.\n", __func__, err);
1958                 case 0:
1959                 case -ENOENT:
1960                 case -EAGAIN:
1961                 case -ESTALE:
1962                         break;
1963                 case -NFS4ERR_BADSESSION:
1964                 case -NFS4ERR_BADSLOT:
1965                 case -NFS4ERR_BAD_HIGH_SLOT:
1966                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1967                 case -NFS4ERR_DEADSESSION:
1968                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1969                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1970                         return -EAGAIN;
1971                 case -NFS4ERR_STALE_CLIENTID:
1972                 case -NFS4ERR_STALE_STATEID:
1973                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1974                         /* Don't recall a delegation if it was lost */
1975                         nfs4_schedule_lease_recovery(server->nfs_client);
1976                         return -EAGAIN;
1977                 case -NFS4ERR_MOVED:
1978                         nfs4_schedule_migration_recovery(server);
1979                         return -EAGAIN;
1980                 case -NFS4ERR_LEASE_MOVED:
1981                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1982                         return -EAGAIN;
1983                 case -NFS4ERR_DELEG_REVOKED:
1984                 case -NFS4ERR_ADMIN_REVOKED:
1985                 case -NFS4ERR_EXPIRED:
1986                 case -NFS4ERR_BAD_STATEID:
1987                 case -NFS4ERR_OPENMODE:
1988                         nfs_inode_find_state_and_recover(state->inode,
1989                                         stateid);
1990                         nfs4_schedule_stateid_recovery(server, state);
1991                         return -EAGAIN;
1992                 case -NFS4ERR_DELAY:
1993                 case -NFS4ERR_GRACE:
1994                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1995                         ssleep(1);
1996                         return -EAGAIN;
1997                 case -ENOMEM:
1998                 case -NFS4ERR_DENIED:
1999                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
2000                         return 0;
2001         }
2002         return err;
2003 }
2004
2005 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2006                 struct nfs4_state *state, const nfs4_stateid *stateid,
2007                 fmode_t type)
2008 {
2009         struct nfs_server *server = NFS_SERVER(state->inode);
2010         struct nfs4_opendata *opendata;
2011         int err = 0;
2012
2013         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2014                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2015         if (IS_ERR(opendata))
2016                 return PTR_ERR(opendata);
2017         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2018         write_seqlock(&state->seqlock);
2019         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2020         write_sequnlock(&state->seqlock);
2021         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2022         switch (type & (FMODE_READ|FMODE_WRITE)) {
2023         case FMODE_READ|FMODE_WRITE:
2024         case FMODE_WRITE:
2025                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2026                 if (err)
2027                         break;
2028                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2029                 if (err)
2030                         break;
2031         case FMODE_READ:
2032                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2033         }
2034         nfs4_opendata_put(opendata);
2035         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
2036 }
2037
2038 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2039 {
2040         struct nfs4_opendata *data = calldata;
2041
2042         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
2043                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
2044 }
2045
2046 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2047 {
2048         struct nfs4_opendata *data = calldata;
2049
2050         nfs40_sequence_done(task, &data->c_res.seq_res);
2051
2052         data->rpc_status = task->tk_status;
2053         if (data->rpc_status == 0) {
2054                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2055                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2056                 renew_lease(data->o_res.server, data->timestamp);
2057                 data->rpc_done = 1;
2058         }
2059 }
2060
2061 static void nfs4_open_confirm_release(void *calldata)
2062 {
2063         struct nfs4_opendata *data = calldata;
2064         struct nfs4_state *state = NULL;
2065
2066         /* If this request hasn't been cancelled, do nothing */
2067         if (data->cancelled == 0)
2068                 goto out_free;
2069         /* In case of error, no cleanup! */
2070         if (!data->rpc_done)
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_confirm_ops = {
2080         .rpc_call_prepare = nfs4_open_confirm_prepare,
2081         .rpc_call_done = nfs4_open_confirm_done,
2082         .rpc_release = nfs4_open_confirm_release,
2083 };
2084
2085 /*
2086  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2087  */
2088 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2089 {
2090         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2091         struct rpc_task *task;
2092         struct  rpc_message msg = {
2093                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2094                 .rpc_argp = &data->c_arg,
2095                 .rpc_resp = &data->c_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_confirm_ops,
2102                 .callback_data = data,
2103                 .workqueue = nfsiod_workqueue,
2104                 .flags = RPC_TASK_ASYNC,
2105         };
2106         int status;
2107
2108         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2109         kref_get(&data->kref);
2110         data->rpc_done = 0;
2111         data->rpc_status = 0;
2112         data->timestamp = jiffies;
2113         if (data->is_recover)
2114                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2115         task = rpc_run_task(&task_setup_data);
2116         if (IS_ERR(task))
2117                 return PTR_ERR(task);
2118         status = nfs4_wait_for_completion_rpc_task(task);
2119         if (status != 0) {
2120                 data->cancelled = 1;
2121                 smp_wmb();
2122         } else
2123                 status = data->rpc_status;
2124         rpc_put_task(task);
2125         return status;
2126 }
2127
2128 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2129 {
2130         struct nfs4_opendata *data = calldata;
2131         struct nfs4_state_owner *sp = data->owner;
2132         struct nfs_client *clp = sp->so_server->nfs_client;
2133         enum open_claim_type4 claim = data->o_arg.claim;
2134
2135         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2136                 goto out_wait;
2137         /*
2138          * Check if we still need to send an OPEN call, or if we can use
2139          * a delegation instead.
2140          */
2141         if (data->state != NULL) {
2142                 struct nfs_delegation *delegation;
2143
2144                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2145                         goto out_no_action;
2146                 rcu_read_lock();
2147                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2148                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2149                         goto unlock_no_action;
2150                 rcu_read_unlock();
2151         }
2152         /* Update client id. */
2153         data->o_arg.clientid = clp->cl_clientid;
2154         switch (claim) {
2155         default:
2156                 break;
2157         case NFS4_OPEN_CLAIM_PREVIOUS:
2158         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2159         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2160                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2161         case NFS4_OPEN_CLAIM_FH:
2162                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2163                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2164         }
2165         data->timestamp = jiffies;
2166         if (nfs4_setup_sequence(data->o_arg.server,
2167                                 &data->o_arg.seq_args,
2168                                 &data->o_res.seq_res,
2169                                 task) != 0)
2170                 nfs_release_seqid(data->o_arg.seqid);
2171
2172         /* Set the create mode (note dependency on the session type) */
2173         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2174         if (data->o_arg.open_flags & O_EXCL) {
2175                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2176                 if (nfs4_has_persistent_session(clp))
2177                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2178                 else if (clp->cl_mvops->minor_version > 0)
2179                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2180         }
2181         return;
2182 unlock_no_action:
2183         trace_nfs4_cached_open(data->state);
2184         rcu_read_unlock();
2185 out_no_action:
2186         task->tk_action = NULL;
2187 out_wait:
2188         nfs4_sequence_done(task, &data->o_res.seq_res);
2189 }
2190
2191 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2192 {
2193         struct nfs4_opendata *data = calldata;
2194
2195         data->rpc_status = task->tk_status;
2196
2197         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2198                 return;
2199
2200         if (task->tk_status == 0) {
2201                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2202                         switch (data->o_res.f_attr->mode & S_IFMT) {
2203                         case S_IFREG:
2204                                 break;
2205                         case S_IFLNK:
2206                                 data->rpc_status = -ELOOP;
2207                                 break;
2208                         case S_IFDIR:
2209                                 data->rpc_status = -EISDIR;
2210                                 break;
2211                         default:
2212                                 data->rpc_status = -ENOTDIR;
2213                         }
2214                 }
2215                 renew_lease(data->o_res.server, data->timestamp);
2216                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2217                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2218         }
2219         data->rpc_done = 1;
2220 }
2221
2222 static void nfs4_open_release(void *calldata)
2223 {
2224         struct nfs4_opendata *data = calldata;
2225         struct nfs4_state *state = NULL;
2226
2227         /* If this request hasn't been cancelled, do nothing */
2228         if (data->cancelled == 0)
2229                 goto out_free;
2230         /* In case of error, no cleanup! */
2231         if (data->rpc_status != 0 || !data->rpc_done)
2232                 goto out_free;
2233         /* In case we need an open_confirm, no cleanup! */
2234         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2235                 goto out_free;
2236         state = nfs4_opendata_to_nfs4_state(data);
2237         if (!IS_ERR(state))
2238                 nfs4_close_state(state, data->o_arg.fmode);
2239 out_free:
2240         nfs4_opendata_put(data);
2241 }
2242
2243 static const struct rpc_call_ops nfs4_open_ops = {
2244         .rpc_call_prepare = nfs4_open_prepare,
2245         .rpc_call_done = nfs4_open_done,
2246         .rpc_release = nfs4_open_release,
2247 };
2248
2249 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2250 {
2251         struct inode *dir = d_inode(data->dir);
2252         struct nfs_server *server = NFS_SERVER(dir);
2253         struct nfs_openargs *o_arg = &data->o_arg;
2254         struct nfs_openres *o_res = &data->o_res;
2255         struct rpc_task *task;
2256         struct rpc_message msg = {
2257                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2258                 .rpc_argp = o_arg,
2259                 .rpc_resp = o_res,
2260                 .rpc_cred = data->owner->so_cred,
2261         };
2262         struct rpc_task_setup task_setup_data = {
2263                 .rpc_client = server->client,
2264                 .rpc_message = &msg,
2265                 .callback_ops = &nfs4_open_ops,
2266                 .callback_data = data,
2267                 .workqueue = nfsiod_workqueue,
2268                 .flags = RPC_TASK_ASYNC,
2269         };
2270         int status;
2271
2272         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2273         kref_get(&data->kref);
2274         data->rpc_done = 0;
2275         data->rpc_status = 0;
2276         data->cancelled = 0;
2277         data->is_recover = 0;
2278         if (isrecover) {
2279                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2280                 data->is_recover = 1;
2281         }
2282         task = rpc_run_task(&task_setup_data);
2283         if (IS_ERR(task))
2284                 return PTR_ERR(task);
2285         status = nfs4_wait_for_completion_rpc_task(task);
2286         if (status != 0) {
2287                 data->cancelled = 1;
2288                 smp_wmb();
2289         } else
2290                 status = data->rpc_status;
2291         rpc_put_task(task);
2292
2293         return status;
2294 }
2295
2296 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2297 {
2298         struct inode *dir = d_inode(data->dir);
2299         struct nfs_openres *o_res = &data->o_res;
2300         int status;
2301
2302         status = nfs4_run_open_task(data, 1);
2303         if (status != 0 || !data->rpc_done)
2304                 return status;
2305
2306         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2307
2308         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2309                 status = _nfs4_proc_open_confirm(data);
2310                 if (status != 0)
2311                         return status;
2312         }
2313
2314         return status;
2315 }
2316
2317 /*
2318  * Additional permission checks in order to distinguish between an
2319  * open for read, and an open for execute. This works around the
2320  * fact that NFSv4 OPEN treats read and execute permissions as being
2321  * the same.
2322  * Note that in the non-execute case, we want to turn off permission
2323  * checking if we just created a new file (POSIX open() semantics).
2324  */
2325 static int nfs4_opendata_access(struct rpc_cred *cred,
2326                                 struct nfs4_opendata *opendata,
2327                                 struct nfs4_state *state, fmode_t fmode,
2328                                 int openflags)
2329 {
2330         struct nfs_access_entry cache;
2331         u32 mask;
2332
2333         /* access call failed or for some reason the server doesn't
2334          * support any access modes -- defer access call until later */
2335         if (opendata->o_res.access_supported == 0)
2336                 return 0;
2337
2338         mask = 0;
2339         /*
2340          * Use openflags to check for exec, because fmode won't
2341          * always have FMODE_EXEC set when file open for exec.
2342          */
2343         if (openflags & __FMODE_EXEC) {
2344                 /* ONLY check for exec rights */
2345                 mask = MAY_EXEC;
2346         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2347                 mask = MAY_READ;
2348
2349         cache.cred = cred;
2350         cache.jiffies = jiffies;
2351         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2352         nfs_access_add_cache(state->inode, &cache);
2353
2354         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2355                 return 0;
2356
2357         /* even though OPEN succeeded, access is denied. Close the file */
2358         nfs4_close_state(state, fmode);
2359         return -EACCES;
2360 }
2361
2362 /*
2363  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2364  */
2365 static int _nfs4_proc_open(struct nfs4_opendata *data)
2366 {
2367         struct inode *dir = d_inode(data->dir);
2368         struct nfs_server *server = NFS_SERVER(dir);
2369         struct nfs_openargs *o_arg = &data->o_arg;
2370         struct nfs_openres *o_res = &data->o_res;
2371         int status;
2372
2373         status = nfs4_run_open_task(data, 0);
2374         if (!data->rpc_done)
2375                 return status;
2376         if (status != 0) {
2377                 if (status == -NFS4ERR_BADNAME &&
2378                                 !(o_arg->open_flags & O_CREAT))
2379                         return -ENOENT;
2380                 return status;
2381         }
2382
2383         nfs_fattr_map_and_free_names(server, &data->f_attr);
2384
2385         if (o_arg->open_flags & O_CREAT) {
2386                 update_changeattr(dir, &o_res->cinfo);
2387                 if (o_arg->open_flags & O_EXCL)
2388                         data->file_created = 1;
2389                 else if (o_res->cinfo.before != o_res->cinfo.after)
2390                         data->file_created = 1;
2391         }
2392         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2393                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2394         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2395                 status = _nfs4_proc_open_confirm(data);
2396                 if (status != 0)
2397                         return status;
2398         }
2399         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2400                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2401         return 0;
2402 }
2403
2404 static int nfs4_recover_expired_lease(struct nfs_server *server)
2405 {
2406         return nfs4_client_recover_expired_lease(server->nfs_client);
2407 }
2408
2409 /*
2410  * OPEN_EXPIRED:
2411  *      reclaim state on the server after a network partition.
2412  *      Assumes caller holds the appropriate lock
2413  */
2414 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2415 {
2416         struct nfs4_opendata *opendata;
2417         int ret;
2418
2419         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2420                         NFS4_OPEN_CLAIM_FH);
2421         if (IS_ERR(opendata))
2422                 return PTR_ERR(opendata);
2423         ret = nfs4_open_recover(opendata, state);
2424         if (ret == -ESTALE)
2425                 d_drop(ctx->dentry);
2426         nfs4_opendata_put(opendata);
2427         return ret;
2428 }
2429
2430 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2431 {
2432         struct nfs_server *server = NFS_SERVER(state->inode);
2433         struct nfs4_exception exception = { };
2434         int err;
2435
2436         do {
2437                 err = _nfs4_open_expired(ctx, state);
2438                 trace_nfs4_open_expired(ctx, 0, err);
2439                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2440                         continue;
2441                 switch (err) {
2442                 default:
2443                         goto out;
2444                 case -NFS4ERR_GRACE:
2445                 case -NFS4ERR_DELAY:
2446                         nfs4_handle_exception(server, err, &exception);
2447                         err = 0;
2448                 }
2449         } while (exception.retry);
2450 out:
2451         return err;
2452 }
2453
2454 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2455 {
2456         struct nfs_open_context *ctx;
2457         int ret;
2458
2459         ctx = nfs4_state_find_open_context(state);
2460         if (IS_ERR(ctx))
2461                 return -EAGAIN;
2462         ret = nfs4_do_open_expired(ctx, state);
2463         put_nfs_open_context(ctx);
2464         return ret;
2465 }
2466
2467 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2468                 const nfs4_stateid *stateid)
2469 {
2470         nfs_remove_bad_delegation(state->inode, stateid);
2471         write_seqlock(&state->seqlock);
2472         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2473         write_sequnlock(&state->seqlock);
2474         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2475 }
2476
2477 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2478 {
2479         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2480                 nfs_finish_clear_delegation_stateid(state, NULL);
2481 }
2482
2483 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2484 {
2485         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2486         nfs40_clear_delegation_stateid(state);
2487         return nfs4_open_expired(sp, state);
2488 }
2489
2490 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2491                 nfs4_stateid *stateid,
2492                 struct rpc_cred *cred)
2493 {
2494         return -NFS4ERR_BAD_STATEID;
2495 }
2496
2497 #if defined(CONFIG_NFS_V4_1)
2498 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2499                 nfs4_stateid *stateid,
2500                 struct rpc_cred *cred)
2501 {
2502         int status;
2503
2504         switch (stateid->type) {
2505         default:
2506                 break;
2507         case NFS4_INVALID_STATEID_TYPE:
2508         case NFS4_SPECIAL_STATEID_TYPE:
2509                 return -NFS4ERR_BAD_STATEID;
2510         case NFS4_REVOKED_STATEID_TYPE:
2511                 goto out_free;
2512         }
2513
2514         status = nfs41_test_stateid(server, stateid, cred);
2515         switch (status) {
2516         case -NFS4ERR_EXPIRED:
2517         case -NFS4ERR_ADMIN_REVOKED:
2518         case -NFS4ERR_DELEG_REVOKED:
2519                 break;
2520         default:
2521                 return status;
2522         }
2523 out_free:
2524         /* Ack the revoked state to the server */
2525         nfs41_free_stateid(server, stateid, cred, true);
2526         return -NFS4ERR_EXPIRED;
2527 }
2528
2529 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2530 {
2531         struct nfs_server *server = NFS_SERVER(state->inode);
2532         nfs4_stateid stateid;
2533         struct nfs_delegation *delegation;
2534         struct rpc_cred *cred;
2535         int status;
2536
2537         /* Get the delegation credential for use by test/free_stateid */
2538         rcu_read_lock();
2539         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2540         if (delegation == NULL) {
2541                 rcu_read_unlock();
2542                 return;
2543         }
2544
2545         nfs4_stateid_copy(&stateid, &delegation->stateid);
2546         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2547                 rcu_read_unlock();
2548                 nfs_finish_clear_delegation_stateid(state, &stateid);
2549                 return;
2550         }
2551
2552         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags)) {
2553                 rcu_read_unlock();
2554                 return;
2555         }
2556
2557         cred = get_rpccred(delegation->cred);
2558         rcu_read_unlock();
2559         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2560         trace_nfs4_test_delegation_stateid(state, NULL, status);
2561         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2562                 nfs_finish_clear_delegation_stateid(state, &stateid);
2563
2564         put_rpccred(cred);
2565 }
2566
2567 /**
2568  * nfs41_check_expired_locks - possibly free a lock stateid
2569  *
2570  * @state: NFSv4 state for an inode
2571  *
2572  * Returns NFS_OK if recovery for this stateid is now finished.
2573  * Otherwise a negative NFS4ERR value is returned.
2574  */
2575 static int nfs41_check_expired_locks(struct nfs4_state *state)
2576 {
2577         int status, ret = NFS_OK;
2578         struct nfs4_lock_state *lsp, *prev = NULL;
2579         struct nfs_server *server = NFS_SERVER(state->inode);
2580
2581         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2582                 goto out;
2583
2584         spin_lock(&state->state_lock);
2585         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2586                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2587                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2588
2589                         atomic_inc(&lsp->ls_count);
2590                         spin_unlock(&state->state_lock);
2591
2592                         nfs4_put_lock_state(prev);
2593                         prev = lsp;
2594
2595                         status = nfs41_test_and_free_expired_stateid(server,
2596                                         &lsp->ls_stateid,
2597                                         cred);
2598                         trace_nfs4_test_lock_stateid(state, lsp, status);
2599                         if (status == -NFS4ERR_EXPIRED ||
2600                             status == -NFS4ERR_BAD_STATEID) {
2601                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2602                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2603                                 if (!recover_lost_locks)
2604                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2605                         } else if (status != NFS_OK) {
2606                                 ret = status;
2607                                 nfs4_put_lock_state(prev);
2608                                 goto out;
2609                         }
2610                         spin_lock(&state->state_lock);
2611                 }
2612         }
2613         spin_unlock(&state->state_lock);
2614         nfs4_put_lock_state(prev);
2615 out:
2616         return ret;
2617 }
2618
2619 /**
2620  * nfs41_check_open_stateid - possibly free an open stateid
2621  *
2622  * @state: NFSv4 state for an inode
2623  *
2624  * Returns NFS_OK if recovery for this stateid is now finished.
2625  * Otherwise a negative NFS4ERR value is returned.
2626  */
2627 static int nfs41_check_open_stateid(struct nfs4_state *state)
2628 {
2629         struct nfs_server *server = NFS_SERVER(state->inode);
2630         nfs4_stateid *stateid = &state->open_stateid;
2631         struct rpc_cred *cred = state->owner->so_cred;
2632         int status;
2633
2634         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2635                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2636                         if (nfs4_have_delegation(state->inode, state->state))
2637                                 return NFS_OK;
2638                         return -NFS4ERR_OPENMODE;
2639                 }
2640                 return -NFS4ERR_BAD_STATEID;
2641         }
2642         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2643         trace_nfs4_test_open_stateid(state, NULL, status);
2644         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2645                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2646                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2647                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2648                 clear_bit(NFS_OPEN_STATE, &state->flags);
2649                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2650         }
2651         if (status != NFS_OK)
2652                 return status;
2653         if (nfs_open_stateid_recover_openmode(state))
2654                 return -NFS4ERR_OPENMODE;
2655         return NFS_OK;
2656 }
2657
2658 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2659 {
2660         int status;
2661
2662         nfs41_check_delegation_stateid(state);
2663         status = nfs41_check_expired_locks(state);
2664         if (status != NFS_OK)
2665                 return status;
2666         status = nfs41_check_open_stateid(state);
2667         if (status != NFS_OK)
2668                 status = nfs4_open_expired(sp, state);
2669         return status;
2670 }
2671 #endif
2672
2673 /*
2674  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2675  * fields corresponding to attributes that were used to store the verifier.
2676  * Make sure we clobber those fields in the later setattr call
2677  */
2678 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2679                                 struct iattr *sattr, struct nfs4_label **label)
2680 {
2681         const u32 *attrset = opendata->o_res.attrset;
2682
2683         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2684             !(sattr->ia_valid & ATTR_ATIME_SET))
2685                 sattr->ia_valid |= ATTR_ATIME;
2686
2687         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2688             !(sattr->ia_valid & ATTR_MTIME_SET))
2689                 sattr->ia_valid |= ATTR_MTIME;
2690
2691         /* Except MODE, it seems harmless of setting twice. */
2692         if ((attrset[1] & FATTR4_WORD1_MODE))
2693                 sattr->ia_valid &= ~ATTR_MODE;
2694
2695         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2696                 *label = NULL;
2697 }
2698
2699 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2700                 fmode_t fmode,
2701                 int flags,
2702                 struct nfs_open_context *ctx)
2703 {
2704         struct nfs4_state_owner *sp = opendata->owner;
2705         struct nfs_server *server = sp->so_server;
2706         struct dentry *dentry;
2707         struct nfs4_state *state;
2708         unsigned int seq;
2709         int ret;
2710
2711         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2712
2713         ret = _nfs4_proc_open(opendata);
2714         if (ret != 0)
2715                 goto out;
2716
2717         state = nfs4_opendata_to_nfs4_state(opendata);
2718         ret = PTR_ERR(state);
2719         if (IS_ERR(state))
2720                 goto out;
2721         if (server->caps & NFS_CAP_POSIX_LOCK)
2722                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2723         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2724                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2725
2726         dentry = opendata->dentry;
2727         if (d_really_is_negative(dentry)) {
2728                 struct dentry *alias;
2729                 d_drop(dentry);
2730                 alias = d_exact_alias(dentry, state->inode);
2731                 if (!alias)
2732                         alias = d_splice_alias(igrab(state->inode), dentry);
2733                 /* d_splice_alias() can't fail here - it's a non-directory */
2734                 if (alias) {
2735                         dput(ctx->dentry);
2736                         ctx->dentry = dentry = alias;
2737                 }
2738                 nfs_set_verifier(dentry,
2739                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2740         }
2741
2742         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2743         if (ret != 0)
2744                 goto out;
2745
2746         ctx->state = state;
2747         if (d_inode(dentry) == state->inode) {
2748                 nfs_inode_attach_open_context(ctx);
2749                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2750                         nfs4_schedule_stateid_recovery(server, state);
2751         }
2752 out:
2753         return ret;
2754 }
2755
2756 /*
2757  * Returns a referenced nfs4_state
2758  */
2759 static int _nfs4_do_open(struct inode *dir,
2760                         struct nfs_open_context *ctx,
2761                         int flags,
2762                         struct iattr *sattr,
2763                         struct nfs4_label *label,
2764                         int *opened)
2765 {
2766         struct nfs4_state_owner  *sp;
2767         struct nfs4_state     *state = NULL;
2768         struct nfs_server       *server = NFS_SERVER(dir);
2769         struct nfs4_opendata *opendata;
2770         struct dentry *dentry = ctx->dentry;
2771         struct rpc_cred *cred = ctx->cred;
2772         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2773         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2774         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2775         struct nfs4_label *olabel = NULL;
2776         int status;
2777
2778         /* Protect against reboot recovery conflicts */
2779         status = -ENOMEM;
2780         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2781         if (sp == NULL) {
2782                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2783                 goto out_err;
2784         }
2785         status = nfs4_recover_expired_lease(server);
2786         if (status != 0)
2787                 goto err_put_state_owner;
2788         if (d_really_is_positive(dentry))
2789                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2790         status = -ENOMEM;
2791         if (d_really_is_positive(dentry))
2792                 claim = NFS4_OPEN_CLAIM_FH;
2793         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2794                         label, claim, GFP_KERNEL);
2795         if (opendata == NULL)
2796                 goto err_put_state_owner;
2797
2798         if (label) {
2799                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2800                 if (IS_ERR(olabel)) {
2801                         status = PTR_ERR(olabel);
2802                         goto err_opendata_put;
2803                 }
2804         }
2805
2806         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2807                 if (!opendata->f_attr.mdsthreshold) {
2808                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2809                         if (!opendata->f_attr.mdsthreshold)
2810                                 goto err_free_label;
2811                 }
2812                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2813         }
2814         if (d_really_is_positive(dentry))
2815                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2816
2817         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2818         if (status != 0)
2819                 goto err_free_label;
2820         state = ctx->state;
2821
2822         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2823             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2824                 nfs4_exclusive_attrset(opendata, sattr, &label);
2825                 /*
2826                  * send create attributes which was not set by open
2827                  * with an extra setattr.
2828                  */
2829                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2830                         nfs_fattr_init(opendata->o_res.f_attr);
2831                         status = nfs4_do_setattr(state->inode, cred,
2832                                         opendata->o_res.f_attr, sattr,
2833                                         ctx, label, olabel);
2834                         if (status == 0) {
2835                                 nfs_setattr_update_inode(state->inode, sattr,
2836                                                 opendata->o_res.f_attr);
2837                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2838                         }
2839                 }
2840         }
2841         if (opened && opendata->file_created)
2842                 *opened |= FILE_CREATED;
2843
2844         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2845                 *ctx_th = opendata->f_attr.mdsthreshold;
2846                 opendata->f_attr.mdsthreshold = NULL;
2847         }
2848
2849         nfs4_label_free(olabel);
2850
2851         nfs4_opendata_put(opendata);
2852         nfs4_put_state_owner(sp);
2853         return 0;
2854 err_free_label:
2855         nfs4_label_free(olabel);
2856 err_opendata_put:
2857         nfs4_opendata_put(opendata);
2858 err_put_state_owner:
2859         nfs4_put_state_owner(sp);
2860 out_err:
2861         return status;
2862 }
2863
2864
2865 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2866                                         struct nfs_open_context *ctx,
2867                                         int flags,
2868                                         struct iattr *sattr,
2869                                         struct nfs4_label *label,
2870                                         int *opened)
2871 {
2872         struct nfs_server *server = NFS_SERVER(dir);
2873         struct nfs4_exception exception = { };
2874         struct nfs4_state *res;
2875         int status;
2876
2877         do {
2878                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2879                 res = ctx->state;
2880                 trace_nfs4_open_file(ctx, flags, status);
2881                 if (status == 0)
2882                         break;
2883                 /* NOTE: BAD_SEQID means the server and client disagree about the
2884                  * book-keeping w.r.t. state-changing operations
2885                  * (OPEN/CLOSE/LOCK/LOCKU...)
