fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 /* file attributes which can be mapped to nfs attributes */
  78 #define NFS4_VALID_ATTRS (ATTR_MODE \
  79         | ATTR_UID \
  80         | ATTR_GID \
  81         | ATTR_SIZE \
  82         | ATTR_ATIME \
  83         | ATTR_MTIME \
  84         | ATTR_CTIME \
  85         | ATTR_ATIME_SET \
  86         | ATTR_MTIME_SET)
  87
  88 struct nfs4_opendata;
  89 static int _nfs4_proc_open(struct nfs4_opendata *data);
  90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs4_state *state, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 101                 struct rpc_cred *);
 102 static int nfs41_free_stateid(struct nfs_server *, 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_CHANGE
 230         | FATTR4_WORD0_FILEID,
 231 };
 232
 233 const u32 nfs4_statfs_bitmap[3] = {
 234         FATTR4_WORD0_FILES_AVAIL
 235         | FATTR4_WORD0_FILES_FREE
 236         | FATTR4_WORD0_FILES_TOTAL,
 237         FATTR4_WORD1_SPACE_AVAIL
 238         | FATTR4_WORD1_SPACE_FREE
 239         | FATTR4_WORD1_SPACE_TOTAL
 240 };
 241
 242 const u32 nfs4_pathconf_bitmap[3] = {
 243         FATTR4_WORD0_MAXLINK
 244         | FATTR4_WORD0_MAXNAME,
 245         0
 246 };
 247
 248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 249                         | FATTR4_WORD0_MAXREAD
 250                         | FATTR4_WORD0_MAXWRITE
 251                         | FATTR4_WORD0_LEASE_TIME,
 252                         FATTR4_WORD1_TIME_DELTA
 253                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 254                         FATTR4_WORD2_LAYOUT_BLKSIZE
 255                         | FATTR4_WORD2_CLONE_BLKSIZE
 256 };
 257
 258 const u32 nfs4_fs_locations_bitmap[3] = {
 259         FATTR4_WORD0_TYPE
 260         | FATTR4_WORD0_CHANGE
 261         | FATTR4_WORD0_SIZE
 262         | FATTR4_WORD0_FSID
 263         | FATTR4_WORD0_FILEID
 264         | FATTR4_WORD0_FS_LOCATIONS,
 265         FATTR4_WORD1_MODE
 266         | FATTR4_WORD1_NUMLINKS
 267         | FATTR4_WORD1_OWNER
 268         | FATTR4_WORD1_OWNER_GROUP
 269         | FATTR4_WORD1_RAWDEV
 270         | FATTR4_WORD1_SPACE_USED
 271         | FATTR4_WORD1_TIME_ACCESS
 272         | FATTR4_WORD1_TIME_METADATA
 273         | FATTR4_WORD1_TIME_MODIFY
 274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 275 };
 276
 277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 278                 struct nfs4_readdir_arg *readdir)
 279 {
 280         __be32 *start, *p;
 281
 282         if (cookie > 2) {
 283                 readdir->cookie = cookie;
 284                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 285                 return;
 286         }
 287
 288         readdir->cookie = 0;
 289         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 290         if (cookie == 2)
 291                 return;
 292
 293         /*
 294          * NFSv4 servers do not return entries for '.' and '..'
 295          * Therefore, we fake these entries here.  We let '.'
 296          * have cookie 0 and '..' have cookie 1.  Note that
 297          * when talking to the server, we always send cookie 0
 298          * instead of 1 or 2.
 299          */
 300         start = p = kmap_atomic(*readdir->pages);
 301
 302         if (cookie == 0) {
 303                 *p++ = xdr_one;                                  /* next */
 304                 *p++ = xdr_zero;                   /* cookie, first word */
 305                 *p++ = xdr_one;                   /* cookie, second word */
 306                 *p++ = xdr_one;                             /* entry len */
 307                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 308                 p++;
 309                 *p++ = xdr_one;                         /* bitmap length */
 310                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 311                 *p++ = htonl(8);              /* attribute buffer length */
 312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 313         }
 314
 315         *p++ = xdr_one;                                  /* next */
 316         *p++ = xdr_zero;                   /* cookie, first word */
 317         *p++ = xdr_two;                   /* cookie, second word */
 318         *p++ = xdr_two;                             /* entry len */
 319         memcpy(p, "..\0\0", 4);                         /* entry */
 320         p++;
 321         *p++ = xdr_one;                         /* bitmap length */
 322         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 323         *p++ = htonl(8);              /* attribute buffer length */
 324         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 325
 326         readdir->pgbase = (char *)p - (char *)start;
 327         readdir->count -= readdir->pgbase;
 328         kunmap_atomic(start);
 329 }
 330
 331 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 332                 nfs4_stateid *stateid,
 333                 struct rpc_cred *cred)
 334 {
 335         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 336
 337         ops->test_and_free_expired(server, stateid, cred);
 338 }
 339
 340 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 341                 nfs4_stateid *stateid,
 342                 struct rpc_cred *cred)
 343 {
 344         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 345         nfs4_test_and_free_stateid(server, stateid, cred);
 346 }
 347
 348 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 349                 const nfs4_stateid *stateid,
 350                 struct rpc_cred *cred)
 351 {
 352         nfs4_stateid tmp;
 353
 354         nfs4_stateid_copy(&tmp, stateid);
 355         __nfs4_free_revoked_stateid(server, &tmp, cred);
 356 }
 357
 358 static long nfs4_update_delay(long *timeout)
 359 {
 360         long ret;
 361         if (!timeout)
 362                 return NFS4_POLL_RETRY_MAX;
 363         if (*timeout <= 0)
 364                 *timeout = NFS4_POLL_RETRY_MIN;
 365         if (*timeout > NFS4_POLL_RETRY_MAX)
 366                 *timeout = NFS4_POLL_RETRY_MAX;
 367         ret = *timeout;
 368         *timeout <<= 1;
 369         return ret;
 370 }
 371
 372 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 373 {
 374         int res = 0;
 375
 376         might_sleep();
 377
 378         freezable_schedule_timeout_killable_unsafe(
 379                 nfs4_update_delay(timeout));
 380         if (fatal_signal_pending(current))
 381                 res = -ERESTARTSYS;
 382         return res;
 383 }
 384
 385 /* This is the error handling routine for processes that are allowed
 386  * to sleep.
 387  */
 388 static int nfs4_do_handle_exception(struct nfs_server *server,
 389                 int errorcode, struct nfs4_exception *exception)
 390 {
 391         struct nfs_client *clp = server->nfs_client;
 392         struct nfs4_state *state = exception->state;
 393         const nfs4_stateid *stateid = exception->stateid;
 394         struct inode *inode = exception->inode;
 395         int ret = errorcode;
 396
 397         exception->delay = 0;
 398         exception->recovering = 0;
 399         exception->retry = 0;
 400
 401         if (stateid == NULL && state != NULL)
 402                 stateid = &state->stateid;
 403
 404         switch(errorcode) {
 405                 case 0:
 406                         return 0;
 407                 case -NFS4ERR_DELEG_REVOKED:
 408                 case -NFS4ERR_ADMIN_REVOKED:
 409                 case -NFS4ERR_EXPIRED:
 410                 case -NFS4ERR_BAD_STATEID:
 411                         if (inode != NULL && stateid != NULL) {
 412                                 nfs_inode_find_state_and_recover(inode,
 413                                                 stateid);
 414                                 goto wait_on_recovery;
 415                         }
 416                 case -NFS4ERR_OPENMODE:
 417                         if (inode) {
 418                                 int err;
 419
 420                                 err = nfs_async_inode_return_delegation(inode,
 421                                                 stateid);
 422                                 if (err == 0)
 423                                         goto wait_on_recovery;
 424                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 425                                         exception->retry = 1;
 426                                         break;
 427                                 }
 428                         }
 429                         if (state == NULL)
 430                                 break;
 431                         ret = nfs4_schedule_stateid_recovery(server, state);
 432                         if (ret < 0)
 433                                 break;
 434                         goto wait_on_recovery;
 435                 case -NFS4ERR_STALE_STATEID:
 436                 case -NFS4ERR_STALE_CLIENTID:
 437                         nfs4_schedule_lease_recovery(clp);
 438                         goto wait_on_recovery;
 439                 case -NFS4ERR_MOVED:
 440                         ret = nfs4_schedule_migration_recovery(server);
 441                         if (ret < 0)
 442                                 break;
 443                         goto wait_on_recovery;
 444                 case -NFS4ERR_LEASE_MOVED:
 445                         nfs4_schedule_lease_moved_recovery(clp);
 446                         goto wait_on_recovery;
 447 #if defined(CONFIG_NFS_V4_1)
 448                 case -NFS4ERR_BADSESSION:
 449                 case -NFS4ERR_BADSLOT:
 450                 case -NFS4ERR_BAD_HIGH_SLOT:
 451                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 452                 case -NFS4ERR_DEADSESSION:
 453                 case -NFS4ERR_SEQ_FALSE_RETRY:
 454                 case -NFS4ERR_SEQ_MISORDERED:
 455                         dprintk("%s ERROR: %d Reset session\n", __func__,
 456                                 errorcode);
 457                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 458                         goto wait_on_recovery;
 459 #endif /* defined(CONFIG_NFS_V4_1) */
 460                 case -NFS4ERR_FILE_OPEN:
 461                         if (exception->timeout > HZ) {
 462                                 /* We have retried a decent amount, time to
 463                                  * fail
 464                                  */
 465                                 ret = -EBUSY;
 466                                 break;
 467                         }
 468                 case -NFS4ERR_DELAY:
 469                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 470                 case -NFS4ERR_GRACE:
 471                 case -NFS4ERR_LAYOUTTRYLATER:
 472                 case -NFS4ERR_RECALLCONFLICT:
 473                         exception->delay = 1;
 474                         return 0;
 475
 476                 case -NFS4ERR_RETRY_UNCACHED_REP:
 477                 case -NFS4ERR_OLD_STATEID:
 478                         exception->retry = 1;
 479                         break;
 480                 case -NFS4ERR_BADOWNER:
 481                         /* The following works around a Linux server bug! */
 482                 case -NFS4ERR_BADNAME:
 483                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 484                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 485                                 exception->retry = 1;
 486                                 printk(KERN_WARNING "NFS: v4 server %s "
 487                                                 "does not accept raw "
 488                                                 "uid/gids. "
 489                                                 "Reenabling the idmapper.\n",
 490                                                 server->nfs_client->cl_hostname);
 491                         }
 492         }
 493         /* We failed to handle the error */
 494         return nfs4_map_errors(ret);
 495 wait_on_recovery:
 496         exception->recovering = 1;
 497         return 0;
 498 }
 499
 500 /* This is the error handling routine for processes that are allowed
 501  * to sleep.
 502  */
 503 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 504 {
 505         struct nfs_client *clp = server->nfs_client;
 506         int ret;
 507
 508         ret = nfs4_do_handle_exception(server, errorcode, exception);
 509         if (exception->delay) {
 510                 ret = nfs4_delay(server->client, &exception->timeout);
 511                 goto out_retry;
 512         }
 513         if (exception->recovering) {
 514                 ret = nfs4_wait_clnt_recover(clp);
 515                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 516                         return -EIO;
 517                 goto out_retry;
 518         }
 519         return ret;
 520 out_retry:
 521         if (ret == 0)
 522                 exception->retry = 1;
 523         return ret;
 524 }
 525
 526 static int
 527 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 528                 int errorcode, struct nfs4_exception *exception)
 529 {
 530         struct nfs_client *clp = server->nfs_client;
 531         int ret;
 532
 533         ret = nfs4_do_handle_exception(server, errorcode, exception);
 534         if (exception->delay) {
 535                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 536                 goto out_retry;
 537         }
 538         if (exception->recovering) {
 539                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 540                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 541                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 542                 goto out_retry;
 543         }
 544         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 545                 ret = -EIO;
 546         return ret;
 547 out_retry:
 548         if (ret == 0)
 549                 exception->retry = 1;
 550         return ret;
 551 }
 552
 553 static int
 554 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 555                         struct nfs4_state *state, long *timeout)
 556 {
 557         struct nfs4_exception exception = {
 558                 .state = state,
 559         };
 560
 561         if (task->tk_status >= 0)
 562                 return 0;
 563         if (timeout)
 564                 exception.timeout = *timeout;
 565         task->tk_status = nfs4_async_handle_exception(task, server,
 566                         task->tk_status,
 567                         &exception);
 568         if (exception.delay && timeout)
 569                 *timeout = exception.timeout;
 570         if (exception.retry)
 571                 return -EAGAIN;
 572         return 0;
 573 }
 574
 575 /*
 576  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 577  * or 'false' otherwise.
 578  */
 579 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 580 {
 581         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 582
 583         if (flavor == RPC_AUTH_GSS_KRB5I ||
 584             flavor == RPC_AUTH_GSS_KRB5P)
 585                 return true;
 586
 587         return false;
 588 }
 589
 590 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 591 {
 592         spin_lock(&clp->cl_lock);
 593         if (time_before(clp->cl_last_renewal,timestamp))
 594                 clp->cl_last_renewal = timestamp;
 595         spin_unlock(&clp->cl_lock);
 596 }
 597
 598 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 599 {
 600         struct nfs_client *clp = server->nfs_client;
 601
 602         if (!nfs4_has_session(clp))
 603                 do_renew_lease(clp, timestamp);
 604 }
 605
 606 struct nfs4_call_sync_data {
 607         const struct nfs_server *seq_server;
 608         struct nfs4_sequence_args *seq_args;
 609         struct nfs4_sequence_res *seq_res;
 610 };
 611
 612 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 613                         struct nfs4_sequence_res *res, int cache_reply)
 614 {
 615         args->sa_slot = NULL;
 616         args->sa_cache_this = cache_reply;
 617         args->sa_privileged = 0;
 618
 619         res->sr_slot = NULL;
 620 }
 621
 622 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 623 {
 624         args->sa_privileged = 1;
 625 }
 626
 627 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 628                          struct nfs4_sequence_args *args,
 629                          struct nfs4_sequence_res *res,
 630                          struct rpc_task *task)
 631 {
 632         struct nfs4_slot *slot;
 633
 634         /* slot already allocated? */
 635         if (res->sr_slot != NULL)
 636                 goto out_start;
 637
 638         spin_lock(&tbl->slot_tbl_lock);
 639         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 640                 goto out_sleep;
 641
 642         slot = nfs4_alloc_slot(tbl);
 643         if (IS_ERR(slot)) {
 644                 if (slot == ERR_PTR(-ENOMEM))
 645                         task->tk_timeout = HZ >> 2;
 646                 goto out_sleep;
 647         }
 648         spin_unlock(&tbl->slot_tbl_lock);
 649
 650         slot->privileged = args->sa_privileged ? 1 : 0;
 651         args->sa_slot = slot;
 652         res->sr_slot = slot;
 653
 654 out_start:
 655         rpc_call_start(task);
 656         return 0;
 657
 658 out_sleep:
 659         if (args->sa_privileged)
 660                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 661                                 NULL, RPC_PRIORITY_PRIVILEGED);
 662         else
 663                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 664         spin_unlock(&tbl->slot_tbl_lock);
 665         return -EAGAIN;
 666 }
 667 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 668
 669 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 670 {
 671         struct nfs4_slot *slot = res->sr_slot;
 672         struct nfs4_slot_table *tbl;
 673
 674         tbl = slot->table;
 675         spin_lock(&tbl->slot_tbl_lock);
 676         if (!nfs41_wake_and_assign_slot(tbl, slot))
 677                 nfs4_free_slot(tbl, slot);
 678         spin_unlock(&tbl->slot_tbl_lock);
 679
 680         res->sr_slot = NULL;
 681 }
 682
 683 static int nfs40_sequence_done(struct rpc_task *task,
 684                                struct nfs4_sequence_res *res)
 685 {
 686         if (res->sr_slot != NULL)
 687                 nfs40_sequence_free_slot(res);
 688         return 1;
 689 }
 690
 691 #if defined(CONFIG_NFS_V4_1)
 692
 693 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 694 {
 695         struct nfs4_session *session;
 696         struct nfs4_slot_table *tbl;
 697         struct nfs4_slot *slot = res->sr_slot;
 698         bool send_new_highest_used_slotid = false;
 699
 700         tbl = slot->table;
 701         session = tbl->session;
 702
 703         /* Bump the slot sequence number */
 704         if (slot->seq_done)
 705                 slot->seq_nr++;
 706         slot->seq_done = 0;
 707
 708         spin_lock(&tbl->slot_tbl_lock);
 709         /* Be nice to the server: try to ensure that the last transmitted
 710          * value for highest_user_slotid <= target_highest_slotid
 711          */
 712         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 713                 send_new_highest_used_slotid = true;
 714
 715         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 716                 send_new_highest_used_slotid = false;
 717                 goto out_unlock;
 718         }
 719         nfs4_free_slot(tbl, slot);
 720
 721         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 722                 send_new_highest_used_slotid = false;
 723 out_unlock:
 724         spin_unlock(&tbl->slot_tbl_lock);
 725         res->sr_slot = NULL;
 726         if (send_new_highest_used_slotid)
 727                 nfs41_notify_server(session->clp);
 728         if (waitqueue_active(&tbl->slot_waitq))
 729                 wake_up_all(&tbl->slot_waitq);
 730 }
 731
 732 static int nfs41_sequence_process(struct rpc_task *task,
 733                 struct nfs4_sequence_res *res)
 734 {
 735         struct nfs4_session *session;
 736         struct nfs4_slot *slot = res->sr_slot;
 737         struct nfs_client *clp;
 738         bool interrupted = false;
 739         int ret = 1;
 740
 741         if (slot == NULL)
 742                 goto out_noaction;
 743         /* don't increment the sequence number if the task wasn't sent */
 744         if (!RPC_WAS_SENT(task))
 745                 goto out;
 746
 747         session = slot->table->session;
 748
 749         if (slot->interrupted) {
 750                 slot->interrupted = 0;
 751                 interrupted = true;
 752         }
 753
 754         trace_nfs4_sequence_done(session, res);
 755         /* Check the SEQUENCE operation status */
 756         switch (res->sr_status) {
 757         case 0:
 758                 /* If previous op on slot was interrupted and we reused
 759                  * the seq# and got a reply from the cache, then retry
 760                  */
 761                 if (task->tk_status == -EREMOTEIO && interrupted) {
 762                         ++slot->seq_nr;
 763                         goto retry_nowait;
 764                 }
 765                 /* Update the slot's sequence and clientid lease timer */
 766                 slot->seq_done = 1;
 767                 clp = session->clp;
 768                 do_renew_lease(clp, res->sr_timestamp);
 769                 /* Check sequence flags */
 770                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 771                                 !!slot->privileged);
 772                 nfs41_update_target_slotid(slot->table, slot, res);
 773                 break;
 774         case 1:
 775                 /*
 776                  * sr_status remains 1 if an RPC level error occurred.
 777                  * The server may or may not have processed the sequence
 778                  * operation..
 779                  * Mark the slot as having hosted an interrupted RPC call.
 780                  */
 781                 slot->interrupted = 1;
 782                 goto out;
 783         case -NFS4ERR_DELAY:
 784                 /* The server detected a resend of the RPC call and
 785                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 786                  * of RFC5661.
 787                  */
 788                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 789                         __func__,
 790                         slot->slot_nr,
 791                         slot->seq_nr);
 792                 goto out_retry;
 793         case -NFS4ERR_BADSLOT:
 794                 /*
 795                  * The slot id we used was probably retired. Try again
 796                  * using a different slot id.
 797                  */
 798                 goto retry_nowait;
 799         case -NFS4ERR_SEQ_MISORDERED:
 800                 /*
 801                  * Was the last operation on this sequence interrupted?
 802                  * If so, retry after bumping the sequence number.
 803                  */
 804                 if (interrupted) {
 805                         ++slot->seq_nr;
 806                         goto retry_nowait;
 807                 }
 808                 /*
 809                  * Could this slot have been previously retired?
 810                  * If so, then the server may be expecting seq_nr = 1!
 811                  */
 812                 if (slot->seq_nr != 1) {
 813                         slot->seq_nr = 1;
 814                         goto retry_nowait;
 815                 }
 816                 break;
 817         case -NFS4ERR_SEQ_FALSE_RETRY:
 818                 ++slot->seq_nr;
 819                 goto retry_nowait;
 820         default:
 821                 /* Just update the slot sequence no. */
 822                 slot->seq_done = 1;
 823         }
 824 out:
 825         /* The session may be reset by one of the error handlers. */
 826         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 827 out_noaction:
 828         return ret;
 829 retry_nowait:
 830         if (rpc_restart_call_prepare(task)) {
 831                 nfs41_sequence_free_slot(res);
 832                 task->tk_status = 0;
 833                 ret = 0;
 834         }
 835         goto out;
 836 out_retry:
 837         if (!rpc_restart_call(task))
 838                 goto out;
 839         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 840         return 0;
 841 }
 842
 843 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 844 {
 845         if (!nfs41_sequence_process(task, res))
 846                 return 0;
 847         if (res->sr_slot != NULL)
 848                 nfs41_sequence_free_slot(res);
 849         return 1;
 850
 851 }
 852 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 853
 854 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 855 {
 856         if (res->sr_slot == NULL)
 857                 return 1;
 858         if (res->sr_slot->table->session != NULL)
 859                 return nfs41_sequence_process(task, res);
 860         return nfs40_sequence_done(task, res);
 861 }
 862
 863 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 864 {
 865         if (res->sr_slot != NULL) {
 866                 if (res->sr_slot->table->session != NULL)
 867                         nfs41_sequence_free_slot(res);
 868                 else
 869                         nfs40_sequence_free_slot(res);
 870         }
 871 }
 872
 873 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 874 {
 875         if (res->sr_slot == NULL)
 876                 return 1;
 877         if (!res->sr_slot->table->session)
 878                 return nfs40_sequence_done(task, res);
 879         return nfs41_sequence_done(task, res);
 880 }
 881 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 882
 883 int nfs41_setup_sequence(struct nfs4_session *session,
 884                                 struct nfs4_sequence_args *args,
 885                                 struct nfs4_sequence_res *res,
 886                                 struct rpc_task *task)
 887 {
 888         struct nfs4_slot *slot;
 889         struct nfs4_slot_table *tbl;
 890
 891         dprintk("--> %s\n", __func__);
 892         /* slot already allocated? */
 893         if (res->sr_slot != NULL)
 894                 goto out_success;
 895
 896         tbl = &session->fc_slot_table;
 897
 898         task->tk_timeout = 0;
 899
 900         spin_lock(&tbl->slot_tbl_lock);
 901         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 902             !args->sa_privileged) {
 903                 /* The state manager will wait until the slot table is empty */
 904                 dprintk("%s session is draining\n", __func__);
 905                 goto out_sleep;
 906         }
 907
 908         slot = nfs4_alloc_slot(tbl);
 909         if (IS_ERR(slot)) {
 910                 /* If out of memory, try again in 1/4 second */
 911                 if (slot == ERR_PTR(-ENOMEM))
 912                         task->tk_timeout = HZ >> 2;
 913                 dprintk("<-- %s: no free slots\n", __func__);
 914                 goto out_sleep;
 915         }
 916         spin_unlock(&tbl->slot_tbl_lock);
 917
 918         slot->privileged = args->sa_privileged ? 1 : 0;
 919         args->sa_slot = slot;
 920
 921         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 922                         slot->slot_nr, slot->seq_nr);
 923
 924         res->sr_slot = slot;
 925         res->sr_timestamp = jiffies;
 926         res->sr_status_flags = 0;
 927         /*
 928          * sr_status is only set in decode_sequence, and so will remain
 929          * set to 1 if an rpc level failure occurs.
 930          */
 931         res->sr_status = 1;
 932         trace_nfs4_setup_sequence(session, args);
 933 out_success:
 934         rpc_call_start(task);
 935         return 0;
 936 out_sleep:
 937         /* Privileged tasks are queued with top priority */
 938         if (args->sa_privileged)
 939                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 940                                 NULL, RPC_PRIORITY_PRIVILEGED);
 941         else
 942                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 943         spin_unlock(&tbl->slot_tbl_lock);
 944         return -EAGAIN;
 945 }
 946 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 947
 948 static int nfs4_setup_sequence(const struct nfs_server *server,
 949                                struct nfs4_sequence_args *args,
 950                                struct nfs4_sequence_res *res,
 951                                struct rpc_task *task)
 952 {
 953         struct nfs4_session *session = nfs4_get_session(server);
 954         int ret = 0;
 955
 956         if (!session)
 957                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 958                                             args, res, task);
 959
 960         dprintk("--> %s clp %p session %p sr_slot %u\n",
 961                 __func__, session->clp, session, res->sr_slot ?