2886                  * It is actually a sign of a bug on the client or on the server.
2887                  *
2888                  * If we receive a BAD_SEQID error in the particular case of
2889                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2890                  * have unhashed the old state_owner for us, and that we can
2891                  * therefore safely retry using a new one. We should still warn
2892                  * the user though...
2893                  */
2894                 if (status == -NFS4ERR_BAD_SEQID) {
2895                         pr_warn_ratelimited("NFS: v4 server %s "
2896                                         " returned a bad sequence-id error!\n",
2897                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2898                         exception.retry = 1;
2899                         continue;
2900                 }
2901                 /*
2902                  * BAD_STATEID on OPEN means that the server cancelled our
2903                  * state before it received the OPEN_CONFIRM.
2904                  * Recover by retrying the request as per the discussion
2905                  * on Page 181 of RFC3530.
2906                  */
2907                 if (status == -NFS4ERR_BAD_STATEID) {
2908                         exception.retry = 1;
2909                         continue;
2910                 }
2911                 if (status == -EAGAIN) {
2912                         /* We must have found a delegation */
2913                         exception.retry = 1;
2914                         continue;
2915                 }
2916                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2917                         continue;
2918                 res = ERR_PTR(nfs4_handle_exception(server,
2919                                         status, &exception));
2920         } while (exception.retry);
2921         return res;
2922 }
2923
2924 static int _nfs4_do_setattr(struct inode *inode,
2925                             struct nfs_setattrargs *arg,
2926                             struct nfs_setattrres *res,
2927                             struct rpc_cred *cred,
2928                             struct nfs_open_context *ctx)
2929 {
2930         struct nfs_server *server = NFS_SERVER(inode);
2931         struct rpc_message msg = {
2932                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2933                 .rpc_argp       = arg,
2934                 .rpc_resp       = res,
2935                 .rpc_cred       = cred,
2936         };
2937         struct rpc_cred *delegation_cred = NULL;
2938         unsigned long timestamp = jiffies;
2939         fmode_t fmode;
2940         bool truncate;
2941         int status;
2942
2943         nfs_fattr_init(res->fattr);
2944
2945         /* Servers should only apply open mode checks for file size changes */
2946         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2947         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2948
2949         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2950                 /* Use that stateid */
2951         } else if (truncate && ctx != NULL) {
2952                 struct nfs_lock_context *l_ctx;
2953                 if (!nfs4_valid_open_stateid(ctx->state))
2954                         return -EBADF;
2955                 l_ctx = nfs_get_lock_context(ctx);
2956                 if (IS_ERR(l_ctx))
2957                         return PTR_ERR(l_ctx);
2958                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2959                                                 &arg->stateid, &delegation_cred);
2960                 nfs_put_lock_context(l_ctx);
2961                 if (status == -EIO)
2962                         return -EBADF;
2963         } else
2964                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2965         if (delegation_cred)
2966                 msg.rpc_cred = delegation_cred;
2967
2968         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2969
2970         put_rpccred(delegation_cred);
2971         if (status == 0 && ctx != NULL)
2972                 renew_lease(server, timestamp);
2973         trace_nfs4_setattr(inode, &arg->stateid, status);
2974         return status;
2975 }
2976
2977 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2978                            struct nfs_fattr *fattr, struct iattr *sattr,
2979                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2980                            struct nfs4_label *olabel)
2981 {
2982         struct nfs_server *server = NFS_SERVER(inode);
2983         struct nfs4_state *state = ctx ? ctx->state : NULL;
2984         struct nfs_setattrargs  arg = {
2985                 .fh             = NFS_FH(inode),
2986                 .iap            = sattr,
2987                 .server         = server,
2988                 .bitmask = server->attr_bitmask,
2989                 .label          = ilabel,
2990         };
2991         struct nfs_setattrres  res = {
2992                 .fattr          = fattr,
2993                 .label          = olabel,
2994                 .server         = server,
2995         };
2996         struct nfs4_exception exception = {
2997                 .state = state,
2998                 .inode = inode,
2999                 .stateid = &arg.stateid,
3000         };
3001         int err;
3002
3003         arg.bitmask = nfs4_bitmask(server, ilabel);
3004         if (ilabel)
3005                 arg.bitmask = nfs4_bitmask(server, olabel);
3006
3007         do {
3008                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3009                 switch (err) {
3010                 case -NFS4ERR_OPENMODE:
3011                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3012                                 pr_warn_once("NFSv4: server %s is incorrectly "
3013                                                 "applying open mode checks to "
3014                                                 "a SETATTR that is not "
3015                                                 "changing file size.\n",
3016                                                 server->nfs_client->cl_hostname);
3017                         }
3018                         if (state && !(state->state & FMODE_WRITE)) {
3019                                 err = -EBADF;
3020                                 if (sattr->ia_valid & ATTR_OPEN)
3021                                         err = -EACCES;
3022                                 goto out;
3023                         }
3024                 }
3025                 err = nfs4_handle_exception(server, err, &exception);
3026         } while (exception.retry);
3027 out:
3028         return err;
3029 }
3030
3031 static bool
3032 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3033 {
3034         if (inode == NULL || !nfs_have_layout(inode))
3035                 return false;
3036
3037         return pnfs_wait_on_layoutreturn(inode, task);
3038 }
3039
3040 struct nfs4_closedata {
3041         struct inode *inode;
3042         struct nfs4_state *state;
3043         struct nfs_closeargs arg;
3044         struct nfs_closeres res;
3045         struct {
3046                 struct nfs4_layoutreturn_args arg;
3047                 struct nfs4_layoutreturn_res res;
3048                 struct nfs4_xdr_opaque_data ld_private;
3049                 u32 roc_barrier;
3050                 bool roc;
3051         } lr;
3052         struct nfs_fattr fattr;
3053         unsigned long timestamp;
3054 };
3055
3056 static void nfs4_free_closedata(void *data)
3057 {
3058         struct nfs4_closedata *calldata = data;
3059         struct nfs4_state_owner *sp = calldata->state->owner;
3060         struct super_block *sb = calldata->state->inode->i_sb;
3061
3062         if (calldata->lr.roc)
3063                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3064                                 calldata->res.lr_ret);
3065         nfs4_put_open_state(calldata->state);
3066         nfs_free_seqid(calldata->arg.seqid);
3067         nfs4_put_state_owner(sp);
3068         nfs_sb_deactive(sb);
3069         kfree(calldata);
3070 }
3071
3072 static void nfs4_close_done(struct rpc_task *task, void *data)
3073 {
3074         struct nfs4_closedata *calldata = data;
3075         struct nfs4_state *state = calldata->state;
3076         struct nfs_server *server = NFS_SERVER(calldata->inode);
3077         nfs4_stateid *res_stateid = NULL;
3078
3079         dprintk("%s: begin!\n", __func__);
3080         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3081                 return;
3082         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3083
3084         /* Handle Layoutreturn errors */
3085         if (calldata->arg.lr_args && task->tk_status != 0) {
3086                 switch (calldata->res.lr_ret) {
3087                 default:
3088                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3089                         break;
3090                 case 0:
3091                         calldata->arg.lr_args = NULL;
3092                         calldata->res.lr_res = NULL;
3093                         break;
3094                 case -NFS4ERR_ADMIN_REVOKED:
3095                 case -NFS4ERR_DELEG_REVOKED:
3096                 case -NFS4ERR_EXPIRED:
3097                 case -NFS4ERR_BAD_STATEID:
3098                 case -NFS4ERR_OLD_STATEID:
3099                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3100                 case -NFS4ERR_WRONG_CRED:
3101                         calldata->arg.lr_args = NULL;
3102                         calldata->res.lr_res = NULL;
3103                         calldata->res.lr_ret = 0;
3104                         rpc_restart_call_prepare(task);
3105                         return;
3106                 }
3107         }
3108
3109         /* hmm. we are done with the inode, and in the process of freeing
3110          * the state_owner. we keep this around to process errors
3111          */
3112         switch (task->tk_status) {
3113                 case 0:
3114                         res_stateid = &calldata->res.stateid;
3115                         renew_lease(server, calldata->timestamp);
3116                         break;
3117                 case -NFS4ERR_ADMIN_REVOKED:
3118                 case -NFS4ERR_STALE_STATEID:
3119                 case -NFS4ERR_EXPIRED:
3120                         nfs4_free_revoked_stateid(server,
3121                                         &calldata->arg.stateid,
3122                                         task->tk_msg.rpc_cred);
3123                 case -NFS4ERR_OLD_STATEID:
3124                 case -NFS4ERR_BAD_STATEID:
3125                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3126                                                 &state->open_stateid)) {
3127                                 rpc_restart_call_prepare(task);
3128                                 goto out_release;
3129                         }
3130                         if (calldata->arg.fmode == 0)
3131                                 break;
3132                 default:
3133                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3134                                 rpc_restart_call_prepare(task);
3135                                 goto out_release;
3136                         }
3137         }
3138         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3139                         res_stateid, calldata->arg.fmode);
3140 out_release:
3141         nfs_release_seqid(calldata->arg.seqid);
3142         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
3143         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3144 }
3145
3146 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3147 {
3148         struct nfs4_closedata *calldata = data;
3149         struct nfs4_state *state = calldata->state;
3150         struct inode *inode = calldata->inode;
3151         bool is_rdonly, is_wronly, is_rdwr;
3152         int call_close = 0;
3153
3154         dprintk("%s: begin!\n", __func__);
3155         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3156                 goto out_wait;
3157
3158         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3159         spin_lock(&state->owner->so_lock);
3160         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3161         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3162         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3163         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3164         /* Calculate the change in open mode */
3165         calldata->arg.fmode = 0;
3166         if (state->n_rdwr == 0) {
3167                 if (state->n_rdonly == 0)
3168                         call_close |= is_rdonly;
3169                 else if (is_rdonly)
3170                         calldata->arg.fmode |= FMODE_READ;
3171                 if (state->n_wronly == 0)
3172                         call_close |= is_wronly;
3173                 else if (is_wronly)
3174                         calldata->arg.fmode |= FMODE_WRITE;
3175                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3176                         call_close |= is_rdwr;
3177         } else if (is_rdwr)
3178                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3179
3180         if (!nfs4_valid_open_stateid(state) ||
3181             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3182                 call_close = 0;
3183         spin_unlock(&state->owner->so_lock);
3184
3185         if (!call_close) {
3186                 /* Note: exit _without_ calling nfs4_close_done */
3187                 goto out_no_action;
3188         }
3189
3190         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3191                 nfs_release_seqid(calldata->arg.seqid);
3192                 goto out_wait;
3193         }
3194
3195         if (calldata->arg.fmode == 0) {
3196                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3197
3198                 /* Close-to-open cache consistency revalidation */
3199                 if (!nfs4_have_delegation(inode, FMODE_READ))
3200                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3201                 else
3202                         calldata->arg.bitmask = NULL;
3203         }
3204
3205         calldata->arg.share_access =
3206                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3207                                 calldata->arg.fmode, 0);
3208
3209         nfs_fattr_init(calldata->res.fattr);
3210         calldata->timestamp = jiffies;
3211         if (nfs4_setup_sequence(NFS_SERVER(inode),
3212                                 &calldata->arg.seq_args,
3213                                 &calldata->res.seq_res,
3214                                 task) != 0)
3215                 nfs_release_seqid(calldata->arg.seqid);
3216         dprintk("%s: done!\n", __func__);
3217         return;
3218 out_no_action:
3219         task->tk_action = NULL;
3220 out_wait:
3221         nfs4_sequence_done(task, &calldata->res.seq_res);
3222 }
3223
3224 static const struct rpc_call_ops nfs4_close_ops = {
3225         .rpc_call_prepare = nfs4_close_prepare,
3226         .rpc_call_done = nfs4_close_done,
3227         .rpc_release = nfs4_free_closedata,
3228 };
3229
3230 /*
3231  * It is possible for data to be read/written from a mem-mapped file
3232  * after the sys_close call (which hits the vfs layer as a flush).
3233  * This means that we can't safely call nfsv4 close on a file until
3234  * the inode is cleared. This in turn means that we are not good
3235  * NFSv4 citizens - we do not indicate to the server to update the file's
3236  * share state even when we are done with one of the three share
3237  * stateid's in the inode.
3238  *
3239  * NOTE: Caller must be holding the sp->so_owner semaphore!
3240  */
3241 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3242 {
3243         struct nfs_server *server = NFS_SERVER(state->inode);
3244         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3245         struct nfs4_closedata *calldata;
3246         struct nfs4_state_owner *sp = state->owner;
3247         struct rpc_task *task;
3248         struct rpc_message msg = {
3249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3250                 .rpc_cred = state->owner->so_cred,
3251         };
3252         struct rpc_task_setup task_setup_data = {
3253                 .rpc_client = server->client,
3254                 .rpc_message = &msg,
3255                 .callback_ops = &nfs4_close_ops,
3256                 .workqueue = nfsiod_workqueue,
3257                 .flags = RPC_TASK_ASYNC,
3258         };
3259         int status = -ENOMEM;
3260
3261         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3262                 &task_setup_data.rpc_client, &msg);
3263
3264         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3265         if (calldata == NULL)
3266                 goto out;
3267         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3268         calldata->inode = state->inode;
3269         calldata->state = state;
3270         calldata->arg.fh = NFS_FH(state->inode);
3271         /* Serialization for the sequence id */
3272         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3273         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3274         if (IS_ERR(calldata->arg.seqid))
3275                 goto out_free_calldata;
3276         calldata->arg.fmode = 0;
3277         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3278         calldata->res.fattr = &calldata->fattr;
3279         calldata->res.seqid = calldata->arg.seqid;
3280         calldata->res.server = server;
3281         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3282         calldata->lr.roc = pnfs_roc(state->inode,
3283                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3284         if (calldata->lr.roc) {
3285                 calldata->arg.lr_args = &calldata->lr.arg;
3286                 calldata->res.lr_res = &calldata->lr.res;
3287         }
3288         nfs_sb_active(calldata->inode->i_sb);
3289
3290         msg.rpc_argp = &calldata->arg;
3291         msg.rpc_resp = &calldata->res;
3292         task_setup_data.callback_data = calldata;
3293         task = rpc_run_task(&task_setup_data);
3294         if (IS_ERR(task))
3295                 return PTR_ERR(task);
3296         status = 0;
3297         if (wait)
3298                 status = rpc_wait_for_completion_task(task);
3299         rpc_put_task(task);
3300         return status;
3301 out_free_calldata:
3302         kfree(calldata);
3303 out:
3304         nfs4_put_open_state(state);
3305         nfs4_put_state_owner(sp);
3306         return status;
3307 }
3308
3309 static struct inode *
3310 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3311                 int open_flags, struct iattr *attr, int *opened)
3312 {
3313         struct nfs4_state *state;
3314         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3315
3316         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3317
3318         /* Protect against concurrent sillydeletes */
3319         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3320
3321         nfs4_label_release_security(label);
3322
3323         if (IS_ERR(state))
3324                 return ERR_CAST(state);
3325         return state->inode;
3326 }
3327
3328 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3329 {
3330         if (ctx->state == NULL)
3331                 return;
3332         if (is_sync)
3333                 nfs4_close_sync(ctx->state, ctx->mode);
3334         else
3335                 nfs4_close_state(ctx->state, ctx->mode);
3336 }
3337
3338 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3339 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3340 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3341
3342 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3343 {
3344         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3345         struct nfs4_server_caps_arg args = {
3346                 .fhandle = fhandle,
3347                 .bitmask = bitmask,
3348         };
3349         struct nfs4_server_caps_res res = {};
3350         struct rpc_message msg = {
3351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3352                 .rpc_argp = &args,
3353                 .rpc_resp = &res,
3354         };
3355         int status;
3356
3357         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3358                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3359                      FATTR4_WORD0_LINK_SUPPORT |
3360                      FATTR4_WORD0_SYMLINK_SUPPORT |
3361                      FATTR4_WORD0_ACLSUPPORT;
3362         if (minorversion)
3363                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3364
3365         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3366         if (status == 0) {
3367                 /* Sanity check the server answers */
3368                 switch (minorversion) {
3369                 case 0:
3370                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3371                         res.attr_bitmask[2] = 0;
3372                         break;
3373                 case 1:
3374                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3375                         break;
3376                 case 2:
3377                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3378                 }
3379                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3380                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3381                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3382                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3383                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3384                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3385                                 NFS_CAP_SECURITY_LABEL);
3386                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3387                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3388                         server->caps |= NFS_CAP_ACLS;
3389                 if (res.has_links != 0)
3390                         server->caps |= NFS_CAP_HARDLINKS;
3391                 if (res.has_symlinks != 0)
3392                         server->caps |= NFS_CAP_SYMLINKS;
3393                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3394                         server->caps |= NFS_CAP_FILEID;
3395                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3396                         server->caps |= NFS_CAP_MODE;
3397                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3398                         server->caps |= NFS_CAP_NLINK;
3399                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3400                         server->caps |= NFS_CAP_OWNER;
3401                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3402                         server->caps |= NFS_CAP_OWNER_GROUP;
3403                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3404                         server->caps |= NFS_CAP_ATIME;
3405                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3406                         server->caps |= NFS_CAP_CTIME;
3407                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3408                         server->caps |= NFS_CAP_MTIME;
3409 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3410                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3411                         server->caps |= NFS_CAP_SECURITY_LABEL;
3412 #endif
3413                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3414                                 sizeof(server->attr_bitmask));
3415                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3416
3417                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3418                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3419                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3420                 server->cache_consistency_bitmask[2] = 0;
3421                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3422                         sizeof(server->exclcreat_bitmask));
3423                 server->acl_bitmask = res.acl_bitmask;
3424                 server->fh_expire_type = res.fh_expire_type;
3425         }
3426
3427         return status;
3428 }
3429
3430 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3431 {
3432         struct nfs4_exception exception = { };
3433         int err;
3434         do {
3435                 err = nfs4_handle_exception(server,
3436                                 _nfs4_server_capabilities(server, fhandle),
3437                                 &exception);
3438         } while (exception.retry);
3439         return err;
3440 }
3441
3442 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3443                 struct nfs_fsinfo *info)
3444 {
3445         u32 bitmask[3];
3446         struct nfs4_lookup_root_arg args = {
3447                 .bitmask = bitmask,
3448         };
3449         struct nfs4_lookup_res res = {
3450                 .server = server,
3451                 .fattr = info->fattr,
3452                 .fh = fhandle,
3453         };
3454         struct rpc_message msg = {
3455                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3456                 .rpc_argp = &args,
3457                 .rpc_resp = &res,
3458         };
3459
3460         bitmask[0] = nfs4_fattr_bitmap[0];
3461         bitmask[1] = nfs4_fattr_bitmap[1];
3462         /*
3463          * Process the label in the upcoming getfattr
3464          */
3465         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3466
3467         nfs_fattr_init(info->fattr);
3468         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3469 }
3470
3471 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3472                 struct nfs_fsinfo *info)
3473 {
3474         struct nfs4_exception exception = { };
3475         int err;
3476         do {
3477                 err = _nfs4_lookup_root(server, fhandle, info);
3478                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3479                 switch (err) {
3480                 case 0:
3481                 case -NFS4ERR_WRONGSEC:
3482                         goto out;
3483                 default:
3484                         err = nfs4_handle_exception(server, err, &exception);
3485                 }
3486         } while (exception.retry);
3487 out:
3488         return err;
3489 }
3490
3491 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3492                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3493 {
3494         struct rpc_auth_create_args auth_args = {
3495                 .pseudoflavor = flavor,
3496         };
3497         struct rpc_auth *auth;
3498         int ret;
3499
3500         auth = rpcauth_create(&auth_args, server->client);
3501         if (IS_ERR(auth)) {
3502                 ret = -EACCES;
3503                 goto out;
3504         }
3505         ret = nfs4_lookup_root(server, fhandle, info);
3506 out:
3507         return ret;
3508 }
3509
3510 /*
3511  * Retry pseudoroot lookup with various security flavors.  We do this when:
3512  *
3513  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3514  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3515  *
3516  * Returns zero on success, or a negative NFS4ERR value, or a
3517  * negative errno value.
3518  */
3519 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3520                               struct nfs_fsinfo *info)
3521 {
3522         /* Per 3530bis 15.33.5 */
3523         static const rpc_authflavor_t flav_array[] = {
3524                 RPC_AUTH_GSS_KRB5P,
3525                 RPC_AUTH_GSS_KRB5I,
3526                 RPC_AUTH_GSS_KRB5,
3527                 RPC_AUTH_UNIX,                  /* courtesy */
3528                 RPC_AUTH_NULL,
3529         };
3530         int status = -EPERM;
3531         size_t i;
3532
3533         if (server->auth_info.flavor_len > 0) {
3534                 /* try each flavor specified by user */
3535                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3536                         status = nfs4_lookup_root_sec(server, fhandle, info,
3537                                                 server->auth_info.flavors[i]);
3538                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3539                                 continue;
3540                         break;
3541                 }
3542         } else {
3543                 /* no flavors specified by user, try default list */
3544                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3545                         status = nfs4_lookup_root_sec(server, fhandle, info,
3546                                                       flav_array[i]);
3547                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3548                                 continue;
3549                         break;
3550                 }
3551         }
3552
3553         /*
3554          * -EACCESS could mean that the user doesn't have correct permissions
3555          * to access the mount.  It could also mean that we tried to mount
3556          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3557          * existing mount programs don't handle -EACCES very well so it should
3558          * be mapped to -EPERM instead.
3559          */
3560         if (status == -EACCES)
3561                 status = -EPERM;
3562         return status;
3563 }
3564
3565 /**
3566  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3567  * @server: initialized nfs_server handle
3568  * @fhandle: we fill in the pseudo-fs root file handle
3569  * @info: we fill in an FSINFO struct
3570  * @auth_probe: probe the auth flavours
3571  *
3572  * Returns zero on success, or a negative errno.
3573  */
3574 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3575                          struct nfs_fsinfo *info,
3576                          bool auth_probe)
3577 {
3578         int status = 0;
3579
3580         if (!auth_probe)
3581                 status = nfs4_lookup_root(server, fhandle, info);
3582
3583         if (auth_probe || status == NFS4ERR_WRONGSEC)
3584                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3585                                 fhandle, info);
3586
3587         if (status == 0)
3588                 status = nfs4_server_capabilities(server, fhandle);
3589         if (status == 0)
3590                 status = nfs4_do_fsinfo(server, fhandle, info);
3591
3592         return nfs4_map_errors(status);
3593 }
3594
3595 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3596                               struct nfs_fsinfo *info)
3597 {
3598         int error;
3599         struct nfs_fattr *fattr = info->fattr;
3600         struct nfs4_label *label = NULL;
3601
3602         error = nfs4_server_capabilities(server, mntfh);
3603         if (error < 0) {
3604                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3605                 return error;
3606         }
3607
3608         label = nfs4_label_alloc(server, GFP_KERNEL);
3609         if (IS_ERR(label))
3610                 return PTR_ERR(label);
3611
3612         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3613         if (error < 0) {
3614                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3615                 goto err_free_label;
3616         }
3617
3618         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3619             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3620                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3621
3622 err_free_label:
3623         nfs4_label_free(label);
3624
3625         return error;
3626 }
3627
3628 /*
3629  * Get locations and (maybe) other attributes of a referral.
3630  * Note that we'll actually follow the referral later when
3631  * we detect fsid mismatch in inode revalidation
3632  */
3633 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3634                              const struct qstr *name, struct nfs_fattr *fattr,
3635                              struct nfs_fh *fhandle)
3636 {
3637         int status = -ENOMEM;
3638         struct page *page = NULL;
3639         struct nfs4_fs_locations *locations = NULL;
3640
3641         page = alloc_page(GFP_KERNEL);
3642         if (page == NULL)
3643                 goto out;
3644         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3645         if (locations == NULL)
3646                 goto out;
3647
3648         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3649         if (status != 0)
3650                 goto out;
3651
3652         /*
3653          * If the fsid didn't change, this is a migration event, not a
3654          * referral.  Cause us to drop into the exception handler, which
3655          * will kick off migration recovery.
3656          */
3657         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3658                 dprintk("%s: server did not return a different fsid for"
3659                         " a referral at %s\n", __func__, name->name);
3660                 status = -NFS4ERR_MOVED;
3661                 goto out;
3662         }
3663         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3664         nfs_fixup_referral_attributes(&locations->fattr);
3665
3666         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3667         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3668         memset(fhandle, 0, sizeof(struct nfs_fh));
3669 out:
3670         if (page)
3671                 __free_page(page);
3672         kfree(locations);
3673         return status;
3674 }
3675
3676 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3677                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3678 {
3679         struct nfs4_getattr_arg args = {
3680                 .fh = fhandle,
3681                 .bitmask = server->attr_bitmask,
3682         };
3683         struct nfs4_getattr_res res = {
3684                 .fattr = fattr,
3685                 .label = label,
3686                 .server = server,
3687         };
3688         struct rpc_message msg = {
3689                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3690                 .rpc_argp = &args,
3691                 .rpc_resp = &res,
3692         };
3693
3694         args.bitmask = nfs4_bitmask(server, label);
3695
3696         nfs_fattr_init(fattr);
3697         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3698 }
3699
3700 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3701                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3702 {
3703         struct nfs4_exception exception = { };
3704         int err;
3705         do {
3706                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3707                 trace_nfs4_getattr(server, fhandle, fattr, err);
3708                 err = nfs4_handle_exception(server, err,
3709                                 &exception);
3710         } while (exception.retry);
3711         return err;
3712 }
3713
3714 /*
3715  * The file is not closed if it is opened due to the a request to change
3716  * the size of the file. The open call will not be needed once the
3717  * VFS layer lookup-intents are implemented.
3718  *
3719  * Close is called when the inode is destroyed.
3720  * If we haven't opened the file for O_WRONLY, we
3721  * need to in the size_change case to obtain a stateid.
3722  *
3723  * Got race?
3724  * Because OPEN is always done by name in nfsv4, it is
3725  * possible that we opened a different file by the same
3726  * name.  We can recognize this race condition, but we
3727  * can't do anything about it besides returning an error.
3728  *
3729  * This will be fixed with VFS changes (lookup-intent).
3730  */
3731 static int
3732 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3733                   struct iattr *sattr)
3734 {
3735         struct inode *inode = d_inode(dentry);
3736         struct rpc_cred *cred = NULL;
3737         struct nfs_open_context *ctx = NULL;
3738         struct nfs4_label *label = NULL;
3739         int status;
3740
3741         if (pnfs_ld_layoutret_on_setattr(inode) &&
3742             sattr->ia_valid & ATTR_SIZE &&
3743             sattr->ia_size < i_size_read(inode))
3744                 pnfs_commit_and_return_layout(inode);
3745
3746         nfs_fattr_init(fattr);
3747
3748         /* Deal with open(O_TRUNC) */
3749         if (sattr->ia_valid & ATTR_OPEN)
3750                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3751
3752         /* Optimization: if the end result is no change, don't RPC */
3753         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3754                 return 0;
3755
3756         /* Search for an existing open(O_WRITE) file */
3757         if (sattr->ia_valid & ATTR_FILE) {
3758
3759                 ctx = nfs_file_open_context(sattr->ia_file);
3760                 if (ctx)
3761                         cred = ctx->cred;
3762         }
3763
3764         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3765         if (IS_ERR(label))
3766                 return PTR_ERR(label);
3767
3768         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3769         if (status == 0) {
3770                 nfs_setattr_update_inode(inode, sattr, fattr);
3771                 nfs_setsecurity(inode, fattr, label);
3772         }
3773         nfs4_label_free(label);
3774         return status;
3775 }
3776
3777 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3778                 const struct qstr *name, struct nfs_fh *fhandle,
3779                 struct nfs_fattr *fattr, struct nfs4_label *label)
3780 {
3781         struct nfs_server *server = NFS_SERVER(dir);
3782         int                    status;
3783         struct nfs4_lookup_arg args = {
3784                 .bitmask = server->attr_bitmask,
3785                 .dir_fh = NFS_FH(dir),
3786                 .name = name,
3787         };
3788         struct nfs4_lookup_res res = {
3789                 .server = server,
3790                 .fattr = fattr,
3791                 .label = label,
3792                 .fh = fhandle,
3793         };
3794         struct rpc_message msg = {
3795                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3796                 .rpc_argp = &args,
3797                 .rpc_resp = &res,
3798         };
3799
3800         args.bitmask = nfs4_bitmask(server, label);
3801
3802         nfs_fattr_init(fattr);
3803
3804         dprintk("NFS call  lookup %s\n", name->name);
3805         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3806         dprintk("NFS reply lookup: %d\n", status);
3807         return status;
3808 }
3809
3810 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3811 {
3812         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3813                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3814         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3815         fattr->nlink = 2;
3816 }
3817
3818 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3819                                    const struct qstr *name, struct nfs_fh *fhandle,
3820                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3821 {
3822         struct nfs4_exception exception = { };
3823         struct rpc_clnt *client = *clnt;
3824         int err;
3825         do {
3826                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3827                 trace_nfs4_lookup(dir, name, err);
3828                 switch (err) {
3829                 case -NFS4ERR_BADNAME:
3830                         err = -ENOENT;
3831                         goto out;
3832                 case -NFS4ERR_MOVED:
3833                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3834                         if (err == -NFS4ERR_MOVED)
3835                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3836                         goto out;
3837                 case -NFS4ERR_WRONGSEC:
3838                         err = -EPERM;
3839                         if (client != *clnt)
3840                                 goto out;
3841                         client = nfs4_negotiate_security(client, dir, name);
3842                         if (IS_ERR(client))
3843                                 return PTR_ERR(client);
3844
3845                         exception.retry = 1;
3846                         break;
3847                 default:
3848                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3849                 }
3850         } while (exception.retry);
3851
3852 out:
3853         if (err == 0)
3854                 *clnt = client;
3855         else if (client != *clnt)
3856                 rpc_shutdown_client(client);
3857
3858         return err;
3859 }
3860
3861 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3862                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3863                             struct nfs4_label *label)
3864 {
3865         int status;
3866         struct rpc_clnt *client = NFS_CLIENT(dir);
3867
3868         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3869         if (client != NFS_CLIENT(dir)) {
3870                 rpc_shutdown_client(client);
3871                 nfs_fixup_secinfo_attributes(fattr);
3872         }
3873         return status;
3874 }
3875
3876 struct rpc_clnt *
3877 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3878                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3879 {
3880         struct rpc_clnt *client = NFS_CLIENT(dir);
3881         int status;
3882
3883         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3884         if (status < 0)
3885                 return ERR_PTR(status);
3886         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3887 }
3888
3889 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3890 {
3891         struct nfs_server *server = NFS_SERVER(inode);
3892         struct nfs4_accessargs args = {
3893                 .fh = NFS_FH(inode),
3894                 .bitmask = server->cache_consistency_bitmask,
3895         };
3896         struct nfs4_accessres res = {
3897                 .server = server,
3898         };
3899         struct rpc_message msg = {
3900                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3901                 .rpc_argp = &args,
3902                 .rpc_resp = &res,
3903                 .rpc_cred = entry->cred,
3904         };
3905         int mode = entry->mask;
3906         int status = 0;
3907
3908         /*
3909          * Determine which access bits we want to ask for...
3910          */
3911         if (mode & MAY_READ)
3912                 args.access |= NFS4_ACCESS_READ;
3913         if (S_ISDIR(inode->i_mode)) {
3914                 if (mode & MAY_WRITE)
3915                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3916                 if (mode & MAY_EXEC)
3917                         args.access |= NFS4_ACCESS_LOOKUP;
3918         } else {
3919                 if (mode & MAY_WRITE)
3920                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3921                 if (mode & MAY_EXEC)
3922                         args.access |= NFS4_ACCESS_EXECUTE;
3923         }
3924
3925         res.fattr = nfs_alloc_fattr();
3926         if (res.fattr == NULL)
3927                 return -ENOMEM;
3928
3929         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3930         if (!status) {
3931                 nfs_access_set_mask(entry, res.access);
3932                 nfs_refresh_inode(inode, res.fattr);
3933         }
3934         nfs_free_fattr(res.fattr);
3935         return status;
3936 }
3937
3938 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3939 {
3940         struct nfs4_exception exception = { };
3941         int err;
3942         do {
3943                 err = _nfs4_proc_access(inode, entry);
3944                 trace_nfs4_access(inode, err);
3945                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3946                                 &exception);
3947         } while (exception.retry);
3948         return err;
3949 }
3950
3951 /*
3952  * TODO: For the time being, we don't try to get any attributes
3953  * along with any of the zero-copy operations READ, READDIR,
3954  * READLINK, WRITE.
3955  *
3956  * In the case of the first three, we want to put the GETATTR
3957  * after the read-type operation -- this is because it is hard
3958  * to predict the length of a GETATTR response in v4, and thus
3959  * align the READ data correctly.  This means that the GETATTR
3960  * may end up partially falling into the page cache, and we should
3961  * shift it into the 'tail' of the xdr_buf before processing.
3962  * To do this efficiently, we need to know the total length
3963  * of data received, which doesn't seem to be available outside
3964  * of the RPC layer.
3965  *
3966  * In the case of WRITE, we also want to put the GETATTR after
3967  * the operation -- in this case because we want to make sure
3968  * we get the post-operation mtime and size.
3969  *
3970  * Both of these changes to the XDR layer would in fact be quite
3971  * minor, but I decided to leave them for a subsequent patch.
3972  */
3973 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3974                 unsigned int pgbase, unsigned int pglen)
3975 {
3976         struct nfs4_readlink args = {
3977                 .fh       = NFS_FH(inode),
3978                 .pgbase   = pgbase,
3979                 .pglen    = pglen,
3980                 .pages    = &page,
3981         };
3982         struct nfs4_readlink_res res;
3983         struct rpc_message msg = {
3984                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3985                 .rpc_argp = &args,
3986                 .rpc_resp = &res,
3987         };
3988
3989         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3990 }
3991
3992 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3993                 unsigned int pgbase, unsigned int pglen)
3994 {
3995         struct nfs4_exception exception = { };
3996         int err;
3997         do {
3998                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3999                 trace_nfs4_readlink(inode, err);
4000                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4001                                 &exception);
4002         } while (exception.retry);
4003         return err;
4004 }
4005
4006 /*
4007  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4008  */
4009 static int
4010 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4011                  int flags)
4012 {
4013         struct nfs4_label l, *ilabel = NULL;
4014         struct nfs_open_context *ctx;
4015         struct nfs4_state *state;
4016         int status = 0;
4017
4018         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4019         if (IS_ERR(ctx))
4020                 return PTR_ERR(ctx);
4021
4022         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4023
4024         sattr->ia_mode &= ~current_umask();
4025         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4026         if (IS_ERR(state)) {
4027                 status = PTR_ERR(state);
4028                 goto out;
4029         }
4030 out:
4031         nfs4_label_release_security(ilabel);
4032         put_nfs_open_context(ctx);
4033         return status;
4034 }
4035
4036 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4037 {
4038         struct nfs_server *server = NFS_SERVER(dir);
4039         struct nfs_removeargs args = {
4040                 .fh = NFS_FH(dir),
4041                 .name = *name,
4042         };
4043         struct nfs_removeres res = {
4044                 .server = server,
4045         };
4046         struct rpc_message msg = {
4047                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4048                 .rpc_argp = &args,
4049                 .rpc_resp = &res,
4050         };
4051         int status;
4052
4053         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4054         if (status == 0)
4055                 update_changeattr(dir, &res.cinfo);
4056         return status;
4057 }
4058
4059 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4060 {
4061         struct nfs4_exception exception = { };
4062         int err;
4063         do {
4064                 err = _nfs4_proc_remove(dir, name);
4065                 trace_nfs4_remove(dir, name, err);
4066                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4067                                 &exception);
4068         } while (exception.retry);
4069         return err;
4070 }
4071
4072 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4073 {
4074         struct nfs_server *server = NFS_SERVER(dir);
4075         struct nfs_removeargs *args = msg->rpc_argp;
4076         struct nfs_removeres *res = msg->rpc_resp;
4077
4078         res->server = server;
4079         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4080         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4081
4082         nfs_fattr_init(res->dir_attr);
4083 }
4084
4085 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4086 {
4087         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4088                         &data->args.seq_args,
4089                         &data->res.seq_res,
4090                         task);
4091 }
4092
4093 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4094 {
4095         struct nfs_unlinkdata *data = task->tk_calldata;
4096         struct nfs_removeres *res = &data->res;
4097
4098         if (!nfs4_sequence_done(task, &res->seq_res))
4099                 return 0;
4100         if (nfs4_async_handle_error(task, res->server, NULL,
4101                                     &data->timeout) == -EAGAIN)
4102                 return 0;
4103         update_changeattr(dir, &res->cinfo);
4104         return 1;
4105 }
4106
4107 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4108 {
4109         struct nfs_server *server = NFS_SERVER(dir);
4110         struct nfs_renameargs *arg = msg->rpc_argp;
4111         struct nfs_renameres *res = msg->rpc_resp;
4112
4113         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4114         res->server = server;
4115         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4116 }
4117
4118 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4119 {
4120         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4121                         &data->args.seq_args,
4122                         &data->res.seq_res,
4123                         task);
4124 }
4125
4126 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4127                                  struct inode *new_dir)
4128 {
4129         struct nfs_renamedata *data = task->tk_calldata;
4130         struct nfs_renameres *res = &data->res;
4131
4132         if (!nfs4_sequence_done(task, &res->seq_res))
4133                 return 0;
4134         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4135                 return 0;
4136
4137         update_changeattr(old_dir, &res->old_cinfo);
4138         update_changeattr(new_dir, &res->new_cinfo);
4139         return 1;
4140 }
4141
4142 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4143 {
4144         struct nfs_server *server = NFS_SERVER(inode);
4145         struct nfs4_link_arg arg = {
4146                 .fh     = NFS_FH(inode),
4147                 .dir_fh = NFS_FH(dir),
4148                 .name   = name,
4149                 .bitmask = server->attr_bitmask,
4150         };
4151         struct nfs4_link_res res = {
4152                 .server = server,
4153                 .label = NULL,
4154         };
4155         struct rpc_message msg = {
4156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4157                 .rpc_argp = &arg,
4158                 .rpc_resp = &res,
4159         };
4160         int status = -ENOMEM;
4161
4162         res.fattr = nfs_alloc_fattr();
4163         if (res.fattr == NULL)
4164                 goto out;
4165
4166         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4167         if (IS_ERR(res.label)) {
4168                 status = PTR_ERR(res.label);
4169                 goto out;
4170         }
4171         arg.bitmask = nfs4_bitmask(server, res.label);
4172
4173         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4174         if (!status) {
4175                 update_changeattr(dir, &res.cinfo);
4176                 status = nfs_post_op_update_inode(inode, res.fattr);
4177                 if (!status)
4178                         nfs_setsecurity(inode, res.fattr, res.label);
4179         }
4180
4181
4182         nfs4_label_free(res.label);
4183
4184 out:
4185         nfs_free_fattr(res.fattr);
4186         return status;
4187 }
4188
4189 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4190 {
4191         struct nfs4_exception exception = { };
4192         int err;
4193         do {
4194                 err = nfs4_handle_exception(NFS_SERVER(inode),
4195                                 _nfs4_proc_link(inode, dir, name),
4196                                 &exception);
4197         } while (exception.retry);
4198         return err;
4199 }
4200
4201 struct nfs4_createdata {
4202         struct rpc_message msg;
4203         struct nfs4_create_arg arg;
4204         struct nfs4_create_res res;
4205         struct nfs_fh fh;
4206         struct nfs_fattr fattr;
4207         struct nfs4_label *label;
4208 };
4209
4210 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4211                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4212 {
4213         struct nfs4_createdata *data;
4214
4215         data = kzalloc(sizeof(*data), GFP_KERNEL);
4216         if (data != NULL) {
4217                 struct nfs_server *server = NFS_SERVER(dir);
4218
4219                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4220                 if (IS_ERR(data->label))
4221                         goto out_free;
4222
4223                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4224                 data->msg.rpc_argp = &data->arg;
4225                 data->msg.rpc_resp = &data->res;
4226                 data->arg.dir_fh = NFS_FH(dir);
4227                 data->arg.server = server;
4228                 data->arg.name = name;
4229                 data->arg.attrs = sattr;
4230                 data->arg.ftype = ftype;
4231                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4232                 data->res.server = server;
4233                 data->res.fh = &data->fh;
4234                 data->res.fattr = &data->fattr;
4235                 data->res.label = data->label;
4236                 nfs_fattr_init(data->res.fattr);
4237         }
4238         return data;
4239 out_free:
4240         kfree(data);
4241         return NULL;
4242 }
4243
4244 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4245 {
4246         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4247                                     &data->arg.seq_args, &data->res.seq_res, 1);
4248         if (status == 0) {
4249                 update_changeattr(dir, &data->res.dir_cinfo);
4250                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4251         }
4252         return status;
4253 }
4254
4255 static void nfs4_free_createdata(struct nfs4_createdata *data)
4256 {
4257         nfs4_label_free(data->label);
4258         kfree(data);
4259 }
4260
4261 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4262                 struct page *page, unsigned int len, struct iattr *sattr,
4263                 struct nfs4_label *label)
4264 {
4265         struct nfs4_createdata *data;
4266         int status = -ENAMETOOLONG;
4267
4268         if (len > NFS4_MAXPATHLEN)
4269                 goto out;
4270
4271         status = -ENOMEM;
4272         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4273         if (data == NULL)
4274                 goto out;
4275
4276         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4277         data->arg.u.symlink.pages = &page;
4278         data->arg.u.symlink.len = len;
4279         data->arg.label = label;
4280
4281         status = nfs4_do_create(dir, dentry, data);
4282
4283         nfs4_free_createdata(data);
4284 out:
4285         return status;
4286 }
4287
4288 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4289                 struct page *page, unsigned int len, struct iattr *sattr)
4290 {
4291         struct nfs4_exception exception = { };
4292         struct nfs4_label l, *label = NULL;
4293         int err;
4294
4295         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4296
4297         do {
4298                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4299                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4300                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4301                                 &exception);
4302         } while (exception.retry);
4303
4304         nfs4_label_release_security(label);
4305         return err;
4306 }
4307
4308 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4309                 struct iattr *sattr, struct nfs4_label *label)
4310 {
4311         struct nfs4_createdata *data;
4312         int status = -ENOMEM;
4313
4314         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4315         if (data == NULL)
4316                 goto out;
4317
4318         data->arg.label = label;
4319         status = nfs4_do_create(dir, dentry, data);
4320
4321         nfs4_free_createdata(data);
4322 out:
4323         return status;
4324 }
4325
4326 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4327                 struct iattr *sattr)
4328 {
4329         struct nfs4_exception exception = { };
4330         struct nfs4_label l, *label = NULL;
4331         int err;
4332
4333         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4334
4335         sattr->ia_mode &= ~current_umask();
4336         do {
4337                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4338                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4339                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4340                                 &exception);
4341         } while (exception.retry);
4342         nfs4_label_release_security(label);
4343
4344         return err;
4345 }
4346
4347 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4348                 u64 cookie, struct page **pages, unsigned int count, int plus)
4349 {
4350         struct inode            *dir = d_inode(dentry);
4351         struct nfs4_readdir_arg args = {
4352                 .fh = NFS_FH(dir),
4353                 .pages = pages,
4354                 .pgbase = 0,
4355                 .count = count,
4356                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4357                 .plus = plus,
4358         };
4359         struct nfs4_readdir_res res;
4360         struct rpc_message msg = {
4361                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4362                 .rpc_argp = &args,
4363                 .rpc_resp = &res,
4364                 .rpc_cred = cred,
4365         };
4366         int                     status;
4367
4368         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4369                         dentry,
4370                         (unsigned long long)cookie);
4371         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4372         res.pgbase = args.pgbase;
4373         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4374         if (status >= 0) {
4375                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4376                 status += args.pgbase;
4377         }
4378
4379         nfs_invalidate_atime(dir);
4380
4381         dprintk("%s: returns %d\n", __func__, status);
4382         return status;
4383 }
4384
4385 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4386                 u64 cookie, struct page **pages, unsigned int count, int plus)
4387 {
4388         struct nfs4_exception exception = { };
4389         int err;
4390         do {
4391                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4392                                 pages, count, plus);
4393                 trace_nfs4_readdir(d_inode(dentry), err);
4394                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4395                                 &exception);
4396         } while (exception.retry);
4397         return err;
4398 }
4399
4400 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4401                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4402 {
4403         struct nfs4_createdata *data;
4404         int mode = sattr->ia_mode;
4405         int status = -ENOMEM;
4406
4407         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4408         if (data == NULL)
4409                 goto out;
4410
4411         if (S_ISFIFO(mode))
4412                 data->arg.ftype = NF4FIFO;
4413         else if (S_ISBLK(mode)) {
4414                 data->arg.ftype = NF4BLK;
4415                 data->arg.u.device.specdata1 = MAJOR(rdev);
4416                 data->arg.u.device.specdata2 = MINOR(rdev);
4417         }
4418         else if (S_ISCHR(mode)) {
4419                 data->arg.ftype = NF4CHR;
4420                 data->arg.u.device.specdata1 = MAJOR(rdev);
4421                 data->arg.u.device.specdata2 = MINOR(rdev);
4422         } else if (!S_ISSOCK(mode)) {
4423                 status = -EINVAL;
4424                 goto out_free;
4425         }
4426
4427         data->arg.label = label;
4428         status = nfs4_do_create(dir, dentry, data);
4429 out_free:
4430         nfs4_free_createdata(data);
4431 out:
4432         return status;
4433 }
4434
4435 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4436                 struct iattr *sattr, dev_t rdev)
4437 {
4438         struct nfs4_exception exception = { };
4439         struct nfs4_label l, *label = NULL;
4440         int err;
4441
4442         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4443
4444         sattr->ia_mode &= ~current_umask();
4445         do {
4446                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4447                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4448                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4449                                 &exception);
4450         } while (exception.retry);
4451
4452         nfs4_label_release_security(label);
4453
4454         return err;
4455 }
4456
4457 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4458                  struct nfs_fsstat *fsstat)
4459 {
4460         struct nfs4_statfs_arg args = {
4461                 .fh = fhandle,
4462                 .bitmask = server->attr_bitmask,
4463         };
4464         struct nfs4_statfs_res res = {
4465                 .fsstat = fsstat,
4466         };
4467         struct rpc_message msg = {
4468                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4469                 .rpc_argp = &args,
4470                 .rpc_resp = &res,
4471         };
4472
4473         nfs_fattr_init(fsstat->fattr);
4474         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4475 }
4476
4477 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4478 {
4479         struct nfs4_exception exception = { };
4480         int err;
4481         do {
4482                 err = nfs4_handle_exception(server,
4483                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4484                                 &exception);
4485         } while (exception.retry);
4486         return err;
4487 }
4488
4489 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4490                 struct nfs_fsinfo *fsinfo)
4491 {
4492         struct nfs4_fsinfo_arg args = {
4493                 .fh = fhandle,
4494                 .bitmask = server->attr_bitmask,
4495         };
4496         struct nfs4_fsinfo_res res = {
4497                 .fsinfo = fsinfo,
4498         };
4499         struct rpc_message msg = {
4500                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4501                 .rpc_argp = &args,
4502                 .rpc_resp = &res,
4503         };
4504
4505         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4506 }
4507
4508 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4509 {
4510         struct nfs4_exception exception = { };
4511         unsigned long now = jiffies;
4512         int err;
4513
4514         do {
4515                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4516                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4517                 if (err == 0) {
4518                         nfs4_set_lease_period(server->nfs_client,
4519                                         fsinfo->lease_time * HZ,
4520                                         now);
4521                         break;
4522                 }
4523                 err = nfs4_handle_exception(server, err, &exception);
4524         } while (exception.retry);
4525         return err;
4526 }
4527
4528 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4529 {
4530         int error;
4531
4532         nfs_fattr_init(fsinfo->fattr);
4533         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4534         if (error == 0) {
4535                 /* block layout checks this! */
4536                 server->pnfs_blksize = fsinfo->blksize;
4537                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4538         }
4539
4540         return error;
4541 }
4542
4543 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4544                 struct nfs_pathconf *pathconf)
4545 {
4546         struct nfs4_pathconf_arg args = {
4547                 .fh = fhandle,
4548                 .bitmask = server->attr_bitmask,
4549         };
4550         struct nfs4_pathconf_res res = {
4551                 .pathconf = pathconf,
4552         };
4553         struct rpc_message msg = {
4554                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4555                 .rpc_argp = &args,
4556                 .rpc_resp = &res,
4557         };
4558
4559         /* None of the pathconf attributes are mandatory to implement */
4560         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4561                 memset(pathconf, 0, sizeof(*pathconf));
4562                 return 0;
4563         }
4564
4565         nfs_fattr_init(pathconf->fattr);
4566         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4567 }
4568
4569 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4570                 struct nfs_pathconf *pathconf)
4571 {
4572         struct nfs4_exception exception = { };
4573         int err;
4574
4575         do {
4576                 err = nfs4_handle_exception(server,
4577                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4578                                 &exception);
4579         } while (exception.retry);
4580         return err;
4581 }
4582
4583 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4584                 const struct nfs_open_context *ctx,
4585                 const struct nfs_lock_context *l_ctx,
4586                 fmode_t fmode)
4587 {
4588         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4589 }
4590 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4591
4592 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4593                 const struct nfs_open_context *ctx,
4594                 const struct nfs_lock_context *l_ctx,
4595                 fmode_t fmode)
4596 {
4597         nfs4_stateid current_stateid;
4598
4599         /* If the current stateid represents a lost lock, then exit */
4600         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4601                 return true;
4602         return nfs4_stateid_match(stateid, &current_stateid);
4603 }
4604
4605 static bool nfs4_error_stateid_expired(int err)
4606 {
4607         switch (err) {
4608         case -NFS4ERR_DELEG_REVOKED:
4609         case -NFS4ERR_ADMIN_REVOKED:
4610         case -NFS4ERR_BAD_STATEID:
4611         case -NFS4ERR_STALE_STATEID:
4612         case -NFS4ERR_OLD_STATEID:
4613         case -NFS4ERR_OPENMODE:
4614         case -NFS4ERR_EXPIRED:
4615                 return true;
4616         }
4617         return false;
4618 }
4619
4620 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4621 {
4622         struct nfs_server *server = NFS_SERVER(hdr->inode);
4623
4624         trace_nfs4_read(hdr, task->tk_status);
4625         if (task->tk_status < 0) {
4626                 struct nfs4_exception exception = {
4627                         .inode = hdr->inode,
4628                         .state = hdr->args.context->state,
4629                         .stateid = &hdr->args.stateid,
4630                 };
4631                 task->tk_status = nfs4_async_handle_exception(task,
4632                                 server, task->tk_status, &exception);
4633                 if (exception.retry) {
4634                         rpc_restart_call_prepare(task);
4635                         return -EAGAIN;
4636                 }
4637         }
4638
4639         if (task->tk_status > 0)
4640                 renew_lease(server, hdr->timestamp);
4641         return 0;
4642 }
4643
4644 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4645                 struct nfs_pgio_args *args)
4646 {
4647
4648         if (!nfs4_error_stateid_expired(task->tk_status) ||
4649                 nfs4_stateid_is_current(&args->stateid,
4650                                 args->context,
4651                                 args->lock_context,
4652                                 FMODE_READ))
4653                 return false;
4654         rpc_restart_call_prepare(task);
4655         return true;
4656 }
4657
4658 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4659 {
4660
4661         dprintk("--> %s\n", __func__);
4662
4663         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4664                 return -EAGAIN;
4665         if (nfs4_read_stateid_changed(task, &hdr->args))
4666                 return -EAGAIN;
4667         if (task->tk_status > 0)
4668                 nfs_invalidate_atime(hdr->inode);
4669         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4670                                     nfs4_read_done_cb(task, hdr);
4671 }
4672
4673 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4674                                  struct rpc_message *msg)
4675 {
4676         hdr->timestamp   = jiffies;
4677         if (!hdr->pgio_done_cb)
4678                 hdr->pgio_done_cb = nfs4_read_done_cb;
4679         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4680         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4681 }
4682
4683 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4684                                       struct nfs_pgio_header *hdr)
4685 {
4686         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4687                         &hdr->args.seq_args,
4688                         &hdr->res.seq_res,
4689                         task))
4690                 return 0;
4691         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4692                                 hdr->args.lock_context,
4693                                 hdr->rw_ops->rw_mode) == -EIO)
4694                 return -EIO;
4695         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4696                 return -EIO;
4697         return 0;
4698 }
4699
4700 static int nfs4_write_done_cb(struct rpc_task *task,
4701                               struct nfs_pgio_header *hdr)
4702 {
4703         struct inode *inode = hdr->inode;
4704
4705         trace_nfs4_write(hdr, task->tk_status);
4706         if (task->tk_status < 0) {
4707                 struct nfs4_exception exception = {
4708                         .inode = hdr->inode,
4709                         .state = hdr->args.context->state,
4710                         .stateid = &hdr->args.stateid,
4711                 };
4712                 task->tk_status = nfs4_async_handle_exception(task,
4713                                 NFS_SERVER(inode), task->tk_status,
4714                                 &exception);
4715                 if (exception.retry) {
4716                         rpc_restart_call_prepare(task);
4717                         return -EAGAIN;
4718                 }
4719         }
4720         if (task->tk_status >= 0) {
4721                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4722                 nfs_writeback_update_inode(hdr);
4723         }
4724         return 0;
4725 }
4726
4727 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4728                 struct nfs_pgio_args *args)
4729 {
4730
4731         if (!nfs4_error_stateid_expired(task->tk_status) ||
4732                 nfs4_stateid_is_current(&args->stateid,
4733                                 args->context,
4734                                 args->lock_context,
4735                                 FMODE_WRITE))
4736                 return false;
4737         rpc_restart_call_prepare(task);
4738         return true;
4739 }
4740
4741 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4742 {
4743         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4744                 return -EAGAIN;
4745         if (nfs4_write_stateid_changed(task, &hdr->args))
4746                 return -EAGAIN;
4747         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4748                 nfs4_write_done_cb(task, hdr);
4749 }
4750
4751 static
4752 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4753 {
4754         /* Don't request attributes for pNFS or O_DIRECT writes */
4755         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4756                 return false;
4757         /* Otherwise, request attributes if and only if we don't hold
4758          * a delegation
4759          */
4760         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4761 }
4762
4763 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4764                                   struct rpc_message *msg)
4765 {
4766         struct nfs_server *server = NFS_SERVER(hdr->inode);
4767
4768         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4769                 hdr->args.bitmask = NULL;
4770                 hdr->res.fattr = NULL;
4771         } else
4772                 hdr->args.bitmask = server->cache_consistency_bitmask;
4773
4774         if (!hdr->pgio_done_cb)
4775                 hdr->pgio_done_cb = nfs4_write_done_cb;
4776         hdr->res.server = server;
4777         hdr->timestamp   = jiffies;
4778
4779         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4780         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4781 }
4782
4783 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4784 {
4785         nfs4_setup_sequence(NFS_SERVER(data->inode),
4786                         &data->args.seq_args,
4787                         &data->res.seq_res,
4788                         task);
4789 }
4790
4791 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4792 {
4793         struct inode *inode = data->inode;
4794
4795         trace_nfs4_commit(data, task->tk_status);
4796         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4797                                     NULL, NULL) == -EAGAIN) {
4798                 rpc_restart_call_prepare(task);
4799                 return -EAGAIN;
4800         }
4801         return 0;
4802 }
4803
4804 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4805 {
4806         if (!nfs4_sequence_done(task, &data->res.seq_res))
4807                 return -EAGAIN;
4808         return data->commit_done_cb(task, data);
4809 }
4810
4811 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4812 {
4813         struct nfs_server *server = NFS_SERVER(data->inode);
4814
4815         if (data->commit_done_cb == NULL)
4816                 data->commit_done_cb = nfs4_commit_done_cb;
4817         data->res.server = server;
4818         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4819         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4820 }
4821
4822 struct nfs4_renewdata {
4823         struct nfs_client       *client;
4824         unsigned long           timestamp;
4825 };
4826
4827 /*
4828  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4829  * standalone procedure for queueing an asynchronous RENEW.