 962                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 963
 964         ret = nfs41_setup_sequence(session, args, res, task);
 965
 966         dprintk("<-- %s status=%d\n", __func__, ret);
 967         return ret;
 968 }
 969
 970 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 971 {
 972         struct nfs4_call_sync_data *data = calldata;
 973         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 974
 975         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 976
 977         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 978 }
 979
 980 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 981 {
 982         struct nfs4_call_sync_data *data = calldata;
 983
 984         nfs41_sequence_done(task, data->seq_res);
 985 }
 986
 987 static const struct rpc_call_ops nfs41_call_sync_ops = {
 988         .rpc_call_prepare = nfs41_call_sync_prepare,
 989         .rpc_call_done = nfs41_call_sync_done,
 990 };
 991
 992 #else   /* !CONFIG_NFS_V4_1 */
 993
 994 static int nfs4_setup_sequence(const struct nfs_server *server,
 995                                struct nfs4_sequence_args *args,
 996                                struct nfs4_sequence_res *res,
 997                                struct rpc_task *task)
 998 {
 999         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
1000                                     args, res, task);
1001 }
1002
1003 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1004 {
1005         return nfs40_sequence_done(task, res);
1006 }
1007
1008 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1009 {
1010         if (res->sr_slot != NULL)
1011                 nfs40_sequence_free_slot(res);
1012 }
1013
1014 int nfs4_sequence_done(struct rpc_task *task,
1015                        struct nfs4_sequence_res *res)
1016 {
1017         return nfs40_sequence_done(task, res);
1018 }
1019 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1020
1021 #endif  /* !CONFIG_NFS_V4_1 */
1022
1023 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1024 {
1025         struct nfs4_call_sync_data *data = calldata;
1026         nfs4_setup_sequence(data->seq_server,
1027                                 data->seq_args, data->seq_res, task);
1028 }
1029
1030 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1031 {
1032         struct nfs4_call_sync_data *data = calldata;
1033         nfs4_sequence_done(task, data->seq_res);
1034 }
1035
1036 static const struct rpc_call_ops nfs40_call_sync_ops = {
1037         .rpc_call_prepare = nfs40_call_sync_prepare,
1038         .rpc_call_done = nfs40_call_sync_done,
1039 };
1040
1041 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1042                                    struct nfs_server *server,
1043                                    struct rpc_message *msg,
1044                                    struct nfs4_sequence_args *args,
1045                                    struct nfs4_sequence_res *res)
1046 {
1047         int ret;
1048         struct rpc_task *task;
1049         struct nfs_client *clp = server->nfs_client;
1050         struct nfs4_call_sync_data data = {
1051                 .seq_server = server,
1052                 .seq_args = args,
1053                 .seq_res = res,
1054         };
1055         struct rpc_task_setup task_setup = {
1056                 .rpc_client = clnt,
1057                 .rpc_message = msg,
1058                 .callback_ops = clp->cl_mvops->call_sync_ops,
1059                 .callback_data = &data
1060         };
1061
1062         task = rpc_run_task(&task_setup);
1063         if (IS_ERR(task))
1064                 ret = PTR_ERR(task);
1065         else {
1066                 ret = task->tk_status;
1067                 rpc_put_task(task);
1068         }
1069         return ret;
1070 }
1071
1072 int nfs4_call_sync(struct rpc_clnt *clnt,
1073                    struct nfs_server *server,
1074                    struct rpc_message *msg,
1075                    struct nfs4_sequence_args *args,
1076                    struct nfs4_sequence_res *res,
1077                    int cache_reply)
1078 {
1079         nfs4_init_sequence(args, res, cache_reply);
1080         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1081 }
1082
1083 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1084 {
1085         struct nfs_inode *nfsi = NFS_I(dir);
1086
1087         spin_lock(&dir->i_lock);
1088         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1089         if (!cinfo->atomic || cinfo->before != dir->i_version)
1090                 nfs_force_lookup_revalidate(dir);
1091         dir->i_version = cinfo->after;
1092         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1093         nfs_fscache_invalidate(dir);
1094         spin_unlock(&dir->i_lock);
1095 }
1096
1097 struct nfs4_opendata {
1098         struct kref kref;
1099         struct nfs_openargs o_arg;
1100         struct nfs_openres o_res;
1101         struct nfs_open_confirmargs c_arg;
1102         struct nfs_open_confirmres c_res;
1103         struct nfs4_string owner_name;
1104         struct nfs4_string group_name;
1105         struct nfs4_label *a_label;
1106         struct nfs_fattr f_attr;
1107         struct nfs4_label *f_label;
1108         struct dentry *dir;
1109         struct dentry *dentry;
1110         struct nfs4_state_owner *owner;
1111         struct nfs4_state *state;
1112         struct iattr attrs;
1113         unsigned long timestamp;
1114         unsigned int rpc_done : 1;
1115         unsigned int file_created : 1;
1116         unsigned int is_recover : 1;
1117         int rpc_status;
1118         int cancelled;
1119 };
1120
1121 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1122                 int err, struct nfs4_exception *exception)
1123 {
1124         if (err != -EINVAL)
1125                 return false;
1126         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1127                 return false;
1128         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1129         exception->retry = 1;
1130         return true;
1131 }
1132
1133 static u32
1134 nfs4_map_atomic_open_share(struct nfs_server *server,
1135                 fmode_t fmode, int openflags)
1136 {
1137         u32 res = 0;
1138
1139         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1140         case FMODE_READ:
1141                 res = NFS4_SHARE_ACCESS_READ;
1142                 break;
1143         case FMODE_WRITE:
1144                 res = NFS4_SHARE_ACCESS_WRITE;
1145                 break;
1146         case FMODE_READ|FMODE_WRITE:
1147                 res = NFS4_SHARE_ACCESS_BOTH;
1148         }
1149         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1150                 goto out;
1151         /* Want no delegation if we're using O_DIRECT */
1152         if (openflags & O_DIRECT)
1153                 res |= NFS4_SHARE_WANT_NO_DELEG;
1154 out:
1155         return res;
1156 }
1157
1158 static enum open_claim_type4
1159 nfs4_map_atomic_open_claim(struct nfs_server *server,
1160                 enum open_claim_type4 claim)
1161 {
1162         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1163                 return claim;
1164         switch (claim) {
1165         default:
1166                 return claim;
1167         case NFS4_OPEN_CLAIM_FH:
1168                 return NFS4_OPEN_CLAIM_NULL;
1169         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1170                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1171         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1172                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1173         }
1174 }
1175
1176 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1177 {
1178         p->o_res.f_attr = &p->f_attr;
1179         p->o_res.f_label = p->f_label;
1180         p->o_res.seqid = p->o_arg.seqid;
1181         p->c_res.seqid = p->c_arg.seqid;
1182         p->o_res.server = p->o_arg.server;
1183         p->o_res.access_request = p->o_arg.access;
1184         nfs_fattr_init(&p->f_attr);
1185         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1186 }
1187
1188 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1189                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1190                 const struct iattr *attrs,
1191                 struct nfs4_label *label,
1192                 enum open_claim_type4 claim,
1193                 gfp_t gfp_mask)
1194 {
1195         struct dentry *parent = dget_parent(dentry);
1196         struct inode *dir = d_inode(parent);
1197         struct nfs_server *server = NFS_SERVER(dir);
1198         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1199         struct nfs4_opendata *p;
1200
1201         p = kzalloc(sizeof(*p), gfp_mask);
1202         if (p == NULL)
1203                 goto err;
1204
1205         p->f_label = nfs4_label_alloc(server, gfp_mask);
1206         if (IS_ERR(p->f_label))
1207                 goto err_free_p;
1208
1209         p->a_label = nfs4_label_alloc(server, gfp_mask);
1210         if (IS_ERR(p->a_label))
1211                 goto err_free_f;
1212
1213         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1214         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1215         if (IS_ERR(p->o_arg.seqid))
1216                 goto err_free_label;
1217         nfs_sb_active(dentry->d_sb);
1218         p->dentry = dget(dentry);
1219         p->dir = parent;
1220         p->owner = sp;
1221         atomic_inc(&sp->so_count);
1222         p->o_arg.open_flags = flags;
1223         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1224         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1225                         fmode, flags);
1226         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1227          * will return permission denied for all bits until close */
1228         if (!(flags & O_EXCL)) {
1229                 /* ask server to check for all possible rights as results
1230                  * are cached */
1231                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1232                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1233         }
1234         p->o_arg.clientid = server->nfs_client->cl_clientid;
1235         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1236         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1237         p->o_arg.name = &dentry->d_name;
1238         p->o_arg.server = server;
1239         p->o_arg.bitmask = nfs4_bitmask(server, label);
1240         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1241         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1242         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1243         switch (p->o_arg.claim) {
1244         case NFS4_OPEN_CLAIM_NULL:
1245         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1246         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1247                 p->o_arg.fh = NFS_FH(dir);
1248                 break;
1249         case NFS4_OPEN_CLAIM_PREVIOUS:
1250         case NFS4_OPEN_CLAIM_FH:
1251         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1252         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1253                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1254         }
1255         if (attrs != NULL && attrs->ia_valid != 0) {
1256                 __u32 verf[2];
1257
1258                 p->o_arg.u.attrs = &p->attrs;
1259                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1260
1261                 verf[0] = jiffies;
1262                 verf[1] = current->pid;
1263                 memcpy(p->o_arg.u.verifier.data, verf,
1264                                 sizeof(p->o_arg.u.verifier.data));
1265         }
1266         p->c_arg.fh = &p->o_res.fh;
1267         p->c_arg.stateid = &p->o_res.stateid;
1268         p->c_arg.seqid = p->o_arg.seqid;
1269         nfs4_init_opendata_res(p);
1270         kref_init(&p->kref);
1271         return p;
1272
1273 err_free_label:
1274         nfs4_label_free(p->a_label);
1275 err_free_f:
1276         nfs4_label_free(p->f_label);
1277 err_free_p:
1278         kfree(p);
1279 err:
1280         dput(parent);
1281         return NULL;
1282 }
1283
1284 static void nfs4_opendata_free(struct kref *kref)
1285 {
1286         struct nfs4_opendata *p = container_of(kref,
1287                         struct nfs4_opendata, kref);
1288         struct super_block *sb = p->dentry->d_sb;
1289
1290         nfs_free_seqid(p->o_arg.seqid);
1291         nfs4_sequence_free_slot(&p->o_res.seq_res);
1292         if (p->state != NULL)
1293                 nfs4_put_open_state(p->state);
1294         nfs4_put_state_owner(p->owner);
1295
1296         nfs4_label_free(p->a_label);
1297         nfs4_label_free(p->f_label);
1298
1299         dput(p->dir);
1300         dput(p->dentry);
1301         nfs_sb_deactive(sb);
1302         nfs_fattr_free_names(&p->f_attr);
1303         kfree(p->f_attr.mdsthreshold);
1304         kfree(p);
1305 }
1306
1307 static void nfs4_opendata_put(struct nfs4_opendata *p)
1308 {
1309         if (p != NULL)
1310                 kref_put(&p->kref, nfs4_opendata_free);
1311 }
1312
1313 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1314 {
1315         int ret;
1316
1317         ret = rpc_wait_for_completion_task(task);
1318         return ret;
1319 }
1320
1321 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1322                 fmode_t fmode)
1323 {
1324         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1325         case FMODE_READ|FMODE_WRITE:
1326                 return state->n_rdwr != 0;
1327         case FMODE_WRITE:
1328                 return state->n_wronly != 0;
1329         case FMODE_READ:
1330                 return state->n_rdonly != 0;
1331         }
1332         WARN_ON_ONCE(1);
1333         return false;
1334 }
1335
1336 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1337 {
1338         int ret = 0;
1339
1340         if (open_mode & (O_EXCL|O_TRUNC))
1341                 goto out;
1342         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1343                 case FMODE_READ:
1344                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1345                                 && state->n_rdonly != 0;
1346                         break;
1347                 case FMODE_WRITE:
1348                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1349                                 && state->n_wronly != 0;
1350                         break;
1351                 case FMODE_READ|FMODE_WRITE:
1352                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1353                                 && state->n_rdwr != 0;
1354         }
1355 out:
1356         return ret;
1357 }
1358
1359 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1360                 enum open_claim_type4 claim)
1361 {
1362         if (delegation == NULL)
1363                 return 0;
1364         if ((delegation->type & fmode) != fmode)
1365                 return 0;
1366         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1367                 return 0;
1368         switch (claim) {
1369         case NFS4_OPEN_CLAIM_NULL:
1370         case NFS4_OPEN_CLAIM_FH:
1371                 break;
1372         case NFS4_OPEN_CLAIM_PREVIOUS:
1373                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1374                         break;
1375         default:
1376                 return 0;
1377         }
1378         nfs_mark_delegation_referenced(delegation);
1379         return 1;
1380 }
1381
1382 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1383 {
1384         switch (fmode) {
1385                 case FMODE_WRITE:
1386                         state->n_wronly++;
1387                         break;
1388                 case FMODE_READ:
1389                         state->n_rdonly++;
1390                         break;
1391                 case FMODE_READ|FMODE_WRITE:
1392                         state->n_rdwr++;
1393         }
1394         nfs4_state_set_mode_locked(state, state->state | fmode);
1395 }
1396
1397 #ifdef CONFIG_NFS_V4_1
1398 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1399 {
1400         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1401                 return true;
1402         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1403                 return true;
1404         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1405                 return true;
1406         return false;
1407 }
1408 #endif /* CONFIG_NFS_V4_1 */
1409
1410 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1411 {
1412         struct nfs_client *clp = state->owner->so_server->nfs_client;
1413         bool need_recover = false;
1414
1415         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1416                 need_recover = true;
1417         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1418                 need_recover = true;
1419         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1420                 need_recover = true;
1421         if (need_recover)
1422                 nfs4_state_mark_reclaim_nograce(clp, state);
1423 }
1424
1425 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1426                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1427 {
1428         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1429                 return true;
1430         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1431                 nfs4_stateid_copy(freeme, &state->open_stateid);
1432                 nfs_test_and_clear_all_open_stateid(state);
1433                 return true;
1434         }
1435         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1436                 return true;
1437         return false;
1438 }
1439
1440 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1441 {
1442         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1443                 return;
1444         if (state->n_wronly)
1445                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1446         if (state->n_rdonly)
1447                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1448         if (state->n_rdwr)
1449                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1450         set_bit(NFS_OPEN_STATE, &state->flags);
1451 }
1452
1453 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1454                 nfs4_stateid *stateid, fmode_t fmode)
1455 {
1456         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1457         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1458         case FMODE_WRITE:
1459                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1460                 break;
1461         case FMODE_READ:
1462                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1463                 break;
1464         case 0:
1465                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1466                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1467                 clear_bit(NFS_OPEN_STATE, &state->flags);
1468         }
1469         if (stateid == NULL)
1470                 return;
1471         /* Handle OPEN+OPEN_DOWNGRADE races */
1472         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1473             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1474                 nfs_resync_open_stateid_locked(state);
1475                 return;
1476         }
1477         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1478                 nfs4_stateid_copy(&state->stateid, stateid);
1479         nfs4_stateid_copy(&state->open_stateid, stateid);
1480 }
1481
1482 static void nfs_clear_open_stateid(struct nfs4_state *state,
1483         nfs4_stateid *arg_stateid,
1484         nfs4_stateid *stateid, fmode_t fmode)
1485 {
1486         write_seqlock(&state->seqlock);
1487         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1488         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1489                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1490         write_sequnlock(&state->seqlock);
1491         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1492                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1493 }
1494
1495 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1496                 const nfs4_stateid *stateid, fmode_t fmode,
1497                 nfs4_stateid *freeme)
1498 {
1499         switch (fmode) {
1500                 case FMODE_READ:
1501                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1502                         break;
1503                 case FMODE_WRITE:
1504                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1505                         break;
1506                 case FMODE_READ|FMODE_WRITE:
1507                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1508         }
1509         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1510                 return;
1511         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1512                 nfs4_stateid_copy(&state->stateid, stateid);
1513         nfs4_stateid_copy(&state->open_stateid, stateid);
1514 }
1515
1516 static void __update_open_stateid(struct nfs4_state *state,
1517                 const nfs4_stateid *open_stateid,
1518                 const nfs4_stateid *deleg_stateid,
1519                 fmode_t fmode,
1520                 nfs4_stateid *freeme)
1521 {
1522         /*
1523          * Protect the call to nfs4_state_set_mode_locked and
1524          * serialise the stateid update
1525          */
1526         spin_lock(&state->owner->so_lock);
1527         write_seqlock(&state->seqlock);
1528         if (deleg_stateid != NULL) {
1529                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1530                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1531         }
1532         if (open_stateid != NULL)
1533                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1534         write_sequnlock(&state->seqlock);
1535         update_open_stateflags(state, fmode);
1536         spin_unlock(&state->owner->so_lock);
1537 }
1538
1539 static int update_open_stateid(struct nfs4_state *state,
1540                 const nfs4_stateid *open_stateid,
1541                 const nfs4_stateid *delegation,
1542                 fmode_t fmode)
1543 {
1544         struct nfs_server *server = NFS_SERVER(state->inode);
1545         struct nfs_client *clp = server->nfs_client;
1546         struct nfs_inode *nfsi = NFS_I(state->inode);
1547         struct nfs_delegation *deleg_cur;
1548         nfs4_stateid freeme = { };
1549         int ret = 0;
1550
1551         fmode &= (FMODE_READ|FMODE_WRITE);
1552
1553         rcu_read_lock();
1554         deleg_cur = rcu_dereference(nfsi->delegation);
1555         if (deleg_cur == NULL)
1556                 goto no_delegation;
1557
1558         spin_lock(&deleg_cur->lock);
1559         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1560            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1561             (deleg_cur->type & fmode) != fmode)
1562                 goto no_delegation_unlock;
1563
1564         if (delegation == NULL)
1565                 delegation = &deleg_cur->stateid;
1566         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1567                 goto no_delegation_unlock;
1568
1569         nfs_mark_delegation_referenced(deleg_cur);
1570         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1571                         fmode, &freeme);
1572         ret = 1;
1573 no_delegation_unlock:
1574         spin_unlock(&deleg_cur->lock);
1575 no_delegation:
1576         rcu_read_unlock();
1577
1578         if (!ret && open_stateid != NULL) {
1579                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1580                 ret = 1;
1581         }
1582         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1583                 nfs4_schedule_state_manager(clp);
1584         if (freeme.type != 0)
1585                 nfs4_test_and_free_stateid(server, &freeme,
1586                                 state->owner->so_cred);
1587
1588         return ret;
1589 }
1590
1591 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1592                 const nfs4_stateid *stateid)
1593 {
1594         struct nfs4_state *state = lsp->ls_state;
1595         bool ret = false;
1596
1597         spin_lock(&state->state_lock);
1598         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1599                 goto out_noupdate;
1600         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1601                 goto out_noupdate;
1602         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1603         ret = true;
1604 out_noupdate:
1605         spin_unlock(&state->state_lock);
1606         return ret;
1607 }
1608
1609 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1610 {
1611         struct nfs_delegation *delegation;
1612
1613         rcu_read_lock();
1614         delegation = rcu_dereference(NFS_I(inode)->delegation);
1615         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1616                 rcu_read_unlock();
1617                 return;
1618         }
1619         rcu_read_unlock();
1620         nfs4_inode_return_delegation(inode);
1621 }
1622
1623 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1624 {
1625         struct nfs4_state *state = opendata->state;
1626         struct nfs_inode *nfsi = NFS_I(state->inode);
1627         struct nfs_delegation *delegation;
1628         int open_mode = opendata->o_arg.open_flags;
1629         fmode_t fmode = opendata->o_arg.fmode;
1630         enum open_claim_type4 claim = opendata->o_arg.claim;
1631         nfs4_stateid stateid;
1632         int ret = -EAGAIN;
1633
1634         for (;;) {
1635                 spin_lock(&state->owner->so_lock);
1636                 if (can_open_cached(state, fmode, open_mode)) {
1637                         update_open_stateflags(state, fmode);
1638                         spin_unlock(&state->owner->so_lock);
1639                         goto out_return_state;
1640                 }
1641                 spin_unlock(&state->owner->so_lock);
1642                 rcu_read_lock();
1643                 delegation = rcu_dereference(nfsi->delegation);
1644                 if (!can_open_delegated(delegation, fmode, claim)) {
1645                         rcu_read_unlock();
1646                         break;
1647                 }
1648                 /* Save the delegation */
1649                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1650                 rcu_read_unlock();
1651                 nfs_release_seqid(opendata->o_arg.seqid);
1652                 if (!opendata->is_recover) {
1653                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1654                         if (ret != 0)
1655                                 goto out;
1656                 }
1657                 ret = -EAGAIN;
1658
1659                 /* Try to update the stateid using the delegation */
1660                 if (update_open_stateid(state, NULL, &stateid, fmode))
1661                         goto out_return_state;
1662         }
1663 out:
1664         return ERR_PTR(ret);
1665 out_return_state:
1666         atomic_inc(&state->count);
1667         return state;
1668 }
1669
1670 static void
1671 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1672 {
1673         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1674         struct nfs_delegation *delegation;
1675         int delegation_flags = 0;
1676
1677         rcu_read_lock();
1678         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1679         if (delegation)
1680                 delegation_flags = delegation->flags;
1681         rcu_read_unlock();
1682         switch (data->o_arg.claim) {
1683         default:
1684                 break;
1685         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1686         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1687                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1688                                    "returning a delegation for "
1689                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1690                                    clp->cl_hostname);
1691                 return;
1692         }
1693         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1694                 nfs_inode_set_delegation(state->inode,
1695                                          data->owner->so_cred,
1696                                          &data->o_res);
1697         else
1698                 nfs_inode_reclaim_delegation(state->inode,
1699                                              data->owner->so_cred,
1700                                              &data->o_res);
1701 }
1702
1703 /*
1704  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1705  * and update the nfs4_state.
1706  */
1707 static struct nfs4_state *
1708 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1709 {
1710         struct inode *inode = data->state->inode;
1711         struct nfs4_state *state = data->state;
1712         int ret;
1713
1714         if (!data->rpc_done) {
1715                 if (data->rpc_status) {
1716                         ret = data->rpc_status;
1717                         goto err;
1718                 }
1719                 /* cached opens have already been processed */
1720                 goto update;
1721         }
1722
1723         ret = nfs_refresh_inode(inode, &data->f_attr);
1724         if (ret)
1725                 goto err;
1726
1727         if (data->o_res.delegation_type != 0)
1728                 nfs4_opendata_check_deleg(data, state);
1729 update:
1730         update_open_stateid(state, &data->o_res.stateid, NULL,
1731                             data->o_arg.fmode);
1732         atomic_inc(&state->count);
1733
1734         return state;
1735 err:
1736         return ERR_PTR(ret);
1737
1738 }
1739
1740 static struct nfs4_state *
1741 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1742 {
1743         struct inode *inode;
1744         struct nfs4_state *state = NULL;
1745         int ret;
1746
1747         if (!data->rpc_done) {
1748                 state = nfs4_try_open_cached(data);
1749                 trace_nfs4_cached_open(data->state);
1750                 goto out;
1751         }
1752
1753         ret = -EAGAIN;
1754         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1755                 goto err;
1756         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1757         ret = PTR_ERR(inode);
1758         if (IS_ERR(inode))
1759                 goto err;
1760         ret = -ENOMEM;
1761         state = nfs4_get_open_state(inode, data->owner);
1762         if (state == NULL)
1763                 goto err_put_inode;
1764         if (data->o_res.delegation_type != 0)
1765                 nfs4_opendata_check_deleg(data, state);
1766         update_open_stateid(state, &data->o_res.stateid, NULL,
1767                         data->o_arg.fmode);
1768         iput(inode);
1769 out:
1770         nfs_release_seqid(data->o_arg.seqid);
1771         return state;
1772 err_put_inode:
1773         iput(inode);
1774 err:
1775         return ERR_PTR(ret);
1776 }
1777
1778 static struct nfs4_state *
1779 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1780 {
1781         struct nfs4_state *ret;
1782
1783         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1784                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1785         else
1786                 ret = _nfs4_opendata_to_nfs4_state(data);
1787         nfs4_sequence_free_slot(&data->o_res.seq_res);
1788         return ret;
1789 }
1790
1791 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1792 {
1793         struct nfs_inode *nfsi = NFS_I(state->inode);
1794         struct nfs_open_context *ctx;
1795
1796         spin_lock(&state->inode->i_lock);
1797         list_for_each_entry(ctx, &nfsi->open_files, list) {
1798                 if (ctx->state != state)
1799                         continue;
1800                 get_nfs_open_context(ctx);
1801                 spin_unlock(&state->inode->i_lock);
1802                 return ctx;
1803         }
1804         spin_unlock(&state->inode->i_lock);
1805         return ERR_PTR(-ENOENT);
1806 }
1807
1808 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1809                 struct nfs4_state *state, enum open_claim_type4 claim)
1810 {
1811         struct nfs4_opendata *opendata;
1812
1813         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1814                         NULL, NULL, claim, GFP_NOFS);
1815         if (opendata == NULL)
1816                 return ERR_PTR(-ENOMEM);
1817         opendata->state = state;
1818         atomic_inc(&state->count);
1819         return opendata;
1820 }
1821
1822 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1823                 fmode_t fmode)
1824 {
1825         struct nfs4_state *newstate;
1826         int ret;
1827
1828         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1829                 return 0;
1830         opendata->o_arg.open_flags = 0;
1831         opendata->o_arg.fmode = fmode;
1832         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1833                         NFS_SB(opendata->dentry->d_sb),
1834                         fmode, 0);
1835         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1836         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1837         nfs4_init_opendata_res(opendata);
1838         ret = _nfs4_recover_proc_open(opendata);
1839         if (ret != 0)
1840                 return ret;
1841         newstate = nfs4_opendata_to_nfs4_state(opendata);
1842         if (IS_ERR(newstate))
1843                 return PTR_ERR(newstate);
1844         if (newstate != opendata->state)
1845                 ret = -ESTALE;
1846         nfs4_close_state(newstate, fmode);
1847         return ret;
1848 }
1849
1850 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1851 {
1852         int ret;
1853
1854         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1855         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1856         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1857         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1858         /* memory barrier prior to reading state->n_* */
1859         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1860         clear_bit(NFS_OPEN_STATE, &state->flags);
1861         smp_rmb();
1862         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1863         if (ret != 0)
1864                 return ret;
1865         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1866         if (ret != 0)
1867                 return ret;
1868         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1869         if (ret != 0)
1870                 return ret;
1871         /*
1872          * We may have performed cached opens for all three recoveries.
1873          * Check if we need to update the current stateid.
1874          */
1875         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1876             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1877                 write_seqlock(&state->seqlock);
1878                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1879                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1880                 write_sequnlock(&state->seqlock);
1881         }
1882         return 0;
1883 }
1884
1885 /*
1886  * OPEN_RECLAIM:
1887  *      reclaim state on the server after a reboot.
1888  */
1889 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1890 {
1891         struct nfs_delegation *delegation;
1892         struct nfs4_opendata *opendata;
1893         fmode_t delegation_type = 0;
1894         int status;
1895
1896         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1897                         NFS4_OPEN_CLAIM_PREVIOUS);
1898         if (IS_ERR(opendata))
1899                 return PTR_ERR(opendata);
1900         rcu_read_lock();
1901         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1902         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1903                 delegation_type = delegation->type;
1904         rcu_read_unlock();
1905         opendata->o_arg.u.delegation_type = delegation_type;
1906         status = nfs4_open_recover(opendata, state);
1907         nfs4_opendata_put(opendata);
1908         return status;
1909 }
1910
1911 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1912 {
1913         struct nfs_server *server = NFS_SERVER(state->inode);
1914         struct nfs4_exception exception = { };
1915         int err;
1916         do {
1917                 err = _nfs4_do_open_reclaim(ctx, state);
1918                 trace_nfs4_open_reclaim(ctx, 0, err);
1919                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1920                         continue;
1921                 if (err != -NFS4ERR_DELAY)
1922                         break;
1923                 nfs4_handle_exception(server, err, &exception);
1924         } while (exception.retry);
1925         return err;
1926 }
1927
1928 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1929 {
1930         struct nfs_open_context *ctx;
1931         int ret;
1932
1933         ctx = nfs4_state_find_open_context(state);
1934         if (IS_ERR(ctx))
1935                 return -EAGAIN;
1936         ret = nfs4_do_open_reclaim(ctx, state);
1937         put_nfs_open_context(ctx);
1938         return ret;
1939 }
1940
1941 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1942 {
1943         switch (err) {
1944                 default:
1945                         printk(KERN_ERR "NFS: %s: unhandled error "
1946                                         "%d.\n", __func__, err);
1947                 case 0:
1948                 case -ENOENT:
1949                 case -EAGAIN:
1950                 case -ESTALE:
1951                         break;
1952                 case -NFS4ERR_BADSESSION:
1953                 case -NFS4ERR_BADSLOT:
1954                 case -NFS4ERR_BAD_HIGH_SLOT:
1955                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1956                 case -NFS4ERR_DEADSESSION:
1957                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1958                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1959                         return -EAGAIN;
1960                 case -NFS4ERR_STALE_CLIENTID:
1961                 case -NFS4ERR_STALE_STATEID:
1962                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1963                         /* Don't recall a delegation if it was lost */
1964                         nfs4_schedule_lease_recovery(server->nfs_client);
1965                         return -EAGAIN;
1966                 case -NFS4ERR_MOVED:
1967                         nfs4_schedule_migration_recovery(server);
1968                         return -EAGAIN;
1969                 case -NFS4ERR_LEASE_MOVED:
1970                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1971                         return -EAGAIN;
1972                 case -NFS4ERR_DELEG_REVOKED:
1973                 case -NFS4ERR_ADMIN_REVOKED:
1974                 case -NFS4ERR_EXPIRED:
1975                 case -NFS4ERR_BAD_STATEID:
1976                 case -NFS4ERR_OPENMODE:
1977                         nfs_inode_find_state_and_recover(state->inode,
1978                                         stateid);
1979                         nfs4_schedule_stateid_recovery(server, state);
1980                         return -EAGAIN;
1981                 case -NFS4ERR_DELAY:
1982                 case -NFS4ERR_GRACE:
1983                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1984                         ssleep(1);
1985                         return -EAGAIN;
1986                 case -ENOMEM:
1987                 case -NFS4ERR_DENIED:
1988                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1989                         return 0;
1990         }
1991         return err;
1992 }
1993
1994 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1995                 struct nfs4_state *state, const nfs4_stateid *stateid,
1996                 fmode_t type)
1997 {
1998         struct nfs_server *server = NFS_SERVER(state->inode);
1999         struct nfs4_opendata *opendata;
2000         int err = 0;
2001
2002         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2003                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2004         if (IS_ERR(opendata))
2005                 return PTR_ERR(opendata);
2006         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2007         write_seqlock(&state->seqlock);
2008         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2009         write_sequnlock(&state->seqlock);
2010         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2011         switch (type & (FMODE_READ|FMODE_WRITE)) {
2012         case FMODE_READ|FMODE_WRITE:
2013         case FMODE_WRITE:
2014                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2015                 if (err)
2016                         break;
2017                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2018                 if (err)
2019                         break;
2020         case FMODE_READ:
2021                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2022         }
2023         nfs4_opendata_put(opendata);
2024         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
2025 }
2026
2027 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2028 {
2029         struct nfs4_opendata *data = calldata;
2030
2031         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
2032                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
2033 }
2034
2035 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2036 {
2037         struct nfs4_opendata *data = calldata;
2038
2039         nfs40_sequence_done(task, &data->c_res.seq_res);
2040
2041         data->rpc_status = task->tk_status;
2042         if (data->rpc_status == 0) {
2043                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2044                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2045                 renew_lease(data->o_res.server, data->timestamp);
2046                 data->rpc_done = 1;
2047         }
2048 }
2049
2050 static void nfs4_open_confirm_release(void *calldata)
2051 {
2052         struct nfs4_opendata *data = calldata;
2053         struct nfs4_state *state = NULL;
2054
2055         /* If this request hasn't been cancelled, do nothing */
2056         if (data->cancelled == 0)
2057                 goto out_free;
2058         /* In case of error, no cleanup! */
2059         if (!data->rpc_done)
2060                 goto out_free;
2061         state = nfs4_opendata_to_nfs4_state(data);
2062         if (!IS_ERR(state))
2063                 nfs4_close_state(state, data->o_arg.fmode);
2064 out_free:
2065         nfs4_opendata_put(data);
2066 }
2067
2068 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2069         .rpc_call_prepare = nfs4_open_confirm_prepare,
2070         .rpc_call_done = nfs4_open_confirm_done,
2071         .rpc_release = nfs4_open_confirm_release,
2072 };
2073
2074 /*
2075  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2076  */
2077 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2078 {
2079         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2080         struct rpc_task *task;
2081         struct  rpc_message msg = {
2082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2083                 .rpc_argp = &data->c_arg,
2084                 .rpc_resp = &data->c_res,
2085                 .rpc_cred = data->owner->so_cred,
2086         };
2087         struct rpc_task_setup task_setup_data = {
2088                 .rpc_client = server->client,
2089                 .rpc_message = &msg,
2090                 .callback_ops = &nfs4_open_confirm_ops,
2091                 .callback_data = data,
2092                 .workqueue = nfsiod_workqueue,
2093                 .flags = RPC_TASK_ASYNC,
2094         };
2095         int status;
2096
2097         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2098         kref_get(&data->kref);
2099         data->rpc_done = 0;
2100         data->rpc_status = 0;
2101         data->timestamp = jiffies;
2102         if (data->is_recover)
2103                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2104         task = rpc_run_task(&task_setup_data);
2105         if (IS_ERR(task))
2106                 return PTR_ERR(task);
2107         status = nfs4_wait_for_completion_rpc_task(task);
2108         if (status != 0) {
2109                 data->cancelled = 1;
2110                 smp_wmb();
2111         } else
2112                 status = data->rpc_status;
2113         rpc_put_task(task);
2114         return status;
2115 }
2116
2117 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2118 {
2119         struct nfs4_opendata *data = calldata;
2120         struct nfs4_state_owner *sp = data->owner;
2121         struct nfs_client *clp = sp->so_server->nfs_client;
2122         enum open_claim_type4 claim = data->o_arg.claim;
2123
2124         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2125                 goto out_wait;
2126         /*
2127          * Check if we still need to send an OPEN call, or if we can use
2128          * a delegation instead.
2129          */
2130         if (data->state != NULL) {
2131                 struct nfs_delegation *delegation;
2132
2133                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2134                         goto out_no_action;
2135                 rcu_read_lock();
2136                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2137                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2138                         goto unlock_no_action;
2139                 rcu_read_unlock();
2140         }
2141         /* Update client id. */
2142         data->o_arg.clientid = clp->cl_clientid;
2143         switch (claim) {
2144         default:
2145                 break;
2146         case NFS4_OPEN_CLAIM_PREVIOUS:
2147         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2148         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2149                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2150         case NFS4_OPEN_CLAIM_FH:
2151                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2152                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2153         }
2154         data->timestamp = jiffies;
2155         if (nfs4_setup_sequence(data->o_arg.server,
2156                                 &data->o_arg.seq_args,
2157                                 &data->o_res.seq_res,
2158                                 task) != 0)
2159                 nfs_release_seqid(data->o_arg.seqid);
2160
2161         /* Set the create mode (note dependency on the session type) */
2162         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2163         if (data->o_arg.open_flags & O_EXCL) {
2164                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2165                 if (nfs4_has_persistent_session(clp))
2166                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2167                 else if (clp->cl_mvops->minor_version > 0)
2168                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2169         }
2170         return;
2171 unlock_no_action:
2172         trace_nfs4_cached_open(data->state);
2173         rcu_read_unlock();
2174 out_no_action:
2175         task->tk_action = NULL;
2176 out_wait:
2177         nfs4_sequence_done(task, &data->o_res.seq_res);
2178 }
2179
2180 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2181 {
2182         struct nfs4_opendata *data = calldata;
2183
2184         data->rpc_status = task->tk_status;
2185
2186         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2187                 return;
2188
2189         if (task->tk_status == 0) {
2190                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2191                         switch (data->o_res.f_attr->mode & S_IFMT) {
2192                         case S_IFREG:
2193                                 break;
2194                         case S_IFLNK:
2195                                 data->rpc_status = -ELOOP;
2196                                 break;
2197                         case S_IFDIR:
2198                                 data->rpc_status = -EISDIR;
2199                                 break;
2200                         default:
2201                                 data->rpc_status = -ENOTDIR;
2202                         }
2203                 }
2204                 renew_lease(data->o_res.server, data->timestamp);
2205                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2206                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2207         }
2208         data->rpc_done = 1;
2209 }
2210
2211 static void nfs4_open_release(void *calldata)
2212 {
2213         struct nfs4_opendata *data = calldata;
2214         struct nfs4_state *state = NULL;
2215
2216         /* If this request hasn't been cancelled, do nothing */
2217         if (data->cancelled == 0)
2218                 goto out_free;
2219         /* In case of error, no cleanup! */
2220         if (data->rpc_status != 0 || !data->rpc_done)
2221                 goto out_free;
2222         /* In case we need an open_confirm, no cleanup! */
2223         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2224                 goto out_free;
2225         state = nfs4_opendata_to_nfs4_state(data);
2226         if (!IS_ERR(state))
2227                 nfs4_close_state(state, data->o_arg.fmode);
2228 out_free:
2229         nfs4_opendata_put(data);
2230 }
2231
2232 static const struct rpc_call_ops nfs4_open_ops = {
2233         .rpc_call_prepare = nfs4_open_prepare,
2234         .rpc_call_done = nfs4_open_done,
2235         .rpc_release = nfs4_open_release,
2236 };
2237
2238 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2239 {
2240         struct inode *dir = d_inode(data->dir);
2241         struct nfs_server *server = NFS_SERVER(dir);
2242         struct nfs_openargs *o_arg = &data->o_arg;
2243         struct nfs_openres *o_res = &data->o_res;
2244         struct rpc_task *task;
2245         struct rpc_message msg = {
2246                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2247                 .rpc_argp = o_arg,
2248                 .rpc_resp = o_res,
2249                 .rpc_cred = data->owner->so_cred,
2250         };
2251         struct rpc_task_setup task_setup_data = {
2252                 .rpc_client = server->client,
2253                 .rpc_message = &msg,
2254                 .callback_ops = &nfs4_open_ops,
2255                 .callback_data = data,
2256                 .workqueue = nfsiod_workqueue,
2257                 .flags = RPC_TASK_ASYNC,
2258         };
2259         int status;
2260
2261         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2262         kref_get(&data->kref);
2263         data->rpc_done = 0;
2264         data->rpc_status = 0;
2265         data->cancelled = 0;
2266         data->is_recover = 0;
2267         if (isrecover) {
2268                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2269                 data->is_recover = 1;
2270         }
2271         task = rpc_run_task(&task_setup_data);
2272         if (IS_ERR(task))
2273                 return PTR_ERR(task);
2274         status = nfs4_wait_for_completion_rpc_task(task);
2275         if (status != 0) {
2276                 data->cancelled = 1;
2277                 smp_wmb();
2278         } else
2279                 status = data->rpc_status;
2280         rpc_put_task(task);
2281
2282         return status;
2283 }
2284
2285 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2286 {
2287         struct inode *dir = d_inode(data->dir);
2288         struct nfs_openres *o_res = &data->o_res;
2289         int status;
2290
2291         status = nfs4_run_open_task(data, 1);
2292         if (status != 0 || !data->rpc_done)
2293                 return status;
2294
2295         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2296
2297         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2298                 status = _nfs4_proc_open_confirm(data);
2299                 if (status != 0)
2300                         return status;
2301         }
2302
2303         return status;
2304 }
2305
2306 /*
2307  * Additional permission checks in order to distinguish between an
2308  * open for read, and an open for execute. This works around the
2309  * fact that NFSv4 OPEN treats read and execute permissions as being
2310  * the same.
2311  * Note that in the non-execute case, we want to turn off permission
2312  * checking if we just created a new file (POSIX open() semantics).
2313  */
2314 static int nfs4_opendata_access(struct rpc_cred *cred,
2315                                 struct nfs4_opendata *opendata,
2316                                 struct nfs4_state *state, fmode_t fmode,
2317                                 int openflags)
2318 {
2319         struct nfs_access_entry cache;
2320         u32 mask;
2321
2322         /* access call failed or for some reason the server doesn't
2323          * support any access modes -- defer access call until later */
2324         if (opendata->o_res.access_supported == 0)
2325                 return 0;
2326
2327         mask = 0;
2328         /*
2329          * Use openflags to check for exec, because fmode won't
2330          * always have FMODE_EXEC set when file open for exec.
2331          */
2332         if (openflags & __FMODE_EXEC) {
2333                 /* ONLY check for exec rights */
2334                 mask = MAY_EXEC;
2335         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2336                 mask = MAY_READ;
2337
2338         cache.cred = cred;
2339         cache.jiffies = jiffies;
2340         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2341         nfs_access_add_cache(state->inode, &cache);
2342
2343         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2344                 return 0;
2345
2346         /* even though OPEN succeeded, access is denied. Close the file */
2347         nfs4_close_state(state, fmode);
2348         return -EACCES;
2349 }
2350
2351 /*
2352  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2353  */
2354 static int _nfs4_proc_open(struct nfs4_opendata *data)
2355 {
2356         struct inode *dir = d_inode(data->dir);
2357         struct nfs_server *server = NFS_SERVER(dir);
2358         struct nfs_openargs *o_arg = &data->o_arg;
2359         struct nfs_openres *o_res = &data->o_res;
2360         int status;
2361
2362         status = nfs4_run_open_task(data, 0);
2363         if (!data->rpc_done)
2364                 return status;
2365         if (status != 0) {
2366                 if (status == -NFS4ERR_BADNAME &&
2367                                 !(o_arg->open_flags & O_CREAT))
2368                         return -ENOENT;
2369                 return status;
2370         }
2371
2372         nfs_fattr_map_and_free_names(server, &data->f_attr);
2373
2374         if (o_arg->open_flags & O_CREAT) {
2375                 update_changeattr(dir, &o_res->cinfo);
2376                 if (o_arg->open_flags & O_EXCL)
2377                         data->file_created = 1;
2378                 else if (o_res->cinfo.before != o_res->cinfo.after)
2379                         data->file_created = 1;
2380         }
2381         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2382                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2383         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2384                 status = _nfs4_proc_open_confirm(data);
2385                 if (status != 0)
2386                         return status;
2387         }
2388         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2389                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2390         return 0;
2391 }
2392
2393 static int nfs4_recover_expired_lease(struct nfs_server *server)
2394 {
2395         return nfs4_client_recover_expired_lease(server->nfs_client);
2396 }
2397
2398 /*
2399  * OPEN_EXPIRED:
2400  *      reclaim state on the server after a network partition.
2401  *      Assumes caller holds the appropriate lock
2402  */
2403 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2404 {
2405         struct nfs4_opendata *opendata;
2406         int ret;
2407
2408         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2409                         NFS4_OPEN_CLAIM_FH);
2410         if (IS_ERR(opendata))
2411                 return PTR_ERR(opendata);
2412         ret = nfs4_open_recover(opendata, state);
2413         if (ret == -ESTALE)
2414                 d_drop(ctx->dentry);
2415         nfs4_opendata_put(opendata);
2416         return ret;
2417 }
2418
2419 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2420 {
2421         struct nfs_server *server = NFS_SERVER(state->inode);
2422         struct nfs4_exception exception = { };
2423         int err;
2424
2425         do {
2426                 err = _nfs4_open_expired(ctx, state);
2427                 trace_nfs4_open_expired(ctx, 0, err);
2428                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2429                         continue;
2430                 switch (err) {
2431                 default:
2432                         goto out;
2433                 case -NFS4ERR_GRACE:
2434                 case -NFS4ERR_DELAY:
2435                         nfs4_handle_exception(server, err, &exception);
2436                         err = 0;
2437                 }
2438         } while (exception.retry);
2439 out:
2440         return err;
2441 }
2442
2443 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2444 {
2445         struct nfs_open_context *ctx;
2446         int ret;
2447
2448         ctx = nfs4_state_find_open_context(state);
2449         if (IS_ERR(ctx))
2450                 return -EAGAIN;
2451         ret = nfs4_do_open_expired(ctx, state);
2452         put_nfs_open_context(ctx);
2453         return ret;
2454 }
2455
2456 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2457                 const nfs4_stateid *stateid)
2458 {
2459         nfs_remove_bad_delegation(state->inode, stateid);
2460         write_seqlock(&state->seqlock);
2461         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2462         write_sequnlock(&state->seqlock);
2463         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2464 }
2465
2466 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2467 {
2468         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2469                 nfs_finish_clear_delegation_stateid(state, NULL);
2470 }
2471
2472 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2473 {
2474         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2475         nfs40_clear_delegation_stateid(state);
2476         return nfs4_open_expired(sp, state);
2477 }
2478
2479 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2480                 nfs4_stateid *stateid,
2481                 struct rpc_cred *cred)
2482 {
2483         return -NFS4ERR_BAD_STATEID;
2484 }
2485
2486 #if defined(CONFIG_NFS_V4_1)
2487 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2488                 nfs4_stateid *stateid,
2489                 struct rpc_cred *cred)
2490 {
2491         int status;
2492
2493         switch (stateid->type) {
2494         default:
2495                 break;
2496         case NFS4_INVALID_STATEID_TYPE:
2497         case NFS4_SPECIAL_STATEID_TYPE:
2498                 return -NFS4ERR_BAD_STATEID;
2499         case NFS4_REVOKED_STATEID_TYPE:
2500                 goto out_free;
2501         }
2502
2503         status = nfs41_test_stateid(server, stateid, cred);
2504         switch (status) {
2505         case -NFS4ERR_EXPIRED:
2506         case -NFS4ERR_ADMIN_REVOKED:
2507         case -NFS4ERR_DELEG_REVOKED:
2508                 break;
2509         default:
2510                 return status;
2511         }
2512 out_free:
2513         /* Ack the revoked state to the server */
2514         nfs41_free_stateid(server, stateid, cred, true);
2515         return -NFS4ERR_EXPIRED;
2516 }
2517
2518 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2519 {
2520         struct nfs_server *server = NFS_SERVER(state->inode);
2521         nfs4_stateid stateid;
2522         struct nfs_delegation *delegation;
2523         struct rpc_cred *cred;
2524         int status;
2525
2526         /* Get the delegation credential for use by test/free_stateid */
2527         rcu_read_lock();
2528         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2529         if (delegation == NULL) {
2530                 rcu_read_unlock();
2531                 return;
2532         }
2533
2534         nfs4_stateid_copy(&stateid, &delegation->stateid);
2535         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2536                 rcu_read_unlock();
2537                 nfs_finish_clear_delegation_stateid(state, &stateid);
2538                 return;
2539         }
2540
2541         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags)) {
2542                 rcu_read_unlock();
2543                 return;
2544         }
2545
2546         cred = get_rpccred(delegation->cred);
2547         rcu_read_unlock();
2548         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2549         trace_nfs4_test_delegation_stateid(state, NULL, status);
2550         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2551                 nfs_finish_clear_delegation_stateid(state, &stateid);
2552
2553         put_rpccred(cred);
2554 }
2555
2556 /**
2557  * nfs41_check_expired_locks - possibly free a lock stateid
2558  *
2559  * @state: NFSv4 state for an inode
2560  *
2561  * Returns NFS_OK if recovery for this stateid is now finished.
2562  * Otherwise a negative NFS4ERR value is returned.
2563  */
2564 static int nfs41_check_expired_locks(struct nfs4_state *state)
2565 {
2566         int status, ret = NFS_OK;
2567         struct nfs4_lock_state *lsp;
2568         struct nfs_server *server = NFS_SERVER(state->inode);
2569
2570         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2571                 goto out;
2572         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2573                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2574                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2575
2576                         status = nfs41_test_and_free_expired_stateid(server,
2577                                         &lsp->ls_stateid,
2578                                         cred);
2579                         trace_nfs4_test_lock_stateid(state, lsp, status);
2580                         if (status == -NFS4ERR_EXPIRED ||
2581                             status == -NFS4ERR_BAD_STATEID) {
2582                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2583                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2584                                 if (!recover_lost_locks)
2585                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2586                         } else if (status != NFS_OK) {
2587                                 ret = status;
2588                                 break;
2589                         }
2590                 }
2591         };
2592 out:
2593         return ret;
2594 }
2595
2596 /**
2597  * nfs41_check_open_stateid - possibly free an open stateid
2598  *
2599  * @state: NFSv4 state for an inode
2600  *
2601  * Returns NFS_OK if recovery for this stateid is now finished.
2602  * Otherwise a negative NFS4ERR value is returned.
2603  */
2604 static int nfs41_check_open_stateid(struct nfs4_state *state)
2605 {
2606         struct nfs_server *server = NFS_SERVER(state->inode);
2607         nfs4_stateid *stateid = &state->open_stateid;
2608         struct rpc_cred *cred = state->owner->so_cred;
2609         int status;
2610
2611         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2612                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2613                         if (nfs4_have_delegation(state->inode, state->state))
2614                                 return NFS_OK;
2615                         return -NFS4ERR_OPENMODE;
2616                 }
2617                 return -NFS4ERR_BAD_STATEID;
2618         }
2619         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2620         trace_nfs4_test_open_stateid(state, NULL, status);
2621         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2622                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2623                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2624                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2625                 clear_bit(NFS_OPEN_STATE, &state->flags);
2626                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2627         }
2628         if (status != NFS_OK)
2629                 return status;
2630         if (nfs_open_stateid_recover_openmode(state))
2631                 return -NFS4ERR_OPENMODE;
2632         return NFS_OK;
2633 }
2634
2635 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2636 {
2637         int status;
2638
2639         nfs41_check_delegation_stateid(state);
2640         status = nfs41_check_expired_locks(state);
2641         if (status != NFS_OK)
2642                 return status;
2643         status = nfs41_check_open_stateid(state);
2644         if (status != NFS_OK)
2645                 status = nfs4_open_expired(sp, state);
2646         return status;
2647 }
2648 #endif
2649
2650 /*
2651  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2652  * fields corresponding to attributes that were used to store the verifier.
2653  * Make sure we clobber those fields in the later setattr call
2654  */
2655 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2656                                 struct iattr *sattr, struct nfs4_label **label)
2657 {
2658         const u32 *attrset = opendata->o_res.attrset;
2659
2660         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2661             !(sattr->ia_valid & ATTR_ATIME_SET))
2662                 sattr->ia_valid |= ATTR_ATIME;
2663
2664         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2665             !(sattr->ia_valid & ATTR_MTIME_SET))
2666                 sattr->ia_valid |= ATTR_MTIME;
2667
2668         /* Except MODE, it seems harmless of setting twice. */
2669         if ((attrset[1] & FATTR4_WORD1_MODE))
2670                 sattr->ia_valid &= ~ATTR_MODE;
2671
2672         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2673                 *label = NULL;
2674 }
2675
2676 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2677                 fmode_t fmode,
2678                 int flags,
2679                 struct nfs_open_context *ctx)
2680 {
2681         struct nfs4_state_owner *sp = opendata->owner;
2682         struct nfs_server *server = sp->so_server;
2683         struct dentry *dentry;
2684         struct nfs4_state *state;
2685         unsigned int seq;
2686         int ret;
2687
2688         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2689
2690         ret = _nfs4_proc_open(opendata);
2691         if (ret != 0)
2692                 goto out;
2693
2694         state = nfs4_opendata_to_nfs4_state(opendata);
2695         ret = PTR_ERR(state);
2696         if (IS_ERR(state))
2697                 goto out;
2698         if (server->caps & NFS_CAP_POSIX_LOCK)
2699                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2700         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2701                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2702
2703         dentry = opendata->dentry;
2704         if (d_really_is_negative(dentry)) {
2705                 struct dentry *alias;
2706                 d_drop(dentry);
2707                 alias = d_exact_alias(dentry, state->inode);
2708                 if (!alias)
2709                         alias = d_splice_alias(igrab(state->inode), dentry);
2710                 /* d_splice_alias() can't fail here - it's a non-directory */
2711                 if (alias) {
2712                         dput(ctx->dentry);
2713                         ctx->dentry = dentry = alias;
2714                 }
2715                 nfs_set_verifier(dentry,
2716                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2717         }
2718
2719         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2720         if (ret != 0)
2721                 goto out;
2722
2723         ctx->state = state;
2724         if (d_inode(dentry) == state->inode) {
2725                 nfs_inode_attach_open_context(ctx);
2726                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2727                         nfs4_schedule_stateid_recovery(server, state);
2728         }
2729 out:
2730         return ret;
2731 }
2732
2733 /*
2734  * Returns a referenced nfs4_state
2735  */
2736 static int _nfs4_do_open(struct inode *dir,
2737                         struct nfs_open_context *ctx,
2738                         int flags,
2739                         struct iattr *sattr,
2740                         struct nfs4_label *label,
2741                         int *opened)
2742 {
2743         struct nfs4_state_owner  *sp;
2744         struct nfs4_state     *state = NULL;
2745         struct nfs_server       *server = NFS_SERVER(dir);
2746         struct nfs4_opendata *opendata;
2747         struct dentry *dentry = ctx->dentry;
2748         struct rpc_cred *cred = ctx->cred;
2749         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2750         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2751         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2752         struct nfs4_label *olabel = NULL;
2753         int status;
2754
2755         /* Protect against reboot recovery conflicts */
2756         status = -ENOMEM;
2757         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2758         if (sp == NULL) {
2759                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2760                 goto out_err;
2761         }
2762         status = nfs4_recover_expired_lease(server);
2763         if (status != 0)
2764                 goto err_put_state_owner;
2765         if (d_really_is_positive(dentry))
2766                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2767         status = -ENOMEM;
2768         if (d_really_is_positive(dentry))
2769                 claim = NFS4_OPEN_CLAIM_FH;
2770         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2771                         label, claim, GFP_KERNEL);
2772         if (opendata == NULL)
2773                 goto err_put_state_owner;
2774
2775         if (label) {
2776                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2777                 if (IS_ERR(olabel)) {
2778                         status = PTR_ERR(olabel);
2779                         goto err_opendata_put;
2780                 }
2781         }
2782
2783         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2784                 if (!opendata->f_attr.mdsthreshold) {
2785                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2786                         if (!opendata->f_attr.mdsthreshold)
2787                                 goto err_free_label;
2788                 }
2789                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2790         }
2791         if (d_really_is_positive(dentry))
2792                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2793
2794         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2795         if (status != 0)
2796                 goto err_free_label;
2797         state = ctx->state;
2798
2799         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2800             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2801                 nfs4_exclusive_attrset(opendata, sattr, &label);
2802                 /*
2803                  * send create attributes which was not set by open
2804                  * with an extra setattr.
2805                  */
2806                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2807                         nfs_fattr_init(opendata->o_res.f_attr);
2808                         status = nfs4_do_setattr(state->inode, cred,
2809                                         opendata->o_res.f_attr, sattr,
2810                                         state, label, olabel);
2811                         if (status == 0) {
2812                                 nfs_setattr_update_inode(state->inode, sattr,
2813                                                 opendata->o_res.f_attr);
2814                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2815                         }
2816                 }
2817         }
2818         if (opened && opendata->file_created)
2819                 *opened |= FILE_CREATED;
2820
2821         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2822                 *ctx_th = opendata->f_attr.mdsthreshold;
2823                 opendata->f_attr.mdsthreshold = NULL;
2824         }
2825
2826         nfs4_label_free(olabel);
2827
2828         nfs4_opendata_put(opendata);
2829         nfs4_put_state_owner(sp);
2830         return 0;
2831 err_free_label:
2832         nfs4_label_free(olabel);
2833 err_opendata_put:
2834         nfs4_opendata_put(opendata);
2835 err_put_state_owner:
2836         nfs4_put_state_owner(sp);
2837 out_err:
2838         return status;
2839 }
2840
2841
2842 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2843                                         struct nfs_open_context *ctx,
2844                                         int flags,
2845                                         struct iattr *sattr,
2846                                         struct nfs4_label *label,
2847                                         int *opened)
2848 {
2849         struct nfs_server *server = NFS_SERVER(dir);
2850         struct nfs4_exception exception = { };
2851         struct nfs4_state *res;
2852         int status;
2853
2854         do {
2855                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2856                 res = ctx->state;
2857                 trace_nfs4_open_file(ctx, flags, status);
2858                 if (status == 0)
2859                         break;
2860                 /* NOTE: BAD_SEQID means the server and client disagree about the
2861                  * book-keeping w.r.t. state-changing operations
2862                  * (OPEN/CLOSE/LOCK/LOCKU...)
2863                  * It is actually a sign of a bug on the client or on the server.
2864                  *
2865                  * If we receive a BAD_SEQID error in the particular case of
2866                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2867                  * have unhashed the old state_owner for us, and that we can
2868                  * therefore safely retry using a new one. We should still warn
2869                  * the user though...
2870                  */
2871                 if (status == -NFS4ERR_BAD_SEQID) {
2872                         pr_warn_ratelimited("NFS: v4 server %s "
2873                                         " returned a bad sequence-id error!\n",
2874                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2875                         exception.retry = 1;
2876                         continue;
2877                 }
2878                 /*
2879                  * BAD_STATEID on OPEN means that the server cancelled our
2880                  * state before it received the OPEN_CONFIRM.
2881                  * Recover by retrying the request as per the discussion
2882                  * on Page 181 of RFC3530.