4830  */
4831 static void nfs4_renew_release(void *calldata)
4832 {
4833         struct nfs4_renewdata *data = calldata;
4834         struct nfs_client *clp = data->client;
4835
4836         if (atomic_read(&clp->cl_count) > 1)
4837                 nfs4_schedule_state_renewal(clp);
4838         nfs_put_client(clp);
4839         kfree(data);
4840 }
4841
4842 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4843 {
4844         struct nfs4_renewdata *data = calldata;
4845         struct nfs_client *clp = data->client;
4846         unsigned long timestamp = data->timestamp;
4847
4848         trace_nfs4_renew_async(clp, task->tk_status);
4849         switch (task->tk_status) {
4850         case 0:
4851                 break;
4852         case -NFS4ERR_LEASE_MOVED:
4853                 nfs4_schedule_lease_moved_recovery(clp);
4854                 break;
4855         default:
4856                 /* Unless we're shutting down, schedule state recovery! */
4857                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4858                         return;
4859                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4860                         nfs4_schedule_lease_recovery(clp);
4861                         return;
4862                 }
4863                 nfs4_schedule_path_down_recovery(clp);
4864         }
4865         do_renew_lease(clp, timestamp);
4866 }
4867
4868 static const struct rpc_call_ops nfs4_renew_ops = {
4869         .rpc_call_done = nfs4_renew_done,
4870         .rpc_release = nfs4_renew_release,
4871 };
4872
4873 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4874 {
4875         struct rpc_message msg = {
4876                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4877                 .rpc_argp       = clp,
4878                 .rpc_cred       = cred,
4879         };
4880         struct nfs4_renewdata *data;
4881
4882         if (renew_flags == 0)
4883                 return 0;
4884         if (!atomic_inc_not_zero(&clp->cl_count))
4885                 return -EIO;
4886         data = kmalloc(sizeof(*data), GFP_NOFS);
4887         if (data == NULL)
4888                 return -ENOMEM;
4889         data->client = clp;
4890         data->timestamp = jiffies;
4891         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4892                         &nfs4_renew_ops, data);
4893 }
4894
4895 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4896 {
4897         struct rpc_message msg = {
4898                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4899                 .rpc_argp       = clp,
4900                 .rpc_cred       = cred,
4901         };
4902         unsigned long now = jiffies;
4903         int status;
4904
4905         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4906         if (status < 0)
4907                 return status;
4908         do_renew_lease(clp, now);
4909         return 0;
4910 }
4911
4912 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4913 {
4914         return server->caps & NFS_CAP_ACLS;
4915 }
4916
4917 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4918  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4919  * the stack.
4920  */
4921 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4922
4923 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4924                 struct page **pages)
4925 {
4926         struct page *newpage, **spages;
4927         int rc = 0;
4928         size_t len;
4929         spages = pages;
4930
4931         do {
4932                 len = min_t(size_t, PAGE_SIZE, buflen);
4933                 newpage = alloc_page(GFP_KERNEL);
4934
4935                 if (newpage == NULL)
4936                         goto unwind;
4937                 memcpy(page_address(newpage), buf, len);
4938                 buf += len;
4939                 buflen -= len;
4940                 *pages++ = newpage;
4941                 rc++;
4942         } while (buflen != 0);
4943
4944         return rc;
4945
4946 unwind:
4947         for(; rc > 0; rc--)
4948                 __free_page(spages[rc-1]);
4949         return -ENOMEM;
4950 }
4951
4952 struct nfs4_cached_acl {
4953         int cached;
4954         size_t len;
4955         char data[0];
4956 };
4957
4958 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4959 {
4960         struct nfs_inode *nfsi = NFS_I(inode);
4961
4962         spin_lock(&inode->i_lock);
4963         kfree(nfsi->nfs4_acl);
4964         nfsi->nfs4_acl = acl;
4965         spin_unlock(&inode->i_lock);
4966 }
4967
4968 static void nfs4_zap_acl_attr(struct inode *inode)
4969 {
4970         nfs4_set_cached_acl(inode, NULL);
4971 }
4972
4973 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4974 {
4975         struct nfs_inode *nfsi = NFS_I(inode);
4976         struct nfs4_cached_acl *acl;
4977         int ret = -ENOENT;
4978
4979         spin_lock(&inode->i_lock);
4980         acl = nfsi->nfs4_acl;
4981         if (acl == NULL)
4982                 goto out;
4983         if (buf == NULL) /* user is just asking for length */
4984                 goto out_len;
4985         if (acl->cached == 0)
4986                 goto out;
4987         ret = -ERANGE; /* see getxattr(2) man page */
4988         if (acl->len > buflen)
4989                 goto out;
4990         memcpy(buf, acl->data, acl->len);
4991 out_len:
4992         ret = acl->len;
4993 out:
4994         spin_unlock(&inode->i_lock);
4995         return ret;
4996 }
4997
4998 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4999 {
5000         struct nfs4_cached_acl *acl;
5001         size_t buflen = sizeof(*acl) + acl_len;
5002
5003         if (buflen <= PAGE_SIZE) {
5004                 acl = kmalloc(buflen, GFP_KERNEL);
5005                 if (acl == NULL)
5006                         goto out;
5007                 acl->cached = 1;
5008                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5009         } else {
5010                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5011                 if (acl == NULL)
5012                         goto out;
5013                 acl->cached = 0;
5014         }
5015         acl->len = acl_len;
5016 out:
5017         nfs4_set_cached_acl(inode, acl);
5018 }
5019
5020 /*
5021  * The getxattr API returns the required buffer length when called with a
5022  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5023  * the required buf.  On a NULL buf, we send a page of data to the server
5024  * guessing that the ACL request can be serviced by a page. If so, we cache
5025  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5026  * the cache. If not so, we throw away the page, and cache the required
5027  * length. The next getxattr call will then produce another round trip to
5028  * the server, this time with the input buf of the required size.
5029  */
5030 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5031 {
5032         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
5033         struct nfs_getaclargs args = {
5034                 .fh = NFS_FH(inode),
5035                 .acl_pages = pages,
5036                 .acl_len = buflen,
5037         };
5038         struct nfs_getaclres res = {
5039                 .acl_len = buflen,
5040         };
5041         struct rpc_message msg = {
5042                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5043                 .rpc_argp = &args,
5044                 .rpc_resp = &res,
5045         };
5046         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5047         int ret = -ENOMEM, i;
5048
5049         /* As long as we're doing a round trip to the server anyway,
5050          * let's be prepared for a page of acl data. */
5051         if (npages == 0)
5052                 npages = 1;
5053         if (npages > ARRAY_SIZE(pages))
5054                 return -ERANGE;
5055
5056         for (i = 0; i < npages; i++) {
5057                 pages[i] = alloc_page(GFP_KERNEL);
5058                 if (!pages[i])
5059                         goto out_free;
5060         }
5061
5062         /* for decoding across pages */
5063         res.acl_scratch = alloc_page(GFP_KERNEL);
5064         if (!res.acl_scratch)
5065                 goto out_free;
5066
5067         args.acl_len = npages * PAGE_SIZE;
5068
5069         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5070                 __func__, buf, buflen, npages, args.acl_len);
5071         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5072                              &msg, &args.seq_args, &res.seq_res, 0);
5073         if (ret)
5074                 goto out_free;
5075
5076         /* Handle the case where the passed-in buffer is too short */
5077         if (res.acl_flags & NFS4_ACL_TRUNC) {
5078                 /* Did the user only issue a request for the acl length? */
5079                 if (buf == NULL)
5080                         goto out_ok;
5081                 ret = -ERANGE;
5082                 goto out_free;
5083         }
5084         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5085         if (buf) {
5086                 if (res.acl_len > buflen) {
5087                         ret = -ERANGE;
5088                         goto out_free;
5089                 }
5090                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5091         }
5092 out_ok:
5093         ret = res.acl_len;
5094 out_free:
5095         for (i = 0; i < npages; i++)
5096                 if (pages[i])
5097                         __free_page(pages[i]);
5098         if (res.acl_scratch)
5099                 __free_page(res.acl_scratch);
5100         return ret;
5101 }
5102
5103 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5104 {
5105         struct nfs4_exception exception = { };
5106         ssize_t ret;
5107         do {
5108                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5109                 trace_nfs4_get_acl(inode, ret);
5110                 if (ret >= 0)
5111                         break;
5112                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5113         } while (exception.retry);
5114         return ret;
5115 }
5116
5117 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5118 {
5119         struct nfs_server *server = NFS_SERVER(inode);
5120         int ret;
5121
5122         if (!nfs4_server_supports_acls(server))
5123                 return -EOPNOTSUPP;
5124         ret = nfs_revalidate_inode(server, inode);
5125         if (ret < 0)
5126                 return ret;
5127         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5128                 nfs_zap_acl_cache(inode);
5129         ret = nfs4_read_cached_acl(inode, buf, buflen);
5130         if (ret != -ENOENT)
5131                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5132                  * but no cached acl data, just the acl length */
5133                 return ret;
5134         return nfs4_get_acl_uncached(inode, buf, buflen);
5135 }
5136
5137 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5138 {
5139         struct nfs_server *server = NFS_SERVER(inode);
5140         struct page *pages[NFS4ACL_MAXPAGES];
5141         struct nfs_setaclargs arg = {
5142                 .fh             = NFS_FH(inode),
5143                 .acl_pages      = pages,
5144                 .acl_len        = buflen,
5145         };
5146         struct nfs_setaclres res;
5147         struct rpc_message msg = {
5148                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5149                 .rpc_argp       = &arg,
5150                 .rpc_resp       = &res,
5151         };
5152         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5153         int ret, i;
5154
5155         if (!nfs4_server_supports_acls(server))
5156                 return -EOPNOTSUPP;
5157         if (npages > ARRAY_SIZE(pages))
5158                 return -ERANGE;
5159         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5160         if (i < 0)
5161                 return i;
5162         nfs4_inode_return_delegation(inode);
5163         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5164
5165         /*
5166          * Free each page after tx, so the only ref left is
5167          * held by the network stack
5168          */
5169         for (; i > 0; i--)
5170                 put_page(pages[i-1]);
5171
5172         /*
5173          * Acl update can result in inode attribute update.
5174          * so mark the attribute cache invalid.
5175          */
5176         spin_lock(&inode->i_lock);
5177         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5178         spin_unlock(&inode->i_lock);
5179         nfs_access_zap_cache(inode);
5180         nfs_zap_acl_cache(inode);
5181         return ret;
5182 }
5183
5184 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5185 {
5186         struct nfs4_exception exception = { };
5187         int err;
5188         do {
5189                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5190                 trace_nfs4_set_acl(inode, err);
5191                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5192                                 &exception);
5193         } while (exception.retry);
5194         return err;
5195 }
5196
5197 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5198 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5199                                         size_t buflen)
5200 {
5201         struct nfs_server *server = NFS_SERVER(inode);
5202         struct nfs_fattr fattr;
5203         struct nfs4_label label = {0, 0, buflen, buf};
5204
5205         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5206         struct nfs4_getattr_arg arg = {
5207                 .fh             = NFS_FH(inode),
5208                 .bitmask        = bitmask,
5209         };
5210         struct nfs4_getattr_res res = {
5211                 .fattr          = &fattr,
5212                 .label          = &label,
5213                 .server         = server,
5214         };
5215         struct rpc_message msg = {
5216                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5217                 .rpc_argp       = &arg,
5218                 .rpc_resp       = &res,
5219         };
5220         int ret;
5221
5222         nfs_fattr_init(&fattr);
5223
5224         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5225         if (ret)
5226                 return ret;
5227         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5228                 return -ENOENT;
5229         if (buflen < label.len)
5230                 return -ERANGE;
5231         return 0;
5232 }
5233
5234 static int nfs4_get_security_label(struct inode *inode, void *buf,
5235                                         size_t buflen)
5236 {
5237         struct nfs4_exception exception = { };
5238         int err;
5239
5240         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5241                 return -EOPNOTSUPP;
5242
5243         do {
5244                 err = _nfs4_get_security_label(inode, buf, buflen);
5245                 trace_nfs4_get_security_label(inode, err);
5246                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5247                                 &exception);
5248         } while (exception.retry);
5249         return err;
5250 }
5251
5252 static int _nfs4_do_set_security_label(struct inode *inode,
5253                 struct nfs4_label *ilabel,
5254                 struct nfs_fattr *fattr,
5255                 struct nfs4_label *olabel)
5256 {
5257
5258         struct iattr sattr = {0};
5259         struct nfs_server *server = NFS_SERVER(inode);
5260         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5261         struct nfs_setattrargs arg = {
5262                 .fh             = NFS_FH(inode),
5263                 .iap            = &sattr,
5264                 .server         = server,
5265                 .bitmask        = bitmask,
5266                 .label          = ilabel,
5267         };
5268         struct nfs_setattrres res = {
5269                 .fattr          = fattr,
5270                 .label          = olabel,
5271                 .server         = server,
5272         };
5273         struct rpc_message msg = {
5274                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5275                 .rpc_argp       = &arg,
5276                 .rpc_resp       = &res,
5277         };
5278         int status;
5279
5280         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5281
5282         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5283         if (status)
5284                 dprintk("%s failed: %d\n", __func__, status);
5285
5286         return status;
5287 }
5288
5289 static int nfs4_do_set_security_label(struct inode *inode,
5290                 struct nfs4_label *ilabel,
5291                 struct nfs_fattr *fattr,
5292                 struct nfs4_label *olabel)
5293 {
5294         struct nfs4_exception exception = { };
5295         int err;
5296
5297         do {
5298                 err = _nfs4_do_set_security_label(inode, ilabel,
5299                                 fattr, olabel);
5300                 trace_nfs4_set_security_label(inode, err);
5301                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5302                                 &exception);
5303         } while (exception.retry);
5304         return err;
5305 }
5306
5307 static int
5308 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5309 {
5310         struct nfs4_label ilabel, *olabel = NULL;
5311         struct nfs_fattr fattr;
5312         struct rpc_cred *cred;
5313         int status;
5314
5315         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5316                 return -EOPNOTSUPP;
5317
5318         nfs_fattr_init(&fattr);
5319
5320         ilabel.pi = 0;
5321         ilabel.lfs = 0;
5322         ilabel.label = (char *)buf;
5323         ilabel.len = buflen;
5324
5325         cred = rpc_lookup_cred();
5326         if (IS_ERR(cred))
5327                 return PTR_ERR(cred);
5328
5329         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5330         if (IS_ERR(olabel)) {
5331                 status = -PTR_ERR(olabel);
5332                 goto out;
5333         }
5334
5335         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5336         if (status == 0)
5337                 nfs_setsecurity(inode, &fattr, olabel);
5338
5339         nfs4_label_free(olabel);
5340 out:
5341         put_rpccred(cred);
5342         return status;
5343 }
5344 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5345
5346
5347 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5348                                     nfs4_verifier *bootverf)
5349 {
5350         __be32 verf[2];
5351
5352         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5353                 /* An impossible timestamp guarantees this value
5354                  * will never match a generated boot time. */
5355                 verf[0] = cpu_to_be32(U32_MAX);
5356                 verf[1] = cpu_to_be32(U32_MAX);
5357         } else {
5358                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5359                 u64 ns = ktime_to_ns(nn->boot_time);
5360
5361                 verf[0] = cpu_to_be32(ns >> 32);
5362                 verf[1] = cpu_to_be32(ns);
5363         }
5364         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5365 }
5366
5367 static int
5368 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5369 {
5370         size_t len;
5371         char *str;
5372
5373         if (clp->cl_owner_id != NULL)
5374                 return 0;
5375
5376         rcu_read_lock();
5377         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5378                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5379                 1 +
5380                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5381                 1;
5382         rcu_read_unlock();
5383
5384         if (len > NFS4_OPAQUE_LIMIT + 1)
5385                 return -EINVAL;
5386
5387         /*
5388          * Since this string is allocated at mount time, and held until the
5389          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5390          * about a memory-reclaim deadlock.
5391          */
5392         str = kmalloc(len, GFP_KERNEL);
5393         if (!str)
5394                 return -ENOMEM;
5395
5396         rcu_read_lock();
5397         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5398                         clp->cl_ipaddr,
5399                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5400                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5401         rcu_read_unlock();
5402
5403         clp->cl_owner_id = str;
5404         return 0;
5405 }
5406
5407 static int
5408 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5409 {
5410         size_t len;
5411         char *str;
5412
5413         len = 10 + 10 + 1 + 10 + 1 +
5414                 strlen(nfs4_client_id_uniquifier) + 1 +
5415                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5416
5417         if (len > NFS4_OPAQUE_LIMIT + 1)
5418                 return -EINVAL;
5419
5420         /*
5421          * Since this string is allocated at mount time, and held until the
5422          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5423          * about a memory-reclaim deadlock.
5424          */
5425         str = kmalloc(len, GFP_KERNEL);
5426         if (!str)
5427                 return -ENOMEM;
5428
5429         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5430                         clp->rpc_ops->version, clp->cl_minorversion,
5431                         nfs4_client_id_uniquifier,
5432                         clp->cl_rpcclient->cl_nodename);
5433         clp->cl_owner_id = str;
5434         return 0;
5435 }
5436
5437 static int
5438 nfs4_init_uniform_client_string(struct nfs_client *clp)
5439 {
5440         size_t len;
5441         char *str;
5442
5443         if (clp->cl_owner_id != NULL)
5444                 return 0;
5445
5446         if (nfs4_client_id_uniquifier[0] != '\0')
5447                 return nfs4_init_uniquifier_client_string(clp);
5448
5449         len = 10 + 10 + 1 + 10 + 1 +
5450                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5451
5452         if (len > NFS4_OPAQUE_LIMIT + 1)
5453                 return -EINVAL;
5454
5455         /*
5456          * Since this string is allocated at mount time, and held until the
5457          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5458          * about a memory-reclaim deadlock.
5459          */
5460         str = kmalloc(len, GFP_KERNEL);
5461         if (!str)
5462                 return -ENOMEM;
5463
5464         scnprintf(str, len, "Linux NFSv%u.%u %s",
5465                         clp->rpc_ops->version, clp->cl_minorversion,
5466                         clp->cl_rpcclient->cl_nodename);
5467         clp->cl_owner_id = str;
5468         return 0;
5469 }
5470
5471 /*
5472  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5473  * services.  Advertise one based on the address family of the
5474  * clientaddr.
5475  */
5476 static unsigned int
5477 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5478 {
5479         if (strchr(clp->cl_ipaddr, ':') != NULL)
5480                 return scnprintf(buf, len, "tcp6");
5481         else
5482                 return scnprintf(buf, len, "tcp");
5483 }
5484
5485 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5486 {
5487         struct nfs4_setclientid *sc = calldata;
5488
5489         if (task->tk_status == 0)
5490                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5491 }
5492
5493 static const struct rpc_call_ops nfs4_setclientid_ops = {
5494         .rpc_call_done = nfs4_setclientid_done,
5495 };
5496
5497 /**
5498  * nfs4_proc_setclientid - Negotiate client ID
5499  * @clp: state data structure
5500  * @program: RPC program for NFSv4 callback service
5501  * @port: IP port number for NFS4 callback service
5502  * @cred: RPC credential to use for this call
5503  * @res: where to place the result
5504  *
5505  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5506  */
5507 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5508                 unsigned short port, struct rpc_cred *cred,
5509                 struct nfs4_setclientid_res *res)
5510 {
5511         nfs4_verifier sc_verifier;
5512         struct nfs4_setclientid setclientid = {
5513                 .sc_verifier = &sc_verifier,
5514                 .sc_prog = program,
5515                 .sc_clnt = clp,
5516         };
5517         struct rpc_message msg = {
5518                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5519                 .rpc_argp = &setclientid,
5520                 .rpc_resp = res,
5521                 .rpc_cred = cred,
5522         };
5523         struct rpc_task *task;
5524         struct rpc_task_setup task_setup_data = {
5525                 .rpc_client = clp->cl_rpcclient,
5526                 .rpc_message = &msg,
5527                 .callback_ops = &nfs4_setclientid_ops,
5528                 .callback_data = &setclientid,
5529                 .flags = RPC_TASK_TIMEOUT,
5530         };
5531         int status;
5532
5533         /* nfs_client_id4 */
5534         nfs4_init_boot_verifier(clp, &sc_verifier);
5535
5536         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5537                 status = nfs4_init_uniform_client_string(clp);
5538         else
5539                 status = nfs4_init_nonuniform_client_string(clp);
5540
5541         if (status)
5542                 goto out;
5543
5544         /* cb_client4 */
5545         setclientid.sc_netid_len =
5546                                 nfs4_init_callback_netid(clp,
5547                                                 setclientid.sc_netid,
5548                                                 sizeof(setclientid.sc_netid));
5549         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5550                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5551                                 clp->cl_ipaddr, port >> 8, port & 255);
5552
5553         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5554                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5555                 clp->cl_owner_id);
5556         task = rpc_run_task(&task_setup_data);
5557         if (IS_ERR(task)) {
5558                 status = PTR_ERR(task);
5559                 goto out;
5560         }
5561         status = task->tk_status;
5562         if (setclientid.sc_cred) {
5563                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5564                 put_rpccred(setclientid.sc_cred);
5565         }
5566         rpc_put_task(task);
5567 out:
5568         trace_nfs4_setclientid(clp, status);
5569         dprintk("NFS reply setclientid: %d\n", status);
5570         return status;
5571 }
5572
5573 /**
5574  * nfs4_proc_setclientid_confirm - Confirm client ID
5575  * @clp: state data structure
5576  * @res: result of a previous SETCLIENTID
5577  * @cred: RPC credential to use for this call
5578  *
5579  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5580  */
5581 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5582                 struct nfs4_setclientid_res *arg,
5583                 struct rpc_cred *cred)
5584 {
5585         struct rpc_message msg = {
5586                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5587                 .rpc_argp = arg,
5588                 .rpc_cred = cred,
5589         };
5590         int status;
5591
5592         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5593                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5594                 clp->cl_clientid);
5595         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5596         trace_nfs4_setclientid_confirm(clp, status);
5597         dprintk("NFS reply setclientid_confirm: %d\n", status);
5598         return status;
5599 }
5600
5601 struct nfs4_delegreturndata {
5602         struct nfs4_delegreturnargs args;
5603         struct nfs4_delegreturnres res;
5604         struct nfs_fh fh;
5605         nfs4_stateid stateid;
5606         unsigned long timestamp;
5607         struct {
5608                 struct nfs4_layoutreturn_args arg;
5609                 struct nfs4_layoutreturn_res res;
5610                 struct nfs4_xdr_opaque_data ld_private;
5611                 u32 roc_barrier;
5612                 bool roc;
5613         } lr;
5614         struct nfs_fattr fattr;
5615         int rpc_status;
5616         struct inode *inode;
5617 };
5618
5619 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5620 {
5621         struct nfs4_delegreturndata *data = calldata;
5622
5623         if (!nfs4_sequence_done(task, &data->res.seq_res))
5624                 return;
5625
5626         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5627
5628         /* Handle Layoutreturn errors */
5629         if (data->args.lr_args && task->tk_status != 0) {
5630                 switch(data->res.lr_ret) {
5631                 default:
5632                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5633                         break;
5634                 case 0:
5635                         data->args.lr_args = NULL;
5636                         data->res.lr_res = NULL;
5637                         break;
5638                 case -NFS4ERR_ADMIN_REVOKED:
5639                 case -NFS4ERR_DELEG_REVOKED:
5640                 case -NFS4ERR_EXPIRED:
5641                 case -NFS4ERR_BAD_STATEID:
5642                 case -NFS4ERR_OLD_STATEID:
5643                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5644                 case -NFS4ERR_WRONG_CRED:
5645                         data->args.lr_args = NULL;
5646                         data->res.lr_res = NULL;
5647                         data->res.lr_ret = 0;
5648                         rpc_restart_call_prepare(task);
5649                         return;
5650                 }
5651         }
5652
5653         switch (task->tk_status) {
5654         case 0:
5655                 renew_lease(data->res.server, data->timestamp);
5656                 break;
5657         case -NFS4ERR_ADMIN_REVOKED:
5658         case -NFS4ERR_DELEG_REVOKED:
5659         case -NFS4ERR_EXPIRED:
5660                 nfs4_free_revoked_stateid(data->res.server,
5661                                 data->args.stateid,
5662                                 task->tk_msg.rpc_cred);
5663         case -NFS4ERR_BAD_STATEID:
5664         case -NFS4ERR_OLD_STATEID:
5665         case -NFS4ERR_STALE_STATEID:
5666                 task->tk_status = 0;
5667                 break;
5668         default:
5669                 if (nfs4_async_handle_error(task, data->res.server,
5670                                             NULL, NULL) == -EAGAIN) {
5671                         rpc_restart_call_prepare(task);
5672                         return;
5673                 }
5674         }
5675         data->rpc_status = task->tk_status;
5676 }
5677
5678 static void nfs4_delegreturn_release(void *calldata)
5679 {
5680         struct nfs4_delegreturndata *data = calldata;
5681         struct inode *inode = data->inode;
5682
5683         if (inode) {
5684                 if (data->lr.roc)
5685                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5686                                         data->res.lr_ret);
5687                 nfs_iput_and_deactive(inode);
5688         }
5689         kfree(calldata);
5690 }
5691
5692 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5693 {
5694         struct nfs4_delegreturndata *d_data;
5695
5696         d_data = (struct nfs4_delegreturndata *)data;
5697
5698         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5699                 return;
5700
5701         nfs4_setup_sequence(d_data->res.server,
5702                         &d_data->args.seq_args,
5703                         &d_data->res.seq_res,
5704                         task);
5705 }
5706
5707 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5708         .rpc_call_prepare = nfs4_delegreturn_prepare,
5709         .rpc_call_done = nfs4_delegreturn_done,
5710         .rpc_release = nfs4_delegreturn_release,
5711 };
5712
5713 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5714 {
5715         struct nfs4_delegreturndata *data;
5716         struct nfs_server *server = NFS_SERVER(inode);
5717         struct rpc_task *task;
5718         struct rpc_message msg = {
5719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5720                 .rpc_cred = cred,
5721         };
5722         struct rpc_task_setup task_setup_data = {
5723                 .rpc_client = server->client,
5724                 .rpc_message = &msg,
5725                 .callback_ops = &nfs4_delegreturn_ops,
5726                 .flags = RPC_TASK_ASYNC,
5727         };
5728         int status = 0;
5729
5730         data = kzalloc(sizeof(*data), GFP_NOFS);
5731         if (data == NULL)
5732                 return -ENOMEM;
5733         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5734
5735         nfs4_state_protect(server->nfs_client,
5736                         NFS_SP4_MACH_CRED_CLEANUP,
5737                         &task_setup_data.rpc_client, &msg);
5738
5739         data->args.fhandle = &data->fh;
5740         data->args.stateid = &data->stateid;
5741         data->args.bitmask = server->cache_consistency_bitmask;
5742         nfs_copy_fh(&data->fh, NFS_FH(inode));
5743         nfs4_stateid_copy(&data->stateid, stateid);
5744         data->res.fattr = &data->fattr;
5745         data->res.server = server;
5746         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5747         data->lr.arg.ld_private = &data->lr.ld_private;
5748         nfs_fattr_init(data->res.fattr);
5749         data->timestamp = jiffies;
5750         data->rpc_status = 0;
5751         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5752         data->inode = nfs_igrab_and_active(inode);
5753         if (data->inode) {
5754                 if (data->lr.roc) {
5755                         data->args.lr_args = &data->lr.arg;
5756                         data->res.lr_res = &data->lr.res;
5757                 }
5758         } else if (data->lr.roc) {
5759                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5760                 data->lr.roc = false;
5761         }
5762
5763         task_setup_data.callback_data = data;
5764         msg.rpc_argp = &data->args;
5765         msg.rpc_resp = &data->res;
5766         task = rpc_run_task(&task_setup_data);
5767         if (IS_ERR(task))
5768                 return PTR_ERR(task);
5769         if (!issync)
5770                 goto out;
5771         status = nfs4_wait_for_completion_rpc_task(task);
5772         if (status != 0)
5773                 goto out;
5774         status = data->rpc_status;
5775         if (status == 0)
5776                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5777         else
5778                 nfs_refresh_inode(inode, &data->fattr);
5779 out:
5780         rpc_put_task(task);
5781         return status;
5782 }
5783
5784 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5785 {
5786         struct nfs_server *server = NFS_SERVER(inode);
5787         struct nfs4_exception exception = { };
5788         int err;
5789         do {
5790                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5791                 trace_nfs4_delegreturn(inode, stateid, err);
5792                 switch (err) {
5793                         case -NFS4ERR_STALE_STATEID:
5794                         case -NFS4ERR_EXPIRED:
5795                         case 0:
5796                                 return 0;
5797                 }
5798                 err = nfs4_handle_exception(server, err, &exception);
5799         } while (exception.retry);
5800         return err;
5801 }
5802
5803 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5804 {
5805         struct inode *inode = state->inode;
5806         struct nfs_server *server = NFS_SERVER(inode);
5807         struct nfs_client *clp = server->nfs_client;
5808         struct nfs_lockt_args arg = {
5809                 .fh = NFS_FH(inode),
5810                 .fl = request,
5811         };
5812         struct nfs_lockt_res res = {
5813                 .denied = request,
5814         };
5815         struct rpc_message msg = {
5816                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5817                 .rpc_argp       = &arg,
5818                 .rpc_resp       = &res,
5819                 .rpc_cred       = state->owner->so_cred,
5820         };
5821         struct nfs4_lock_state *lsp;
5822         int status;
5823
5824         arg.lock_owner.clientid = clp->cl_clientid;
5825         status = nfs4_set_lock_state(state, request);
5826         if (status != 0)
5827                 goto out;
5828         lsp = request->fl_u.nfs4_fl.owner;
5829         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5830         arg.lock_owner.s_dev = server->s_dev;
5831         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5832         switch (status) {
5833                 case 0:
5834                         request->fl_type = F_UNLCK;
5835                         break;
5836                 case -NFS4ERR_DENIED:
5837                         status = 0;
5838         }
5839         request->fl_ops->fl_release_private(request);
5840         request->fl_ops = NULL;
5841 out:
5842         return status;
5843 }
5844
5845 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5846 {
5847         struct nfs4_exception exception = { };
5848         int err;
5849
5850         do {
5851                 err = _nfs4_proc_getlk(state, cmd, request);
5852                 trace_nfs4_get_lock(request, state, cmd, err);
5853                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5854                                 &exception);
5855         } while (exception.retry);
5856         return err;
5857 }
5858
5859 struct nfs4_unlockdata {
5860         struct nfs_locku_args arg;
5861         struct nfs_locku_res res;
5862         struct nfs4_lock_state *lsp;
5863         struct nfs_open_context *ctx;
5864         struct file_lock fl;
5865         struct nfs_server *server;
5866         unsigned long timestamp;
5867 };
5868
5869 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5870                 struct nfs_open_context *ctx,
5871                 struct nfs4_lock_state *lsp,
5872                 struct nfs_seqid *seqid)
5873 {
5874         struct nfs4_unlockdata *p;
5875         struct inode *inode = lsp->ls_state->inode;
5876
5877         p = kzalloc(sizeof(*p), GFP_NOFS);
5878         if (p == NULL)
5879                 return NULL;
5880         p->arg.fh = NFS_FH(inode);
5881         p->arg.fl = &p->fl;
5882         p->arg.seqid = seqid;
5883         p->res.seqid = seqid;
5884         p->lsp = lsp;
5885         atomic_inc(&lsp->ls_count);
5886         /* Ensure we don't close file until we're done freeing locks! */
5887         p->ctx = get_nfs_open_context(ctx);
5888         memcpy(&p->fl, fl, sizeof(p->fl));
5889         p->server = NFS_SERVER(inode);
5890         return p;
5891 }
5892
5893 static void nfs4_locku_release_calldata(void *data)
5894 {
5895         struct nfs4_unlockdata *calldata = data;
5896         nfs_free_seqid(calldata->arg.seqid);
5897         nfs4_put_lock_state(calldata->lsp);
5898         put_nfs_open_context(calldata->ctx);
5899         kfree(calldata);
5900 }
5901
5902 static void nfs4_locku_done(struct rpc_task *task, void *data)
5903 {
5904         struct nfs4_unlockdata *calldata = data;
5905
5906         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5907                 return;
5908         switch (task->tk_status) {
5909                 case 0:
5910                         renew_lease(calldata->server, calldata->timestamp);
5911                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5912                         if (nfs4_update_lock_stateid(calldata->lsp,
5913                                         &calldata->res.stateid))
5914                                 break;
5915                 case -NFS4ERR_ADMIN_REVOKED:
5916                 case -NFS4ERR_EXPIRED:
5917                         nfs4_free_revoked_stateid(calldata->server,
5918                                         &calldata->arg.stateid,
5919                                         task->tk_msg.rpc_cred);
5920                 case -NFS4ERR_BAD_STATEID:
5921                 case -NFS4ERR_OLD_STATEID:
5922                 case -NFS4ERR_STALE_STATEID:
5923                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5924                                                 &calldata->lsp->ls_stateid))
5925                                 rpc_restart_call_prepare(task);
5926                         break;
5927                 default:
5928                         if (nfs4_async_handle_error(task, calldata->server,
5929                                                     NULL, NULL) == -EAGAIN)
5930                                 rpc_restart_call_prepare(task);
5931         }
5932         nfs_release_seqid(calldata->arg.seqid);
5933 }
5934
5935 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5936 {
5937         struct nfs4_unlockdata *calldata = data;
5938
5939         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5940                 goto out_wait;
5941         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5942         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5943                 /* Note: exit _without_ running nfs4_locku_done */
5944                 goto out_no_action;
5945         }
5946         calldata->timestamp = jiffies;
5947         if (nfs4_setup_sequence(calldata->server,
5948                                 &calldata->arg.seq_args,
5949                                 &calldata->res.seq_res,
5950                                 task) != 0)
5951                 nfs_release_seqid(calldata->arg.seqid);
5952         return;
5953 out_no_action:
5954         task->tk_action = NULL;
5955 out_wait:
5956         nfs4_sequence_done(task, &calldata->res.seq_res);
5957 }
5958
5959 static const struct rpc_call_ops nfs4_locku_ops = {
5960         .rpc_call_prepare = nfs4_locku_prepare,
5961         .rpc_call_done = nfs4_locku_done,
5962         .rpc_release = nfs4_locku_release_calldata,
5963 };
5964
5965 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5966                 struct nfs_open_context *ctx,
5967                 struct nfs4_lock_state *lsp,
5968                 struct nfs_seqid *seqid)
5969 {
5970         struct nfs4_unlockdata *data;
5971         struct rpc_message msg = {
5972                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5973                 .rpc_cred = ctx->cred,
5974         };
5975         struct rpc_task_setup task_setup_data = {
5976                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5977                 .rpc_message = &msg,
5978                 .callback_ops = &nfs4_locku_ops,
5979                 .workqueue = nfsiod_workqueue,
5980                 .flags = RPC_TASK_ASYNC,
5981         };
5982
5983         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5984                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5985
5986         /* Ensure this is an unlock - when canceling a lock, the
5987          * canceled lock is passed in, and it won't be an unlock.