2883                  */
2884                 if (status == -NFS4ERR_BAD_STATEID) {
2885                         exception.retry = 1;
2886                         continue;
2887                 }
2888                 if (status == -EAGAIN) {
2889                         /* We must have found a delegation */
2890                         exception.retry = 1;
2891                         continue;
2892                 }
2893                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2894                         continue;
2895                 res = ERR_PTR(nfs4_handle_exception(server,
2896                                         status, &exception));
2897         } while (exception.retry);
2898         return res;
2899 }
2900
2901 static int _nfs4_do_setattr(struct inode *inode,
2902                             struct nfs_setattrargs *arg,
2903                             struct nfs_setattrres *res,
2904                             struct rpc_cred *cred,
2905                             struct nfs4_state *state)
2906 {
2907         struct nfs_server *server = NFS_SERVER(inode);
2908         struct rpc_message msg = {
2909                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2910                 .rpc_argp       = arg,
2911                 .rpc_resp       = res,
2912                 .rpc_cred       = cred,
2913         };
2914         struct rpc_cred *delegation_cred = NULL;
2915         unsigned long timestamp = jiffies;
2916         fmode_t fmode;
2917         bool truncate;
2918         int status;
2919
2920         nfs_fattr_init(res->fattr);
2921
2922         /* Servers should only apply open mode checks for file size changes */
2923         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2924         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2925
2926         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2927                 /* Use that stateid */
2928         } else if (truncate && state != NULL) {
2929                 struct nfs_lockowner lockowner = {
2930                         .l_owner = current->files,
2931                         .l_pid = current->tgid,
2932                 };
2933                 if (!nfs4_valid_open_stateid(state))
2934                         return -EBADF;
2935                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2936                                 &arg->stateid, &delegation_cred) == -EIO)
2937                         return -EBADF;
2938         } else
2939                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2940         if (delegation_cred)
2941                 msg.rpc_cred = delegation_cred;
2942
2943         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2944
2945         put_rpccred(delegation_cred);
2946         if (status == 0 && state != NULL)
2947                 renew_lease(server, timestamp);
2948         trace_nfs4_setattr(inode, &arg->stateid, status);
2949         return status;
2950 }
2951
2952 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2953                            struct nfs_fattr *fattr, struct iattr *sattr,
2954                            struct nfs4_state *state, struct nfs4_label *ilabel,
2955                            struct nfs4_label *olabel)
2956 {
2957         struct nfs_server *server = NFS_SERVER(inode);
2958         struct nfs_setattrargs  arg = {
2959                 .fh             = NFS_FH(inode),
2960                 .iap            = sattr,
2961                 .server         = server,
2962                 .bitmask = server->attr_bitmask,
2963                 .label          = ilabel,
2964         };
2965         struct nfs_setattrres  res = {
2966                 .fattr          = fattr,
2967                 .label          = olabel,
2968                 .server         = server,
2969         };
2970         struct nfs4_exception exception = {
2971                 .state = state,
2972                 .inode = inode,
2973                 .stateid = &arg.stateid,
2974         };
2975         int err;
2976
2977         arg.bitmask = nfs4_bitmask(server, ilabel);
2978         if (ilabel)
2979                 arg.bitmask = nfs4_bitmask(server, olabel);
2980
2981         do {
2982                 err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
2983                 switch (err) {
2984                 case -NFS4ERR_OPENMODE:
2985                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2986                                 pr_warn_once("NFSv4: server %s is incorrectly "
2987                                                 "applying open mode checks to "
2988                                                 "a SETATTR that is not "
2989                                                 "changing file size.\n",
2990                                                 server->nfs_client->cl_hostname);
2991                         }
2992                         if (state && !(state->state & FMODE_WRITE)) {
2993                                 err = -EBADF;
2994                                 if (sattr->ia_valid & ATTR_OPEN)
2995                                         err = -EACCES;
2996                                 goto out;
2997                         }
2998                 }
2999                 err = nfs4_handle_exception(server, err, &exception);
3000         } while (exception.retry);
3001 out:
3002         return err;
3003 }
3004
3005 static bool
3006 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3007 {
3008         if (inode == NULL || !nfs_have_layout(inode))
3009                 return false;
3010
3011         return pnfs_wait_on_layoutreturn(inode, task);
3012 }
3013
3014 struct nfs4_closedata {
3015         struct inode *inode;
3016         struct nfs4_state *state;
3017         struct nfs_closeargs arg;
3018         struct nfs_closeres res;
3019         struct nfs_fattr fattr;
3020         unsigned long timestamp;
3021         bool roc;
3022         u32 roc_barrier;
3023 };
3024
3025 static void nfs4_free_closedata(void *data)
3026 {
3027         struct nfs4_closedata *calldata = data;
3028         struct nfs4_state_owner *sp = calldata->state->owner;
3029         struct super_block *sb = calldata->state->inode->i_sb;
3030
3031         if (calldata->roc)
3032                 pnfs_roc_release(calldata->state->inode);
3033         nfs4_put_open_state(calldata->state);
3034         nfs_free_seqid(calldata->arg.seqid);
3035         nfs4_put_state_owner(sp);
3036         nfs_sb_deactive(sb);
3037         kfree(calldata);
3038 }
3039
3040 static void nfs4_close_done(struct rpc_task *task, void *data)
3041 {
3042         struct nfs4_closedata *calldata = data;
3043         struct nfs4_state *state = calldata->state;
3044         struct nfs_server *server = NFS_SERVER(calldata->inode);
3045         nfs4_stateid *res_stateid = NULL;
3046
3047         dprintk("%s: begin!\n", __func__);
3048         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3049                 return;
3050         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3051         /* hmm. we are done with the inode, and in the process of freeing
3052          * the state_owner. we keep this around to process errors
3053          */
3054         switch (task->tk_status) {
3055                 case 0:
3056                         res_stateid = &calldata->res.stateid;
3057                         if (calldata->roc)
3058                                 pnfs_roc_set_barrier(state->inode,
3059                                                      calldata->roc_barrier);
3060                         renew_lease(server, calldata->timestamp);
3061                         break;
3062                 case -NFS4ERR_ADMIN_REVOKED:
3063                 case -NFS4ERR_STALE_STATEID:
3064                 case -NFS4ERR_EXPIRED:
3065                         nfs4_free_revoked_stateid(server,
3066                                         &calldata->arg.stateid,
3067                                         task->tk_msg.rpc_cred);
3068                 case -NFS4ERR_OLD_STATEID:
3069                 case -NFS4ERR_BAD_STATEID:
3070                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3071                                                 &state->open_stateid)) {
3072                                 rpc_restart_call_prepare(task);
3073                                 goto out_release;
3074                         }
3075                         if (calldata->arg.fmode == 0)
3076                                 break;
3077                 default:
3078                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3079                                 rpc_restart_call_prepare(task);
3080                                 goto out_release;
3081                         }
3082         }
3083         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3084                         res_stateid, calldata->arg.fmode);
3085 out_release:
3086         nfs_release_seqid(calldata->arg.seqid);
3087         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
3088         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3089 }
3090
3091 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3092 {
3093         struct nfs4_closedata *calldata = data;
3094         struct nfs4_state *state = calldata->state;
3095         struct inode *inode = calldata->inode;
3096         bool is_rdonly, is_wronly, is_rdwr;
3097         int call_close = 0;
3098
3099         dprintk("%s: begin!\n", __func__);
3100         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3101                 goto out_wait;
3102
3103         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3104         spin_lock(&state->owner->so_lock);
3105         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3106         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3107         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3108         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3109         /* Calculate the change in open mode */
3110         calldata->arg.fmode = 0;
3111         if (state->n_rdwr == 0) {
3112                 if (state->n_rdonly == 0)
3113                         call_close |= is_rdonly;
3114                 else if (is_rdonly)
3115                         calldata->arg.fmode |= FMODE_READ;
3116                 if (state->n_wronly == 0)
3117                         call_close |= is_wronly;
3118                 else if (is_wronly)
3119                         calldata->arg.fmode |= FMODE_WRITE;
3120                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3121                         call_close |= is_rdwr;
3122         } else if (is_rdwr)
3123                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3124
3125         if (!nfs4_valid_open_stateid(state))
3126                 call_close = 0;
3127         spin_unlock(&state->owner->so_lock);
3128
3129         if (!call_close) {
3130                 /* Note: exit _without_ calling nfs4_close_done */
3131                 goto out_no_action;
3132         }
3133
3134         if (nfs4_wait_on_layoutreturn(inode, task)) {
3135                 nfs_release_seqid(calldata->arg.seqid);
3136                 goto out_wait;
3137         }
3138
3139         if (calldata->arg.fmode == 0)
3140                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3141         if (calldata->roc)
3142                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
3143
3144         calldata->arg.share_access =
3145                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3146                                 calldata->arg.fmode, 0);
3147
3148         nfs_fattr_init(calldata->res.fattr);
3149         calldata->timestamp = jiffies;
3150         if (nfs4_setup_sequence(NFS_SERVER(inode),
3151                                 &calldata->arg.seq_args,
3152                                 &calldata->res.seq_res,
3153                                 task) != 0)
3154                 nfs_release_seqid(calldata->arg.seqid);
3155         dprintk("%s: done!\n", __func__);
3156         return;
3157 out_no_action:
3158         task->tk_action = NULL;
3159 out_wait:
3160         nfs4_sequence_done(task, &calldata->res.seq_res);
3161 }
3162
3163 static const struct rpc_call_ops nfs4_close_ops = {
3164         .rpc_call_prepare = nfs4_close_prepare,
3165         .rpc_call_done = nfs4_close_done,
3166         .rpc_release = nfs4_free_closedata,
3167 };
3168
3169 static bool nfs4_roc(struct inode *inode)
3170 {
3171         if (!nfs_have_layout(inode))
3172                 return false;
3173         return pnfs_roc(inode);
3174 }
3175
3176 /*
3177  * It is possible for data to be read/written from a mem-mapped file
3178  * after the sys_close call (which hits the vfs layer as a flush).
3179  * This means that we can't safely call nfsv4 close on a file until
3180  * the inode is cleared. This in turn means that we are not good
3181  * NFSv4 citizens - we do not indicate to the server to update the file's
3182  * share state even when we are done with one of the three share
3183  * stateid's in the inode.
3184  *
3185  * NOTE: Caller must be holding the sp->so_owner semaphore!
3186  */
3187 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3188 {
3189         struct nfs_server *server = NFS_SERVER(state->inode);
3190         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3191         struct nfs4_closedata *calldata;
3192         struct nfs4_state_owner *sp = state->owner;
3193         struct rpc_task *task;
3194         struct rpc_message msg = {
3195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3196                 .rpc_cred = state->owner->so_cred,
3197         };
3198         struct rpc_task_setup task_setup_data = {
3199                 .rpc_client = server->client,
3200                 .rpc_message = &msg,
3201                 .callback_ops = &nfs4_close_ops,
3202                 .workqueue = nfsiod_workqueue,
3203                 .flags = RPC_TASK_ASYNC,
3204         };
3205         int status = -ENOMEM;
3206
3207         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3208                 &task_setup_data.rpc_client, &msg);
3209
3210         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3211         if (calldata == NULL)
3212                 goto out;
3213         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3214         calldata->inode = state->inode;
3215         calldata->state = state;
3216         calldata->arg.fh = NFS_FH(state->inode);
3217         /* Serialization for the sequence id */
3218         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3219         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3220         if (IS_ERR(calldata->arg.seqid))
3221                 goto out_free_calldata;
3222         calldata->arg.fmode = 0;
3223         calldata->arg.bitmask = server->cache_consistency_bitmask;
3224         calldata->res.fattr = &calldata->fattr;
3225         calldata->res.seqid = calldata->arg.seqid;
3226         calldata->res.server = server;
3227         calldata->roc = nfs4_roc(state->inode);
3228         nfs_sb_active(calldata->inode->i_sb);
3229
3230         msg.rpc_argp = &calldata->arg;
3231         msg.rpc_resp = &calldata->res;
3232         task_setup_data.callback_data = calldata;
3233         task = rpc_run_task(&task_setup_data);
3234         if (IS_ERR(task))
3235                 return PTR_ERR(task);
3236         status = 0;
3237         if (wait)
3238                 status = rpc_wait_for_completion_task(task);
3239         rpc_put_task(task);
3240         return status;
3241 out_free_calldata:
3242         kfree(calldata);
3243 out:
3244         nfs4_put_open_state(state);
3245         nfs4_put_state_owner(sp);
3246         return status;
3247 }
3248
3249 static struct inode *
3250 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3251                 int open_flags, struct iattr *attr, int *opened)
3252 {
3253         struct nfs4_state *state;
3254         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3255
3256         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3257
3258         /* Protect against concurrent sillydeletes */
3259         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3260
3261         nfs4_label_release_security(label);
3262
3263         if (IS_ERR(state))
3264                 return ERR_CAST(state);
3265         return state->inode;
3266 }
3267
3268 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3269 {
3270         if (ctx->state == NULL)
3271                 return;
3272         if (is_sync)
3273                 nfs4_close_sync(ctx->state, ctx->mode);
3274         else
3275                 nfs4_close_state(ctx->state, ctx->mode);
3276 }
3277
3278 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3279 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3280 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3281
3282 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3283 {
3284         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3285         struct nfs4_server_caps_arg args = {
3286                 .fhandle = fhandle,
3287                 .bitmask = bitmask,
3288         };
3289         struct nfs4_server_caps_res res = {};
3290         struct rpc_message msg = {
3291                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3292                 .rpc_argp = &args,
3293                 .rpc_resp = &res,
3294         };
3295         int status;
3296
3297         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3298                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3299                      FATTR4_WORD0_LINK_SUPPORT |
3300                      FATTR4_WORD0_SYMLINK_SUPPORT |
3301                      FATTR4_WORD0_ACLSUPPORT;
3302         if (minorversion)
3303                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3304
3305         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3306         if (status == 0) {
3307                 /* Sanity check the server answers */
3308                 switch (minorversion) {
3309                 case 0:
3310                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3311                         res.attr_bitmask[2] = 0;
3312                         break;
3313                 case 1:
3314                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3315                         break;
3316                 case 2:
3317                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3318                 }
3319                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3320                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3321                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3322                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3323                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3324                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3325                                 NFS_CAP_SECURITY_LABEL);
3326                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3327                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3328                         server->caps |= NFS_CAP_ACLS;
3329                 if (res.has_links != 0)
3330                         server->caps |= NFS_CAP_HARDLINKS;
3331                 if (res.has_symlinks != 0)
3332                         server->caps |= NFS_CAP_SYMLINKS;
3333                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3334                         server->caps |= NFS_CAP_FILEID;
3335                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3336                         server->caps |= NFS_CAP_MODE;
3337                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3338                         server->caps |= NFS_CAP_NLINK;
3339                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3340                         server->caps |= NFS_CAP_OWNER;
3341                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3342                         server->caps |= NFS_CAP_OWNER_GROUP;
3343                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3344                         server->caps |= NFS_CAP_ATIME;
3345                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3346                         server->caps |= NFS_CAP_CTIME;
3347                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3348                         server->caps |= NFS_CAP_MTIME;
3349 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3350                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3351                         server->caps |= NFS_CAP_SECURITY_LABEL;
3352 #endif
3353                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3354                                 sizeof(server->attr_bitmask));
3355                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3356
3357                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3358                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3359                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3360                 server->cache_consistency_bitmask[2] = 0;
3361                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3362                         sizeof(server->exclcreat_bitmask));
3363                 server->acl_bitmask = res.acl_bitmask;
3364                 server->fh_expire_type = res.fh_expire_type;
3365         }
3366
3367         return status;
3368 }
3369
3370 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3371 {
3372         struct nfs4_exception exception = { };
3373         int err;
3374         do {
3375                 err = nfs4_handle_exception(server,
3376                                 _nfs4_server_capabilities(server, fhandle),
3377                                 &exception);
3378         } while (exception.retry);
3379         return err;
3380 }
3381
3382 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3383                 struct nfs_fsinfo *info)
3384 {
3385         u32 bitmask[3];
3386         struct nfs4_lookup_root_arg args = {
3387                 .bitmask = bitmask,
3388         };
3389         struct nfs4_lookup_res res = {
3390                 .server = server,
3391                 .fattr = info->fattr,
3392                 .fh = fhandle,
3393         };
3394         struct rpc_message msg = {
3395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3396                 .rpc_argp = &args,
3397                 .rpc_resp = &res,
3398         };
3399
3400         bitmask[0] = nfs4_fattr_bitmap[0];
3401         bitmask[1] = nfs4_fattr_bitmap[1];
3402         /*
3403          * Process the label in the upcoming getfattr
3404          */
3405         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3406
3407         nfs_fattr_init(info->fattr);
3408         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3409 }
3410
3411 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3412                 struct nfs_fsinfo *info)
3413 {
3414         struct nfs4_exception exception = { };
3415         int err;
3416         do {
3417                 err = _nfs4_lookup_root(server, fhandle, info);
3418                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3419                 switch (err) {
3420                 case 0:
3421                 case -NFS4ERR_WRONGSEC:
3422                         goto out;
3423                 default:
3424                         err = nfs4_handle_exception(server, err, &exception);
3425                 }
3426         } while (exception.retry);
3427 out:
3428         return err;
3429 }
3430
3431 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3432                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3433 {
3434         struct rpc_auth_create_args auth_args = {
3435                 .pseudoflavor = flavor,
3436         };
3437         struct rpc_auth *auth;
3438         int ret;
3439
3440         auth = rpcauth_create(&auth_args, server->client);
3441         if (IS_ERR(auth)) {
3442                 ret = -EACCES;
3443                 goto out;
3444         }
3445         ret = nfs4_lookup_root(server, fhandle, info);
3446 out:
3447         return ret;
3448 }
3449
3450 /*
3451  * Retry pseudoroot lookup with various security flavors.  We do this when:
3452  *
3453  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3454  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3455  *
3456  * Returns zero on success, or a negative NFS4ERR value, or a
3457  * negative errno value.
3458  */
3459 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3460                               struct nfs_fsinfo *info)
3461 {
3462         /* Per 3530bis 15.33.5 */
3463         static const rpc_authflavor_t flav_array[] = {
3464                 RPC_AUTH_GSS_KRB5P,
3465                 RPC_AUTH_GSS_KRB5I,
3466                 RPC_AUTH_GSS_KRB5,
3467                 RPC_AUTH_UNIX,                  /* courtesy */
3468                 RPC_AUTH_NULL,
3469         };
3470         int status = -EPERM;
3471         size_t i;
3472
3473         if (server->auth_info.flavor_len > 0) {
3474                 /* try each flavor specified by user */
3475                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3476                         status = nfs4_lookup_root_sec(server, fhandle, info,
3477                                                 server->auth_info.flavors[i]);
3478                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3479                                 continue;
3480                         break;
3481                 }
3482         } else {
3483                 /* no flavors specified by user, try default list */
3484                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3485                         status = nfs4_lookup_root_sec(server, fhandle, info,
3486                                                       flav_array[i]);
3487                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3488                                 continue;
3489                         break;
3490                 }
3491         }
3492
3493         /*
3494          * -EACCESS could mean that the user doesn't have correct permissions
3495          * to access the mount.  It could also mean that we tried to mount
3496          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3497          * existing mount programs don't handle -EACCES very well so it should
3498          * be mapped to -EPERM instead.
3499          */
3500         if (status == -EACCES)
3501                 status = -EPERM;
3502         return status;
3503 }
3504
3505 /**
3506  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3507  * @server: initialized nfs_server handle
3508  * @fhandle: we fill in the pseudo-fs root file handle
3509  * @info: we fill in an FSINFO struct
3510  * @auth_probe: probe the auth flavours
3511  *
3512  * Returns zero on success, or a negative errno.
3513  */
3514 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3515                          struct nfs_fsinfo *info,
3516                          bool auth_probe)
3517 {
3518         int status = 0;
3519
3520         if (!auth_probe)
3521                 status = nfs4_lookup_root(server, fhandle, info);
3522
3523         if (auth_probe || status == NFS4ERR_WRONGSEC)
3524                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3525                                 fhandle, info);
3526
3527         if (status == 0)
3528                 status = nfs4_server_capabilities(server, fhandle);
3529         if (status == 0)
3530                 status = nfs4_do_fsinfo(server, fhandle, info);
3531
3532         return nfs4_map_errors(status);
3533 }
3534
3535 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3536                               struct nfs_fsinfo *info)
3537 {
3538         int error;
3539         struct nfs_fattr *fattr = info->fattr;
3540         struct nfs4_label *label = NULL;
3541
3542         error = nfs4_server_capabilities(server, mntfh);
3543         if (error < 0) {
3544                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3545                 return error;
3546         }
3547
3548         label = nfs4_label_alloc(server, GFP_KERNEL);
3549         if (IS_ERR(label))
3550                 return PTR_ERR(label);
3551
3552         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3553         if (error < 0) {
3554                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3555                 goto err_free_label;
3556         }
3557
3558         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3559             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3560                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3561
3562 err_free_label:
3563         nfs4_label_free(label);
3564
3565         return error;
3566 }
3567
3568 /*
3569  * Get locations and (maybe) other attributes of a referral.
3570  * Note that we'll actually follow the referral later when
3571  * we detect fsid mismatch in inode revalidation
3572  */
3573 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3574                              const struct qstr *name, struct nfs_fattr *fattr,
3575                              struct nfs_fh *fhandle)
3576 {
3577         int status = -ENOMEM;
3578         struct page *page = NULL;
3579         struct nfs4_fs_locations *locations = NULL;
3580
3581         page = alloc_page(GFP_KERNEL);
3582         if (page == NULL)
3583                 goto out;
3584         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3585         if (locations == NULL)
3586                 goto out;
3587
3588         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3589         if (status != 0)
3590                 goto out;
3591
3592         /*
3593          * If the fsid didn't change, this is a migration event, not a
3594          * referral.  Cause us to drop into the exception handler, which
3595          * will kick off migration recovery.
3596          */
3597         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3598                 dprintk("%s: server did not return a different fsid for"
3599                         " a referral at %s\n", __func__, name->name);
3600                 status = -NFS4ERR_MOVED;
3601                 goto out;
3602         }
3603         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3604         nfs_fixup_referral_attributes(&locations->fattr);
3605
3606         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3607         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3608         memset(fhandle, 0, sizeof(struct nfs_fh));
3609 out:
3610         if (page)
3611                 __free_page(page);
3612         kfree(locations);
3613         return status;
3614 }
3615
3616 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3617                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3618 {
3619         struct nfs4_getattr_arg args = {
3620                 .fh = fhandle,
3621                 .bitmask = server->attr_bitmask,
3622         };
3623         struct nfs4_getattr_res res = {
3624                 .fattr = fattr,
3625                 .label = label,
3626                 .server = server,
3627         };
3628         struct rpc_message msg = {
3629                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3630                 .rpc_argp = &args,
3631                 .rpc_resp = &res,
3632         };
3633
3634         args.bitmask = nfs4_bitmask(server, label);
3635
3636         nfs_fattr_init(fattr);
3637         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3638 }
3639
3640 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3641                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3642 {
3643         struct nfs4_exception exception = { };
3644         int err;
3645         do {
3646                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3647                 trace_nfs4_getattr(server, fhandle, fattr, err);
3648                 err = nfs4_handle_exception(server, err,
3649                                 &exception);
3650         } while (exception.retry);
3651         return err;
3652 }
3653
3654 /*
3655  * The file is not closed if it is opened due to the a request to change
3656  * the size of the file. The open call will not be needed once the
3657  * VFS layer lookup-intents are implemented.
3658  *
3659  * Close is called when the inode is destroyed.
3660  * If we haven't opened the file for O_WRONLY, we
3661  * need to in the size_change case to obtain a stateid.
3662  *
3663  * Got race?
3664  * Because OPEN is always done by name in nfsv4, it is
3665  * possible that we opened a different file by the same
3666  * name.  We can recognize this race condition, but we
3667  * can't do anything about it besides returning an error.
3668  *
3669  * This will be fixed with VFS changes (lookup-intent).
3670  */
3671 static int
3672 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3673                   struct iattr *sattr)
3674 {
3675         struct inode *inode = d_inode(dentry);
3676         struct rpc_cred *cred = NULL;
3677         struct nfs4_state *state = NULL;
3678         struct nfs4_label *label = NULL;
3679         int status;
3680
3681         if (pnfs_ld_layoutret_on_setattr(inode) &&
3682             sattr->ia_valid & ATTR_SIZE &&
3683             sattr->ia_size < i_size_read(inode))
3684                 pnfs_commit_and_return_layout(inode);
3685
3686         nfs_fattr_init(fattr);
3687
3688         /* Deal with open(O_TRUNC) */
3689         if (sattr->ia_valid & ATTR_OPEN)
3690                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3691
3692         /* Optimization: if the end result is no change, don't RPC */
3693         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3694                 return 0;
3695
3696         /* Search for an existing open(O_WRITE) file */
3697         if (sattr->ia_valid & ATTR_FILE) {
3698                 struct nfs_open_context *ctx;
3699
3700                 ctx = nfs_file_open_context(sattr->ia_file);
3701                 if (ctx) {
3702                         cred = ctx->cred;
3703                         state = ctx->state;
3704                 }
3705         }
3706
3707         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3708         if (IS_ERR(label))
3709                 return PTR_ERR(label);
3710
3711         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3712         if (status == 0) {
3713                 nfs_setattr_update_inode(inode, sattr, fattr);
3714                 nfs_setsecurity(inode, fattr, label);
3715         }
3716         nfs4_label_free(label);
3717         return status;
3718 }
3719
3720 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3721                 const struct qstr *name, struct nfs_fh *fhandle,
3722                 struct nfs_fattr *fattr, struct nfs4_label *label)
3723 {
3724         struct nfs_server *server = NFS_SERVER(dir);
3725         int                    status;
3726         struct nfs4_lookup_arg args = {
3727                 .bitmask = server->attr_bitmask,
3728                 .dir_fh = NFS_FH(dir),
3729                 .name = name,
3730         };
3731         struct nfs4_lookup_res res = {
3732                 .server = server,
3733                 .fattr = fattr,
3734                 .label = label,
3735                 .fh = fhandle,
3736         };
3737         struct rpc_message msg = {
3738                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3739                 .rpc_argp = &args,
3740                 .rpc_resp = &res,
3741         };
3742
3743         args.bitmask = nfs4_bitmask(server, label);
3744
3745         nfs_fattr_init(fattr);
3746
3747         dprintk("NFS call  lookup %s\n", name->name);
3748         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3749         dprintk("NFS reply lookup: %d\n", status);
3750         return status;
3751 }
3752
3753 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3754 {
3755         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3756                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3757         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3758         fattr->nlink = 2;
3759 }
3760
3761 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3762                                    const struct qstr *name, struct nfs_fh *fhandle,
3763                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3764 {
3765         struct nfs4_exception exception = { };
3766         struct rpc_clnt *client = *clnt;
3767         int err;
3768         do {
3769                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3770                 trace_nfs4_lookup(dir, name, err);
3771                 switch (err) {
3772                 case -NFS4ERR_BADNAME:
3773                         err = -ENOENT;
3774                         goto out;
3775                 case -NFS4ERR_MOVED:
3776                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3777                         if (err == -NFS4ERR_MOVED)
3778                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3779                         goto out;
3780                 case -NFS4ERR_WRONGSEC:
3781                         err = -EPERM;
3782                         if (client != *clnt)
3783                                 goto out;
3784                         client = nfs4_negotiate_security(client, dir, name);
3785                         if (IS_ERR(client))
3786                                 return PTR_ERR(client);
3787
3788                         exception.retry = 1;
3789                         break;
3790                 default:
3791                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3792                 }
3793         } while (exception.retry);
3794
3795 out:
3796         if (err == 0)
3797                 *clnt = client;
3798         else if (client != *clnt)
3799                 rpc_shutdown_client(client);
3800
3801         return err;
3802 }
3803
3804 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3805                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3806                             struct nfs4_label *label)
3807 {
3808         int status;
3809         struct rpc_clnt *client = NFS_CLIENT(dir);
3810
3811         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3812         if (client != NFS_CLIENT(dir)) {
3813                 rpc_shutdown_client(client);
3814                 nfs_fixup_secinfo_attributes(fattr);
3815         }
3816         return status;
3817 }
3818
3819 struct rpc_clnt *
3820 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3821                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3822 {
3823         struct rpc_clnt *client = NFS_CLIENT(dir);
3824         int status;
3825
3826         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3827         if (status < 0)
3828                 return ERR_PTR(status);
3829         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3830 }
3831
3832 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3833 {
3834         struct nfs_server *server = NFS_SERVER(inode);
3835         struct nfs4_accessargs args = {
3836                 .fh = NFS_FH(inode),
3837                 .bitmask = server->cache_consistency_bitmask,
3838         };
3839         struct nfs4_accessres res = {
3840                 .server = server,
3841         };
3842         struct rpc_message msg = {
3843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3844                 .rpc_argp = &args,
3845                 .rpc_resp = &res,
3846                 .rpc_cred = entry->cred,
3847         };
3848         int mode = entry->mask;
3849         int status = 0;
3850
3851         /*
3852          * Determine which access bits we want to ask for...
3853          */
3854         if (mode & MAY_READ)
3855                 args.access |= NFS4_ACCESS_READ;
3856         if (S_ISDIR(inode->i_mode)) {
3857                 if (mode & MAY_WRITE)
3858                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3859                 if (mode & MAY_EXEC)
3860                         args.access |= NFS4_ACCESS_LOOKUP;
3861         } else {
3862                 if (mode & MAY_WRITE)
3863                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3864                 if (mode & MAY_EXEC)
3865                         args.access |= NFS4_ACCESS_EXECUTE;
3866         }
3867
3868         res.fattr = nfs_alloc_fattr();
3869         if (res.fattr == NULL)
3870                 return -ENOMEM;
3871
3872         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3873         if (!status) {
3874                 nfs_access_set_mask(entry, res.access);
3875                 nfs_refresh_inode(inode, res.fattr);
3876         }
3877         nfs_free_fattr(res.fattr);
3878         return status;
3879 }
3880
3881 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3882 {
3883         struct nfs4_exception exception = { };
3884         int err;
3885         do {
3886                 err = _nfs4_proc_access(inode, entry);
3887                 trace_nfs4_access(inode, err);
3888                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3889                                 &exception);
3890         } while (exception.retry);
3891         return err;
3892 }
3893
3894 /*
3895  * TODO: For the time being, we don't try to get any attributes
3896  * along with any of the zero-copy operations READ, READDIR,
3897  * READLINK, WRITE.
3898  *
3899  * In the case of the first three, we want to put the GETATTR
3900  * after the read-type operation -- this is because it is hard
3901  * to predict the length of a GETATTR response in v4, and thus
3902  * align the READ data correctly.  This means that the GETATTR
3903  * may end up partially falling into the page cache, and we should
3904  * shift it into the 'tail' of the xdr_buf before processing.
3905  * To do this efficiently, we need to know the total length
3906  * of data received, which doesn't seem to be available outside
3907  * of the RPC layer.
3908  *
3909  * In the case of WRITE, we also want to put the GETATTR after
3910  * the operation -- in this case because we want to make sure
3911  * we get the post-operation mtime and size.
3912  *
3913  * Both of these changes to the XDR layer would in fact be quite
3914  * minor, but I decided to leave them for a subsequent patch.
3915  */
3916 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3917                 unsigned int pgbase, unsigned int pglen)
3918 {
3919         struct nfs4_readlink args = {
3920                 .fh       = NFS_FH(inode),
3921                 .pgbase   = pgbase,
3922                 .pglen    = pglen,
3923                 .pages    = &page,
3924         };
3925         struct nfs4_readlink_res res;
3926         struct rpc_message msg = {
3927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3928                 .rpc_argp = &args,
3929                 .rpc_resp = &res,
3930         };
3931
3932         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3933 }
3934
3935 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3936                 unsigned int pgbase, unsigned int pglen)
3937 {
3938         struct nfs4_exception exception = { };
3939         int err;
3940         do {
3941                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3942                 trace_nfs4_readlink(inode, err);
3943                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3944                                 &exception);
3945         } while (exception.retry);
3946         return err;
3947 }
3948
3949 /*
3950  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3951  */
3952 static int
3953 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3954                  int flags)
3955 {
3956         struct nfs4_label l, *ilabel = NULL;
3957         struct nfs_open_context *ctx;
3958         struct nfs4_state *state;
3959         int status = 0;
3960
3961         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3962         if (IS_ERR(ctx))
3963                 return PTR_ERR(ctx);
3964
3965         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3966
3967         sattr->ia_mode &= ~current_umask();
3968         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3969         if (IS_ERR(state)) {
3970                 status = PTR_ERR(state);
3971                 goto out;
3972         }
3973 out:
3974         nfs4_label_release_security(ilabel);
3975         put_nfs_open_context(ctx);
3976         return status;
3977 }
3978
3979 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3980 {
3981         struct nfs_server *server = NFS_SERVER(dir);
3982         struct nfs_removeargs args = {
3983                 .fh = NFS_FH(dir),
3984                 .name = *name,
3985         };
3986         struct nfs_removeres res = {
3987                 .server = server,
3988         };
3989         struct rpc_message msg = {
3990                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3991                 .rpc_argp = &args,
3992                 .rpc_resp = &res,
3993         };
3994         int status;
3995
3996         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3997         if (status == 0)
3998                 update_changeattr(dir, &res.cinfo);
3999         return status;
4000 }
4001
4002 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4003 {
4004         struct nfs4_exception exception = { };
4005         int err;
4006         do {
4007                 err = _nfs4_proc_remove(dir, name);
4008                 trace_nfs4_remove(dir, name, err);
4009                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4010                                 &exception);
4011         } while (exception.retry);
4012         return err;
4013 }
4014
4015 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4016 {
4017         struct nfs_server *server = NFS_SERVER(dir);
4018         struct nfs_removeargs *args = msg->rpc_argp;
4019         struct nfs_removeres *res = msg->rpc_resp;
4020
4021         res->server = server;
4022         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4023         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4024
4025         nfs_fattr_init(res->dir_attr);
4026 }
4027
4028 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4029 {
4030         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4031                         &data->args.seq_args,
4032                         &data->res.seq_res,
4033                         task);
4034 }
4035
4036 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4037 {
4038         struct nfs_unlinkdata *data = task->tk_calldata;
4039         struct nfs_removeres *res = &data->res;
4040
4041         if (!nfs4_sequence_done(task, &res->seq_res))
4042                 return 0;
4043         if (nfs4_async_handle_error(task, res->server, NULL,
4044                                     &data->timeout) == -EAGAIN)
4045                 return 0;
4046         update_changeattr(dir, &res->cinfo);
4047         return 1;
4048 }
4049
4050 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4051 {
4052         struct nfs_server *server = NFS_SERVER(dir);
4053         struct nfs_renameargs *arg = msg->rpc_argp;
4054         struct nfs_renameres *res = msg->rpc_resp;
4055
4056         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4057         res->server = server;
4058         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4059 }
4060
4061 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4062 {
4063         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4064                         &data->args.seq_args,
4065                         &data->res.seq_res,
4066                         task);
4067 }
4068
4069 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4070                                  struct inode *new_dir)
4071 {
4072         struct nfs_renamedata *data = task->tk_calldata;
4073         struct nfs_renameres *res = &data->res;
4074
4075         if (!nfs4_sequence_done(task, &res->seq_res))
4076                 return 0;
4077         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4078                 return 0;
4079
4080         update_changeattr(old_dir, &res->old_cinfo);
4081         update_changeattr(new_dir, &res->new_cinfo);
4082         return 1;
4083 }
4084
4085 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4086 {
4087         struct nfs_server *server = NFS_SERVER(inode);
4088         struct nfs4_link_arg arg = {
4089                 .fh     = NFS_FH(inode),
4090                 .dir_fh = NFS_FH(dir),
4091                 .name   = name,
4092                 .bitmask = server->attr_bitmask,
4093         };
4094         struct nfs4_link_res res = {
4095                 .server = server,
4096                 .label = NULL,
4097         };
4098         struct rpc_message msg = {
4099                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4100                 .rpc_argp = &arg,
4101                 .rpc_resp = &res,
4102         };
4103         int status = -ENOMEM;
4104
4105         res.fattr = nfs_alloc_fattr();
4106         if (res.fattr == NULL)
4107                 goto out;
4108
4109         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4110         if (IS_ERR(res.label)) {
4111                 status = PTR_ERR(res.label);
4112                 goto out;
4113         }
4114         arg.bitmask = nfs4_bitmask(server, res.label);
4115
4116         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4117         if (!status) {
4118                 update_changeattr(dir, &res.cinfo);
4119                 status = nfs_post_op_update_inode(inode, res.fattr);
4120                 if (!status)
4121                         nfs_setsecurity(inode, res.fattr, res.label);
4122         }
4123
4124
4125         nfs4_label_free(res.label);
4126
4127 out:
4128         nfs_free_fattr(res.fattr);
4129         return status;
4130 }
4131
4132 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4133 {
4134         struct nfs4_exception exception = { };
4135         int err;
4136         do {
4137                 err = nfs4_handle_exception(NFS_SERVER(inode),
4138                                 _nfs4_proc_link(inode, dir, name),
4139                                 &exception);
4140         } while (exception.retry);
4141         return err;
4142 }
4143
4144 struct nfs4_createdata {
4145         struct rpc_message msg;
4146         struct nfs4_create_arg arg;
4147         struct nfs4_create_res res;
4148         struct nfs_fh fh;
4149         struct nfs_fattr fattr;
4150         struct nfs4_label *label;
4151 };
4152
4153 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4154                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4155 {
4156         struct nfs4_createdata *data;
4157
4158         data = kzalloc(sizeof(*data), GFP_KERNEL);
4159         if (data != NULL) {
4160                 struct nfs_server *server = NFS_SERVER(dir);
4161
4162                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4163                 if (IS_ERR(data->label))
4164                         goto out_free;
4165
4166                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4167                 data->msg.rpc_argp = &data->arg;
4168                 data->msg.rpc_resp = &data->res;
4169                 data->arg.dir_fh = NFS_FH(dir);
4170                 data->arg.server = server;
4171                 data->arg.name = name;
4172                 data->arg.attrs = sattr;
4173                 data->arg.ftype = ftype;
4174                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4175                 data->res.server = server;
4176                 data->res.fh = &data->fh;
4177                 data->res.fattr = &data->fattr;
4178                 data->res.label = data->label;
4179                 nfs_fattr_init(data->res.fattr);
4180         }
4181         return data;
4182 out_free:
4183         kfree(data);
4184         return NULL;
4185 }
4186
4187 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4188 {
4189         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4190                                     &data->arg.seq_args, &data->res.seq_res, 1);
4191         if (status == 0) {
4192                 update_changeattr(dir, &data->res.dir_cinfo);
4193                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4194         }
4195         return status;
4196 }
4197
4198 static void nfs4_free_createdata(struct nfs4_createdata *data)
4199 {
4200         nfs4_label_free(data->label);
4201         kfree(data);
4202 }
4203
4204 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4205                 struct page *page, unsigned int len, struct iattr *sattr,
4206                 struct nfs4_label *label)
4207 {
4208         struct nfs4_createdata *data;
4209         int status = -ENAMETOOLONG;
4210
4211         if (len > NFS4_MAXPATHLEN)
4212                 goto out;
4213
4214         status = -ENOMEM;
4215         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4216         if (data == NULL)
4217                 goto out;
4218
4219         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4220         data->arg.u.symlink.pages = &page;
4221         data->arg.u.symlink.len = len;
4222         data->arg.label = label;
4223
4224         status = nfs4_do_create(dir, dentry, data);
4225
4226         nfs4_free_createdata(data);
4227 out:
4228         return status;
4229 }
4230
4231 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4232                 struct page *page, unsigned int len, struct iattr *sattr)
4233 {
4234         struct nfs4_exception exception = { };
4235         struct nfs4_label l, *label = NULL;
4236         int err;
4237
4238         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4239
4240         do {
4241                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4242                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4243                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4244                                 &exception);
4245         } while (exception.retry);
4246
4247         nfs4_label_release_security(label);
4248         return err;
4249 }
4250
4251 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4252                 struct iattr *sattr, struct nfs4_label *label)
4253 {
4254         struct nfs4_createdata *data;
4255         int status = -ENOMEM;
4256
4257         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4258         if (data == NULL)
4259                 goto out;
4260
4261         data->arg.label = label;
4262         status = nfs4_do_create(dir, dentry, data);
4263
4264         nfs4_free_createdata(data);
4265 out:
4266         return status;
4267 }
4268
4269 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4270                 struct iattr *sattr)
4271 {
4272         struct nfs4_exception exception = { };
4273         struct nfs4_label l, *label = NULL;
4274         int err;
4275
4276         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4277
4278         sattr->ia_mode &= ~current_umask();
4279         do {
4280                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4281                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4282                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4283                                 &exception);
4284         } while (exception.retry);
4285         nfs4_label_release_security(label);
4286
4287         return err;
4288 }
4289
4290 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4291                 u64 cookie, struct page **pages, unsigned int count, int plus)
4292 {
4293         struct inode            *dir = d_inode(dentry);
4294         struct nfs4_readdir_arg args = {
4295                 .fh = NFS_FH(dir),
4296                 .pages = pages,
4297                 .pgbase = 0,
4298                 .count = count,
4299                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4300                 .plus = plus,
4301         };
4302         struct nfs4_readdir_res res;
4303         struct rpc_message msg = {
4304                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4305                 .rpc_argp = &args,
4306                 .rpc_resp = &res,
4307                 .rpc_cred = cred,
4308         };
4309         int                     status;
4310
4311         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4312                         dentry,
4313                         (unsigned long long)cookie);
4314         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4315         res.pgbase = args.pgbase;
4316         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4317         if (status >= 0) {
4318                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4319                 status += args.pgbase;
4320         }
4321
4322         nfs_invalidate_atime(dir);
4323
4324         dprintk("%s: returns %d\n", __func__, status);
4325         return status;
4326 }
4327
4328 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4329                 u64 cookie, struct page **pages, unsigned int count, int plus)
4330 {
4331         struct nfs4_exception exception = { };
4332         int err;
4333         do {
4334                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4335                                 pages, count, plus);
4336                 trace_nfs4_readdir(d_inode(dentry), err);
4337                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4338                                 &exception);
4339         } while (exception.retry);
4340         return err;
4341 }
4342
4343 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4344                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4345 {
4346         struct nfs4_createdata *data;
4347         int mode = sattr->ia_mode;
4348         int status = -ENOMEM;
4349
4350         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4351         if (data == NULL)
4352                 goto out;
4353
4354         if (S_ISFIFO(mode))
4355                 data->arg.ftype = NF4FIFO;
4356         else if (S_ISBLK(mode)) {
4357                 data->arg.ftype = NF4BLK;
4358                 data->arg.u.device.specdata1 = MAJOR(rdev);
4359                 data->arg.u.device.specdata2 = MINOR(rdev);
4360         }
4361         else if (S_ISCHR(mode)) {
4362                 data->arg.ftype = NF4CHR;
4363                 data->arg.u.device.specdata1 = MAJOR(rdev);
4364                 data->arg.u.device.specdata2 = MINOR(rdev);
4365         } else if (!S_ISSOCK(mode)) {
4366                 status = -EINVAL;
4367                 goto out_free;
4368         }
4369
4370         data->arg.label = label;
4371         status = nfs4_do_create(dir, dentry, data);
4372 out_free:
4373         nfs4_free_createdata(data);
4374 out:
4375         return status;
4376 }
4377
4378 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4379                 struct iattr *sattr, dev_t rdev)
4380 {
4381         struct nfs4_exception exception = { };
4382         struct nfs4_label l, *label = NULL;
4383         int err;
4384
4385         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4386
4387         sattr->ia_mode &= ~current_umask();
4388         do {
4389                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4390                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4391                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4392                                 &exception);
4393         } while (exception.retry);
4394
4395         nfs4_label_release_security(label);
4396
4397         return err;
4398 }
4399
4400 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4401                  struct nfs_fsstat *fsstat)
4402 {
4403         struct nfs4_statfs_arg args = {
4404                 .fh = fhandle,
4405                 .bitmask = server->attr_bitmask,
4406         };
4407         struct nfs4_statfs_res res = {
4408                 .fsstat = fsstat,
4409         };
4410         struct rpc_message msg = {
4411                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4412                 .rpc_argp = &args,
4413                 .rpc_resp = &res,
4414         };
4415
4416         nfs_fattr_init(fsstat->fattr);
4417         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4418 }
4419
4420 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4421 {
4422         struct nfs4_exception exception = { };
4423         int err;
4424         do {
4425                 err = nfs4_handle_exception(server,
4426                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4427                                 &exception);
4428         } while (exception.retry);
4429         return err;
4430 }
4431
4432 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4433                 struct nfs_fsinfo *fsinfo)
4434 {
4435         struct nfs4_fsinfo_arg args = {
4436                 .fh = fhandle,
4437                 .bitmask = server->attr_bitmask,
4438         };
4439         struct nfs4_fsinfo_res res = {
4440                 .fsinfo = fsinfo,
4441         };
4442         struct rpc_message msg = {
4443                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4444                 .rpc_argp = &args,
4445                 .rpc_resp = &res,
4446         };
4447
4448         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4449 }
4450
4451 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4452 {
4453         struct nfs4_exception exception = { };
4454         unsigned long now = jiffies;
4455         int err;
4456
4457         do {
4458                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4459                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4460                 if (err == 0) {
4461                         nfs4_set_lease_period(server->nfs_client,
4462                                         fsinfo->lease_time * HZ,
4463                                         now);
4464                         break;
4465                 }
4466                 err = nfs4_handle_exception(server, err, &exception);
4467         } while (exception.retry);
4468         return err;
4469 }
4470
4471 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4472 {
4473         int error;
4474
4475         nfs_fattr_init(fsinfo->fattr);
4476         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4477         if (error == 0) {
4478                 /* block layout checks this! */
4479                 server->pnfs_blksize = fsinfo->blksize;
4480                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4481         }
4482
4483         return error;
4484 }
4485
4486 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4487                 struct nfs_pathconf *pathconf)
4488 {
4489         struct nfs4_pathconf_arg args = {
4490                 .fh = fhandle,
4491                 .bitmask = server->attr_bitmask,
4492         };
4493         struct nfs4_pathconf_res res = {
4494                 .pathconf = pathconf,
4495         };
4496         struct rpc_message msg = {
4497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4498                 .rpc_argp = &args,
4499                 .rpc_resp = &res,
4500         };
4501
4502         /* None of the pathconf attributes are mandatory to implement */
4503         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4504                 memset(pathconf, 0, sizeof(*pathconf));
4505                 return 0;
4506         }
4507
4508         nfs_fattr_init(pathconf->fattr);
4509         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4510 }
4511
4512 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4513                 struct nfs_pathconf *pathconf)
4514 {
4515         struct nfs4_exception exception = { };
4516         int err;
4517
4518         do {
4519                 err = nfs4_handle_exception(server,
4520                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4521                                 &exception);
4522         } while (exception.retry);
4523         return err;
4524 }
4525
4526 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4527                 const struct nfs_open_context *ctx,
4528                 const struct nfs_lock_context *l_ctx,
4529                 fmode_t fmode)
4530 {
4531         const struct nfs_lockowner *lockowner = NULL;
4532
4533         if (l_ctx != NULL)
4534                 lockowner = &l_ctx->lockowner;
4535         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4536 }
4537 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4538
4539 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4540                 const struct nfs_open_context *ctx,
4541                 const struct nfs_lock_context *l_ctx,
4542                 fmode_t fmode)
4543 {
4544         nfs4_stateid current_stateid;
4545
4546         /* If the current stateid represents a lost lock, then exit */
4547         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4548                 return true;
4549         return nfs4_stateid_match(stateid, &current_stateid);
4550 }
4551
4552 static bool nfs4_error_stateid_expired(int err)
4553 {
4554         switch (err) {
4555         case -NFS4ERR_DELEG_REVOKED:
4556         case -NFS4ERR_ADMIN_REVOKED:
4557         case -NFS4ERR_BAD_STATEID:
4558         case -NFS4ERR_STALE_STATEID:
4559         case -NFS4ERR_OLD_STATEID:
4560         case -NFS4ERR_OPENMODE:
4561         case -NFS4ERR_EXPIRED:
4562                 return true;
4563         }
4564         return false;
4565 }
4566
4567 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4568 {
4569         struct nfs_server *server = NFS_SERVER(hdr->inode);
4570
4571         trace_nfs4_read(hdr, task->tk_status);
4572         if (task->tk_status < 0) {
4573                 struct nfs4_exception exception = {
4574                         .inode = hdr->inode,
4575                         .state = hdr->args.context->state,
4576                         .stateid = &hdr->args.stateid,
4577                 };
4578                 task->tk_status = nfs4_async_handle_exception(task,
4579                                 server, task->tk_status, &exception);
4580                 if (exception.retry) {
4581                         rpc_restart_call_prepare(task);
4582                         return -EAGAIN;
4583                 }
4584         }
4585
4586         if (task->tk_status > 0)
4587                 renew_lease(server, hdr->timestamp);
4588         return 0;
4589 }
4590
4591 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4592                 struct nfs_pgio_args *args)
4593 {
4594
4595         if (!nfs4_error_stateid_expired(task->tk_status) ||
4596                 nfs4_stateid_is_current(&args->stateid,
4597                                 args->context,
4598                                 args->lock_context,
4599                                 FMODE_READ))
4600                 return false;
4601         rpc_restart_call_prepare(task);
4602         return true;
4603 }
4604
4605 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4606 {
4607
4608         dprintk("--> %s\n", __func__);
4609
4610         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4611                 return -EAGAIN;
4612         if (nfs4_read_stateid_changed(task, &hdr->args))
4613                 return -EAGAIN;
4614         if (task->tk_status > 0)
4615                 nfs_invalidate_atime(hdr->inode);
4616         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4617                                     nfs4_read_done_cb(task, hdr);
4618 }
4619
4620 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4621                                  struct rpc_message *msg)
4622 {
4623         hdr->timestamp   = jiffies;
4624         if (!hdr->pgio_done_cb)
4625                 hdr->pgio_done_cb = nfs4_read_done_cb;
4626         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4627         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4628 }
4629
4630 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4631                                       struct nfs_pgio_header *hdr)
4632 {
4633         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4634                         &hdr->args.seq_args,
4635                         &hdr->res.seq_res,
4636                         task))
4637                 return 0;
4638         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4639                                 hdr->args.lock_context,
4640                                 hdr->rw_ops->rw_mode) == -EIO)
4641                 return -EIO;
4642         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4643                 return -EIO;
4644         return 0;
4645 }
4646
4647 static int nfs4_write_done_cb(struct rpc_task *task,
4648                               struct nfs_pgio_header *hdr)
4649 {
4650         struct inode *inode = hdr->inode;
4651
4652         trace_nfs4_write(hdr, task->tk_status);
4653         if (task->tk_status < 0) {
4654                 struct nfs4_exception exception = {
4655                         .inode = hdr->inode,
4656                         .state = hdr->args.context->state,
4657                         .stateid = &hdr->args.stateid,
4658                 };
4659                 task->tk_status = nfs4_async_handle_exception(task,
4660                                 NFS_SERVER(inode), task->tk_status,
4661                                 &exception);
4662                 if (exception.retry) {
4663                         rpc_restart_call_prepare(task);
4664                         return -EAGAIN;
4665                 }
4666         }
4667         if (task->tk_status >= 0) {
4668                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4669                 nfs_writeback_update_inode(hdr);
4670         }
4671         return 0;
4672 }
4673
4674 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4675                 struct nfs_pgio_args *args)
4676 {
4677
4678         if (!nfs4_error_stateid_expired(task->tk_status) ||
4679                 nfs4_stateid_is_current(&args->stateid,
4680                                 args->context,
4681                                 args->lock_context,
4682                                 FMODE_WRITE))
4683                 return false;
4684         rpc_restart_call_prepare(task);
4685         return true;
4686 }
4687
4688 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4689 {
4690         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4691                 return -EAGAIN;
4692         if (nfs4_write_stateid_changed(task, &hdr->args))
4693                 return -EAGAIN;
4694         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4695                 nfs4_write_done_cb(task, hdr);
4696 }
4697
4698 static
4699 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4700 {
4701         /* Don't request attributes for pNFS or O_DIRECT writes */
4702         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4703                 return false;
4704         /* Otherwise, request attributes if and only if we don't hold
4705          * a delegation
4706          */
4707         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4708 }
4709
4710 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4711                                   struct rpc_message *msg)
4712 {
4713         struct nfs_server *server = NFS_SERVER(hdr->inode);
4714
4715         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4716                 hdr->args.bitmask = NULL;
4717                 hdr->res.fattr = NULL;
4718         } else
4719                 hdr->args.bitmask = server->cache_consistency_bitmask;
4720
4721         if (!hdr->pgio_done_cb)
4722                 hdr->pgio_done_cb = nfs4_write_done_cb;
4723         hdr->res.server = server;
4724         hdr->timestamp   = jiffies;
4725
4726         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4727         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4728 }
4729
4730 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4731 {
4732         nfs4_setup_sequence(NFS_SERVER(data->inode),
4733                         &data->args.seq_args,
4734                         &data->res.seq_res,
4735                         task);
4736 }
4737
4738 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4739 {
4740         struct inode *inode = data->inode;
4741
4742         trace_nfs4_commit(data, task->tk_status);
4743         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4744                                     NULL, NULL) == -EAGAIN) {
4745                 rpc_restart_call_prepare(task);
4746                 return -EAGAIN;
4747         }
4748         return 0;
4749 }
4750
4751 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4752 {
4753         if (!nfs4_sequence_done(task, &data->res.seq_res))
4754                 return -EAGAIN;
4755         return data->commit_done_cb(task, data);
4756 }
4757
4758 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4759 {
4760         struct nfs_server *server = NFS_SERVER(data->inode);
4761
4762         if (data->commit_done_cb == NULL)
4763                 data->commit_done_cb = nfs4_commit_done_cb;
4764         data->res.server = server;
4765         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4766         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4767 }
4768
4769 struct nfs4_renewdata {
4770         struct nfs_client       *client;
4771         unsigned long           timestamp;
4772 };
4773
4774 /*
4775  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4776  * standalone procedure for queueing an asynchronous RENEW.
4777  */
4778 static void nfs4_renew_release(void *calldata)
4779 {
4780         struct nfs4_renewdata *data = calldata;
4781         struct nfs_client *clp = data->client;
4782
4783         if (atomic_read(&clp->cl_count) > 1)
4784                 nfs4_schedule_state_renewal(clp);
4785         nfs_put_client(clp);
4786         kfree(data);
4787 }
4788
4789 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4790 {
4791         struct nfs4_renewdata *data = calldata;
4792         struct nfs_client *clp = data->client;
4793         unsigned long timestamp = data->timestamp;
4794
4795         trace_nfs4_renew_async(clp, task->tk_status);
4796         switch (task->tk_status) {
4797         case 0:
4798                 break;
4799         case -NFS4ERR_LEASE_MOVED:
4800                 nfs4_schedule_lease_moved_recovery(clp);
4801                 break;
4802         default:
4803                 /* Unless we're shutting down, schedule state recovery! */
4804                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4805                         return;
4806                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4807                         nfs4_schedule_lease_recovery(clp);
4808                         return;
4809                 }
4810                 nfs4_schedule_path_down_recovery(clp);
4811         }
4812         do_renew_lease(clp, timestamp);
4813 }
4814
4815 static const struct rpc_call_ops nfs4_renew_ops = {
4816         .rpc_call_done = nfs4_renew_done,
4817         .rpc_release = nfs4_renew_release,
4818 };
4819
4820 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4821 {
4822         struct rpc_message msg = {
4823                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4824                 .rpc_argp       = clp,
4825                 .rpc_cred       = cred,
4826         };
4827         struct nfs4_renewdata *data;
4828
4829         if (renew_flags == 0)
4830                 return 0;
4831         if (!atomic_inc_not_zero(&clp->cl_count))
4832                 return -EIO;
4833         data = kmalloc(sizeof(*data), GFP_NOFS);
4834         if (data == NULL)
4835                 return -ENOMEM;
4836         data->client = clp;
4837         data->timestamp = jiffies;
4838         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4839                         &nfs4_renew_ops, data);
4840 }
4841
4842 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4843 {
4844         struct rpc_message msg = {
4845                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4846                 .rpc_argp       = clp,
4847                 .rpc_cred       = cred,
4848         };
4849         unsigned long now = jiffies;
4850         int status;
4851
4852         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4853         if (status < 0)
4854                 return status;
4855         do_renew_lease(clp, now);
4856         return 0;
4857 }
4858
4859 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4860 {
4861         return server->caps & NFS_CAP_ACLS;
4862 }
4863
4864 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4865  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4866  * the stack.