5988          */
5989         fl->fl_type = F_UNLCK;
5990
5991         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5992         if (data == NULL) {
5993                 nfs_free_seqid(seqid);
5994                 return ERR_PTR(-ENOMEM);
5995         }
5996
5997         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5998         msg.rpc_argp = &data->arg;
5999         msg.rpc_resp = &data->res;
6000         task_setup_data.callback_data = data;
6001         return rpc_run_task(&task_setup_data);
6002 }
6003
6004 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6005 {
6006         struct inode *inode = state->inode;
6007         struct nfs4_state_owner *sp = state->owner;
6008         struct nfs_inode *nfsi = NFS_I(inode);
6009         struct nfs_seqid *seqid;
6010         struct nfs4_lock_state *lsp;
6011         struct rpc_task *task;
6012         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6013         int status = 0;
6014         unsigned char fl_flags = request->fl_flags;
6015
6016         status = nfs4_set_lock_state(state, request);
6017         /* Unlock _before_ we do the RPC call */
6018         request->fl_flags |= FL_EXISTS;
6019         /* Exclude nfs_delegation_claim_locks() */
6020         mutex_lock(&sp->so_delegreturn_mutex);
6021         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6022         down_read(&nfsi->rwsem);
6023         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6024                 up_read(&nfsi->rwsem);
6025                 mutex_unlock(&sp->so_delegreturn_mutex);
6026                 goto out;
6027         }
6028         up_read(&nfsi->rwsem);
6029         mutex_unlock(&sp->so_delegreturn_mutex);
6030         if (status != 0)
6031                 goto out;
6032         /* Is this a delegated lock? */
6033         lsp = request->fl_u.nfs4_fl.owner;
6034         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6035                 goto out;
6036         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6037         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6038         status = -ENOMEM;
6039         if (IS_ERR(seqid))
6040                 goto out;
6041         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6042         status = PTR_ERR(task);
6043         if (IS_ERR(task))
6044                 goto out;
6045         status = nfs4_wait_for_completion_rpc_task(task);
6046         rpc_put_task(task);
6047 out:
6048         request->fl_flags = fl_flags;
6049         trace_nfs4_unlock(request, state, F_SETLK, status);
6050         return status;
6051 }
6052
6053 struct nfs4_lockdata {
6054         struct nfs_lock_args arg;
6055         struct nfs_lock_res res;
6056         struct nfs4_lock_state *lsp;
6057         struct nfs_open_context *ctx;
6058         struct file_lock fl;
6059         unsigned long timestamp;
6060         int rpc_status;
6061         int cancelled;
6062         struct nfs_server *server;
6063 };
6064
6065 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6066                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6067                 gfp_t gfp_mask)
6068 {
6069         struct nfs4_lockdata *p;
6070         struct inode *inode = lsp->ls_state->inode;
6071         struct nfs_server *server = NFS_SERVER(inode);
6072         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6073
6074         p = kzalloc(sizeof(*p), gfp_mask);
6075         if (p == NULL)
6076                 return NULL;
6077
6078         p->arg.fh = NFS_FH(inode);
6079         p->arg.fl = &p->fl;
6080         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6081         if (IS_ERR(p->arg.open_seqid))
6082                 goto out_free;
6083         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6084         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6085         if (IS_ERR(p->arg.lock_seqid))
6086                 goto out_free_seqid;
6087         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6088         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6089         p->arg.lock_owner.s_dev = server->s_dev;
6090         p->res.lock_seqid = p->arg.lock_seqid;
6091         p->lsp = lsp;
6092         p->server = server;
6093         atomic_inc(&lsp->ls_count);
6094         p->ctx = get_nfs_open_context(ctx);
6095         get_file(fl->fl_file);
6096         memcpy(&p->fl, fl, sizeof(p->fl));
6097         return p;
6098 out_free_seqid:
6099         nfs_free_seqid(p->arg.open_seqid);
6100 out_free:
6101         kfree(p);
6102         return NULL;
6103 }
6104
6105 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6106 {
6107         struct nfs4_lockdata *data = calldata;
6108         struct nfs4_state *state = data->lsp->ls_state;
6109
6110         dprintk("%s: begin!\n", __func__);
6111         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6112                 goto out_wait;
6113         /* Do we need to do an open_to_lock_owner? */
6114         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6115                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6116                         goto out_release_lock_seqid;
6117                 }
6118                 nfs4_stateid_copy(&data->arg.open_stateid,
6119                                 &state->open_stateid);
6120                 data->arg.new_lock_owner = 1;
6121                 data->res.open_seqid = data->arg.open_seqid;
6122         } else {
6123                 data->arg.new_lock_owner = 0;
6124                 nfs4_stateid_copy(&data->arg.lock_stateid,
6125                                 &data->lsp->ls_stateid);
6126         }
6127         if (!nfs4_valid_open_stateid(state)) {
6128                 data->rpc_status = -EBADF;
6129                 task->tk_action = NULL;
6130                 goto out_release_open_seqid;
6131         }
6132         data->timestamp = jiffies;
6133         if (nfs4_setup_sequence(data->server,
6134                                 &data->arg.seq_args,
6135                                 &data->res.seq_res,
6136                                 task) == 0)
6137                 return;
6138 out_release_open_seqid:
6139         nfs_release_seqid(data->arg.open_seqid);
6140 out_release_lock_seqid:
6141         nfs_release_seqid(data->arg.lock_seqid);
6142 out_wait:
6143         nfs4_sequence_done(task, &data->res.seq_res);
6144         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6145 }
6146
6147 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6148 {
6149         struct nfs4_lockdata *data = calldata;
6150         struct nfs4_lock_state *lsp = data->lsp;
6151
6152         dprintk("%s: begin!\n", __func__);
6153
6154         if (!nfs4_sequence_done(task, &data->res.seq_res))
6155                 return;
6156
6157         data->rpc_status = task->tk_status;
6158         switch (task->tk_status) {
6159         case 0:
6160                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6161                                 data->timestamp);
6162                 if (data->arg.new_lock) {
6163                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6164                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6165                                 rpc_restart_call_prepare(task);
6166                                 break;
6167                         }
6168                 }
6169                 if (data->arg.new_lock_owner != 0) {
6170                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6171                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6172                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6173                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6174                         rpc_restart_call_prepare(task);
6175                 break;
6176         case -NFS4ERR_BAD_STATEID:
6177         case -NFS4ERR_OLD_STATEID:
6178         case -NFS4ERR_STALE_STATEID:
6179         case -NFS4ERR_EXPIRED:
6180                 if (data->arg.new_lock_owner != 0) {
6181                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6182                                                 &lsp->ls_state->open_stateid))
6183                                 rpc_restart_call_prepare(task);
6184                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6185                                                 &lsp->ls_stateid))
6186                                 rpc_restart_call_prepare(task);
6187         }
6188         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6189 }
6190
6191 static void nfs4_lock_release(void *calldata)
6192 {
6193         struct nfs4_lockdata *data = calldata;
6194
6195         dprintk("%s: begin!\n", __func__);
6196         nfs_free_seqid(data->arg.open_seqid);
6197         if (data->cancelled != 0) {
6198                 struct rpc_task *task;
6199                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6200                                 data->arg.lock_seqid);
6201                 if (!IS_ERR(task))
6202                         rpc_put_task_async(task);
6203                 dprintk("%s: cancelling lock!\n", __func__);
6204         } else
6205                 nfs_free_seqid(data->arg.lock_seqid);
6206         nfs4_put_lock_state(data->lsp);
6207         put_nfs_open_context(data->ctx);
6208         fput(data->fl.fl_file);
6209         kfree(data);
6210         dprintk("%s: done!\n", __func__);
6211 }
6212
6213 static const struct rpc_call_ops nfs4_lock_ops = {
6214         .rpc_call_prepare = nfs4_lock_prepare,
6215         .rpc_call_done = nfs4_lock_done,
6216         .rpc_release = nfs4_lock_release,
6217 };
6218
6219 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6220 {
6221         switch (error) {
6222         case -NFS4ERR_ADMIN_REVOKED:
6223         case -NFS4ERR_EXPIRED:
6224         case -NFS4ERR_BAD_STATEID:
6225                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6226                 if (new_lock_owner != 0 ||
6227                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6228                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6229                 break;
6230         case -NFS4ERR_STALE_STATEID:
6231                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6232                 nfs4_schedule_lease_recovery(server->nfs_client);
6233         };
6234 }
6235
6236 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6237 {
6238         struct nfs4_lockdata *data;
6239         struct rpc_task *task;
6240         struct rpc_message msg = {
6241                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6242                 .rpc_cred = state->owner->so_cred,
6243         };
6244         struct rpc_task_setup task_setup_data = {
6245                 .rpc_client = NFS_CLIENT(state->inode),
6246                 .rpc_message = &msg,
6247                 .callback_ops = &nfs4_lock_ops,
6248                 .workqueue = nfsiod_workqueue,
6249                 .flags = RPC_TASK_ASYNC,
6250         };
6251         int ret;
6252
6253         dprintk("%s: begin!\n", __func__);
6254         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6255                         fl->fl_u.nfs4_fl.owner,
6256                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6257         if (data == NULL)
6258                 return -ENOMEM;
6259         if (IS_SETLKW(cmd))
6260                 data->arg.block = 1;
6261         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6262         msg.rpc_argp = &data->arg;
6263         msg.rpc_resp = &data->res;
6264         task_setup_data.callback_data = data;
6265         if (recovery_type > NFS_LOCK_NEW) {
6266                 if (recovery_type == NFS_LOCK_RECLAIM)
6267                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6268                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6269         } else
6270                 data->arg.new_lock = 1;
6271         task = rpc_run_task(&task_setup_data);
6272         if (IS_ERR(task))
6273                 return PTR_ERR(task);
6274         ret = nfs4_wait_for_completion_rpc_task(task);
6275         if (ret == 0) {
6276                 ret = data->rpc_status;
6277                 if (ret)
6278                         nfs4_handle_setlk_error(data->server, data->lsp,
6279                                         data->arg.new_lock_owner, ret);
6280         } else
6281                 data->cancelled = 1;
6282         rpc_put_task(task);
6283         dprintk("%s: done, ret = %d!\n", __func__, ret);
6284         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6285         return ret;
6286 }
6287
6288 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6289 {
6290         struct nfs_server *server = NFS_SERVER(state->inode);
6291         struct nfs4_exception exception = {
6292                 .inode = state->inode,
6293         };
6294         int err;
6295
6296         do {
6297                 /* Cache the lock if possible... */
6298                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6299                         return 0;
6300                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6301                 if (err != -NFS4ERR_DELAY)
6302                         break;
6303                 nfs4_handle_exception(server, err, &exception);
6304         } while (exception.retry);
6305         return err;
6306 }
6307
6308 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6309 {
6310         struct nfs_server *server = NFS_SERVER(state->inode);
6311         struct nfs4_exception exception = {
6312                 .inode = state->inode,
6313         };
6314         int err;
6315
6316         err = nfs4_set_lock_state(state, request);
6317         if (err != 0)
6318                 return err;
6319         if (!recover_lost_locks) {
6320                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6321                 return 0;
6322         }
6323         do {
6324                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6325                         return 0;
6326                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6327                 switch (err) {
6328                 default:
6329                         goto out;
6330                 case -NFS4ERR_GRACE:
6331                 case -NFS4ERR_DELAY:
6332                         nfs4_handle_exception(server, err, &exception);
6333                         err = 0;
6334                 }
6335         } while (exception.retry);
6336 out:
6337         return err;
6338 }
6339
6340 #if defined(CONFIG_NFS_V4_1)
6341 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6342 {
6343         struct nfs4_lock_state *lsp;
6344         int status;
6345
6346         status = nfs4_set_lock_state(state, request);
6347         if (status != 0)
6348                 return status;
6349         lsp = request->fl_u.nfs4_fl.owner;
6350         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6351             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6352                 return 0;
6353         status = nfs4_lock_expired(state, request);
6354         return status;
6355 }
6356 #endif
6357
6358 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6359 {
6360         struct nfs_inode *nfsi = NFS_I(state->inode);
6361         struct nfs4_state_owner *sp = state->owner;
6362         unsigned char fl_flags = request->fl_flags;
6363         int status;
6364
6365         request->fl_flags |= FL_ACCESS;
6366         status = locks_lock_inode_wait(state->inode, request);
6367         if (status < 0)
6368                 goto out;
6369         mutex_lock(&sp->so_delegreturn_mutex);
6370         down_read(&nfsi->rwsem);
6371         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6372                 /* Yes: cache locks! */
6373                 /* ...but avoid races with delegation recall... */
6374                 request->fl_flags = fl_flags & ~FL_SLEEP;
6375                 status = locks_lock_inode_wait(state->inode, request);
6376                 up_read(&nfsi->rwsem);
6377                 mutex_unlock(&sp->so_delegreturn_mutex);
6378                 goto out;
6379         }
6380         up_read(&nfsi->rwsem);
6381         mutex_unlock(&sp->so_delegreturn_mutex);
6382         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6383 out:
6384         request->fl_flags = fl_flags;
6385         return status;
6386 }
6387
6388 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6389 {
6390         struct nfs4_exception exception = {
6391                 .state = state,
6392                 .inode = state->inode,
6393         };
6394         int err;
6395
6396         do {
6397                 err = _nfs4_proc_setlk(state, cmd, request);
6398                 if (err == -NFS4ERR_DENIED)
6399                         err = -EAGAIN;
6400                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6401                                 err, &exception);
6402         } while (exception.retry);
6403         return err;
6404 }
6405
6406 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6407 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6408
6409 static int
6410 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6411                         struct file_lock *request)
6412 {
6413         int             status = -ERESTARTSYS;
6414         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6415
6416         while(!signalled()) {
6417                 status = nfs4_proc_setlk(state, cmd, request);
6418                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6419                         break;
6420                 freezable_schedule_timeout_interruptible(timeout);
6421                 timeout *= 2;
6422                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6423                 status = -ERESTARTSYS;
6424         }
6425         return status;
6426 }
6427
6428 #ifdef CONFIG_NFS_V4_1
6429 struct nfs4_lock_waiter {
6430         struct task_struct      *task;
6431         struct inode            *inode;
6432         struct nfs_lowner       *owner;
6433         bool                    notified;
6434 };
6435
6436 static int
6437 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6438 {
6439         int ret;
6440         struct cb_notify_lock_args *cbnl = key;
6441         struct nfs4_lock_waiter *waiter = wait->private;
6442         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6443                                 *wowner = waiter->owner;
6444
6445         /* Only wake if the callback was for the same owner */
6446         if (lowner->clientid != wowner->clientid ||
6447             lowner->id != wowner->id             ||
6448             lowner->s_dev != wowner->s_dev)
6449                 return 0;
6450
6451         /* Make sure it's for the right inode */
6452         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6453                 return 0;
6454
6455         waiter->notified = true;
6456
6457         /* override "private" so we can use default_wake_function */
6458         wait->private = waiter->task;
6459         ret = autoremove_wake_function(wait, mode, flags, key);
6460         wait->private = waiter;
6461         return ret;
6462 }
6463
6464 static int
6465 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6466 {
6467         int status = -ERESTARTSYS;
6468         unsigned long flags;
6469         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6470         struct nfs_server *server = NFS_SERVER(state->inode);
6471         struct nfs_client *clp = server->nfs_client;
6472         wait_queue_head_t *q = &clp->cl_lock_waitq;
6473         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6474                                     .id = lsp->ls_seqid.owner_id,
6475                                     .s_dev = server->s_dev };
6476         struct nfs4_lock_waiter waiter = { .task  = current,
6477                                            .inode = state->inode,
6478                                            .owner = &owner,
6479                                            .notified = false };
6480         wait_queue_t wait;
6481
6482         /* Don't bother with waitqueue if we don't expect a callback */
6483         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6484                 return nfs4_retry_setlk_simple(state, cmd, request);
6485
6486         init_wait(&wait);
6487         wait.private = &waiter;
6488         wait.func = nfs4_wake_lock_waiter;
6489         add_wait_queue(q, &wait);
6490
6491         while(!signalled()) {
6492                 status = nfs4_proc_setlk(state, cmd, request);
6493                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6494                         break;
6495
6496                 status = -ERESTARTSYS;
6497                 spin_lock_irqsave(&q->lock, flags);
6498                 if (waiter.notified) {
6499                         spin_unlock_irqrestore(&q->lock, flags);
6500                         continue;
6501                 }
6502                 set_current_state(TASK_INTERRUPTIBLE);
6503                 spin_unlock_irqrestore(&q->lock, flags);
6504
6505                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6506         }
6507
6508         finish_wait(q, &wait);
6509         return status;
6510 }
6511 #else /* !CONFIG_NFS_V4_1 */
6512 static inline int
6513 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6514 {
6515         return nfs4_retry_setlk_simple(state, cmd, request);
6516 }
6517 #endif
6518
6519 static int
6520 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6521 {
6522         struct nfs_open_context *ctx;
6523         struct nfs4_state *state;
6524         int status;
6525
6526         /* verify open state */
6527         ctx = nfs_file_open_context(filp);
6528         state = ctx->state;
6529
6530         if (request->fl_start < 0 || request->fl_end < 0)
6531                 return -EINVAL;
6532
6533         if (IS_GETLK(cmd)) {
6534                 if (state != NULL)
6535                         return nfs4_proc_getlk(state, F_GETLK, request);
6536                 return 0;
6537         }
6538
6539         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6540                 return -EINVAL;
6541
6542         if (request->fl_type == F_UNLCK) {
6543                 if (state != NULL)
6544                         return nfs4_proc_unlck(state, cmd, request);
6545                 return 0;
6546         }
6547
6548         if (state == NULL)
6549                 return -ENOLCK;
6550
6551         if ((request->fl_flags & FL_POSIX) &&
6552             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6553                 return -ENOLCK;
6554
6555         /*
6556          * Don't rely on the VFS having checked the file open mode,
6557          * since it won't do this for flock() locks.
6558          */
6559         switch (request->fl_type) {
6560         case F_RDLCK:
6561                 if (!(filp->f_mode & FMODE_READ))
6562                         return -EBADF;
6563                 break;
6564         case F_WRLCK:
6565                 if (!(filp->f_mode & FMODE_WRITE))
6566                         return -EBADF;
6567         }
6568
6569         status = nfs4_set_lock_state(state, request);
6570         if (status != 0)
6571                 return status;
6572
6573         return nfs4_retry_setlk(state, cmd, request);
6574 }
6575
6576 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6577 {
6578         struct nfs_server *server = NFS_SERVER(state->inode);
6579         int err;
6580
6581         err = nfs4_set_lock_state(state, fl);
6582         if (err != 0)
6583                 return err;
6584         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6585         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6586 }
6587
6588 struct nfs_release_lockowner_data {
6589         struct nfs4_lock_state *lsp;
6590         struct nfs_server *server;
6591         struct nfs_release_lockowner_args args;
6592         struct nfs_release_lockowner_res res;
6593         unsigned long timestamp;
6594 };
6595
6596 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6597 {
6598         struct nfs_release_lockowner_data *data = calldata;
6599         struct nfs_server *server = data->server;
6600         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6601                              &data->args.seq_args, &data->res.seq_res, task);
6602         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6603         data->timestamp = jiffies;
6604 }
6605
6606 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6607 {
6608         struct nfs_release_lockowner_data *data = calldata;
6609         struct nfs_server *server = data->server;
6610
6611         nfs40_sequence_done(task, &data->res.seq_res);
6612
6613         switch (task->tk_status) {
6614         case 0:
6615                 renew_lease(server, data->timestamp);
6616                 break;
6617         case -NFS4ERR_STALE_CLIENTID:
6618         case -NFS4ERR_EXPIRED:
6619                 nfs4_schedule_lease_recovery(server->nfs_client);
6620                 break;
6621         case -NFS4ERR_LEASE_MOVED:
6622         case -NFS4ERR_DELAY:
6623                 if (nfs4_async_handle_error(task, server,
6624                                             NULL, NULL) == -EAGAIN)
6625                         rpc_restart_call_prepare(task);
6626         }
6627 }
6628
6629 static void nfs4_release_lockowner_release(void *calldata)
6630 {
6631         struct nfs_release_lockowner_data *data = calldata;
6632         nfs4_free_lock_state(data->server, data->lsp);
6633         kfree(calldata);
6634 }
6635
6636 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6637         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6638         .rpc_call_done = nfs4_release_lockowner_done,
6639         .rpc_release = nfs4_release_lockowner_release,
6640 };
6641
6642 static void
6643 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6644 {
6645         struct nfs_release_lockowner_data *data;
6646         struct rpc_message msg = {
6647                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6648         };
6649
6650         if (server->nfs_client->cl_mvops->minor_version != 0)
6651                 return;
6652
6653         data = kmalloc(sizeof(*data), GFP_NOFS);
6654         if (!data)
6655                 return;
6656         data->lsp = lsp;
6657         data->server = server;
6658         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6659         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6660         data->args.lock_owner.s_dev = server->s_dev;
6661
6662         msg.rpc_argp = &data->args;
6663         msg.rpc_resp = &data->res;
6664         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6665         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6666 }
6667
6668 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6669
6670 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6671                                    struct dentry *unused, struct inode *inode,
6672                                    const char *key, const void *buf,
6673                                    size_t buflen, int flags)
6674 {
6675         return nfs4_proc_set_acl(inode, buf, buflen);
6676 }
6677
6678 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6679                                    struct dentry *unused, struct inode *inode,
6680                                    const char *key, void *buf, size_t buflen)
6681 {
6682         return nfs4_proc_get_acl(inode, buf, buflen);
6683 }
6684
6685 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6686 {
6687         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6688 }
6689
6690 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6691
6692 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6693                                      struct dentry *unused, struct inode *inode,
6694                                      const char *key, const void *buf,
6695                                      size_t buflen, int flags)
6696 {
6697         if (security_ismaclabel(key))
6698                 return nfs4_set_security_label(inode, buf, buflen);
6699
6700         return -EOPNOTSUPP;
6701 }
6702
6703 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6704                                      struct dentry *unused, struct inode *inode,
6705                                      const char *key, void *buf, size_t buflen)
6706 {
6707         if (security_ismaclabel(key))
6708                 return nfs4_get_security_label(inode, buf, buflen);
6709         return -EOPNOTSUPP;
6710 }
6711
6712 static ssize_t
6713 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6714 {
6715         int len = 0;
6716
6717         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6718                 len = security_inode_listsecurity(inode, list, list_len);
6719                 if (list_len && len > list_len)
6720                         return -ERANGE;
6721         }
6722         return len;
6723 }
6724
6725 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6726         .prefix = XATTR_SECURITY_PREFIX,
6727         .get    = nfs4_xattr_get_nfs4_label,
6728         .set    = nfs4_xattr_set_nfs4_label,
6729 };
6730
6731 #else
6732
6733 static ssize_t
6734 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6735 {
6736         return 0;
6737 }
6738
6739 #endif
6740
6741 /*
6742  * nfs_fhget will use either the mounted_on_fileid or the fileid
6743  */
6744 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6745 {
6746         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6747                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6748               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6749               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6750                 return;
6751
6752         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6753                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6754         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6755         fattr->nlink = 2;
6756 }
6757
6758 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6759                                    const struct qstr *name,
6760                                    struct nfs4_fs_locations *fs_locations,
6761                                    struct page *page)
6762 {
6763         struct nfs_server *server = NFS_SERVER(dir);
6764         u32 bitmask[3] = {
6765                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6766         };
6767         struct nfs4_fs_locations_arg args = {
6768                 .dir_fh = NFS_FH(dir),
6769                 .name = name,
6770                 .page = page,
6771                 .bitmask = bitmask,
6772         };
6773         struct nfs4_fs_locations_res res = {
6774                 .fs_locations = fs_locations,
6775         };
6776         struct rpc_message msg = {
6777                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6778                 .rpc_argp = &args,
6779                 .rpc_resp = &res,
6780         };
6781         int status;
6782
6783         dprintk("%s: start\n", __func__);
6784
6785         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6786          * is not supported */
6787         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6788                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6789         else
6790                 bitmask[0] |= FATTR4_WORD0_FILEID;
6791
6792         nfs_fattr_init(&fs_locations->fattr);
6793         fs_locations->server = server;
6794         fs_locations->nlocations = 0;
6795         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6796         dprintk("%s: returned status = %d\n", __func__, status);
6797         return status;
6798 }
6799
6800 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6801                            const struct qstr *name,
6802                            struct nfs4_fs_locations *fs_locations,
6803                            struct page *page)
6804 {
6805         struct nfs4_exception exception = { };
6806         int err;
6807         do {
6808                 err = _nfs4_proc_fs_locations(client, dir, name,
6809                                 fs_locations, page);
6810                 trace_nfs4_get_fs_locations(dir, name, err);
6811                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6812                                 &exception);
6813         } while (exception.retry);
6814         return err;
6815 }
6816
6817 /*
6818  * This operation also signals the server that this client is
6819  * performing migration recovery.  The server can stop returning
6820  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6821  * appended to this compound to identify the client ID which is
6822  * performing recovery.
6823  */
6824 static int _nfs40_proc_get_locations(struct inode *inode,
6825                                      struct nfs4_fs_locations *locations,
6826                                      struct page *page, struct rpc_cred *cred)
6827 {
6828         struct nfs_server *server = NFS_SERVER(inode);
6829         struct rpc_clnt *clnt = server->client;
6830         u32 bitmask[2] = {
6831                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6832         };
6833         struct nfs4_fs_locations_arg args = {
6834                 .clientid       = server->nfs_client->cl_clientid,
6835                 .fh             = NFS_FH(inode),
6836                 .page           = page,
6837                 .bitmask        = bitmask,
6838                 .migration      = 1,            /* skip LOOKUP */
6839                 .renew          = 1,            /* append RENEW */
6840         };
6841         struct nfs4_fs_locations_res res = {
6842                 .fs_locations   = locations,
6843                 .migration      = 1,
6844                 .renew          = 1,
6845         };
6846         struct rpc_message msg = {
6847                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6848                 .rpc_argp       = &args,
6849                 .rpc_resp       = &res,
6850                 .rpc_cred       = cred,
6851         };
6852         unsigned long now = jiffies;
6853         int status;
6854
6855         nfs_fattr_init(&locations->fattr);
6856         locations->server = server;
6857         locations->nlocations = 0;
6858
6859         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6860         nfs4_set_sequence_privileged(&args.seq_args);
6861         status = nfs4_call_sync_sequence(clnt, server, &msg,
6862                                         &args.seq_args, &res.seq_res);
6863         if (status)
6864                 return status;
6865
6866         renew_lease(server, now);
6867         return 0;
6868 }
6869
6870 #ifdef CONFIG_NFS_V4_1
6871
6872 /*
6873  * This operation also signals the server that this client is
6874  * performing migration recovery.  The server can stop asserting
6875  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6876  * performing this operation is identified in the SEQUENCE
6877  * operation in this compound.
6878  *
6879  * When the client supports GETATTR(fs_locations_info), it can
6880  * be plumbed in here.
6881  */
6882 static int _nfs41_proc_get_locations(struct inode *inode,
6883                                      struct nfs4_fs_locations *locations,
6884                                      struct page *page, struct rpc_cred *cred)
6885 {
6886         struct nfs_server *server = NFS_SERVER(inode);
6887         struct rpc_clnt *clnt = server->client;
6888         u32 bitmask[2] = {
6889                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6890         };
6891         struct nfs4_fs_locations_arg args = {
6892                 .fh             = NFS_FH(inode),
6893                 .page           = page,
6894                 .bitmask        = bitmask,
6895                 .migration      = 1,            /* skip LOOKUP */
6896         };
6897         struct nfs4_fs_locations_res res = {
6898                 .fs_locations   = locations,
6899                 .migration      = 1,
6900         };
6901         struct rpc_message msg = {
6902                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6903                 .rpc_argp       = &args,
6904                 .rpc_resp       = &res,
6905                 .rpc_cred       = cred,
6906         };
6907         int status;
6908
6909         nfs_fattr_init(&locations->fattr);
6910         locations->server = server;
6911         locations->nlocations = 0;
6912
6913         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6914         nfs4_set_sequence_privileged(&args.seq_args);
6915         status = nfs4_call_sync_sequence(clnt, server, &msg,
6916                                         &args.seq_args, &res.seq_res);
6917         if (status == NFS4_OK &&
6918             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6919                 status = -NFS4ERR_LEASE_MOVED;
6920         return status;
6921 }
6922
6923 #endif  /* CONFIG_NFS_V4_1 */
6924
6925 /**
6926  * nfs4_proc_get_locations - discover locations for a migrated FSID
6927  * @inode: inode on FSID that is migrating
6928  * @locations: result of query
6929  * @page: buffer
6930  * @cred: credential to use for this operation
6931  *
6932  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6933  * operation failed, or a negative errno if a local error occurred.
6934  *
6935  * On success, "locations" is filled in, but if the server has
6936  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6937  * asserted.
6938  *
6939  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6940  * from this client that require migration recovery.
6941  */
6942 int nfs4_proc_get_locations(struct inode *inode,
6943                             struct nfs4_fs_locations *locations,
6944                             struct page *page, struct rpc_cred *cred)
6945 {
6946         struct nfs_server *server = NFS_SERVER(inode);
6947         struct nfs_client *clp = server->nfs_client;
6948         const struct nfs4_mig_recovery_ops *ops =
6949                                         clp->cl_mvops->mig_recovery_ops;
6950         struct nfs4_exception exception = { };
6951         int status;
6952
6953         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6954                 (unsigned long long)server->fsid.major,
6955                 (unsigned long long)server->fsid.minor,
6956                 clp->cl_hostname);
6957         nfs_display_fhandle(NFS_FH(inode), __func__);
6958
6959         do {
6960                 status = ops->get_locations(inode, locations, page, cred);
6961                 if (status != -NFS4ERR_DELAY)
6962                         break;
6963                 nfs4_handle_exception(server, status, &exception);
6964         } while (exception.retry);
6965         return status;
6966 }
6967
6968 /*
6969  * This operation also signals the server that this client is
6970  * performing "lease moved" recovery.  The server can stop
6971  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6972  * is appended to this compound to identify the client ID which is
6973  * performing recovery.
6974  */
6975 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6976 {
6977         struct nfs_server *server = NFS_SERVER(inode);
6978         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6979         struct rpc_clnt *clnt = server->client;
6980         struct nfs4_fsid_present_arg args = {
6981                 .fh             = NFS_FH(inode),
6982                 .clientid       = clp->cl_clientid,
6983                 .renew          = 1,            /* append RENEW */
6984         };
6985         struct nfs4_fsid_present_res res = {
6986                 .renew          = 1,
6987         };
6988         struct rpc_message msg = {
6989                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6990                 .rpc_argp       = &args,
6991                 .rpc_resp       = &res,
6992                 .rpc_cred       = cred,
6993         };
6994         unsigned long now = jiffies;
6995         int status;
6996
6997         res.fh = nfs_alloc_fhandle();
6998         if (res.fh == NULL)
6999                 return -ENOMEM;
7000
7001         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7002         nfs4_set_sequence_privileged(&args.seq_args);
7003         status = nfs4_call_sync_sequence(clnt, server, &msg,
7004                                                 &args.seq_args, &res.seq_res);
7005         nfs_free_fhandle(res.fh);
7006         if (status)
7007                 return status;
7008
7009         do_renew_lease(clp, now);
7010         return 0;
7011 }
7012
7013 #ifdef CONFIG_NFS_V4_1
7014
7015 /*
7016  * This operation also signals the server that this client is
7017  * performing "lease moved" recovery.  The server can stop asserting
7018  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7019  * this operation is identified in the SEQUENCE operation in this
7020  * compound.
7021  */
7022 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7023 {
7024         struct nfs_server *server = NFS_SERVER(inode);
7025         struct rpc_clnt *clnt = server->client;
7026         struct nfs4_fsid_present_arg args = {
7027                 .fh             = NFS_FH(inode),
7028         };
7029         struct nfs4_fsid_present_res res = {
7030         };
7031         struct rpc_message msg = {
7032                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7033                 .rpc_argp       = &args,
7034                 .rpc_resp       = &res,
7035                 .rpc_cred       = cred,
7036         };
7037         int status;
7038
7039         res.fh = nfs_alloc_fhandle();
7040         if (res.fh == NULL)
7041                 return -ENOMEM;
7042
7043         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7044         nfs4_set_sequence_privileged(&args.seq_args);
7045         status = nfs4_call_sync_sequence(clnt, server, &msg,
7046                                                 &args.seq_args, &res.seq_res);
7047         nfs_free_fhandle(res.fh);
7048         if (status == NFS4_OK &&
7049             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7050                 status = -NFS4ERR_LEASE_MOVED;
7051         return status;
7052 }
7053
7054 #endif  /* CONFIG_NFS_V4_1 */
7055
7056 /**
7057  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7058  * @inode: inode on FSID to check
7059  * @cred: credential to use for this operation
7060  *
7061  * Server indicates whether the FSID is present, moved, or not
7062  * recognized.  This operation is necessary to clear a LEASE_MOVED
7063  * condition for this client ID.
7064  *
7065  * Returns NFS4_OK if the FSID is present on this server,
7066  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7067  *  NFS4ERR code if some error occurred on the server, or a
7068  *  negative errno if a local failure occurred.
7069  */
7070 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7071 {
7072         struct nfs_server *server = NFS_SERVER(inode);
7073         struct nfs_client *clp = server->nfs_client;
7074         const struct nfs4_mig_recovery_ops *ops =
7075                                         clp->cl_mvops->mig_recovery_ops;
7076         struct nfs4_exception exception = { };
7077         int status;
7078
7079         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7080                 (unsigned long long)server->fsid.major,
7081                 (unsigned long long)server->fsid.minor,
7082                 clp->cl_hostname);
7083         nfs_display_fhandle(NFS_FH(inode), __func__);
7084
7085         do {
7086                 status = ops->fsid_present(inode, cred);
7087                 if (status != -NFS4ERR_DELAY)
7088                         break;
7089                 nfs4_handle_exception(server, status, &exception);
7090         } while (exception.retry);
7091         return status;
7092 }
7093
7094 /**
7095  * If 'use_integrity' is true and the state managment nfs_client
7096  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7097  * and the machine credential as per RFC3530bis and RFC5661 Security
7098  * Considerations sections. Otherwise, just use the user cred with the
7099  * filesystem's rpc_client.
7100  */
7101 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7102 {
7103         int status;
7104         struct nfs4_secinfo_arg args = {
7105                 .dir_fh = NFS_FH(dir),
7106                 .name   = name,
7107         };
7108         struct nfs4_secinfo_res res = {
7109                 .flavors     = flavors,
7110         };
7111         struct rpc_message msg = {
7112                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7113                 .rpc_argp = &args,
7114                 .rpc_resp = &res,
7115         };
7116         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7117         struct rpc_cred *cred = NULL;
7118
7119         if (use_integrity) {
7120                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7121                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7122                 msg.rpc_cred = cred;
7123         }
7124
7125         dprintk("NFS call  secinfo %s\n", name->name);
7126
7127         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7128                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7129
7130         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7131                                 &res.seq_res, 0);
7132         dprintk("NFS reply  secinfo: %d\n", status);
7133
7134         if (cred)
7135                 put_rpccred(cred);
7136
7137         return status;
7138 }
7139
7140 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7141                       struct nfs4_secinfo_flavors *flavors)
7142 {
7143         struct nfs4_exception exception = { };
7144         int err;
7145         do {
7146                 err = -NFS4ERR_WRONGSEC;
7147
7148                 /* try to use integrity protection with machine cred */
7149                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7150                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7151
7152                 /*
7153                  * if unable to use integrity protection, or SECINFO with
7154                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7155                  * disallowed by spec, but exists in deployed servers) use
7156                  * the current filesystem's rpc_client and the user cred.
7157                  */
7158                 if (err == -NFS4ERR_WRONGSEC)
7159                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7160
7161                 trace_nfs4_secinfo(dir, name, err);
7162                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7163                                 &exception);
7164         } while (exception.retry);
7165         return err;
7166 }
7167
7168 #ifdef CONFIG_NFS_V4_1
7169 /*
7170  * Check the exchange flags returned by the server for invalid flags, having
7171  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7172  * DS flags set.
7173  */
7174 static int nfs4_check_cl_exchange_flags(u32 flags)
7175 {
7176         if (flags & ~EXCHGID4_FLAG_MASK_R)
7177                 goto out_inval;
7178         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7179             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7180                 goto out_inval;
7181         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7182                 goto out_inval;
7183         return NFS_OK;
7184 out_inval:
7185         return -NFS4ERR_INVAL;
7186 }
7187
7188 static bool
7189 nfs41_same_server_scope(struct nfs41_server_scope *a,
7190                         struct nfs41_server_scope *b)
7191 {
7192         if (a->server_scope_sz == b->server_scope_sz &&
7193             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7194                 return true;
7195
7196         return false;
7197 }
7198
7199 static void
7200 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7201 {
7202 }
7203
7204 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7205         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7206 };
7207
7208 /*
7209  * nfs4_proc_bind_one_conn_to_session()
7210  *
7211  * The 4.1 client currently uses the same TCP connection for the
7212  * fore and backchannel.
7213  */
7214 static
7215 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7216                 struct rpc_xprt *xprt,
7217                 struct nfs_client *clp,
7218                 struct rpc_cred *cred)
7219 {
7220         int status;
7221         struct nfs41_bind_conn_to_session_args args = {
7222                 .client = clp,
7223                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7224         };
7225         struct nfs41_bind_conn_to_session_res res;
7226         struct rpc_message msg = {
7227                 .rpc_proc =
7228                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7229                 .rpc_argp = &args,
7230                 .rpc_resp = &res,
7231                 .rpc_cred = cred,
7232         };
7233         struct rpc_task_setup task_setup_data = {
7234                 .rpc_client = clnt,
7235                 .rpc_xprt = xprt,
7236                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7237                 .rpc_message = &msg,
7238                 .flags = RPC_TASK_TIMEOUT,
7239         };
7240         struct rpc_task *task;
7241
7242         dprintk("--> %s\n", __func__);
7243
7244         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7245         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7246                 args.dir = NFS4_CDFC4_FORE;
7247
7248         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7249         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7250                 args.dir = NFS4_CDFC4_FORE;
7251
7252         task = rpc_run_task(&task_setup_data);
7253         if (!IS_ERR(task)) {
7254                 status = task->tk_status;
7255                 rpc_put_task(task);
7256         } else
7257                 status = PTR_ERR(task);
7258         trace_nfs4_bind_conn_to_session(clp, status);
7259         if (status == 0) {
7260                 if (memcmp(res.sessionid.data,
7261                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7262                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7263                         status = -EIO;
7264                         goto out;
7265                 }
7266                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7267                         dprintk("NFS: %s: Unexpected direction from server\n",
7268                                 __func__);
7269                         status = -EIO;
7270                         goto out;
7271                 }
7272                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7273                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7274                                 __func__);
7275                         status = -EIO;
7276                         goto out;
7277                 }
7278         }
7279 out:
7280         dprintk("<-- %s status= %d\n", __func__, status);
7281         return status;
7282 }
7283
7284 struct rpc_bind_conn_calldata {
7285         struct nfs_client *clp;
7286         struct rpc_cred *cred;
7287 };
7288
7289 static int
7290 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7291                 struct rpc_xprt *xprt,
7292                 void *calldata)
7293 {
7294         struct rpc_bind_conn_calldata *p = calldata;
7295
7296         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7297 }
7298
7299 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7300 {
7301         struct rpc_bind_conn_calldata data = {
7302                 .clp = clp,
7303                 .cred = cred,
7304         };
7305         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7306                         nfs4_proc_bind_conn_to_session_callback, &data);
7307 }
7308
7309 /*
7310  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7311  * and operations we'd like to see to enable certain features in the allow map
7312  */
7313 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7314         .how = SP4_MACH_CRED,
7315         .enforce.u.words = {
7316                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7317                       1 << (OP_EXCHANGE_ID - 32) |
7318                       1 << (OP_CREATE_SESSION - 32) |
7319                       1 << (OP_DESTROY_SESSION - 32) |
7320                       1 << (OP_DESTROY_CLIENTID - 32)
7321         },
7322         .allow.u.words = {
7323                 [0] = 1 << (OP_CLOSE) |
7324                       1 << (OP_OPEN_DOWNGRADE) |
7325                       1 << (OP_LOCKU) |
7326                       1 << (OP_DELEGRETURN) |
7327                       1 << (OP_COMMIT),
7328                 [1] = 1 << (OP_SECINFO - 32) |
7329                       1 << (OP_SECINFO_NO_NAME - 32) |
7330                       1 << (OP_LAYOUTRETURN - 32) |
7331                       1 << (OP_TEST_STATEID - 32) |
7332                       1 << (OP_FREE_STATEID - 32) |
7333                       1 << (OP_WRITE - 32)
7334         }
7335 };
7336
7337 /*
7338  * Select the state protection mode for client `clp' given the server results
7339  * from exchange_id in `sp'.
7340  *
7341  * Returns 0 on success, negative errno otherwise.
7342  */
7343 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7344                                  struct nfs41_state_protection *sp)
7345 {
7346         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7347                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7348                       1 << (OP_EXCHANGE_ID - 32) |
7349                       1 << (OP_CREATE_SESSION - 32) |
7350                       1 << (OP_DESTROY_SESSION - 32) |
7351                       1 << (OP_DESTROY_CLIENTID - 32)
7352         };
7353         unsigned int i;
7354
7355         if (sp->how == SP4_MACH_CRED) {
7356                 /* Print state protect result */
7357                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7358                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7359                         if (test_bit(i, sp->enforce.u.longs))
7360                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7361                         if (test_bit(i, sp->allow.u.longs))
7362                                 dfprintk(MOUNT, "  allow op %d\n", i);
7363                 }
7364
7365                 /* make sure nothing is on enforce list that isn't supported */
7366                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7367                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7368                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7369                                 return -EINVAL;
7370                         }
7371                 }
7372
7373                 /*
7374                  * Minimal mode - state operations are allowed to use machine
7375                  * credential.  Note this already happens by default, so the
7376                  * client doesn't have to do anything more than the negotiation.
7377                  *
7378                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7379                  *       we're already using the machine cred for exchange_id
7380                  *       and will never use a different cred.
7381                  */
7382                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7383                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7384                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7385                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7386                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7387                         dfprintk(MOUNT, "  minimal mode enabled\n");
7388                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7389                 } else {
7390                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7391                         return -EINVAL;
7392                 }
7393
7394                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7395                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7396                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7397                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7398                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7399                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7400                 }
7401
7402                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7403                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7404                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7405                                 &clp->cl_sp4_flags);
7406                 }
7407
7408                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7409                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7410                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7411                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7412                 }
7413
7414                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7415                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7416                         dfprintk(MOUNT, "  stateid mode enabled\n");
7417                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7418                 }
7419
7420                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7421                         dfprintk(MOUNT, "  write mode enabled\n");
7422                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7423                 }
7424
7425                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7426                         dfprintk(MOUNT, "  commit mode enabled\n");
7427                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7428                 }
7429         }
7430
7431         return 0;
7432 }
7433
7434 struct nfs41_exchange_id_data {
7435         struct nfs41_exchange_id_res res;
7436         struct nfs41_exchange_id_args args;
7437         struct rpc_xprt *xprt;
7438         int rpc_status;
7439 };
7440
7441 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7442 {
7443         struct nfs41_exchange_id_data *cdata =
7444                                         (struct nfs41_exchange_id_data *)data;
7445         struct nfs_client *clp = cdata->args.client;
7446         int status = task->tk_status;
7447
7448         trace_nfs4_exchange_id(clp, status);
7449
7450         if (status == 0)
7451                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7452
7453         if (cdata->xprt && status == 0) {
7454                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7455                                                       cdata->xprt);
7456                 goto out;
7457         }
7458
7459         if (status  == 0)
7460                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7461
7462         if (status == 0) {
7463                 clp->cl_clientid = cdata->res.clientid;
7464                 clp->cl_exchange_flags = cdata->res.flags;
7465                 /* Client ID is not confirmed */
7466                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7467                         clear_bit(NFS4_SESSION_ESTABLISHED,
7468                         &clp->cl_session->session_state);
7469                         clp->cl_seqid = cdata->res.seqid;
7470                 }
7471
7472                 kfree(clp->cl_serverowner);
7473                 clp->cl_serverowner = cdata->res.server_owner;
7474                 cdata->res.server_owner = NULL;
7475
7476                 /* use the most recent implementation id */
7477                 kfree(clp->cl_implid);
7478                 clp->cl_implid = cdata->res.impl_id;
7479                 cdata->res.impl_id = NULL;
7480
7481                 if (clp->cl_serverscope != NULL &&
7482                     !nfs41_same_server_scope(clp->cl_serverscope,
7483                                         cdata->res.server_scope)) {
7484                         dprintk("%s: server_scope mismatch detected\n",
7485                                 __func__);
7486                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7487                         kfree(clp->cl_serverscope);
7488                         clp->cl_serverscope = NULL;
7489                 }
7490
7491                 if (clp->cl_serverscope == NULL) {
7492                         clp->cl_serverscope = cdata->res.server_scope;
7493                         cdata->res.server_scope = NULL;
7494                 }
7495                 /* Save the EXCHANGE_ID verifier session trunk tests */
7496                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7497                        sizeof(clp->cl_confirm.data));
7498         }
7499 out:
7500         cdata->rpc_status = status;
7501         return;
7502 }
7503
7504 static void nfs4_exchange_id_release(void *data)
7505 {
7506         struct nfs41_exchange_id_data *cdata =
7507                                         (struct nfs41_exchange_id_data *)data;
7508
7509         nfs_put_client(cdata->args.client);
7510         if (cdata->xprt) {
7511                 xprt_put(cdata->xprt);
7512                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7513         }
7514         kfree(cdata->res.impl_id);
7515         kfree(cdata->res.server_scope);
7516         kfree(cdata->res.server_owner);
7517         kfree(cdata);
7518 }
7519
7520 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7521         .rpc_call_done = nfs4_exchange_id_done,
7522         .rpc_release = nfs4_exchange_id_release,
7523 };
7524
7525 /*
7526  * _nfs4_proc_exchange_id()
7527  *
7528  * Wrapper for EXCHANGE_ID operation.
7529  */
7530 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7531                         u32 sp4_how, struct rpc_xprt *xprt)
7532 {
7533         nfs4_verifier verifier;
7534         struct rpc_message msg = {
7535                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7536                 .rpc_cred = cred,
7537         };
7538         struct rpc_task_setup task_setup_data = {
7539                 .rpc_client = clp->cl_rpcclient,
7540                 .callback_ops = &nfs4_exchange_id_call_ops,
7541                 .rpc_message = &msg,
7542                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7543         };
7544         struct nfs41_exchange_id_data *calldata;
7545         struct rpc_task *task;
7546         int status = -EIO;
7547
7548         if (!atomic_inc_not_zero(&clp->cl_count))
7549                 goto out;
7550
7551         status = -ENOMEM;
7552         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7553         if (!calldata)
7554                 goto out;
7555
7556         if (!xprt)
7557                 nfs4_init_boot_verifier(clp, &verifier);
7558
7559         status = nfs4_init_uniform_client_string(clp);
7560         if (status)
7561                 goto out_calldata;
7562
7563         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7564                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7565                 clp->cl_owner_id);
7566
7567         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7568                                                 GFP_NOFS);
7569         status = -ENOMEM;
7570         if (unlikely(calldata->res.server_owner == NULL))
7571                 goto out_calldata;
7572
7573         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7574                                         GFP_NOFS);
7575         if (unlikely(calldata->res.server_scope == NULL))
7576                 goto out_server_owner;
7577
7578         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7579         if (unlikely(calldata->res.impl_id == NULL))
7580                 goto out_server_scope;
7581
7582         switch (sp4_how) {
7583         case SP4_NONE:
7584                 calldata->args.state_protect.how = SP4_NONE;
7585                 break;
7586
7587         case SP4_MACH_CRED:
7588                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7589                 break;
7590
7591         default:
7592                 /* unsupported! */
7593                 WARN_ON_ONCE(1);
7594                 status = -EINVAL;
7595                 goto out_impl_id;
7596         }
7597         if (xprt) {
7598                 calldata->xprt = xprt;
7599                 task_setup_data.rpc_xprt = xprt;
7600                 task_setup_data.flags =
7601                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7602                 calldata->args.verifier = &clp->cl_confirm;
7603         } else {
7604                 calldata->args.verifier = &verifier;
7605         }
7606         calldata->args.client = clp;
7607 #ifdef CONFIG_NFS_V4_1_MIGRATION
7608         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7609         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7610         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7611 #else
7612         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7613         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7614 #endif
7615         msg.rpc_argp = &calldata->args;
7616         msg.rpc_resp = &calldata->res;
7617         task_setup_data.callback_data = calldata;
7618
7619         task = rpc_run_task(&task_setup_data);
7620         if (IS_ERR(task)) {
7621         status = PTR_ERR(task);
7622                 goto out_impl_id;
7623         }
7624
7625         if (!xprt) {
7626                 status = rpc_wait_for_completion_task(task);
7627                 if (!status)
7628                         status = calldata->rpc_status;
7629         } else  /* session trunking test */
7630                 status = calldata->rpc_status;
7631
7632         rpc_put_task(task);
7633 out:
7634         if (clp->cl_implid != NULL)
7635                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7636                         "domain: %s, name: %s, date: %llu,%u\n",
7637                         clp->cl_implid->domain, clp->cl_implid->name,
7638                         clp->cl_implid->date.seconds,
7639                         clp->cl_implid->date.nseconds);
7640         dprintk("NFS reply exchange_id: %d\n", status);
7641         return status;
7642
7643 out_impl_id:
7644         kfree(calldata->res.impl_id);
7645 out_server_scope:
7646         kfree(calldata->res.server_scope);
7647 out_server_owner:
7648         kfree(calldata->res.server_owner);
7649 out_calldata:
7650         kfree(calldata);
7651         goto out;
7652 }
7653
7654 /*
7655  * nfs4_proc_exchange_id()
7656  *
7657  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7658  *
7659  * Since the clientid has expired, all compounds using sessions
7660  * associated with the stale clientid will be returning
7661  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7662  * be in some phase of session reset.