4867  */
4868 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4869
4870 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4871                 struct page **pages)
4872 {
4873         struct page *newpage, **spages;
4874         int rc = 0;
4875         size_t len;
4876         spages = pages;
4877
4878         do {
4879                 len = min_t(size_t, PAGE_SIZE, buflen);
4880                 newpage = alloc_page(GFP_KERNEL);
4881
4882                 if (newpage == NULL)
4883                         goto unwind;
4884                 memcpy(page_address(newpage), buf, len);
4885                 buf += len;
4886                 buflen -= len;
4887                 *pages++ = newpage;
4888                 rc++;
4889         } while (buflen != 0);
4890
4891         return rc;
4892
4893 unwind:
4894         for(; rc > 0; rc--)
4895                 __free_page(spages[rc-1]);
4896         return -ENOMEM;
4897 }
4898
4899 struct nfs4_cached_acl {
4900         int cached;
4901         size_t len;
4902         char data[0];
4903 };
4904
4905 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4906 {
4907         struct nfs_inode *nfsi = NFS_I(inode);
4908
4909         spin_lock(&inode->i_lock);
4910         kfree(nfsi->nfs4_acl);
4911         nfsi->nfs4_acl = acl;
4912         spin_unlock(&inode->i_lock);
4913 }
4914
4915 static void nfs4_zap_acl_attr(struct inode *inode)
4916 {
4917         nfs4_set_cached_acl(inode, NULL);
4918 }
4919
4920 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4921 {
4922         struct nfs_inode *nfsi = NFS_I(inode);
4923         struct nfs4_cached_acl *acl;
4924         int ret = -ENOENT;
4925
4926         spin_lock(&inode->i_lock);
4927         acl = nfsi->nfs4_acl;
4928         if (acl == NULL)
4929                 goto out;
4930         if (buf == NULL) /* user is just asking for length */
4931                 goto out_len;
4932         if (acl->cached == 0)
4933                 goto out;
4934         ret = -ERANGE; /* see getxattr(2) man page */
4935         if (acl->len > buflen)
4936                 goto out;
4937         memcpy(buf, acl->data, acl->len);
4938 out_len:
4939         ret = acl->len;
4940 out:
4941         spin_unlock(&inode->i_lock);
4942         return ret;
4943 }
4944
4945 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4946 {
4947         struct nfs4_cached_acl *acl;
4948         size_t buflen = sizeof(*acl) + acl_len;
4949
4950         if (buflen <= PAGE_SIZE) {
4951                 acl = kmalloc(buflen, GFP_KERNEL);
4952                 if (acl == NULL)
4953                         goto out;
4954                 acl->cached = 1;
4955                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4956         } else {
4957                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4958                 if (acl == NULL)
4959                         goto out;
4960                 acl->cached = 0;
4961         }
4962         acl->len = acl_len;
4963 out:
4964         nfs4_set_cached_acl(inode, acl);
4965 }
4966
4967 /*
4968  * The getxattr API returns the required buffer length when called with a
4969  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4970  * the required buf.  On a NULL buf, we send a page of data to the server
4971  * guessing that the ACL request can be serviced by a page. If so, we cache
4972  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4973  * the cache. If not so, we throw away the page, and cache the required
4974  * length. The next getxattr call will then produce another round trip to
4975  * the server, this time with the input buf of the required size.
4976  */
4977 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4978 {
4979         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4980         struct nfs_getaclargs args = {
4981                 .fh = NFS_FH(inode),
4982                 .acl_pages = pages,
4983                 .acl_len = buflen,
4984         };
4985         struct nfs_getaclres res = {
4986                 .acl_len = buflen,
4987         };
4988         struct rpc_message msg = {
4989                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4990                 .rpc_argp = &args,
4991                 .rpc_resp = &res,
4992         };
4993         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4994         int ret = -ENOMEM, i;
4995
4996         /* As long as we're doing a round trip to the server anyway,
4997          * let's be prepared for a page of acl data. */
4998         if (npages == 0)
4999                 npages = 1;
5000         if (npages > ARRAY_SIZE(pages))
5001                 return -ERANGE;
5002
5003         for (i = 0; i < npages; i++) {
5004                 pages[i] = alloc_page(GFP_KERNEL);
5005                 if (!pages[i])
5006                         goto out_free;
5007         }
5008
5009         /* for decoding across pages */
5010         res.acl_scratch = alloc_page(GFP_KERNEL);
5011         if (!res.acl_scratch)
5012                 goto out_free;
5013
5014         args.acl_len = npages * PAGE_SIZE;
5015
5016         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5017                 __func__, buf, buflen, npages, args.acl_len);
5018         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5019                              &msg, &args.seq_args, &res.seq_res, 0);
5020         if (ret)
5021                 goto out_free;
5022
5023         /* Handle the case where the passed-in buffer is too short */
5024         if (res.acl_flags & NFS4_ACL_TRUNC) {
5025                 /* Did the user only issue a request for the acl length? */
5026                 if (buf == NULL)
5027                         goto out_ok;
5028                 ret = -ERANGE;
5029                 goto out_free;
5030         }
5031         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5032         if (buf) {
5033                 if (res.acl_len > buflen) {
5034                         ret = -ERANGE;
5035                         goto out_free;
5036                 }
5037                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5038         }
5039 out_ok:
5040         ret = res.acl_len;
5041 out_free:
5042         for (i = 0; i < npages; i++)
5043                 if (pages[i])
5044                         __free_page(pages[i]);
5045         if (res.acl_scratch)
5046                 __free_page(res.acl_scratch);
5047         return ret;
5048 }
5049
5050 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5051 {
5052         struct nfs4_exception exception = { };
5053         ssize_t ret;
5054         do {
5055                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5056                 trace_nfs4_get_acl(inode, ret);
5057                 if (ret >= 0)
5058                         break;
5059                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5060         } while (exception.retry);
5061         return ret;
5062 }
5063
5064 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5065 {
5066         struct nfs_server *server = NFS_SERVER(inode);
5067         int ret;
5068
5069         if (!nfs4_server_supports_acls(server))
5070                 return -EOPNOTSUPP;
5071         ret = nfs_revalidate_inode(server, inode);
5072         if (ret < 0)
5073                 return ret;
5074         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5075                 nfs_zap_acl_cache(inode);
5076         ret = nfs4_read_cached_acl(inode, buf, buflen);
5077         if (ret != -ENOENT)
5078                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5079                  * but no cached acl data, just the acl length */
5080                 return ret;
5081         return nfs4_get_acl_uncached(inode, buf, buflen);
5082 }
5083
5084 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5085 {
5086         struct nfs_server *server = NFS_SERVER(inode);
5087         struct page *pages[NFS4ACL_MAXPAGES];
5088         struct nfs_setaclargs arg = {
5089                 .fh             = NFS_FH(inode),
5090                 .acl_pages      = pages,
5091                 .acl_len        = buflen,
5092         };
5093         struct nfs_setaclres res;
5094         struct rpc_message msg = {
5095                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5096                 .rpc_argp       = &arg,
5097                 .rpc_resp       = &res,
5098         };
5099         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5100         int ret, i;
5101
5102         if (!nfs4_server_supports_acls(server))
5103                 return -EOPNOTSUPP;
5104         if (npages > ARRAY_SIZE(pages))
5105                 return -ERANGE;
5106         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5107         if (i < 0)
5108                 return i;
5109         nfs4_inode_return_delegation(inode);
5110         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5111
5112         /*
5113          * Free each page after tx, so the only ref left is
5114          * held by the network stack
5115          */
5116         for (; i > 0; i--)
5117                 put_page(pages[i-1]);
5118
5119         /*
5120          * Acl update can result in inode attribute update.
5121          * so mark the attribute cache invalid.
5122          */
5123         spin_lock(&inode->i_lock);
5124         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5125         spin_unlock(&inode->i_lock);
5126         nfs_access_zap_cache(inode);
5127         nfs_zap_acl_cache(inode);
5128         return ret;
5129 }
5130
5131 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5132 {
5133         struct nfs4_exception exception = { };
5134         int err;
5135         do {
5136                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5137                 trace_nfs4_set_acl(inode, err);
5138                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5139                                 &exception);
5140         } while (exception.retry);
5141         return err;
5142 }
5143
5144 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5145 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5146                                         size_t buflen)
5147 {
5148         struct nfs_server *server = NFS_SERVER(inode);
5149         struct nfs_fattr fattr;
5150         struct nfs4_label label = {0, 0, buflen, buf};
5151
5152         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5153         struct nfs4_getattr_arg arg = {
5154                 .fh             = NFS_FH(inode),
5155                 .bitmask        = bitmask,
5156         };
5157         struct nfs4_getattr_res res = {
5158                 .fattr          = &fattr,
5159                 .label          = &label,
5160                 .server         = server,
5161         };
5162         struct rpc_message msg = {
5163                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5164                 .rpc_argp       = &arg,
5165                 .rpc_resp       = &res,
5166         };
5167         int ret;
5168
5169         nfs_fattr_init(&fattr);
5170
5171         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5172         if (ret)
5173                 return ret;
5174         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5175                 return -ENOENT;
5176         if (buflen < label.len)
5177                 return -ERANGE;
5178         return 0;
5179 }
5180
5181 static int nfs4_get_security_label(struct inode *inode, void *buf,
5182                                         size_t buflen)
5183 {
5184         struct nfs4_exception exception = { };
5185         int err;
5186
5187         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5188                 return -EOPNOTSUPP;
5189
5190         do {
5191                 err = _nfs4_get_security_label(inode, buf, buflen);
5192                 trace_nfs4_get_security_label(inode, err);
5193                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5194                                 &exception);
5195         } while (exception.retry);
5196         return err;
5197 }
5198
5199 static int _nfs4_do_set_security_label(struct inode *inode,
5200                 struct nfs4_label *ilabel,
5201                 struct nfs_fattr *fattr,
5202                 struct nfs4_label *olabel)
5203 {
5204
5205         struct iattr sattr = {0};
5206         struct nfs_server *server = NFS_SERVER(inode);
5207         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5208         struct nfs_setattrargs arg = {
5209                 .fh             = NFS_FH(inode),
5210                 .iap            = &sattr,
5211                 .server         = server,
5212                 .bitmask        = bitmask,
5213                 .label          = ilabel,
5214         };
5215         struct nfs_setattrres res = {
5216                 .fattr          = fattr,
5217                 .label          = olabel,
5218                 .server         = server,
5219         };
5220         struct rpc_message msg = {
5221                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5222                 .rpc_argp       = &arg,
5223                 .rpc_resp       = &res,
5224         };
5225         int status;
5226
5227         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5228
5229         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5230         if (status)
5231                 dprintk("%s failed: %d\n", __func__, status);
5232
5233         return status;
5234 }
5235
5236 static int nfs4_do_set_security_label(struct inode *inode,
5237                 struct nfs4_label *ilabel,
5238                 struct nfs_fattr *fattr,
5239                 struct nfs4_label *olabel)
5240 {
5241         struct nfs4_exception exception = { };
5242         int err;
5243
5244         do {
5245                 err = _nfs4_do_set_security_label(inode, ilabel,
5246                                 fattr, olabel);
5247                 trace_nfs4_set_security_label(inode, err);
5248                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5249                                 &exception);
5250         } while (exception.retry);
5251         return err;
5252 }
5253
5254 static int
5255 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5256 {
5257         struct nfs4_label ilabel, *olabel = NULL;
5258         struct nfs_fattr fattr;
5259         struct rpc_cred *cred;
5260         int status;
5261
5262         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5263                 return -EOPNOTSUPP;
5264
5265         nfs_fattr_init(&fattr);
5266
5267         ilabel.pi = 0;
5268         ilabel.lfs = 0;
5269         ilabel.label = (char *)buf;
5270         ilabel.len = buflen;
5271
5272         cred = rpc_lookup_cred();
5273         if (IS_ERR(cred))
5274                 return PTR_ERR(cred);
5275
5276         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5277         if (IS_ERR(olabel)) {
5278                 status = -PTR_ERR(olabel);
5279                 goto out;
5280         }
5281
5282         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5283         if (status == 0)
5284                 nfs_setsecurity(inode, &fattr, olabel);
5285
5286         nfs4_label_free(olabel);
5287 out:
5288         put_rpccred(cred);
5289         return status;
5290 }
5291 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5292
5293
5294 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5295                                     nfs4_verifier *bootverf)
5296 {
5297         __be32 verf[2];
5298
5299         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5300                 /* An impossible timestamp guarantees this value
5301                  * will never match a generated boot time. */
5302                 verf[0] = cpu_to_be32(U32_MAX);
5303                 verf[1] = cpu_to_be32(U32_MAX);
5304         } else {
5305                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5306                 u64 ns = ktime_to_ns(nn->boot_time);
5307
5308                 verf[0] = cpu_to_be32(ns >> 32);
5309                 verf[1] = cpu_to_be32(ns);
5310         }
5311         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5312 }
5313
5314 static int
5315 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5316 {
5317         size_t len;
5318         char *str;
5319
5320         if (clp->cl_owner_id != NULL)
5321                 return 0;
5322
5323         rcu_read_lock();
5324         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5325                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5326                 1 +
5327                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5328                 1;
5329         rcu_read_unlock();
5330
5331         if (len > NFS4_OPAQUE_LIMIT + 1)
5332                 return -EINVAL;
5333
5334         /*
5335          * Since this string is allocated at mount time, and held until the
5336          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5337          * about a memory-reclaim deadlock.
5338          */
5339         str = kmalloc(len, GFP_KERNEL);
5340         if (!str)
5341                 return -ENOMEM;
5342
5343         rcu_read_lock();
5344         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5345                         clp->cl_ipaddr,
5346                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5347                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5348         rcu_read_unlock();
5349
5350         clp->cl_owner_id = str;
5351         return 0;
5352 }
5353
5354 static int
5355 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5356 {
5357         size_t len;
5358         char *str;
5359
5360         len = 10 + 10 + 1 + 10 + 1 +
5361                 strlen(nfs4_client_id_uniquifier) + 1 +
5362                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5363
5364         if (len > NFS4_OPAQUE_LIMIT + 1)
5365                 return -EINVAL;
5366
5367         /*
5368          * Since this string is allocated at mount time, and held until the
5369          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5370          * about a memory-reclaim deadlock.
5371          */
5372         str = kmalloc(len, GFP_KERNEL);
5373         if (!str)
5374                 return -ENOMEM;
5375
5376         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5377                         clp->rpc_ops->version, clp->cl_minorversion,
5378                         nfs4_client_id_uniquifier,
5379                         clp->cl_rpcclient->cl_nodename);
5380         clp->cl_owner_id = str;
5381         return 0;
5382 }
5383
5384 static int
5385 nfs4_init_uniform_client_string(struct nfs_client *clp)
5386 {
5387         size_t len;
5388         char *str;
5389
5390         if (clp->cl_owner_id != NULL)
5391                 return 0;
5392
5393         if (nfs4_client_id_uniquifier[0] != '\0')
5394                 return nfs4_init_uniquifier_client_string(clp);
5395
5396         len = 10 + 10 + 1 + 10 + 1 +
5397                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5398
5399         if (len > NFS4_OPAQUE_LIMIT + 1)
5400                 return -EINVAL;
5401
5402         /*
5403          * Since this string is allocated at mount time, and held until the
5404          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5405          * about a memory-reclaim deadlock.
5406          */
5407         str = kmalloc(len, GFP_KERNEL);
5408         if (!str)
5409                 return -ENOMEM;
5410
5411         scnprintf(str, len, "Linux NFSv%u.%u %s",
5412                         clp->rpc_ops->version, clp->cl_minorversion,
5413                         clp->cl_rpcclient->cl_nodename);
5414         clp->cl_owner_id = str;
5415         return 0;
5416 }
5417
5418 /*
5419  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5420  * services.  Advertise one based on the address family of the
5421  * clientaddr.
5422  */
5423 static unsigned int
5424 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5425 {
5426         if (strchr(clp->cl_ipaddr, ':') != NULL)
5427                 return scnprintf(buf, len, "tcp6");
5428         else
5429                 return scnprintf(buf, len, "tcp");
5430 }
5431
5432 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5433 {
5434         struct nfs4_setclientid *sc = calldata;
5435
5436         if (task->tk_status == 0)
5437                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5438 }
5439
5440 static const struct rpc_call_ops nfs4_setclientid_ops = {
5441         .rpc_call_done = nfs4_setclientid_done,
5442 };
5443
5444 /**
5445  * nfs4_proc_setclientid - Negotiate client ID
5446  * @clp: state data structure
5447  * @program: RPC program for NFSv4 callback service
5448  * @port: IP port number for NFS4 callback service
5449  * @cred: RPC credential to use for this call
5450  * @res: where to place the result
5451  *
5452  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5453  */
5454 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5455                 unsigned short port, struct rpc_cred *cred,
5456                 struct nfs4_setclientid_res *res)
5457 {
5458         nfs4_verifier sc_verifier;
5459         struct nfs4_setclientid setclientid = {
5460                 .sc_verifier = &sc_verifier,
5461                 .sc_prog = program,
5462                 .sc_clnt = clp,
5463         };
5464         struct rpc_message msg = {
5465                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5466                 .rpc_argp = &setclientid,
5467                 .rpc_resp = res,
5468                 .rpc_cred = cred,
5469         };
5470         struct rpc_task *task;
5471         struct rpc_task_setup task_setup_data = {
5472                 .rpc_client = clp->cl_rpcclient,
5473                 .rpc_message = &msg,
5474                 .callback_ops = &nfs4_setclientid_ops,
5475                 .callback_data = &setclientid,
5476                 .flags = RPC_TASK_TIMEOUT,
5477         };
5478         int status;
5479
5480         /* nfs_client_id4 */
5481         nfs4_init_boot_verifier(clp, &sc_verifier);
5482
5483         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5484                 status = nfs4_init_uniform_client_string(clp);
5485         else
5486                 status = nfs4_init_nonuniform_client_string(clp);
5487
5488         if (status)
5489                 goto out;
5490
5491         /* cb_client4 */
5492         setclientid.sc_netid_len =
5493                                 nfs4_init_callback_netid(clp,
5494                                                 setclientid.sc_netid,
5495                                                 sizeof(setclientid.sc_netid));
5496         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5497                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5498                                 clp->cl_ipaddr, port >> 8, port & 255);
5499
5500         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5501                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5502                 clp->cl_owner_id);
5503         task = rpc_run_task(&task_setup_data);
5504         if (IS_ERR(task)) {
5505                 status = PTR_ERR(task);
5506                 goto out;
5507         }
5508         status = task->tk_status;
5509         if (setclientid.sc_cred) {
5510                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5511                 put_rpccred(setclientid.sc_cred);
5512         }
5513         rpc_put_task(task);
5514 out:
5515         trace_nfs4_setclientid(clp, status);
5516         dprintk("NFS reply setclientid: %d\n", status);
5517         return status;
5518 }
5519
5520 /**
5521  * nfs4_proc_setclientid_confirm - Confirm client ID
5522  * @clp: state data structure
5523  * @res: result of a previous SETCLIENTID
5524  * @cred: RPC credential to use for this call
5525  *
5526  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5527  */
5528 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5529                 struct nfs4_setclientid_res *arg,
5530                 struct rpc_cred *cred)
5531 {
5532         struct rpc_message msg = {
5533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5534                 .rpc_argp = arg,
5535                 .rpc_cred = cred,
5536         };
5537         int status;
5538
5539         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5540                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5541                 clp->cl_clientid);
5542         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5543         trace_nfs4_setclientid_confirm(clp, status);
5544         dprintk("NFS reply setclientid_confirm: %d\n", status);
5545         return status;
5546 }
5547
5548 struct nfs4_delegreturndata {
5549         struct nfs4_delegreturnargs args;
5550         struct nfs4_delegreturnres res;
5551         struct nfs_fh fh;
5552         nfs4_stateid stateid;
5553         unsigned long timestamp;
5554         struct nfs_fattr fattr;
5555         int rpc_status;
5556         struct inode *inode;
5557         bool roc;
5558         u32 roc_barrier;
5559 };
5560
5561 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5562 {
5563         struct nfs4_delegreturndata *data = calldata;
5564
5565         if (!nfs4_sequence_done(task, &data->res.seq_res))
5566                 return;
5567
5568         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5569         switch (task->tk_status) {
5570         case 0:
5571                 renew_lease(data->res.server, data->timestamp);
5572                 break;
5573         case -NFS4ERR_ADMIN_REVOKED:
5574         case -NFS4ERR_DELEG_REVOKED:
5575         case -NFS4ERR_EXPIRED:
5576                 nfs4_free_revoked_stateid(data->res.server,
5577                                 data->args.stateid,
5578                                 task->tk_msg.rpc_cred);
5579         case -NFS4ERR_BAD_STATEID:
5580         case -NFS4ERR_OLD_STATEID:
5581         case -NFS4ERR_STALE_STATEID:
5582                 task->tk_status = 0;
5583                 break;
5584         default:
5585                 if (nfs4_async_handle_error(task, data->res.server,
5586                                             NULL, NULL) == -EAGAIN) {
5587                         rpc_restart_call_prepare(task);
5588                         return;
5589                 }
5590         }
5591         data->rpc_status = task->tk_status;
5592         if (data->roc && data->rpc_status == 0)
5593                 pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5594 }
5595
5596 static void nfs4_delegreturn_release(void *calldata)
5597 {
5598         struct nfs4_delegreturndata *data = calldata;
5599         struct inode *inode = data->inode;
5600
5601         if (inode) {
5602                 if (data->roc)
5603                         pnfs_roc_release(inode);
5604                 nfs_iput_and_deactive(inode);
5605         }
5606         kfree(calldata);
5607 }
5608
5609 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5610 {
5611         struct nfs4_delegreturndata *d_data;
5612
5613         d_data = (struct nfs4_delegreturndata *)data;
5614
5615         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5616                 return;
5617
5618         if (d_data->roc)
5619                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5620
5621         nfs4_setup_sequence(d_data->res.server,
5622                         &d_data->args.seq_args,
5623                         &d_data->res.seq_res,
5624                         task);
5625 }
5626
5627 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5628         .rpc_call_prepare = nfs4_delegreturn_prepare,
5629         .rpc_call_done = nfs4_delegreturn_done,
5630         .rpc_release = nfs4_delegreturn_release,
5631 };
5632
5633 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5634 {
5635         struct nfs4_delegreturndata *data;
5636         struct nfs_server *server = NFS_SERVER(inode);
5637         struct rpc_task *task;
5638         struct rpc_message msg = {
5639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5640                 .rpc_cred = cred,
5641         };
5642         struct rpc_task_setup task_setup_data = {
5643                 .rpc_client = server->client,
5644                 .rpc_message = &msg,
5645                 .callback_ops = &nfs4_delegreturn_ops,
5646                 .flags = RPC_TASK_ASYNC,
5647         };
5648         int status = 0;
5649
5650         data = kzalloc(sizeof(*data), GFP_NOFS);
5651         if (data == NULL)
5652                 return -ENOMEM;
5653         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5654
5655         nfs4_state_protect(server->nfs_client,
5656                         NFS_SP4_MACH_CRED_CLEANUP,
5657                         &task_setup_data.rpc_client, &msg);
5658
5659         data->args.fhandle = &data->fh;
5660         data->args.stateid = &data->stateid;
5661         data->args.bitmask = server->cache_consistency_bitmask;
5662         nfs_copy_fh(&data->fh, NFS_FH(inode));
5663         nfs4_stateid_copy(&data->stateid, stateid);
5664         data->res.fattr = &data->fattr;
5665         data->res.server = server;
5666         nfs_fattr_init(data->res.fattr);
5667         data->timestamp = jiffies;
5668         data->rpc_status = 0;
5669         data->inode = nfs_igrab_and_active(inode);
5670         if (data->inode)
5671                 data->roc = nfs4_roc(inode);
5672
5673         task_setup_data.callback_data = data;
5674         msg.rpc_argp = &data->args;
5675         msg.rpc_resp = &data->res;
5676         task = rpc_run_task(&task_setup_data);
5677         if (IS_ERR(task))
5678                 return PTR_ERR(task);
5679         if (!issync)
5680                 goto out;
5681         status = nfs4_wait_for_completion_rpc_task(task);
5682         if (status != 0)
5683                 goto out;
5684         status = data->rpc_status;
5685         if (status == 0)
5686                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5687         else
5688                 nfs_refresh_inode(inode, &data->fattr);
5689 out:
5690         rpc_put_task(task);
5691         return status;
5692 }
5693
5694 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5695 {
5696         struct nfs_server *server = NFS_SERVER(inode);
5697         struct nfs4_exception exception = { };
5698         int err;
5699         do {
5700                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5701                 trace_nfs4_delegreturn(inode, stateid, err);
5702                 switch (err) {
5703                         case -NFS4ERR_STALE_STATEID:
5704                         case -NFS4ERR_EXPIRED:
5705                         case 0:
5706                                 return 0;
5707                 }
5708                 err = nfs4_handle_exception(server, err, &exception);
5709         } while (exception.retry);
5710         return err;
5711 }
5712
5713 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5714 {
5715         struct inode *inode = state->inode;
5716         struct nfs_server *server = NFS_SERVER(inode);
5717         struct nfs_client *clp = server->nfs_client;
5718         struct nfs_lockt_args arg = {
5719                 .fh = NFS_FH(inode),
5720                 .fl = request,
5721         };
5722         struct nfs_lockt_res res = {
5723                 .denied = request,
5724         };
5725         struct rpc_message msg = {
5726                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5727                 .rpc_argp       = &arg,
5728                 .rpc_resp       = &res,
5729                 .rpc_cred       = state->owner->so_cred,
5730         };
5731         struct nfs4_lock_state *lsp;
5732         int status;
5733
5734         arg.lock_owner.clientid = clp->cl_clientid;
5735         status = nfs4_set_lock_state(state, request);
5736         if (status != 0)
5737                 goto out;
5738         lsp = request->fl_u.nfs4_fl.owner;
5739         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5740         arg.lock_owner.s_dev = server->s_dev;
5741         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5742         switch (status) {
5743                 case 0:
5744                         request->fl_type = F_UNLCK;
5745                         break;
5746                 case -NFS4ERR_DENIED:
5747                         status = 0;
5748         }
5749         request->fl_ops->fl_release_private(request);
5750         request->fl_ops = NULL;
5751 out:
5752         return status;
5753 }
5754
5755 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5756 {
5757         struct nfs4_exception exception = { };
5758         int err;
5759
5760         do {
5761                 err = _nfs4_proc_getlk(state, cmd, request);
5762                 trace_nfs4_get_lock(request, state, cmd, err);
5763                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5764                                 &exception);
5765         } while (exception.retry);
5766         return err;
5767 }
5768
5769 struct nfs4_unlockdata {
5770         struct nfs_locku_args arg;
5771         struct nfs_locku_res res;
5772         struct nfs4_lock_state *lsp;
5773         struct nfs_open_context *ctx;
5774         struct file_lock fl;
5775         struct nfs_server *server;
5776         unsigned long timestamp;
5777 };
5778
5779 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5780                 struct nfs_open_context *ctx,
5781                 struct nfs4_lock_state *lsp,
5782                 struct nfs_seqid *seqid)
5783 {
5784         struct nfs4_unlockdata *p;
5785         struct inode *inode = lsp->ls_state->inode;
5786
5787         p = kzalloc(sizeof(*p), GFP_NOFS);
5788         if (p == NULL)
5789                 return NULL;
5790         p->arg.fh = NFS_FH(inode);
5791         p->arg.fl = &p->fl;
5792         p->arg.seqid = seqid;
5793         p->res.seqid = seqid;
5794         p->lsp = lsp;
5795         atomic_inc(&lsp->ls_count);
5796         /* Ensure we don't close file until we're done freeing locks! */
5797         p->ctx = get_nfs_open_context(ctx);
5798         memcpy(&p->fl, fl, sizeof(p->fl));
5799         p->server = NFS_SERVER(inode);
5800         return p;
5801 }
5802
5803 static void nfs4_locku_release_calldata(void *data)
5804 {
5805         struct nfs4_unlockdata *calldata = data;
5806         nfs_free_seqid(calldata->arg.seqid);
5807         nfs4_put_lock_state(calldata->lsp);
5808         put_nfs_open_context(calldata->ctx);
5809         kfree(calldata);
5810 }
5811
5812 static void nfs4_locku_done(struct rpc_task *task, void *data)
5813 {
5814         struct nfs4_unlockdata *calldata = data;
5815
5816         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5817                 return;
5818         switch (task->tk_status) {
5819                 case 0:
5820                         renew_lease(calldata->server, calldata->timestamp);
5821                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5822                         if (nfs4_update_lock_stateid(calldata->lsp,
5823                                         &calldata->res.stateid))
5824                                 break;
5825                 case -NFS4ERR_ADMIN_REVOKED:
5826                 case -NFS4ERR_EXPIRED:
5827                         nfs4_free_revoked_stateid(calldata->server,
5828                                         &calldata->arg.stateid,
5829                                         task->tk_msg.rpc_cred);
5830                 case -NFS4ERR_BAD_STATEID:
5831                 case -NFS4ERR_OLD_STATEID:
5832                 case -NFS4ERR_STALE_STATEID:
5833                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5834                                                 &calldata->lsp->ls_stateid))
5835                                 rpc_restart_call_prepare(task);
5836                         break;
5837                 default:
5838                         if (nfs4_async_handle_error(task, calldata->server,
5839                                                     NULL, NULL) == -EAGAIN)
5840                                 rpc_restart_call_prepare(task);
5841         }
5842         nfs_release_seqid(calldata->arg.seqid);
5843 }
5844
5845 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5846 {
5847         struct nfs4_unlockdata *calldata = data;
5848
5849         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5850                 goto out_wait;
5851         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5852         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5853                 /* Note: exit _without_ running nfs4_locku_done */
5854                 goto out_no_action;
5855         }
5856         calldata->timestamp = jiffies;
5857         if (nfs4_setup_sequence(calldata->server,
5858                                 &calldata->arg.seq_args,
5859                                 &calldata->res.seq_res,
5860                                 task) != 0)
5861                 nfs_release_seqid(calldata->arg.seqid);
5862         return;
5863 out_no_action:
5864         task->tk_action = NULL;
5865 out_wait:
5866         nfs4_sequence_done(task, &calldata->res.seq_res);
5867 }
5868
5869 static const struct rpc_call_ops nfs4_locku_ops = {
5870         .rpc_call_prepare = nfs4_locku_prepare,
5871         .rpc_call_done = nfs4_locku_done,
5872         .rpc_release = nfs4_locku_release_calldata,
5873 };
5874
5875 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5876                 struct nfs_open_context *ctx,
5877                 struct nfs4_lock_state *lsp,
5878                 struct nfs_seqid *seqid)
5879 {
5880         struct nfs4_unlockdata *data;
5881         struct rpc_message msg = {
5882                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5883                 .rpc_cred = ctx->cred,
5884         };
5885         struct rpc_task_setup task_setup_data = {
5886                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5887                 .rpc_message = &msg,
5888                 .callback_ops = &nfs4_locku_ops,
5889                 .workqueue = nfsiod_workqueue,
5890                 .flags = RPC_TASK_ASYNC,
5891         };
5892
5893         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5894                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5895
5896         /* Ensure this is an unlock - when canceling a lock, the
5897          * canceled lock is passed in, and it won't be an unlock.