7663  *
7664  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7665  */
7666 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7667 {
7668         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7669         int status;
7670
7671         /* try SP4_MACH_CRED if krb5i/p */
7672         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7673             authflavor == RPC_AUTH_GSS_KRB5P) {
7674                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7675                 if (!status)
7676                         return 0;
7677         }
7678
7679         /* try SP4_NONE */
7680         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7681 }
7682
7683 /**
7684  * nfs4_test_session_trunk
7685  *
7686  * This is an add_xprt_test() test function called from
7687  * rpc_clnt_setup_test_and_add_xprt.
7688  *
7689  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7690  * and is dereferrenced in nfs4_exchange_id_release
7691  *
7692  * Upon success, add the new transport to the rpc_clnt
7693  *
7694  * @clnt: struct rpc_clnt to get new transport
7695  * @xprt: the rpc_xprt to test
7696  * @data: call data for _nfs4_proc_exchange_id.
7697  */
7698 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7699                             void *data)
7700 {
7701         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7702         u32 sp4_how;
7703
7704         dprintk("--> %s try %s\n", __func__,
7705                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7706
7707         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7708
7709         /* Test connection for session trunking. Async exchange_id call */
7710         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7711 }
7712 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7713
7714 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7715                 struct rpc_cred *cred)
7716 {
7717         struct rpc_message msg = {
7718                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7719                 .rpc_argp = clp,
7720                 .rpc_cred = cred,
7721         };
7722         int status;
7723
7724         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7725         trace_nfs4_destroy_clientid(clp, status);
7726         if (status)
7727                 dprintk("NFS: Got error %d from the server %s on "
7728                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7729         return status;
7730 }
7731
7732 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7733                 struct rpc_cred *cred)
7734 {
7735         unsigned int loop;
7736         int ret;
7737
7738         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7739                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7740                 switch (ret) {
7741                 case -NFS4ERR_DELAY:
7742                 case -NFS4ERR_CLIENTID_BUSY:
7743                         ssleep(1);
7744                         break;
7745                 default:
7746                         return ret;
7747                 }
7748         }
7749         return 0;
7750 }
7751
7752 int nfs4_destroy_clientid(struct nfs_client *clp)
7753 {
7754         struct rpc_cred *cred;
7755         int ret = 0;
7756
7757         if (clp->cl_mvops->minor_version < 1)
7758                 goto out;
7759         if (clp->cl_exchange_flags == 0)
7760                 goto out;
7761         if (clp->cl_preserve_clid)
7762                 goto out;
7763         cred = nfs4_get_clid_cred(clp);
7764         ret = nfs4_proc_destroy_clientid(clp, cred);
7765         if (cred)
7766                 put_rpccred(cred);
7767         switch (ret) {
7768         case 0:
7769         case -NFS4ERR_STALE_CLIENTID:
7770                 clp->cl_exchange_flags = 0;
7771         }
7772 out:
7773         return ret;
7774 }
7775
7776 struct nfs4_get_lease_time_data {
7777         struct nfs4_get_lease_time_args *args;
7778         struct nfs4_get_lease_time_res *res;
7779         struct nfs_client *clp;
7780 };
7781
7782 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7783                                         void *calldata)
7784 {
7785         struct nfs4_get_lease_time_data *data =
7786                         (struct nfs4_get_lease_time_data *)calldata;
7787
7788         dprintk("--> %s\n", __func__);
7789         /* just setup sequence, do not trigger session recovery
7790            since we're invoked within one */
7791         nfs41_setup_sequence(data->clp->cl_session,
7792                         &data->args->la_seq_args,
7793                         &data->res->lr_seq_res,
7794                         task);
7795         dprintk("<-- %s\n", __func__);
7796 }
7797
7798 /*
7799  * Called from nfs4_state_manager thread for session setup, so don't recover
7800  * from sequence operation or clientid errors.
7801  */
7802 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7803 {
7804         struct nfs4_get_lease_time_data *data =
7805                         (struct nfs4_get_lease_time_data *)calldata;
7806
7807         dprintk("--> %s\n", __func__);
7808         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7809                 return;
7810         switch (task->tk_status) {
7811         case -NFS4ERR_DELAY:
7812         case -NFS4ERR_GRACE:
7813                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7814                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7815                 task->tk_status = 0;
7816                 /* fall through */
7817         case -NFS4ERR_RETRY_UNCACHED_REP:
7818                 rpc_restart_call_prepare(task);
7819                 return;
7820         }
7821         dprintk("<-- %s\n", __func__);
7822 }
7823
7824 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7825         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7826         .rpc_call_done = nfs4_get_lease_time_done,
7827 };
7828
7829 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7830 {
7831         struct rpc_task *task;
7832         struct nfs4_get_lease_time_args args;
7833         struct nfs4_get_lease_time_res res = {
7834                 .lr_fsinfo = fsinfo,
7835         };
7836         struct nfs4_get_lease_time_data data = {
7837                 .args = &args,
7838                 .res = &res,
7839                 .clp = clp,
7840         };
7841         struct rpc_message msg = {
7842                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7843                 .rpc_argp = &args,
7844                 .rpc_resp = &res,
7845         };
7846         struct rpc_task_setup task_setup = {
7847                 .rpc_client = clp->cl_rpcclient,
7848                 .rpc_message = &msg,
7849                 .callback_ops = &nfs4_get_lease_time_ops,
7850                 .callback_data = &data,
7851                 .flags = RPC_TASK_TIMEOUT,
7852         };
7853         int status;
7854
7855         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7856         nfs4_set_sequence_privileged(&args.la_seq_args);
7857         dprintk("--> %s\n", __func__);
7858         task = rpc_run_task(&task_setup);
7859
7860         if (IS_ERR(task))
7861                 status = PTR_ERR(task);
7862         else {
7863                 status = task->tk_status;
7864                 rpc_put_task(task);
7865         }
7866         dprintk("<-- %s return %d\n", __func__, status);
7867
7868         return status;
7869 }
7870
7871 /*
7872  * Initialize the values to be used by the client in CREATE_SESSION
7873  * If nfs4_init_session set the fore channel request and response sizes,
7874  * use them.
7875  *
7876  * Set the back channel max_resp_sz_cached to zero to force the client to
7877  * always set csa_cachethis to FALSE because the current implementation
7878  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7879  */
7880 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7881                                     struct rpc_clnt *clnt)
7882 {
7883         unsigned int max_rqst_sz, max_resp_sz;
7884         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7885
7886         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7887         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7888
7889         /* Fore channel attributes */
7890         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7891         args->fc_attrs.max_resp_sz = max_resp_sz;
7892         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7893         args->fc_attrs.max_reqs = max_session_slots;
7894
7895         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7896                 "max_ops=%u max_reqs=%u\n",
7897                 __func__,
7898                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7899                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7900
7901         /* Back channel attributes */
7902         args->bc_attrs.max_rqst_sz = max_bc_payload;
7903         args->bc_attrs.max_resp_sz = max_bc_payload;
7904         args->bc_attrs.max_resp_sz_cached = 0;
7905         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7906         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7907
7908         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7909                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7910                 __func__,
7911                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7912                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7913                 args->bc_attrs.max_reqs);
7914 }
7915
7916 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7917                 struct nfs41_create_session_res *res)
7918 {
7919         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7920         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7921
7922         if (rcvd->max_resp_sz > sent->max_resp_sz)
7923                 return -EINVAL;
7924         /*
7925          * Our requested max_ops is the minimum we need; we're not
7926          * prepared to break up compounds into smaller pieces than that.
7927          * So, no point even trying to continue if the server won't
7928          * cooperate:
7929          */
7930         if (rcvd->max_ops < sent->max_ops)
7931                 return -EINVAL;
7932         if (rcvd->max_reqs == 0)
7933                 return -EINVAL;
7934         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7935                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7936         return 0;
7937 }
7938
7939 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7940                 struct nfs41_create_session_res *res)
7941 {
7942         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7943         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7944
7945         if (!(res->flags & SESSION4_BACK_CHAN))
7946                 goto out;
7947         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7948                 return -EINVAL;
7949         if (rcvd->max_resp_sz < sent->max_resp_sz)
7950                 return -EINVAL;
7951         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7952                 return -EINVAL;
7953         if (rcvd->max_ops > sent->max_ops)
7954                 return -EINVAL;
7955         if (rcvd->max_reqs > sent->max_reqs)
7956                 return -EINVAL;
7957 out:
7958         return 0;
7959 }
7960
7961 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7962                                      struct nfs41_create_session_res *res)
7963 {
7964         int ret;
7965
7966         ret = nfs4_verify_fore_channel_attrs(args, res);
7967         if (ret)
7968                 return ret;
7969         return nfs4_verify_back_channel_attrs(args, res);
7970 }
7971
7972 static void nfs4_update_session(struct nfs4_session *session,
7973                 struct nfs41_create_session_res *res)
7974 {
7975         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7976         /* Mark client id and session as being confirmed */
7977         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7978         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7979         session->flags = res->flags;
7980         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7981         if (res->flags & SESSION4_BACK_CHAN)
7982                 memcpy(&session->bc_attrs, &res->bc_attrs,
7983                                 sizeof(session->bc_attrs));
7984 }
7985
7986 static int _nfs4_proc_create_session(struct nfs_client *clp,
7987                 struct rpc_cred *cred)
7988 {
7989         struct nfs4_session *session = clp->cl_session;
7990         struct nfs41_create_session_args args = {
7991                 .client = clp,
7992                 .clientid = clp->cl_clientid,
7993                 .seqid = clp->cl_seqid,
7994                 .cb_program = NFS4_CALLBACK,
7995         };
7996         struct nfs41_create_session_res res;
7997
7998         struct rpc_message msg = {
7999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8000                 .rpc_argp = &args,
8001                 .rpc_resp = &res,
8002                 .rpc_cred = cred,
8003         };
8004         int status;
8005
8006         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8007         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8008
8009         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8010         trace_nfs4_create_session(clp, status);
8011
8012         switch (status) {
8013         case -NFS4ERR_STALE_CLIENTID:
8014         case -NFS4ERR_DELAY:
8015         case -ETIMEDOUT:
8016         case -EACCES:
8017         case -EAGAIN:
8018                 goto out;
8019         };
8020
8021         clp->cl_seqid++;
8022         if (!status) {
8023                 /* Verify the session's negotiated channel_attrs values */
8024                 status = nfs4_verify_channel_attrs(&args, &res);
8025                 /* Increment the clientid slot sequence id */
8026                 if (status)
8027                         goto out;
8028                 nfs4_update_session(session, &res);
8029         }
8030 out:
8031         return status;
8032 }
8033
8034 /*
8035  * Issues a CREATE_SESSION operation to the server.
8036  * It is the responsibility of the caller to verify the session is
8037  * expired before calling this routine.
8038  */
8039 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8040 {
8041         int status;
8042         unsigned *ptr;
8043         struct nfs4_session *session = clp->cl_session;
8044
8045         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8046
8047         status = _nfs4_proc_create_session(clp, cred);
8048         if (status)
8049                 goto out;
8050
8051         /* Init or reset the session slot tables */
8052         status = nfs4_setup_session_slot_tables(session);
8053         dprintk("slot table setup returned %d\n", status);
8054         if (status)
8055                 goto out;
8056
8057         ptr = (unsigned *)&session->sess_id.data[0];
8058         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8059                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8060 out:
8061         dprintk("<-- %s\n", __func__);
8062         return status;
8063 }
8064
8065 /*
8066  * Issue the over-the-wire RPC DESTROY_SESSION.
8067  * The caller must serialize access to this routine.
8068  */
8069 int nfs4_proc_destroy_session(struct nfs4_session *session,
8070                 struct rpc_cred *cred)
8071 {
8072         struct rpc_message msg = {
8073                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8074                 .rpc_argp = session,
8075                 .rpc_cred = cred,
8076         };
8077         int status = 0;
8078
8079         dprintk("--> nfs4_proc_destroy_session\n");
8080
8081         /* session is still being setup */
8082         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8083                 return 0;
8084
8085         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8086         trace_nfs4_destroy_session(session->clp, status);
8087
8088         if (status)
8089                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8090                         "Session has been destroyed regardless...\n", status);
8091
8092         dprintk("<-- nfs4_proc_destroy_session\n");
8093         return status;
8094 }
8095
8096 /*
8097  * Renew the cl_session lease.
8098  */
8099 struct nfs4_sequence_data {
8100         struct nfs_client *clp;
8101         struct nfs4_sequence_args args;
8102         struct nfs4_sequence_res res;
8103 };
8104
8105 static void nfs41_sequence_release(void *data)
8106 {
8107         struct nfs4_sequence_data *calldata = data;
8108         struct nfs_client *clp = calldata->clp;
8109
8110         if (atomic_read(&clp->cl_count) > 1)
8111                 nfs4_schedule_state_renewal(clp);
8112         nfs_put_client(clp);
8113         kfree(calldata);
8114 }
8115
8116 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8117 {
8118         switch(task->tk_status) {
8119         case -NFS4ERR_DELAY:
8120                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8121                 return -EAGAIN;
8122         default:
8123                 nfs4_schedule_lease_recovery(clp);
8124         }
8125         return 0;
8126 }
8127
8128 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8129 {
8130         struct nfs4_sequence_data *calldata = data;
8131         struct nfs_client *clp = calldata->clp;
8132
8133         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8134                 return;
8135
8136         trace_nfs4_sequence(clp, task->tk_status);
8137         if (task->tk_status < 0) {
8138                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8139                 if (atomic_read(&clp->cl_count) == 1)
8140                         goto out;
8141
8142                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8143                         rpc_restart_call_prepare(task);
8144                         return;
8145                 }
8146         }
8147         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8148 out:
8149         dprintk("<-- %s\n", __func__);
8150 }
8151
8152 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8153 {
8154         struct nfs4_sequence_data *calldata = data;
8155         struct nfs_client *clp = calldata->clp;
8156         struct nfs4_sequence_args *args;
8157         struct nfs4_sequence_res *res;
8158
8159         args = task->tk_msg.rpc_argp;
8160         res = task->tk_msg.rpc_resp;
8161
8162         nfs41_setup_sequence(clp->cl_session, args, res, task);
8163 }
8164
8165 static const struct rpc_call_ops nfs41_sequence_ops = {
8166         .rpc_call_done = nfs41_sequence_call_done,
8167         .rpc_call_prepare = nfs41_sequence_prepare,
8168         .rpc_release = nfs41_sequence_release,
8169 };
8170
8171 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8172                 struct rpc_cred *cred,
8173                 bool is_privileged)
8174 {
8175         struct nfs4_sequence_data *calldata;
8176         struct rpc_message msg = {
8177                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8178                 .rpc_cred = cred,
8179         };
8180         struct rpc_task_setup task_setup_data = {
8181                 .rpc_client = clp->cl_rpcclient,
8182                 .rpc_message = &msg,
8183                 .callback_ops = &nfs41_sequence_ops,
8184                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8185         };
8186
8187         if (!atomic_inc_not_zero(&clp->cl_count))
8188                 return ERR_PTR(-EIO);
8189         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8190         if (calldata == NULL) {
8191                 nfs_put_client(clp);
8192                 return ERR_PTR(-ENOMEM);
8193         }
8194         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8195         if (is_privileged)
8196                 nfs4_set_sequence_privileged(&calldata->args);
8197         msg.rpc_argp = &calldata->args;
8198         msg.rpc_resp = &calldata->res;
8199         calldata->clp = clp;
8200         task_setup_data.callback_data = calldata;
8201
8202         return rpc_run_task(&task_setup_data);
8203 }
8204
8205 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8206 {
8207         struct rpc_task *task;
8208         int ret = 0;
8209
8210         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8211                 return -EAGAIN;
8212         task = _nfs41_proc_sequence(clp, cred, false);
8213         if (IS_ERR(task))
8214                 ret = PTR_ERR(task);
8215         else
8216                 rpc_put_task_async(task);
8217         dprintk("<-- %s status=%d\n", __func__, ret);
8218         return ret;
8219 }
8220
8221 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8222 {
8223         struct rpc_task *task;
8224         int ret;
8225
8226         task = _nfs41_proc_sequence(clp, cred, true);
8227         if (IS_ERR(task)) {
8228                 ret = PTR_ERR(task);
8229                 goto out;
8230         }
8231         ret = rpc_wait_for_completion_task(task);
8232         if (!ret)
8233                 ret = task->tk_status;
8234         rpc_put_task(task);
8235 out:
8236         dprintk("<-- %s status=%d\n", __func__, ret);
8237         return ret;
8238 }
8239
8240 struct nfs4_reclaim_complete_data {
8241         struct nfs_client *clp;
8242         struct nfs41_reclaim_complete_args arg;
8243         struct nfs41_reclaim_complete_res res;
8244 };
8245
8246 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8247 {
8248         struct nfs4_reclaim_complete_data *calldata = data;
8249
8250         nfs41_setup_sequence(calldata->clp->cl_session,
8251                         &calldata->arg.seq_args,
8252                         &calldata->res.seq_res,
8253                         task);
8254 }
8255
8256 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8257 {
8258         switch(task->tk_status) {
8259         case 0:
8260         case -NFS4ERR_COMPLETE_ALREADY:
8261         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8262                 break;
8263         case -NFS4ERR_DELAY:
8264                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8265                 /* fall through */
8266         case -NFS4ERR_RETRY_UNCACHED_REP:
8267                 return -EAGAIN;
8268         default:
8269                 nfs4_schedule_lease_recovery(clp);
8270         }
8271         return 0;
8272 }
8273
8274 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8275 {
8276         struct nfs4_reclaim_complete_data *calldata = data;
8277         struct nfs_client *clp = calldata->clp;
8278         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8279
8280         dprintk("--> %s\n", __func__);
8281         if (!nfs41_sequence_done(task, res))
8282                 return;
8283
8284         trace_nfs4_reclaim_complete(clp, task->tk_status);
8285         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8286                 rpc_restart_call_prepare(task);
8287                 return;
8288         }
8289         dprintk("<-- %s\n", __func__);
8290 }
8291
8292 static void nfs4_free_reclaim_complete_data(void *data)
8293 {
8294         struct nfs4_reclaim_complete_data *calldata = data;
8295
8296         kfree(calldata);
8297 }
8298
8299 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8300         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8301         .rpc_call_done = nfs4_reclaim_complete_done,
8302         .rpc_release = nfs4_free_reclaim_complete_data,
8303 };
8304
8305 /*
8306  * Issue a global reclaim complete.
8307  */
8308 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8309                 struct rpc_cred *cred)
8310 {
8311         struct nfs4_reclaim_complete_data *calldata;
8312         struct rpc_task *task;
8313         struct rpc_message msg = {
8314                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8315                 .rpc_cred = cred,
8316         };
8317         struct rpc_task_setup task_setup_data = {
8318                 .rpc_client = clp->cl_rpcclient,
8319                 .rpc_message = &msg,
8320                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8321                 .flags = RPC_TASK_ASYNC,
8322         };
8323         int status = -ENOMEM;
8324
8325         dprintk("--> %s\n", __func__);
8326         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8327         if (calldata == NULL)
8328                 goto out;
8329         calldata->clp = clp;
8330         calldata->arg.one_fs = 0;
8331
8332         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8333         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8334         msg.rpc_argp = &calldata->arg;
8335         msg.rpc_resp = &calldata->res;
8336         task_setup_data.callback_data = calldata;
8337         task = rpc_run_task(&task_setup_data);
8338         if (IS_ERR(task)) {
8339                 status = PTR_ERR(task);
8340                 goto out;
8341         }
8342         status = nfs4_wait_for_completion_rpc_task(task);
8343         if (status == 0)
8344                 status = task->tk_status;
8345         rpc_put_task(task);
8346         return 0;
8347 out:
8348         dprintk("<-- %s status=%d\n", __func__, status);
8349         return status;
8350 }
8351
8352 static void
8353 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8354 {
8355         struct nfs4_layoutget *lgp = calldata;
8356         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8357         struct nfs4_session *session = nfs4_get_session(server);
8358
8359         dprintk("--> %s\n", __func__);
8360         nfs41_setup_sequence(session, &lgp->args.seq_args,
8361                                 &lgp->res.seq_res, task);
8362         dprintk("<-- %s\n", __func__);
8363 }
8364
8365 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8366 {
8367         struct nfs4_layoutget *lgp = calldata;
8368
8369         dprintk("--> %s\n", __func__);
8370         nfs41_sequence_process(task, &lgp->res.seq_res);
8371         dprintk("<-- %s\n", __func__);
8372 }
8373
8374 static int
8375 nfs4_layoutget_handle_exception(struct rpc_task *task,
8376                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8377 {
8378         struct inode *inode = lgp->args.inode;
8379         struct nfs_server *server = NFS_SERVER(inode);
8380         struct pnfs_layout_hdr *lo;
8381         int nfs4err = task->tk_status;
8382         int err, status = 0;
8383         LIST_HEAD(head);
8384
8385         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8386
8387         switch (nfs4err) {
8388         case 0:
8389                 goto out;
8390
8391         /*
8392          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8393          * on the file. set tk_status to -ENODATA to tell upper layer to
8394          * retry go inband.
8395          */
8396         case -NFS4ERR_LAYOUTUNAVAILABLE:
8397                 status = -ENODATA;
8398                 goto out;
8399         /*
8400          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8401          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8402          */
8403         case -NFS4ERR_BADLAYOUT:
8404                 status = -EOVERFLOW;
8405                 goto out;
8406         /*
8407          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8408          * (or clients) writing to the same RAID stripe except when
8409          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8410          *
8411          * Treat it like we would RECALLCONFLICT -- we retry for a little
8412          * while, and then eventually give up.