5898          */
5899         fl->fl_type = F_UNLCK;
5900
5901         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5902         if (data == NULL) {
5903                 nfs_free_seqid(seqid);
5904                 return ERR_PTR(-ENOMEM);
5905         }
5906
5907         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5908         msg.rpc_argp = &data->arg;
5909         msg.rpc_resp = &data->res;
5910         task_setup_data.callback_data = data;
5911         return rpc_run_task(&task_setup_data);
5912 }
5913
5914 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5915 {
5916         struct inode *inode = state->inode;
5917         struct nfs4_state_owner *sp = state->owner;
5918         struct nfs_inode *nfsi = NFS_I(inode);
5919         struct nfs_seqid *seqid;
5920         struct nfs4_lock_state *lsp;
5921         struct rpc_task *task;
5922         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5923         int status = 0;
5924         unsigned char fl_flags = request->fl_flags;
5925
5926         status = nfs4_set_lock_state(state, request);
5927         /* Unlock _before_ we do the RPC call */
5928         request->fl_flags |= FL_EXISTS;
5929         /* Exclude nfs_delegation_claim_locks() */
5930         mutex_lock(&sp->so_delegreturn_mutex);
5931         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5932         down_read(&nfsi->rwsem);
5933         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
5934                 up_read(&nfsi->rwsem);
5935                 mutex_unlock(&sp->so_delegreturn_mutex);
5936                 goto out;
5937         }
5938         up_read(&nfsi->rwsem);
5939         mutex_unlock(&sp->so_delegreturn_mutex);
5940         if (status != 0)
5941                 goto out;
5942         /* Is this a delegated lock? */
5943         lsp = request->fl_u.nfs4_fl.owner;
5944         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5945                 goto out;
5946         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5947         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5948         status = -ENOMEM;
5949         if (IS_ERR(seqid))
5950                 goto out;
5951         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5952         status = PTR_ERR(task);
5953         if (IS_ERR(task))
5954                 goto out;
5955         status = nfs4_wait_for_completion_rpc_task(task);
5956         rpc_put_task(task);
5957 out:
5958         request->fl_flags = fl_flags;
5959         trace_nfs4_unlock(request, state, F_SETLK, status);
5960         return status;
5961 }
5962
5963 struct nfs4_lockdata {
5964         struct nfs_lock_args arg;
5965         struct nfs_lock_res res;
5966         struct nfs4_lock_state *lsp;
5967         struct nfs_open_context *ctx;
5968         struct file_lock fl;
5969         unsigned long timestamp;
5970         int rpc_status;
5971         int cancelled;
5972         struct nfs_server *server;
5973 };
5974
5975 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5976                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5977                 gfp_t gfp_mask)
5978 {
5979         struct nfs4_lockdata *p;
5980         struct inode *inode = lsp->ls_state->inode;
5981         struct nfs_server *server = NFS_SERVER(inode);
5982         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5983
5984         p = kzalloc(sizeof(*p), gfp_mask);
5985         if (p == NULL)
5986                 return NULL;
5987
5988         p->arg.fh = NFS_FH(inode);
5989         p->arg.fl = &p->fl;
5990         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5991         if (IS_ERR(p->arg.open_seqid))
5992                 goto out_free;
5993         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5994         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5995         if (IS_ERR(p->arg.lock_seqid))
5996                 goto out_free_seqid;
5997         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5998         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5999         p->arg.lock_owner.s_dev = server->s_dev;
6000         p->res.lock_seqid = p->arg.lock_seqid;
6001         p->lsp = lsp;
6002         p->server = server;
6003         atomic_inc(&lsp->ls_count);
6004         p->ctx = get_nfs_open_context(ctx);
6005         get_file(fl->fl_file);
6006         memcpy(&p->fl, fl, sizeof(p->fl));
6007         return p;
6008 out_free_seqid:
6009         nfs_free_seqid(p->arg.open_seqid);
6010 out_free:
6011         kfree(p);
6012         return NULL;
6013 }
6014
6015 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6016 {
6017         struct nfs4_lockdata *data = calldata;
6018         struct nfs4_state *state = data->lsp->ls_state;
6019
6020         dprintk("%s: begin!\n", __func__);
6021         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6022                 goto out_wait;
6023         /* Do we need to do an open_to_lock_owner? */
6024         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6025                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6026                         goto out_release_lock_seqid;
6027                 }
6028                 nfs4_stateid_copy(&data->arg.open_stateid,
6029                                 &state->open_stateid);
6030                 data->arg.new_lock_owner = 1;
6031                 data->res.open_seqid = data->arg.open_seqid;
6032         } else {
6033                 data->arg.new_lock_owner = 0;
6034                 nfs4_stateid_copy(&data->arg.lock_stateid,
6035                                 &data->lsp->ls_stateid);
6036         }
6037         if (!nfs4_valid_open_stateid(state)) {
6038                 data->rpc_status = -EBADF;
6039                 task->tk_action = NULL;
6040                 goto out_release_open_seqid;
6041         }
6042         data->timestamp = jiffies;
6043         if (nfs4_setup_sequence(data->server,
6044                                 &data->arg.seq_args,
6045                                 &data->res.seq_res,
6046                                 task) == 0)
6047                 return;
6048 out_release_open_seqid:
6049         nfs_release_seqid(data->arg.open_seqid);
6050 out_release_lock_seqid:
6051         nfs_release_seqid(data->arg.lock_seqid);
6052 out_wait:
6053         nfs4_sequence_done(task, &data->res.seq_res);
6054         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6055 }
6056
6057 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6058 {
6059         struct nfs4_lockdata *data = calldata;
6060         struct nfs4_lock_state *lsp = data->lsp;
6061
6062         dprintk("%s: begin!\n", __func__);
6063
6064         if (!nfs4_sequence_done(task, &data->res.seq_res))
6065                 return;
6066
6067         data->rpc_status = task->tk_status;
6068         switch (task->tk_status) {
6069         case 0:
6070                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6071                                 data->timestamp);
6072                 if (data->arg.new_lock) {
6073                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6074                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6075                                 rpc_restart_call_prepare(task);
6076                                 break;
6077                         }
6078                 }
6079                 if (data->arg.new_lock_owner != 0) {
6080                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6081                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6082                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6083                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6084                         rpc_restart_call_prepare(task);
6085                 break;
6086         case -NFS4ERR_BAD_STATEID:
6087         case -NFS4ERR_OLD_STATEID:
6088         case -NFS4ERR_STALE_STATEID:
6089         case -NFS4ERR_EXPIRED:
6090                 if (data->arg.new_lock_owner != 0) {
6091                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6092                                                 &lsp->ls_state->open_stateid))
6093                                 rpc_restart_call_prepare(task);
6094                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6095                                                 &lsp->ls_stateid))
6096                                 rpc_restart_call_prepare(task);
6097         }
6098         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6099 }
6100
6101 static void nfs4_lock_release(void *calldata)
6102 {
6103         struct nfs4_lockdata *data = calldata;
6104
6105         dprintk("%s: begin!\n", __func__);
6106         nfs_free_seqid(data->arg.open_seqid);
6107         if (data->cancelled != 0) {
6108                 struct rpc_task *task;
6109                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6110                                 data->arg.lock_seqid);
6111                 if (!IS_ERR(task))
6112                         rpc_put_task_async(task);
6113                 dprintk("%s: cancelling lock!\n", __func__);
6114         } else
6115                 nfs_free_seqid(data->arg.lock_seqid);
6116         nfs4_put_lock_state(data->lsp);
6117         put_nfs_open_context(data->ctx);
6118         fput(data->fl.fl_file);
6119         kfree(data);
6120         dprintk("%s: done!\n", __func__);
6121 }
6122
6123 static const struct rpc_call_ops nfs4_lock_ops = {
6124         .rpc_call_prepare = nfs4_lock_prepare,
6125         .rpc_call_done = nfs4_lock_done,
6126         .rpc_release = nfs4_lock_release,
6127 };
6128
6129 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6130 {
6131         switch (error) {
6132         case -NFS4ERR_ADMIN_REVOKED:
6133         case -NFS4ERR_EXPIRED:
6134         case -NFS4ERR_BAD_STATEID:
6135                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6136                 if (new_lock_owner != 0 ||
6137                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6138                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6139                 break;
6140         case -NFS4ERR_STALE_STATEID:
6141                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6142                 nfs4_schedule_lease_recovery(server->nfs_client);
6143         };
6144 }
6145
6146 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6147 {
6148         struct nfs4_lockdata *data;
6149         struct rpc_task *task;
6150         struct rpc_message msg = {
6151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6152                 .rpc_cred = state->owner->so_cred,
6153         };
6154         struct rpc_task_setup task_setup_data = {
6155                 .rpc_client = NFS_CLIENT(state->inode),
6156                 .rpc_message = &msg,
6157                 .callback_ops = &nfs4_lock_ops,
6158                 .workqueue = nfsiod_workqueue,
6159                 .flags = RPC_TASK_ASYNC,
6160         };
6161         int ret;
6162
6163         dprintk("%s: begin!\n", __func__);
6164         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6165                         fl->fl_u.nfs4_fl.owner,
6166                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6167         if (data == NULL)
6168                 return -ENOMEM;
6169         if (IS_SETLKW(cmd))
6170                 data->arg.block = 1;
6171         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6172         msg.rpc_argp = &data->arg;
6173         msg.rpc_resp = &data->res;
6174         task_setup_data.callback_data = data;
6175         if (recovery_type > NFS_LOCK_NEW) {
6176                 if (recovery_type == NFS_LOCK_RECLAIM)
6177                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6178                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6179         } else
6180                 data->arg.new_lock = 1;
6181         task = rpc_run_task(&task_setup_data);
6182         if (IS_ERR(task))
6183                 return PTR_ERR(task);
6184         ret = nfs4_wait_for_completion_rpc_task(task);
6185         if (ret == 0) {
6186                 ret = data->rpc_status;
6187                 if (ret)
6188                         nfs4_handle_setlk_error(data->server, data->lsp,
6189                                         data->arg.new_lock_owner, ret);
6190         } else
6191                 data->cancelled = 1;
6192         rpc_put_task(task);
6193         dprintk("%s: done, ret = %d!\n", __func__, ret);
6194         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6195         return ret;
6196 }
6197
6198 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6199 {
6200         struct nfs_server *server = NFS_SERVER(state->inode);
6201         struct nfs4_exception exception = {
6202                 .inode = state->inode,
6203         };
6204         int err;
6205
6206         do {
6207                 /* Cache the lock if possible... */
6208                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6209                         return 0;
6210                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6211                 if (err != -NFS4ERR_DELAY)
6212                         break;
6213                 nfs4_handle_exception(server, err, &exception);
6214         } while (exception.retry);
6215         return err;
6216 }
6217
6218 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6219 {
6220         struct nfs_server *server = NFS_SERVER(state->inode);
6221         struct nfs4_exception exception = {
6222                 .inode = state->inode,
6223         };
6224         int err;
6225
6226         err = nfs4_set_lock_state(state, request);
6227         if (err != 0)
6228                 return err;
6229         if (!recover_lost_locks) {
6230                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6231                 return 0;
6232         }
6233         do {
6234                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6235                         return 0;
6236                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6237                 switch (err) {
6238                 default:
6239                         goto out;
6240                 case -NFS4ERR_GRACE:
6241                 case -NFS4ERR_DELAY:
6242                         nfs4_handle_exception(server, err, &exception);
6243                         err = 0;
6244                 }
6245         } while (exception.retry);
6246 out:
6247         return err;
6248 }
6249
6250 #if defined(CONFIG_NFS_V4_1)
6251 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6252 {
6253         struct nfs4_lock_state *lsp;
6254         int status;
6255
6256         status = nfs4_set_lock_state(state, request);
6257         if (status != 0)
6258                 return status;
6259         lsp = request->fl_u.nfs4_fl.owner;
6260         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6261             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6262                 return 0;
6263         status = nfs4_lock_expired(state, request);
6264         return status;
6265 }
6266 #endif
6267
6268 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6269 {
6270         struct nfs_inode *nfsi = NFS_I(state->inode);
6271         struct nfs4_state_owner *sp = state->owner;
6272         unsigned char fl_flags = request->fl_flags;
6273         int status;
6274
6275         request->fl_flags |= FL_ACCESS;
6276         status = locks_lock_inode_wait(state->inode, request);
6277         if (status < 0)
6278                 goto out;
6279         mutex_lock(&sp->so_delegreturn_mutex);
6280         down_read(&nfsi->rwsem);
6281         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6282                 /* Yes: cache locks! */
6283                 /* ...but avoid races with delegation recall... */
6284                 request->fl_flags = fl_flags & ~FL_SLEEP;
6285                 status = locks_lock_inode_wait(state->inode, request);
6286                 up_read(&nfsi->rwsem);
6287                 mutex_unlock(&sp->so_delegreturn_mutex);
6288                 goto out;
6289         }
6290         up_read(&nfsi->rwsem);
6291         mutex_unlock(&sp->so_delegreturn_mutex);
6292         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6293 out:
6294         request->fl_flags = fl_flags;
6295         return status;
6296 }
6297
6298 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6299 {
6300         struct nfs4_exception exception = {
6301                 .state = state,
6302                 .inode = state->inode,
6303         };
6304         int err;
6305
6306         do {
6307                 err = _nfs4_proc_setlk(state, cmd, request);
6308                 if (err == -NFS4ERR_DENIED)
6309                         err = -EAGAIN;
6310                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6311                                 err, &exception);
6312         } while (exception.retry);
6313         return err;
6314 }
6315
6316 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6317 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6318
6319 static int
6320 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6321                         struct file_lock *request)
6322 {
6323         int             status = -ERESTARTSYS;
6324         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6325
6326         while(!signalled()) {
6327                 status = nfs4_proc_setlk(state, cmd, request);
6328                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6329                         break;
6330                 freezable_schedule_timeout_interruptible(timeout);
6331                 timeout *= 2;
6332                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6333                 status = -ERESTARTSYS;
6334         }
6335         return status;
6336 }
6337
6338 #ifdef CONFIG_NFS_V4_1
6339 struct nfs4_lock_waiter {
6340         struct task_struct      *task;
6341         struct inode            *inode;
6342         struct nfs_lowner       *owner;
6343         bool                    notified;
6344 };
6345
6346 static int
6347 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6348 {
6349         int ret;
6350         struct cb_notify_lock_args *cbnl = key;
6351         struct nfs4_lock_waiter *waiter = wait->private;
6352         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6353                                 *wowner = waiter->owner;
6354
6355         /* Only wake if the callback was for the same owner */
6356         if (lowner->clientid != wowner->clientid ||
6357             lowner->id != wowner->id             ||
6358             lowner->s_dev != wowner->s_dev)
6359                 return 0;
6360
6361         /* Make sure it's for the right inode */
6362         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6363                 return 0;
6364
6365         waiter->notified = true;
6366
6367         /* override "private" so we can use default_wake_function */
6368         wait->private = waiter->task;
6369         ret = autoremove_wake_function(wait, mode, flags, key);
6370         wait->private = waiter;
6371         return ret;
6372 }
6373
6374 static int
6375 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6376 {
6377         int status = -ERESTARTSYS;
6378         unsigned long flags;
6379         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6380         struct nfs_server *server = NFS_SERVER(state->inode);
6381         struct nfs_client *clp = server->nfs_client;
6382         wait_queue_head_t *q = &clp->cl_lock_waitq;
6383         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6384                                     .id = lsp->ls_seqid.owner_id,
6385                                     .s_dev = server->s_dev };
6386         struct nfs4_lock_waiter waiter = { .task  = current,
6387                                            .inode = state->inode,
6388                                            .owner = &owner,
6389                                            .notified = false };
6390         wait_queue_t wait;
6391
6392         /* Don't bother with waitqueue if we don't expect a callback */
6393         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6394                 return nfs4_retry_setlk_simple(state, cmd, request);
6395
6396         init_wait(&wait);
6397         wait.private = &waiter;
6398         wait.func = nfs4_wake_lock_waiter;
6399         add_wait_queue(q, &wait);
6400
6401         while(!signalled()) {
6402                 status = nfs4_proc_setlk(state, cmd, request);
6403                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6404                         break;
6405
6406                 status = -ERESTARTSYS;
6407                 spin_lock_irqsave(&q->lock, flags);
6408                 if (waiter.notified) {
6409                         spin_unlock_irqrestore(&q->lock, flags);
6410                         continue;
6411                 }
6412                 set_current_state(TASK_INTERRUPTIBLE);
6413                 spin_unlock_irqrestore(&q->lock, flags);
6414
6415                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6416         }
6417
6418         finish_wait(q, &wait);
6419         return status;
6420 }
6421 #else /* !CONFIG_NFS_V4_1 */
6422 static inline int
6423 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6424 {
6425         return nfs4_retry_setlk_simple(state, cmd, request);
6426 }
6427 #endif
6428
6429 static int
6430 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6431 {
6432         struct nfs_open_context *ctx;
6433         struct nfs4_state *state;
6434         int status;
6435
6436         /* verify open state */
6437         ctx = nfs_file_open_context(filp);
6438         state = ctx->state;
6439
6440         if (request->fl_start < 0 || request->fl_end < 0)
6441                 return -EINVAL;
6442
6443         if (IS_GETLK(cmd)) {
6444                 if (state != NULL)
6445                         return nfs4_proc_getlk(state, F_GETLK, request);
6446                 return 0;
6447         }
6448
6449         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6450                 return -EINVAL;
6451
6452         if (request->fl_type == F_UNLCK) {
6453                 if (state != NULL)
6454                         return nfs4_proc_unlck(state, cmd, request);
6455                 return 0;
6456         }
6457
6458         if (state == NULL)
6459                 return -ENOLCK;
6460
6461         if ((request->fl_flags & FL_POSIX) &&
6462             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6463                 return -ENOLCK;
6464
6465         /*
6466          * Don't rely on the VFS having checked the file open mode,
6467          * since it won't do this for flock() locks.
6468          */
6469         switch (request->fl_type) {
6470         case F_RDLCK:
6471                 if (!(filp->f_mode & FMODE_READ))
6472                         return -EBADF;
6473                 break;
6474         case F_WRLCK:
6475                 if (!(filp->f_mode & FMODE_WRITE))
6476                         return -EBADF;
6477         }
6478
6479         status = nfs4_set_lock_state(state, request);
6480         if (status != 0)
6481                 return status;
6482
6483         return nfs4_retry_setlk(state, cmd, request);
6484 }
6485
6486 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6487 {
6488         struct nfs_server *server = NFS_SERVER(state->inode);
6489         int err;
6490
6491         err = nfs4_set_lock_state(state, fl);
6492         if (err != 0)
6493                 return err;
6494         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6495         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6496 }
6497
6498 struct nfs_release_lockowner_data {
6499         struct nfs4_lock_state *lsp;
6500         struct nfs_server *server;
6501         struct nfs_release_lockowner_args args;
6502         struct nfs_release_lockowner_res res;
6503         unsigned long timestamp;
6504 };
6505
6506 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6507 {
6508         struct nfs_release_lockowner_data *data = calldata;
6509         struct nfs_server *server = data->server;
6510         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6511                              &data->args.seq_args, &data->res.seq_res, task);
6512         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6513         data->timestamp = jiffies;
6514 }
6515
6516 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6517 {
6518         struct nfs_release_lockowner_data *data = calldata;
6519         struct nfs_server *server = data->server;
6520
6521         nfs40_sequence_done(task, &data->res.seq_res);
6522
6523         switch (task->tk_status) {
6524         case 0:
6525                 renew_lease(server, data->timestamp);
6526                 break;
6527         case -NFS4ERR_STALE_CLIENTID:
6528         case -NFS4ERR_EXPIRED:
6529                 nfs4_schedule_lease_recovery(server->nfs_client);
6530                 break;
6531         case -NFS4ERR_LEASE_MOVED:
6532         case -NFS4ERR_DELAY:
6533                 if (nfs4_async_handle_error(task, server,
6534                                             NULL, NULL) == -EAGAIN)
6535                         rpc_restart_call_prepare(task);
6536         }
6537 }
6538
6539 static void nfs4_release_lockowner_release(void *calldata)
6540 {
6541         struct nfs_release_lockowner_data *data = calldata;
6542         nfs4_free_lock_state(data->server, data->lsp);
6543         kfree(calldata);
6544 }
6545
6546 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6547         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6548         .rpc_call_done = nfs4_release_lockowner_done,
6549         .rpc_release = nfs4_release_lockowner_release,
6550 };
6551
6552 static void
6553 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6554 {
6555         struct nfs_release_lockowner_data *data;
6556         struct rpc_message msg = {
6557                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6558         };
6559
6560         if (server->nfs_client->cl_mvops->minor_version != 0)
6561                 return;
6562
6563         data = kmalloc(sizeof(*data), GFP_NOFS);
6564         if (!data)
6565                 return;
6566         data->lsp = lsp;
6567         data->server = server;
6568         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6569         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6570         data->args.lock_owner.s_dev = server->s_dev;
6571
6572         msg.rpc_argp = &data->args;
6573         msg.rpc_resp = &data->res;
6574         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6575         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6576 }
6577
6578 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6579
6580 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6581                                    struct dentry *unused, struct inode *inode,
6582                                    const char *key, const void *buf,
6583                                    size_t buflen, int flags)
6584 {
6585         return nfs4_proc_set_acl(inode, buf, buflen);
6586 }
6587
6588 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6589                                    struct dentry *unused, struct inode *inode,
6590                                    const char *key, void *buf, size_t buflen)
6591 {
6592         return nfs4_proc_get_acl(inode, buf, buflen);
6593 }
6594
6595 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6596 {
6597         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6598 }
6599
6600 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6601
6602 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6603                                      struct dentry *unused, struct inode *inode,
6604                                      const char *key, const void *buf,
6605                                      size_t buflen, int flags)
6606 {
6607         if (security_ismaclabel(key))
6608                 return nfs4_set_security_label(inode, buf, buflen);
6609
6610         return -EOPNOTSUPP;
6611 }
6612
6613 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6614                                      struct dentry *unused, struct inode *inode,
6615                                      const char *key, void *buf, size_t buflen)
6616 {
6617         if (security_ismaclabel(key))
6618                 return nfs4_get_security_label(inode, buf, buflen);
6619         return -EOPNOTSUPP;
6620 }
6621
6622 static ssize_t
6623 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6624 {
6625         int len = 0;
6626
6627         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6628                 len = security_inode_listsecurity(inode, list, list_len);
6629                 if (list_len && len > list_len)
6630                         return -ERANGE;
6631         }
6632         return len;
6633 }
6634
6635 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6636         .prefix = XATTR_SECURITY_PREFIX,
6637         .get    = nfs4_xattr_get_nfs4_label,
6638         .set    = nfs4_xattr_set_nfs4_label,
6639 };
6640
6641 #else
6642
6643 static ssize_t
6644 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6645 {
6646         return 0;
6647 }
6648
6649 #endif
6650
6651 /*
6652  * nfs_fhget will use either the mounted_on_fileid or the fileid
6653  */
6654 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6655 {
6656         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6657                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6658               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6659               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6660                 return;
6661
6662         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6663                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6664         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6665         fattr->nlink = 2;
6666 }
6667
6668 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6669                                    const struct qstr *name,
6670                                    struct nfs4_fs_locations *fs_locations,
6671                                    struct page *page)
6672 {
6673         struct nfs_server *server = NFS_SERVER(dir);
6674         u32 bitmask[3] = {
6675                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6676         };
6677         struct nfs4_fs_locations_arg args = {
6678                 .dir_fh = NFS_FH(dir),
6679                 .name = name,
6680                 .page = page,
6681                 .bitmask = bitmask,
6682         };
6683         struct nfs4_fs_locations_res res = {
6684                 .fs_locations = fs_locations,
6685         };
6686         struct rpc_message msg = {
6687                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6688                 .rpc_argp = &args,
6689                 .rpc_resp = &res,
6690         };
6691         int status;
6692
6693         dprintk("%s: start\n", __func__);
6694
6695         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6696          * is not supported */
6697         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6698                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6699         else
6700                 bitmask[0] |= FATTR4_WORD0_FILEID;
6701
6702         nfs_fattr_init(&fs_locations->fattr);
6703         fs_locations->server = server;
6704         fs_locations->nlocations = 0;
6705         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6706         dprintk("%s: returned status = %d\n", __func__, status);
6707         return status;
6708 }
6709
6710 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6711                            const struct qstr *name,
6712                            struct nfs4_fs_locations *fs_locations,
6713                            struct page *page)
6714 {
6715         struct nfs4_exception exception = { };
6716         int err;
6717         do {
6718                 err = _nfs4_proc_fs_locations(client, dir, name,
6719                                 fs_locations, page);
6720                 trace_nfs4_get_fs_locations(dir, name, err);
6721                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6722                                 &exception);
6723         } while (exception.retry);
6724         return err;
6725 }
6726
6727 /*
6728  * This operation also signals the server that this client is
6729  * performing migration recovery.  The server can stop returning
6730  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6731  * appended to this compound to identify the client ID which is
6732  * performing recovery.