8413          */
8414         case -NFS4ERR_LAYOUTTRYLATER:
8415                 if (lgp->args.minlength == 0) {
8416                         status = -EOVERFLOW;
8417                         goto out;
8418                 }
8419                 status = -EBUSY;
8420                 break;
8421         case -NFS4ERR_RECALLCONFLICT:
8422                 status = -ERECALLCONFLICT;
8423                 break;
8424         case -NFS4ERR_DELEG_REVOKED:
8425         case -NFS4ERR_ADMIN_REVOKED:
8426         case -NFS4ERR_EXPIRED:
8427         case -NFS4ERR_BAD_STATEID:
8428                 exception->timeout = 0;
8429                 spin_lock(&inode->i_lock);
8430                 lo = NFS_I(inode)->layout;
8431                 /* If the open stateid was bad, then recover it. */
8432                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8433                     nfs4_stateid_match_other(&lgp->args.stateid,
8434                                         &lgp->args.ctx->state->stateid)) {
8435                         spin_unlock(&inode->i_lock);
8436                         exception->state = lgp->args.ctx->state;
8437                         exception->stateid = &lgp->args.stateid;
8438                         break;
8439                 }
8440
8441                 /*
8442                  * Mark the bad layout state as invalid, then retry
8443                  */
8444                 pnfs_mark_layout_stateid_invalid(lo, &head);
8445                 spin_unlock(&inode->i_lock);
8446                 pnfs_free_lseg_list(&head);
8447                 status = -EAGAIN;
8448                 goto out;
8449         }
8450
8451         err = nfs4_handle_exception(server, nfs4err, exception);
8452         if (!status) {
8453                 if (exception->retry)
8454                         status = -EAGAIN;
8455                 else
8456                         status = err;
8457         }
8458 out:
8459         dprintk("<-- %s\n", __func__);
8460         return status;
8461 }
8462
8463 static size_t max_response_pages(struct nfs_server *server)
8464 {
8465         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8466         return nfs_page_array_len(0, max_resp_sz);
8467 }
8468
8469 static void nfs4_free_pages(struct page **pages, size_t size)
8470 {
8471         int i;
8472
8473         if (!pages)
8474                 return;
8475
8476         for (i = 0; i < size; i++) {
8477                 if (!pages[i])
8478                         break;
8479                 __free_page(pages[i]);
8480         }
8481         kfree(pages);
8482 }
8483
8484 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8485 {
8486         struct page **pages;
8487         int i;
8488
8489         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8490         if (!pages) {
8491                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8492                 return NULL;
8493         }
8494
8495         for (i = 0; i < size; i++) {
8496                 pages[i] = alloc_page(gfp_flags);
8497                 if (!pages[i]) {
8498                         dprintk("%s: failed to allocate page\n", __func__);
8499                         nfs4_free_pages(pages, size);
8500                         return NULL;
8501                 }
8502         }
8503
8504         return pages;
8505 }
8506
8507 static void nfs4_layoutget_release(void *calldata)
8508 {
8509         struct nfs4_layoutget *lgp = calldata;
8510         struct inode *inode = lgp->args.inode;
8511         struct nfs_server *server = NFS_SERVER(inode);
8512         size_t max_pages = max_response_pages(server);
8513
8514         dprintk("--> %s\n", __func__);
8515         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8516         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8517         put_nfs_open_context(lgp->args.ctx);
8518         kfree(calldata);
8519         dprintk("<-- %s\n", __func__);
8520 }
8521
8522 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8523         .rpc_call_prepare = nfs4_layoutget_prepare,
8524         .rpc_call_done = nfs4_layoutget_done,
8525         .rpc_release = nfs4_layoutget_release,
8526 };
8527
8528 struct pnfs_layout_segment *
8529 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8530 {
8531         struct inode *inode = lgp->args.inode;
8532         struct nfs_server *server = NFS_SERVER(inode);
8533         size_t max_pages = max_response_pages(server);
8534         struct rpc_task *task;
8535         struct rpc_message msg = {
8536                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8537                 .rpc_argp = &lgp->args,
8538                 .rpc_resp = &lgp->res,
8539                 .rpc_cred = lgp->cred,
8540         };
8541         struct rpc_task_setup task_setup_data = {
8542                 .rpc_client = server->client,
8543                 .rpc_message = &msg,
8544                 .callback_ops = &nfs4_layoutget_call_ops,
8545                 .callback_data = lgp,
8546                 .flags = RPC_TASK_ASYNC,
8547         };
8548         struct pnfs_layout_segment *lseg = NULL;
8549         struct nfs4_exception exception = {
8550                 .inode = inode,
8551                 .timeout = *timeout,
8552         };
8553         int status = 0;
8554
8555         dprintk("--> %s\n", __func__);
8556
8557         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8558         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8559
8560         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8561         if (!lgp->args.layout.pages) {
8562                 nfs4_layoutget_release(lgp);
8563                 return ERR_PTR(-ENOMEM);
8564         }
8565         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8566
8567         lgp->res.layoutp = &lgp->args.layout;
8568         lgp->res.seq_res.sr_slot = NULL;
8569         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8570
8571         task = rpc_run_task(&task_setup_data);
8572         if (IS_ERR(task))
8573                 return ERR_CAST(task);
8574         status = nfs4_wait_for_completion_rpc_task(task);
8575         if (status == 0) {
8576                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8577                 *timeout = exception.timeout;
8578         }
8579
8580         trace_nfs4_layoutget(lgp->args.ctx,
8581                         &lgp->args.range,
8582                         &lgp->res.range,
8583                         &lgp->res.stateid,
8584                         status);
8585
8586         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8587         if (status == 0 && lgp->res.layoutp->len)
8588                 lseg = pnfs_layout_process(lgp);
8589         nfs4_sequence_free_slot(&lgp->res.seq_res);
8590         rpc_put_task(task);
8591         dprintk("<-- %s status=%d\n", __func__, status);
8592         if (status)
8593                 return ERR_PTR(status);
8594         return lseg;
8595 }
8596
8597 static void
8598 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8599 {
8600         struct nfs4_layoutreturn *lrp = calldata;
8601
8602         dprintk("--> %s\n", __func__);
8603         nfs41_setup_sequence(lrp->clp->cl_session,
8604                         &lrp->args.seq_args,
8605                         &lrp->res.seq_res,
8606                         task);
8607 }
8608
8609 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8610 {
8611         struct nfs4_layoutreturn *lrp = calldata;
8612         struct nfs_server *server;
8613
8614         dprintk("--> %s\n", __func__);
8615
8616         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8617                 return;
8618
8619         server = NFS_SERVER(lrp->args.inode);
8620         switch (task->tk_status) {
8621         default:
8622                 task->tk_status = 0;
8623         case 0:
8624                 break;
8625         case -NFS4ERR_DELAY:
8626                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8627                         break;
8628                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8629                 rpc_restart_call_prepare(task);
8630                 return;
8631         }
8632         dprintk("<-- %s\n", __func__);
8633 }
8634
8635 static void nfs4_layoutreturn_release(void *calldata)
8636 {
8637         struct nfs4_layoutreturn *lrp = calldata;
8638         struct pnfs_layout_hdr *lo = lrp->args.layout;
8639
8640         dprintk("--> %s\n", __func__);
8641         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8642                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8643         nfs4_sequence_free_slot(&lrp->res.seq_res);
8644         pnfs_put_layout_hdr(lrp->args.layout);
8645         nfs_iput_and_deactive(lrp->inode);
8646         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8647                 lrp->ld_private.ops->free(&lrp->ld_private);
8648         kfree(calldata);
8649         dprintk("<-- %s\n", __func__);
8650 }
8651
8652 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8653         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8654         .rpc_call_done = nfs4_layoutreturn_done,
8655         .rpc_release = nfs4_layoutreturn_release,
8656 };
8657
8658 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8659 {
8660         struct rpc_task *task;
8661         struct rpc_message msg = {
8662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8663                 .rpc_argp = &lrp->args,
8664                 .rpc_resp = &lrp->res,
8665                 .rpc_cred = lrp->cred,
8666         };
8667         struct rpc_task_setup task_setup_data = {
8668                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8669                 .rpc_message = &msg,
8670                 .callback_ops = &nfs4_layoutreturn_call_ops,
8671                 .callback_data = lrp,
8672         };
8673         int status = 0;
8674
8675         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8676                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8677                         &task_setup_data.rpc_client, &msg);
8678
8679         dprintk("--> %s\n", __func__);
8680         if (!sync) {
8681                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8682                 if (!lrp->inode) {
8683                         nfs4_layoutreturn_release(lrp);
8684                         return -EAGAIN;
8685                 }
8686                 task_setup_data.flags |= RPC_TASK_ASYNC;
8687         }
8688         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8689         task = rpc_run_task(&task_setup_data);
8690         if (IS_ERR(task))
8691                 return PTR_ERR(task);
8692         if (sync)
8693                 status = task->tk_status;
8694         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8695         dprintk("<-- %s status=%d\n", __func__, status);
8696         rpc_put_task(task);
8697         return status;
8698 }
8699
8700 static int
8701 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8702                 struct pnfs_device *pdev,
8703                 struct rpc_cred *cred)
8704 {
8705         struct nfs4_getdeviceinfo_args args = {
8706                 .pdev = pdev,
8707                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8708                         NOTIFY_DEVICEID4_DELETE,
8709         };
8710         struct nfs4_getdeviceinfo_res res = {
8711                 .pdev = pdev,
8712         };
8713         struct rpc_message msg = {
8714                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8715                 .rpc_argp = &args,
8716                 .rpc_resp = &res,
8717                 .rpc_cred = cred,
8718         };
8719         int status;
8720
8721         dprintk("--> %s\n", __func__);
8722         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8723         if (res.notification & ~args.notify_types)
8724                 dprintk("%s: unsupported notification\n", __func__);
8725         if (res.notification != args.notify_types)
8726                 pdev->nocache = 1;
8727
8728         dprintk("<-- %s status=%d\n", __func__, status);
8729
8730         return status;
8731 }
8732
8733 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8734                 struct pnfs_device *pdev,
8735                 struct rpc_cred *cred)
8736 {
8737         struct nfs4_exception exception = { };
8738         int err;
8739
8740         do {
8741                 err = nfs4_handle_exception(server,
8742                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8743                                         &exception);
8744         } while (exception.retry);
8745         return err;
8746 }
8747 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8748
8749 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8750 {
8751         struct nfs4_layoutcommit_data *data = calldata;
8752         struct nfs_server *server = NFS_SERVER(data->args.inode);
8753         struct nfs4_session *session = nfs4_get_session(server);
8754
8755         nfs41_setup_sequence(session,
8756                         &data->args.seq_args,
8757                         &data->res.seq_res,
8758                         task);
8759 }
8760
8761 static void
8762 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8763 {
8764         struct nfs4_layoutcommit_data *data = calldata;
8765         struct nfs_server *server = NFS_SERVER(data->args.inode);
8766
8767         if (!nfs41_sequence_done(task, &data->res.seq_res))
8768                 return;
8769
8770         switch (task->tk_status) { /* Just ignore these failures */
8771         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8772         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8773         case -NFS4ERR_BADLAYOUT:     /* no layout */
8774         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8775                 task->tk_status = 0;
8776         case 0:
8777                 break;
8778         default:
8779                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8780                         rpc_restart_call_prepare(task);
8781                         return;
8782                 }
8783         }
8784 }
8785
8786 static void nfs4_layoutcommit_release(void *calldata)
8787 {
8788         struct nfs4_layoutcommit_data *data = calldata;
8789
8790         pnfs_cleanup_layoutcommit(data);
8791         nfs_post_op_update_inode_force_wcc(data->args.inode,
8792                                            data->res.fattr);
8793         put_rpccred(data->cred);
8794         nfs_iput_and_deactive(data->inode);
8795         kfree(data);
8796 }
8797
8798 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8799         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8800         .rpc_call_done = nfs4_layoutcommit_done,
8801         .rpc_release = nfs4_layoutcommit_release,
8802 };
8803
8804 int
8805 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8806 {
8807         struct rpc_message msg = {
8808                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8809                 .rpc_argp = &data->args,
8810                 .rpc_resp = &data->res,
8811                 .rpc_cred = data->cred,
8812         };
8813         struct rpc_task_setup task_setup_data = {
8814                 .task = &data->task,
8815                 .rpc_client = NFS_CLIENT(data->args.inode),
8816                 .rpc_message = &msg,
8817                 .callback_ops = &nfs4_layoutcommit_ops,
8818                 .callback_data = data,
8819         };
8820         struct rpc_task *task;
8821         int status = 0;
8822
8823         dprintk("NFS: initiating layoutcommit call. sync %d "
8824                 "lbw: %llu inode %lu\n", sync,
8825                 data->args.lastbytewritten,
8826                 data->args.inode->i_ino);
8827
8828         if (!sync) {
8829                 data->inode = nfs_igrab_and_active(data->args.inode);
8830                 if (data->inode == NULL) {
8831                         nfs4_layoutcommit_release(data);
8832                         return -EAGAIN;
8833                 }
8834                 task_setup_data.flags = RPC_TASK_ASYNC;
8835         }
8836         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8837         task = rpc_run_task(&task_setup_data);
8838         if (IS_ERR(task))
8839                 return PTR_ERR(task);
8840         if (sync)
8841                 status = task->tk_status;
8842         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8843         dprintk("%s: status %d\n", __func__, status);
8844         rpc_put_task(task);
8845         return status;
8846 }
8847
8848 /**
8849  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8850  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8851  */
8852 static int
8853 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8854                     struct nfs_fsinfo *info,
8855                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8856 {
8857         struct nfs41_secinfo_no_name_args args = {
8858                 .style = SECINFO_STYLE_CURRENT_FH,
8859         };
8860         struct nfs4_secinfo_res res = {
8861                 .flavors = flavors,
8862         };
8863         struct rpc_message msg = {
8864                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8865                 .rpc_argp = &args,
8866                 .rpc_resp = &res,
8867         };
8868         struct rpc_clnt *clnt = server->client;
8869         struct rpc_cred *cred = NULL;
8870         int status;
8871
8872         if (use_integrity) {
8873                 clnt = server->nfs_client->cl_rpcclient;
8874                 cred = nfs4_get_clid_cred(server->nfs_client);
8875                 msg.rpc_cred = cred;
8876         }
8877
8878         dprintk("--> %s\n", __func__);
8879         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8880                                 &res.seq_res, 0);
8881         dprintk("<-- %s status=%d\n", __func__, status);
8882
8883         if (cred)
8884                 put_rpccred(cred);
8885
8886         return status;
8887 }
8888
8889 static int
8890 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8891                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8892 {
8893         struct nfs4_exception exception = { };
8894         int err;
8895         do {
8896                 /* first try using integrity protection */
8897                 err = -NFS4ERR_WRONGSEC;
8898
8899                 /* try to use integrity protection with machine cred */
8900                 if (_nfs4_is_integrity_protected(server->nfs_client))
8901                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8902                                                           flavors, true);
8903
8904                 /*
8905                  * if unable to use integrity protection, or SECINFO with
8906                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8907                  * disallowed by spec, but exists in deployed servers) use
8908                  * the current filesystem's rpc_client and the user cred.
8909                  */
8910                 if (err == -NFS4ERR_WRONGSEC)
8911                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8912                                                           flavors, false);
8913
8914                 switch (err) {
8915                 case 0:
8916                 case -NFS4ERR_WRONGSEC:
8917                 case -ENOTSUPP:
8918                         goto out;
8919                 default:
8920                         err = nfs4_handle_exception(server, err, &exception);
8921                 }
8922         } while (exception.retry);
8923 out:
8924         return err;
8925 }
8926
8927 static int
8928 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8929                     struct nfs_fsinfo *info)
8930 {
8931         int err;
8932         struct page *page;
8933         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8934         struct nfs4_secinfo_flavors *flavors;
8935         struct nfs4_secinfo4 *secinfo;
8936         int i;
8937
8938         page = alloc_page(GFP_KERNEL);
8939         if (!page) {
8940                 err = -ENOMEM;
8941                 goto out;
8942         }
8943
8944         flavors = page_address(page);
8945         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8946
8947         /*
8948          * Fall back on "guess and check" method if
8949          * the server doesn't support SECINFO_NO_NAME
8950          */
8951         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8952                 err = nfs4_find_root_sec(server, fhandle, info);
8953                 goto out_freepage;
8954         }
8955         if (err)
8956                 goto out_freepage;
8957
8958         for (i = 0; i < flavors->num_flavors; i++) {
8959                 secinfo = &flavors->flavors[i];
8960
8961                 switch (secinfo->flavor) {
8962                 case RPC_AUTH_NULL:
8963                 case RPC_AUTH_UNIX:
8964                 case RPC_AUTH_GSS:
8965                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8966                                         &secinfo->flavor_info);
8967                         break;
8968                 default:
8969                         flavor = RPC_AUTH_MAXFLAVOR;
8970                         break;
8971                 }
8972
8973                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8974                         flavor = RPC_AUTH_MAXFLAVOR;
8975
8976                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8977                         err = nfs4_lookup_root_sec(server, fhandle,
8978                                                    info, flavor);
8979                         if (!err)
8980                                 break;
8981                 }
8982         }
8983
8984         if (flavor == RPC_AUTH_MAXFLAVOR)
8985                 err = -EPERM;
8986
8987 out_freepage:
8988         put_page(page);
8989         if (err == -EACCES)
8990                 return -EPERM;
8991 out:
8992         return err;
8993 }
8994
8995 static int _nfs41_test_stateid(struct nfs_server *server,
8996                 nfs4_stateid *stateid,
8997                 struct rpc_cred *cred)
8998 {
8999         int status;
9000         struct nfs41_test_stateid_args args = {
9001                 .stateid = stateid,
9002         };
9003         struct nfs41_test_stateid_res res;
9004         struct rpc_message msg = {
9005                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9006                 .rpc_argp = &args,
9007                 .rpc_resp = &res,
9008                 .rpc_cred = cred,
9009         };
9010         struct rpc_clnt *rpc_client = server->client;
9011
9012         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9013                 &rpc_client, &msg);
9014
9015         dprintk("NFS call  test_stateid %p\n", stateid);
9016         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
9017         nfs4_set_sequence_privileged(&args.seq_args);
9018         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9019                         &args.seq_args, &res.seq_res);
9020         if (status != NFS_OK) {
9021                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9022                 return status;
9023         }
9024         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9025         return -res.status;
9026 }
9027
9028 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9029                 int err, struct nfs4_exception *exception)
9030 {
9031         exception->retry = 0;
9032         switch(err) {
9033         case -NFS4ERR_DELAY:
9034         case -NFS4ERR_RETRY_UNCACHED_REP:
9035                 nfs4_handle_exception(server, err, exception);
9036                 break;
9037         case -NFS4ERR_BADSESSION:
9038         case -NFS4ERR_BADSLOT:
9039         case -NFS4ERR_BAD_HIGH_SLOT:
9040         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9041         case -NFS4ERR_DEADSESSION:
9042                 nfs4_do_handle_exception(server, err, exception);
9043         }
9044 }
9045
9046 /**
9047  * nfs41_test_stateid - perform a TEST_STATEID operation
9048  *
9049  * @server: server / transport on which to perform the operation
9050  * @stateid: state ID to test
9051  * @cred: credential
9052  *
9053  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9054  * Otherwise a negative NFS4ERR value is returned if the operation
9055  * failed or the state ID is not currently valid.
9056  */
9057 static int nfs41_test_stateid(struct nfs_server *server,
9058                 nfs4_stateid *stateid,
9059                 struct rpc_cred *cred)
9060 {
9061         struct nfs4_exception exception = { };
9062         int err;
9063         do {
9064                 err = _nfs41_test_stateid(server, stateid, cred);
9065                 nfs4_handle_delay_or_session_error(server, err, &exception);
9066         } while (exception.retry);
9067         return err;
9068 }
9069
9070 struct nfs_free_stateid_data {
9071         struct nfs_server *server;
9072         struct nfs41_free_stateid_args args;
9073         struct nfs41_free_stateid_res res;
9074 };
9075
9076 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9077 {
9078         struct nfs_free_stateid_data *data = calldata;
9079         nfs41_setup_sequence(nfs4_get_session(data->server),
9080                         &data->args.seq_args,
9081                         &data->res.seq_res,
9082                         task);
9083 }
9084
9085 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9086 {
9087         struct nfs_free_stateid_data *data = calldata;
9088
9089         nfs41_sequence_done(task, &data->res.seq_res);
9090
9091         switch (task->tk_status) {
9092         case -NFS4ERR_DELAY:
9093                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9094                         rpc_restart_call_prepare(task);
9095         }
9096 }
9097
9098 static void nfs41_free_stateid_release(void *calldata)
9099 {
9100         kfree(calldata);
9101 }
9102
9103 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9104         .rpc_call_prepare = nfs41_free_stateid_prepare,
9105         .rpc_call_done = nfs41_free_stateid_done,
9106         .rpc_release = nfs41_free_stateid_release,
9107 };
9108
9109 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9110                 const nfs4_stateid *stateid,
9111                 struct rpc_cred *cred,
9112                 bool privileged)
9113 {
9114         struct rpc_message msg = {
9115                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9116                 .rpc_cred = cred,
9117         };
9118         struct rpc_task_setup task_setup = {
9119                 .rpc_client = server->client,
9120                 .rpc_message = &msg,
9121                 .callback_ops = &nfs41_free_stateid_ops,
9122                 .flags = RPC_TASK_ASYNC,
9123         };
9124         struct nfs_free_stateid_data *data;
9125
9126         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9127                 &task_setup.rpc_client, &msg);
9128
9129         dprintk("NFS call  free_stateid %p\n", stateid);
9130         data = kmalloc(sizeof(*data), GFP_NOFS);
9131         if (!data)
9132                 return ERR_PTR(-ENOMEM);
9133         data->server = server;
9134         nfs4_stateid_copy(&data->args.stateid, stateid);
9135
9136         task_setup.callback_data = data;
9137
9138         msg.rpc_argp = &data->args;
9139         msg.rpc_resp = &data->res;
9140         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9141         if (privileged)
9142                 nfs4_set_sequence_privileged(&data->args.seq_args);
9143
9144         return rpc_run_task(&task_setup);
9145 }
9146
9147 /**
9148  * nfs41_free_stateid - perform a FREE_STATEID operation
9149  *
9150  * @server: server / transport on which to perform the operation
9151  * @stateid: state ID to release
9152  * @cred: credential
9153  * @is_recovery: set to true if this call needs to be privileged
9154  *
9155  * Note: this function is always asynchronous.
9156  */
9157 static int nfs41_free_stateid(struct nfs_server *server,
9158                 const nfs4_stateid *stateid,
9159                 struct rpc_cred *cred,
9160                 bool is_recovery)
9161 {
9162         struct rpc_task *task;
9163
9164         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9165         if (IS_ERR(task))
9166                 return PTR_ERR(task);
9167         rpc_put_task(task);
9168         return 0;
9169 }
9170
9171 static void
9172 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9173 {
9174         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9175
9176         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9177         nfs4_free_lock_state(server, lsp);
9178 }
9179
9180 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9181                 const nfs4_stateid *s2)
9182 {
9183         if (s1->type != s2->type)
9184                 return false;
9185
9186         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9187                 return false;
9188
9189         if (s1->seqid == s2->seqid)
9190                 return true;
9191         if (s1->seqid == 0 || s2->seqid == 0)
9192                 return true;
9193
9194         return false;
9195 }
9196
9197 #endif /* CONFIG_NFS_V4_1 */
9198
9199 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9200                 const nfs4_stateid *s2)
9201 {
9202         return nfs4_stateid_match(s1, s2);
9203 }
9204
9205
9206 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9207         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9208         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9209         .recover_open   = nfs4_open_reclaim,
9210         .recover_lock   = nfs4_lock_reclaim,
9211         .establish_clid = nfs4_init_clientid,
9212         .detect_trunking = nfs40_discover_server_trunking,
9213 };
9214
9215 #if defined(CONFIG_NFS_V4_1)
9216 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9217         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9218         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9219         .recover_open   = nfs4_open_reclaim,
9220         .recover_lock   = nfs4_lock_reclaim,
9221         .establish_clid = nfs41_init_clientid,
9222         .reclaim_complete = nfs41_proc_reclaim_complete,
9223         .detect_trunking = nfs41_discover_server_trunking,
9224 };
9225 #endif /* CONFIG_NFS_V4_1 */
9226
9227 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9228         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9229         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9230         .recover_open   = nfs40_open_expired,
9231         .recover_lock   = nfs4_lock_expired,
9232         .establish_clid = nfs4_init_clientid,
9233 };
9234
9235 #if defined(CONFIG_NFS_V4_1)
9236 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9237         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9238         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9239         .recover_open   = nfs41_open_expired,
9240         .recover_lock   = nfs41_lock_expired,
9241         .establish_clid = nfs41_init_clientid,
9242 };
9243 #endif /* CONFIG_NFS_V4_1 */
9244
9245 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9246         .sched_state_renewal = nfs4_proc_async_renew,
9247         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9248         .renew_lease = nfs4_proc_renew,
9249 };
9250
9251 #if defined(CONFIG_NFS_V4_1)
9252 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9253         .sched_state_renewal = nfs41_proc_async_sequence,
9254         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9255         .renew_lease = nfs4_proc_sequence,
9256 };
9257 #endif
9258
9259 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9260         .get_locations = _nfs40_proc_get_locations,
9261         .fsid_present = _nfs40_proc_fsid_present,
9262 };
9263
9264 #if defined(CONFIG_NFS_V4_1)
9265 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9266         .get_locations = _nfs41_proc_get_locations,
9267         .fsid_present = _nfs41_proc_fsid_present,
9268 };
9269 #endif  /* CONFIG_NFS_V4_1 */
9270
9271 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9272         .minor_version = 0,
9273         .init_caps = NFS_CAP_READDIRPLUS
9274                 | NFS_CAP_ATOMIC_OPEN
9275                 | NFS_CAP_POSIX_LOCK,
9276         .init_client = nfs40_init_client,
9277         .shutdown_client = nfs40_shutdown_client,
9278         .match_stateid = nfs4_match_stateid,
9279         .find_root_sec = nfs4_find_root_sec,
9280         .free_lock_state = nfs4_release_lockowner,
9281         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9282         .alloc_seqid = nfs_alloc_seqid,
9283         .call_sync_ops = &nfs40_call_sync_ops,
9284         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9285         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9286         .state_renewal_ops = &nfs40_state_renewal_ops,
9287         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9288 };
9289
9290 #if defined(CONFIG_NFS_V4_1)
9291 static struct nfs_seqid *
9292 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9293 {
9294         return NULL;
9295 }
9296
9297 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9298         .minor_version = 1,
9299         .init_caps = NFS_CAP_READDIRPLUS
9300                 | NFS_CAP_ATOMIC_OPEN
9301                 | NFS_CAP_POSIX_LOCK
9302                 | NFS_CAP_STATEID_NFSV41
9303                 | NFS_CAP_ATOMIC_OPEN_V1,
9304         .init_client = nfs41_init_client,
9305         .shutdown_client = nfs41_shutdown_client,
9306         .match_stateid = nfs41_match_stateid,
9307         .find_root_sec = nfs41_find_root_sec,
9308         .free_lock_state = nfs41_free_lock_state,
9309         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9310         .alloc_seqid = nfs_alloc_no_seqid,
9311         .session_trunk = nfs4_test_session_trunk,
9312         .call_sync_ops = &nfs41_call_sync_ops,
9313         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9314         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9315         .state_renewal_ops = &nfs41_state_renewal_ops,
9316         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9317 };
9318 #endif
9319
9320 #if defined(CONFIG_NFS_V4_2)
9321 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9322         .minor_version = 2,
9323         .init_caps = NFS_CAP_READDIRPLUS
9324                 | NFS_CAP_ATOMIC_OPEN
9325                 | NFS_CAP_POSIX_LOCK
9326                 | NFS_CAP_STATEID_NFSV41
9327                 | NFS_CAP_ATOMIC_OPEN_V1
9328                 | NFS_CAP_ALLOCATE
9329                 | NFS_CAP_COPY
9330                 | NFS_CAP_DEALLOCATE
9331                 | NFS_CAP_SEEK
9332                 | NFS_CAP_LAYOUTSTATS
9333                 | NFS_CAP_CLONE,
9334         .init_client = nfs41_init_client,
9335         .shutdown_client = nfs41_shutdown_client,
9336         .match_stateid = nfs41_match_stateid,
9337         .find_root_sec = nfs41_find_root_sec,
9338         .free_lock_state = nfs41_free_lock_state,
9339         .call_sync_ops = &nfs41_call_sync_ops,
9340         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9341         .alloc_seqid = nfs_alloc_no_seqid,
9342         .session_trunk = nfs4_test_session_trunk,
9343         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9344         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9345         .state_renewal_ops = &nfs41_state_renewal_ops,
9346         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9347 };
9348 #endif
9349
9350 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9351         [0] = &nfs_v4_0_minor_ops,
9352 #if defined(CONFIG_NFS_V4_1)
9353         [1] = &nfs_v4_1_minor_ops,
9354 #endif
9355 #if defined(CONFIG_NFS_V4_2)
9356         [2] = &nfs_v4_2_minor_ops,
9357 #endif
9358 };
9359
9360 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9361 {
9362         ssize_t error, error2;
9363
9364         error = generic_listxattr(dentry, list, size);
9365         if (error < 0)
9366                 return error;
9367         if (list) {
9368                 list += error;
9369                 size -= error;
9370         }
9371
9372         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9373         if (error2 < 0)
9374                 return error2;
9375         return error + error2;
9376 }
9377
9378 static const struct inode_operations nfs4_dir_inode_operations = {
9379         .create         = nfs_create,
9380         .lookup         = nfs_lookup,
9381         .atomic_open    = nfs_atomic_open,
9382         .link           = nfs_link,
9383         .unlink         = nfs_unlink,
9384         .symlink        = nfs_symlink,
9385         .mkdir          = nfs_mkdir,
9386         .rmdir          = nfs_rmdir,
9387         .mknod          = nfs_mknod,
9388         .rename         = nfs_rename,
9389         .permission     = nfs_permission,
9390         .getattr        = nfs_getattr,
9391         .setattr        = nfs_setattr,
9392         .listxattr      = nfs4_listxattr,
9393 };
9394
9395 static const struct inode_operations nfs4_file_inode_operations = {
9396         .permission     = nfs_permission,
9397         .getattr        = nfs_getattr,
9398         .setattr        = nfs_setattr,
9399         .listxattr      = nfs4_listxattr,
9400 };
9401
9402 const struct nfs_rpc_ops nfs_v4_clientops = {
9403         .version        = 4,                    /* protocol version */
9404         .dentry_ops     = &nfs4_dentry_operations,
9405         .dir_inode_ops  = &nfs4_dir_inode_operations,
9406         .file_inode_ops = &nfs4_file_inode_operations,
9407         .file_ops       = &nfs4_file_operations,
9408         .getroot        = nfs4_proc_get_root,
9409         .submount       = nfs4_submount,
9410         .try_mount      = nfs4_try_mount,
9411         .getattr        = nfs4_proc_getattr,
9412         .setattr        = nfs4_proc_setattr,
9413         .lookup         = nfs4_proc_lookup,
9414         .access         = nfs4_proc_access,
9415         .readlink       = nfs4_proc_readlink,
9416         .create         = nfs4_proc_create,
9417         .remove         = nfs4_proc_remove,
9418         .unlink_setup   = nfs4_proc_unlink_setup,
9419         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9420         .unlink_done    = nfs4_proc_unlink_done,
9421         .rename_setup   = nfs4_proc_rename_setup,
9422         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9423         .rename_done    = nfs4_proc_rename_done,
9424         .link           = nfs4_proc_link,
9425         .symlink        = nfs4_proc_symlink,
9426         .mkdir          = nfs4_proc_mkdir,
9427         .rmdir          = nfs4_proc_remove,
9428         .readdir        = nfs4_proc_readdir,
9429         .mknod          = nfs4_proc_mknod,
9430         .statfs         = nfs4_proc_statfs,
9431         .fsinfo         = nfs4_proc_fsinfo,
9432         .pathconf       = nfs4_proc_pathconf,
9433         .set_capabilities = nfs4_server_capabilities,
9434         .decode_dirent  = nfs4_decode_dirent,
9435         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9436         .read_setup     = nfs4_proc_read_setup,
9437         .read_done      = nfs4_read_done,
9438         .write_setup    = nfs4_proc_write_setup,
9439         .write_done     = nfs4_write_done,
9440         .commit_setup   = nfs4_proc_commit_setup,
9441         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9442         .commit_done    = nfs4_commit_done,
9443         .lock           = nfs4_proc_lock,
9444         .clear_acl_cache = nfs4_zap_acl_attr,
9445         .close_context  = nfs4_close_context,
9446         .open_context   = nfs4_atomic_open,
9447         .have_delegation = nfs4_have_delegation,
9448         .return_delegation = nfs4_inode_return_delegation,
9449         .alloc_client   = nfs4_alloc_client,
9450         .init_client    = nfs4_init_client,
9451         .free_client    = nfs4_free_client,
9452         .create_server  = nfs4_create_server,
9453         .clone_server   = nfs_clone_server,
9454 };
9455
9456 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9457         .name   = XATTR_NAME_NFSV4_ACL,
9458         .list   = nfs4_xattr_list_nfs4_acl,
9459         .get    = nfs4_xattr_get_nfs4_acl,
9460         .set    = nfs4_xattr_set_nfs4_acl,
9461 };
9462
9463 const struct xattr_handler *nfs4_xattr_handlers[] = {
9464         &nfs4_xattr_nfs4_acl_handler,
9465 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9466         &nfs4_xattr_nfs4_label_handler,
9467 #endif
9468         NULL
9469 };
9470
9471 /*
9472  * Local variables:
9473  *  c-basic-offset: 8
9474  * End:
9475  */