6733  */
6734 static int _nfs40_proc_get_locations(struct inode *inode,
6735                                      struct nfs4_fs_locations *locations,
6736                                      struct page *page, struct rpc_cred *cred)
6737 {
6738         struct nfs_server *server = NFS_SERVER(inode);
6739         struct rpc_clnt *clnt = server->client;
6740         u32 bitmask[2] = {
6741                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6742         };
6743         struct nfs4_fs_locations_arg args = {
6744                 .clientid       = server->nfs_client->cl_clientid,
6745                 .fh             = NFS_FH(inode),
6746                 .page           = page,
6747                 .bitmask        = bitmask,
6748                 .migration      = 1,            /* skip LOOKUP */
6749                 .renew          = 1,            /* append RENEW */
6750         };
6751         struct nfs4_fs_locations_res res = {
6752                 .fs_locations   = locations,
6753                 .migration      = 1,
6754                 .renew          = 1,
6755         };
6756         struct rpc_message msg = {
6757                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6758                 .rpc_argp       = &args,
6759                 .rpc_resp       = &res,
6760                 .rpc_cred       = cred,
6761         };
6762         unsigned long now = jiffies;
6763         int status;
6764
6765         nfs_fattr_init(&locations->fattr);
6766         locations->server = server;
6767         locations->nlocations = 0;
6768
6769         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6770         nfs4_set_sequence_privileged(&args.seq_args);
6771         status = nfs4_call_sync_sequence(clnt, server, &msg,
6772                                         &args.seq_args, &res.seq_res);
6773         if (status)
6774                 return status;
6775
6776         renew_lease(server, now);
6777         return 0;
6778 }
6779
6780 #ifdef CONFIG_NFS_V4_1
6781
6782 /*
6783  * This operation also signals the server that this client is
6784  * performing migration recovery.  The server can stop asserting
6785  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6786  * performing this operation is identified in the SEQUENCE
6787  * operation in this compound.
6788  *
6789  * When the client supports GETATTR(fs_locations_info), it can
6790  * be plumbed in here.
6791  */
6792 static int _nfs41_proc_get_locations(struct inode *inode,
6793                                      struct nfs4_fs_locations *locations,
6794                                      struct page *page, struct rpc_cred *cred)
6795 {
6796         struct nfs_server *server = NFS_SERVER(inode);
6797         struct rpc_clnt *clnt = server->client;
6798         u32 bitmask[2] = {
6799                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6800         };
6801         struct nfs4_fs_locations_arg args = {
6802                 .fh             = NFS_FH(inode),
6803                 .page           = page,
6804                 .bitmask        = bitmask,
6805                 .migration      = 1,            /* skip LOOKUP */
6806         };
6807         struct nfs4_fs_locations_res res = {
6808                 .fs_locations   = locations,
6809                 .migration      = 1,
6810         };
6811         struct rpc_message msg = {
6812                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6813                 .rpc_argp       = &args,
6814                 .rpc_resp       = &res,
6815                 .rpc_cred       = cred,
6816         };
6817         int status;
6818
6819         nfs_fattr_init(&locations->fattr);
6820         locations->server = server;
6821         locations->nlocations = 0;
6822
6823         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6824         nfs4_set_sequence_privileged(&args.seq_args);
6825         status = nfs4_call_sync_sequence(clnt, server, &msg,
6826                                         &args.seq_args, &res.seq_res);
6827         if (status == NFS4_OK &&
6828             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6829                 status = -NFS4ERR_LEASE_MOVED;
6830         return status;
6831 }
6832
6833 #endif  /* CONFIG_NFS_V4_1 */
6834
6835 /**
6836  * nfs4_proc_get_locations - discover locations for a migrated FSID
6837  * @inode: inode on FSID that is migrating
6838  * @locations: result of query
6839  * @page: buffer
6840  * @cred: credential to use for this operation
6841  *
6842  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6843  * operation failed, or a negative errno if a local error occurred.
6844  *
6845  * On success, "locations" is filled in, but if the server has
6846  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6847  * asserted.
6848  *
6849  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6850  * from this client that require migration recovery.
6851  */
6852 int nfs4_proc_get_locations(struct inode *inode,
6853                             struct nfs4_fs_locations *locations,
6854                             struct page *page, struct rpc_cred *cred)
6855 {
6856         struct nfs_server *server = NFS_SERVER(inode);
6857         struct nfs_client *clp = server->nfs_client;
6858         const struct nfs4_mig_recovery_ops *ops =
6859                                         clp->cl_mvops->mig_recovery_ops;
6860         struct nfs4_exception exception = { };
6861         int status;
6862
6863         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6864                 (unsigned long long)server->fsid.major,
6865                 (unsigned long long)server->fsid.minor,
6866                 clp->cl_hostname);
6867         nfs_display_fhandle(NFS_FH(inode), __func__);
6868
6869         do {
6870                 status = ops->get_locations(inode, locations, page, cred);
6871                 if (status != -NFS4ERR_DELAY)
6872                         break;
6873                 nfs4_handle_exception(server, status, &exception);
6874         } while (exception.retry);
6875         return status;
6876 }
6877
6878 /*
6879  * This operation also signals the server that this client is
6880  * performing "lease moved" recovery.  The server can stop
6881  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6882  * is appended to this compound to identify the client ID which is
6883  * performing recovery.
6884  */
6885 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6886 {
6887         struct nfs_server *server = NFS_SERVER(inode);
6888         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6889         struct rpc_clnt *clnt = server->client;
6890         struct nfs4_fsid_present_arg args = {
6891                 .fh             = NFS_FH(inode),
6892                 .clientid       = clp->cl_clientid,
6893                 .renew          = 1,            /* append RENEW */
6894         };
6895         struct nfs4_fsid_present_res res = {
6896                 .renew          = 1,
6897         };
6898         struct rpc_message msg = {
6899                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6900                 .rpc_argp       = &args,
6901                 .rpc_resp       = &res,
6902                 .rpc_cred       = cred,
6903         };
6904         unsigned long now = jiffies;
6905         int status;
6906
6907         res.fh = nfs_alloc_fhandle();
6908         if (res.fh == NULL)
6909                 return -ENOMEM;
6910
6911         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6912         nfs4_set_sequence_privileged(&args.seq_args);
6913         status = nfs4_call_sync_sequence(clnt, server, &msg,
6914                                                 &args.seq_args, &res.seq_res);
6915         nfs_free_fhandle(res.fh);
6916         if (status)
6917                 return status;
6918
6919         do_renew_lease(clp, now);
6920         return 0;
6921 }
6922
6923 #ifdef CONFIG_NFS_V4_1
6924
6925 /*
6926  * This operation also signals the server that this client is
6927  * performing "lease moved" recovery.  The server can stop asserting
6928  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6929  * this operation is identified in the SEQUENCE operation in this
6930  * compound.
6931  */
6932 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6933 {
6934         struct nfs_server *server = NFS_SERVER(inode);
6935         struct rpc_clnt *clnt = server->client;
6936         struct nfs4_fsid_present_arg args = {
6937                 .fh             = NFS_FH(inode),
6938         };
6939         struct nfs4_fsid_present_res res = {
6940         };
6941         struct rpc_message msg = {
6942                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6943                 .rpc_argp       = &args,
6944                 .rpc_resp       = &res,
6945                 .rpc_cred       = cred,
6946         };
6947         int status;
6948
6949         res.fh = nfs_alloc_fhandle();
6950         if (res.fh == NULL)
6951                 return -ENOMEM;
6952
6953         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6954         nfs4_set_sequence_privileged(&args.seq_args);
6955         status = nfs4_call_sync_sequence(clnt, server, &msg,
6956                                                 &args.seq_args, &res.seq_res);
6957         nfs_free_fhandle(res.fh);
6958         if (status == NFS4_OK &&
6959             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6960                 status = -NFS4ERR_LEASE_MOVED;
6961         return status;
6962 }
6963
6964 #endif  /* CONFIG_NFS_V4_1 */
6965
6966 /**
6967  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6968  * @inode: inode on FSID to check
6969  * @cred: credential to use for this operation
6970  *
6971  * Server indicates whether the FSID is present, moved, or not
6972  * recognized.  This operation is necessary to clear a LEASE_MOVED
6973  * condition for this client ID.
6974  *
6975  * Returns NFS4_OK if the FSID is present on this server,
6976  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6977  *  NFS4ERR code if some error occurred on the server, or a
6978  *  negative errno if a local failure occurred.
6979  */
6980 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6981 {
6982         struct nfs_server *server = NFS_SERVER(inode);
6983         struct nfs_client *clp = server->nfs_client;
6984         const struct nfs4_mig_recovery_ops *ops =
6985                                         clp->cl_mvops->mig_recovery_ops;
6986         struct nfs4_exception exception = { };
6987         int status;
6988
6989         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6990                 (unsigned long long)server->fsid.major,
6991                 (unsigned long long)server->fsid.minor,
6992                 clp->cl_hostname);
6993         nfs_display_fhandle(NFS_FH(inode), __func__);
6994
6995         do {
6996                 status = ops->fsid_present(inode, cred);
6997                 if (status != -NFS4ERR_DELAY)
6998                         break;
6999                 nfs4_handle_exception(server, status, &exception);
7000         } while (exception.retry);
7001         return status;
7002 }
7003
7004 /**
7005  * If 'use_integrity' is true and the state managment nfs_client
7006  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7007  * and the machine credential as per RFC3530bis and RFC5661 Security
7008  * Considerations sections. Otherwise, just use the user cred with the
7009  * filesystem's rpc_client.
7010  */
7011 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7012 {
7013         int status;
7014         struct nfs4_secinfo_arg args = {
7015                 .dir_fh = NFS_FH(dir),
7016                 .name   = name,
7017         };
7018         struct nfs4_secinfo_res res = {
7019                 .flavors     = flavors,
7020         };
7021         struct rpc_message msg = {
7022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7023                 .rpc_argp = &args,
7024                 .rpc_resp = &res,
7025         };
7026         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7027         struct rpc_cred *cred = NULL;
7028
7029         if (use_integrity) {
7030                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7031                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7032                 msg.rpc_cred = cred;
7033         }
7034
7035         dprintk("NFS call  secinfo %s\n", name->name);
7036
7037         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7038                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7039
7040         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7041                                 &res.seq_res, 0);
7042         dprintk("NFS reply  secinfo: %d\n", status);
7043
7044         if (cred)
7045                 put_rpccred(cred);
7046
7047         return status;
7048 }
7049
7050 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7051                       struct nfs4_secinfo_flavors *flavors)
7052 {
7053         struct nfs4_exception exception = { };
7054         int err;
7055         do {
7056                 err = -NFS4ERR_WRONGSEC;
7057
7058                 /* try to use integrity protection with machine cred */
7059                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7060                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7061
7062                 /*
7063                  * if unable to use integrity protection, or SECINFO with
7064                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7065                  * disallowed by spec, but exists in deployed servers) use
7066                  * the current filesystem's rpc_client and the user cred.
7067                  */
7068                 if (err == -NFS4ERR_WRONGSEC)
7069                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7070
7071                 trace_nfs4_secinfo(dir, name, err);
7072                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7073                                 &exception);
7074         } while (exception.retry);
7075         return err;
7076 }
7077
7078 #ifdef CONFIG_NFS_V4_1
7079 /*
7080  * Check the exchange flags returned by the server for invalid flags, having
7081  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7082  * DS flags set.
7083  */
7084 static int nfs4_check_cl_exchange_flags(u32 flags)
7085 {
7086         if (flags & ~EXCHGID4_FLAG_MASK_R)
7087                 goto out_inval;
7088         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7089             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7090                 goto out_inval;
7091         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7092                 goto out_inval;
7093         return NFS_OK;
7094 out_inval:
7095         return -NFS4ERR_INVAL;
7096 }
7097
7098 static bool
7099 nfs41_same_server_scope(struct nfs41_server_scope *a,
7100                         struct nfs41_server_scope *b)
7101 {
7102         if (a->server_scope_sz == b->server_scope_sz &&
7103             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7104                 return true;
7105
7106         return false;
7107 }
7108
7109 static void
7110 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7111 {
7112 }
7113
7114 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7115         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7116 };
7117
7118 /*
7119  * nfs4_proc_bind_one_conn_to_session()
7120  *
7121  * The 4.1 client currently uses the same TCP connection for the
7122  * fore and backchannel.
7123  */
7124 static
7125 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7126                 struct rpc_xprt *xprt,
7127                 struct nfs_client *clp,
7128                 struct rpc_cred *cred)
7129 {
7130         int status;
7131         struct nfs41_bind_conn_to_session_args args = {
7132                 .client = clp,
7133                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7134         };
7135         struct nfs41_bind_conn_to_session_res res;
7136         struct rpc_message msg = {
7137                 .rpc_proc =
7138                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7139                 .rpc_argp = &args,
7140                 .rpc_resp = &res,
7141                 .rpc_cred = cred,
7142         };
7143         struct rpc_task_setup task_setup_data = {
7144                 .rpc_client = clnt,
7145                 .rpc_xprt = xprt,
7146                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7147                 .rpc_message = &msg,
7148                 .flags = RPC_TASK_TIMEOUT,
7149         };
7150         struct rpc_task *task;
7151
7152         dprintk("--> %s\n", __func__);
7153
7154         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7155         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7156                 args.dir = NFS4_CDFC4_FORE;
7157
7158         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7159         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7160                 args.dir = NFS4_CDFC4_FORE;
7161
7162         task = rpc_run_task(&task_setup_data);
7163         if (!IS_ERR(task)) {
7164                 status = task->tk_status;
7165                 rpc_put_task(task);
7166         } else
7167                 status = PTR_ERR(task);
7168         trace_nfs4_bind_conn_to_session(clp, status);
7169         if (status == 0) {
7170                 if (memcmp(res.sessionid.data,
7171                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7172                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7173                         status = -EIO;
7174                         goto out;
7175                 }
7176                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7177                         dprintk("NFS: %s: Unexpected direction from server\n",
7178                                 __func__);
7179                         status = -EIO;
7180                         goto out;
7181                 }
7182                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7183                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7184                                 __func__);
7185                         status = -EIO;
7186                         goto out;
7187                 }
7188         }
7189 out:
7190         dprintk("<-- %s status= %d\n", __func__, status);
7191         return status;
7192 }
7193
7194 struct rpc_bind_conn_calldata {
7195         struct nfs_client *clp;
7196         struct rpc_cred *cred;
7197 };
7198
7199 static int
7200 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7201                 struct rpc_xprt *xprt,
7202                 void *calldata)
7203 {
7204         struct rpc_bind_conn_calldata *p = calldata;
7205
7206         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7207 }
7208
7209 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7210 {
7211         struct rpc_bind_conn_calldata data = {
7212                 .clp = clp,
7213                 .cred = cred,
7214         };
7215         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7216                         nfs4_proc_bind_conn_to_session_callback, &data);
7217 }
7218
7219 /*
7220  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7221  * and operations we'd like to see to enable certain features in the allow map
7222  */
7223 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7224         .how = SP4_MACH_CRED,
7225         .enforce.u.words = {
7226                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7227                       1 << (OP_EXCHANGE_ID - 32) |
7228                       1 << (OP_CREATE_SESSION - 32) |
7229                       1 << (OP_DESTROY_SESSION - 32) |
7230                       1 << (OP_DESTROY_CLIENTID - 32)
7231         },
7232         .allow.u.words = {
7233                 [0] = 1 << (OP_CLOSE) |
7234                       1 << (OP_OPEN_DOWNGRADE) |
7235                       1 << (OP_LOCKU) |
7236                       1 << (OP_DELEGRETURN) |
7237                       1 << (OP_COMMIT),
7238                 [1] = 1 << (OP_SECINFO - 32) |
7239                       1 << (OP_SECINFO_NO_NAME - 32) |
7240                       1 << (OP_LAYOUTRETURN - 32) |
7241                       1 << (OP_TEST_STATEID - 32) |
7242                       1 << (OP_FREE_STATEID - 32) |
7243                       1 << (OP_WRITE - 32)
7244         }
7245 };
7246
7247 /*
7248  * Select the state protection mode for client `clp' given the server results
7249  * from exchange_id in `sp'.
7250  *
7251  * Returns 0 on success, negative errno otherwise.
7252  */
7253 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7254                                  struct nfs41_state_protection *sp)
7255 {
7256         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7257                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7258                       1 << (OP_EXCHANGE_ID - 32) |
7259                       1 << (OP_CREATE_SESSION - 32) |
7260                       1 << (OP_DESTROY_SESSION - 32) |
7261                       1 << (OP_DESTROY_CLIENTID - 32)
7262         };
7263         unsigned int i;
7264
7265         if (sp->how == SP4_MACH_CRED) {
7266                 /* Print state protect result */
7267                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7268                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7269                         if (test_bit(i, sp->enforce.u.longs))
7270                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7271                         if (test_bit(i, sp->allow.u.longs))
7272                                 dfprintk(MOUNT, "  allow op %d\n", i);
7273                 }
7274
7275                 /* make sure nothing is on enforce list that isn't supported */
7276                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7277                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7278                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7279                                 return -EINVAL;
7280                         }
7281                 }
7282
7283                 /*
7284                  * Minimal mode - state operations are allowed to use machine
7285                  * credential.  Note this already happens by default, so the
7286                  * client doesn't have to do anything more than the negotiation.
7287                  *
7288                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7289                  *       we're already using the machine cred for exchange_id
7290                  *       and will never use a different cred.
7291                  */
7292                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7293                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7294                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7295                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7296                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7297                         dfprintk(MOUNT, "  minimal mode enabled\n");
7298                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7299                 } else {
7300                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7301                         return -EINVAL;
7302                 }
7303
7304                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7305                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7306                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7307                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7308                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7309                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7310                 }
7311
7312                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7313                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7314                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7315                                 &clp->cl_sp4_flags);
7316                 }
7317
7318                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7319                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7320                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7321                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7322                 }
7323
7324                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7325                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7326                         dfprintk(MOUNT, "  stateid mode enabled\n");
7327                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7328                 }
7329
7330                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7331                         dfprintk(MOUNT, "  write mode enabled\n");
7332                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7333                 }
7334
7335                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7336                         dfprintk(MOUNT, "  commit mode enabled\n");
7337                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7338                 }
7339         }
7340
7341         return 0;
7342 }
7343
7344 struct nfs41_exchange_id_data {
7345         struct nfs41_exchange_id_res res;
7346         struct nfs41_exchange_id_args args;
7347         struct rpc_xprt *xprt;
7348         int rpc_status;
7349 };
7350
7351 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7352 {
7353         struct nfs41_exchange_id_data *cdata =
7354                                         (struct nfs41_exchange_id_data *)data;
7355         struct nfs_client *clp = cdata->args.client;
7356         int status = task->tk_status;
7357
7358         trace_nfs4_exchange_id(clp, status);
7359
7360         if (status == 0)
7361                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7362
7363         if (cdata->xprt && status == 0) {
7364                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7365                                                       cdata->xprt);
7366                 goto out;
7367         }
7368
7369         if (status  == 0)
7370                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7371
7372         if (status == 0) {
7373                 clp->cl_clientid = cdata->res.clientid;
7374                 clp->cl_exchange_flags = cdata->res.flags;
7375                 /* Client ID is not confirmed */
7376                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7377                         clear_bit(NFS4_SESSION_ESTABLISHED,
7378                         &clp->cl_session->session_state);
7379                         clp->cl_seqid = cdata->res.seqid;
7380                 }
7381
7382                 kfree(clp->cl_serverowner);
7383                 clp->cl_serverowner = cdata->res.server_owner;
7384                 cdata->res.server_owner = NULL;
7385
7386                 /* use the most recent implementation id */
7387                 kfree(clp->cl_implid);
7388                 clp->cl_implid = cdata->res.impl_id;
7389                 cdata->res.impl_id = NULL;
7390
7391                 if (clp->cl_serverscope != NULL &&
7392                     !nfs41_same_server_scope(clp->cl_serverscope,
7393                                         cdata->res.server_scope)) {
7394                         dprintk("%s: server_scope mismatch detected\n",
7395                                 __func__);
7396                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7397                         kfree(clp->cl_serverscope);
7398                         clp->cl_serverscope = NULL;
7399                 }
7400
7401                 if (clp->cl_serverscope == NULL) {
7402                         clp->cl_serverscope = cdata->res.server_scope;
7403                         cdata->res.server_scope = NULL;
7404                 }
7405                 /* Save the EXCHANGE_ID verifier session trunk tests */
7406                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7407                        sizeof(clp->cl_confirm.data));
7408         }
7409 out:
7410         cdata->rpc_status = status;
7411         return;
7412 }
7413
7414 static void nfs4_exchange_id_release(void *data)
7415 {
7416         struct nfs41_exchange_id_data *cdata =
7417                                         (struct nfs41_exchange_id_data *)data;
7418
7419         nfs_put_client(cdata->args.client);
7420         if (cdata->xprt) {
7421                 xprt_put(cdata->xprt);
7422                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7423         }
7424         kfree(cdata->res.impl_id);
7425         kfree(cdata->res.server_scope);
7426         kfree(cdata->res.server_owner);
7427         kfree(cdata);
7428 }
7429
7430 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7431         .rpc_call_done = nfs4_exchange_id_done,
7432         .rpc_release = nfs4_exchange_id_release,
7433 };
7434
7435 /*
7436  * _nfs4_proc_exchange_id()
7437  *
7438  * Wrapper for EXCHANGE_ID operation.
7439  */
7440 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7441                         u32 sp4_how, struct rpc_xprt *xprt)
7442 {
7443         nfs4_verifier verifier;
7444         struct rpc_message msg = {
7445                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7446                 .rpc_cred = cred,
7447         };
7448         struct rpc_task_setup task_setup_data = {
7449                 .rpc_client = clp->cl_rpcclient,
7450                 .callback_ops = &nfs4_exchange_id_call_ops,
7451                 .rpc_message = &msg,
7452                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7453         };
7454         struct nfs41_exchange_id_data *calldata;
7455         struct rpc_task *task;
7456         int status = -EIO;
7457
7458         if (!atomic_inc_not_zero(&clp->cl_count))
7459                 goto out;
7460
7461         status = -ENOMEM;
7462         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7463         if (!calldata)
7464                 goto out;
7465
7466         if (!xprt)
7467                 nfs4_init_boot_verifier(clp, &verifier);
7468
7469         status = nfs4_init_uniform_client_string(clp);
7470         if (status)
7471                 goto out_calldata;
7472
7473         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7474                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7475                 clp->cl_owner_id);
7476
7477         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7478                                                 GFP_NOFS);
7479         status = -ENOMEM;
7480         if (unlikely(calldata->res.server_owner == NULL))
7481                 goto out_calldata;
7482
7483         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7484                                         GFP_NOFS);
7485         if (unlikely(calldata->res.server_scope == NULL))
7486                 goto out_server_owner;
7487
7488         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7489         if (unlikely(calldata->res.impl_id == NULL))
7490                 goto out_server_scope;
7491
7492         switch (sp4_how) {
7493         case SP4_NONE:
7494                 calldata->args.state_protect.how = SP4_NONE;
7495                 break;
7496
7497         case SP4_MACH_CRED:
7498                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7499                 break;
7500
7501         default:
7502                 /* unsupported! */
7503                 WARN_ON_ONCE(1);
7504                 status = -EINVAL;
7505                 goto out_impl_id;
7506         }
7507         if (xprt) {
7508                 calldata->xprt = xprt;
7509                 task_setup_data.rpc_xprt = xprt;
7510                 task_setup_data.flags =
7511                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7512                 calldata->args.verifier = &clp->cl_confirm;
7513         } else {
7514                 calldata->args.verifier = &verifier;
7515         }
7516         calldata->args.client = clp;
7517 #ifdef CONFIG_NFS_V4_1_MIGRATION
7518         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7519         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7520         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7521 #else
7522         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7523         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7524 #endif
7525         msg.rpc_argp = &calldata->args;
7526         msg.rpc_resp = &calldata->res;
7527         task_setup_data.callback_data = calldata;
7528
7529         task = rpc_run_task(&task_setup_data);
7530         if (IS_ERR(task)) {
7531         status = PTR_ERR(task);
7532                 goto out_impl_id;
7533         }
7534
7535         if (!xprt) {
7536                 status = rpc_wait_for_completion_task(task);
7537                 if (!status)
7538                         status = calldata->rpc_status;
7539         } else  /* session trunking test */
7540                 status = calldata->rpc_status;
7541
7542         rpc_put_task(task);
7543 out:
7544         if (clp->cl_implid != NULL)
7545                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7546                         "domain: %s, name: %s, date: %llu,%u\n",
7547                         clp->cl_implid->domain, clp->cl_implid->name,
7548                         clp->cl_implid->date.seconds,
7549                         clp->cl_implid->date.nseconds);
7550         dprintk("NFS reply exchange_id: %d\n", status);
7551         return status;
7552
7553 out_impl_id:
7554         kfree(calldata->res.impl_id);
7555 out_server_scope:
7556         kfree(calldata->res.server_scope);
7557 out_server_owner:
7558         kfree(calldata->res.server_owner);
7559 out_calldata:
7560         kfree(calldata);
7561         goto out;
7562 }
7563
7564 /*
7565  * nfs4_proc_exchange_id()
7566  *
7567  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7568  *
7569  * Since the clientid has expired, all compounds using sessions
7570  * associated with the stale clientid will be returning
7571  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7572  * be in some phase of session reset.
7573  *
7574  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7575  */
7576 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7577 {
7578         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7579         int status;
7580
7581         /* try SP4_MACH_CRED if krb5i/p */
7582         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7583             authflavor == RPC_AUTH_GSS_KRB5P) {
7584                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7585                 if (!status)
7586                         return 0;
7587         }
7588
7589         /* try SP4_NONE */
7590         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7591 }
7592
7593 /**
7594  * nfs4_test_session_trunk
7595  *
7596  * This is an add_xprt_test() test function called from
7597  * rpc_clnt_setup_test_and_add_xprt.
7598  *
7599  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7600  * and is dereferrenced in nfs4_exchange_id_release
7601  *
7602  * Upon success, add the new transport to the rpc_clnt
7603  *
7604  * @clnt: struct rpc_clnt to get new transport
7605  * @xprt: the rpc_xprt to test
7606  * @data: call data for _nfs4_proc_exchange_id.
7607  */
7608 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7609                             void *data)
7610 {
7611         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7612         u32 sp4_how;
7613
7614         dprintk("--> %s try %s\n", __func__,
7615                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7616
7617         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7618
7619         /* Test connection for session trunking. Async exchange_id call */
7620         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7621 }
7622 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7623
7624 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7625                 struct rpc_cred *cred)
7626 {
7627         struct rpc_message msg = {
7628                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7629                 .rpc_argp = clp,
7630                 .rpc_cred = cred,
7631         };
7632         int status;
7633
7634         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7635         trace_nfs4_destroy_clientid(clp, status);
7636         if (status)
7637                 dprintk("NFS: Got error %d from the server %s on "
7638                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7639         return status;
7640 }
7641
7642 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7643                 struct rpc_cred *cred)
7644 {
7645         unsigned int loop;
7646         int ret;
7647
7648         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7649                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7650                 switch (ret) {
7651                 case -NFS4ERR_DELAY:
7652                 case -NFS4ERR_CLIENTID_BUSY:
7653                         ssleep(1);
7654                         break;
7655                 default:
7656                         return ret;
7657                 }
7658         }
7659         return 0;
7660 }
7661
7662 int nfs4_destroy_clientid(struct nfs_client *clp)
7663 {
7664         struct rpc_cred *cred;
7665         int ret = 0;
7666
7667         if (clp->cl_mvops->minor_version < 1)
7668                 goto out;
7669         if (clp->cl_exchange_flags == 0)
7670                 goto out;
7671         if (clp->cl_preserve_clid)
7672                 goto out;
7673         cred = nfs4_get_clid_cred(clp);
7674         ret = nfs4_proc_destroy_clientid(clp, cred);
7675         if (cred)
7676                 put_rpccred(cred);
7677         switch (ret) {
7678         case 0:
7679         case -NFS4ERR_STALE_CLIENTID:
7680                 clp->cl_exchange_flags = 0;
7681         }
7682 out:
7683         return ret;
7684 }
7685
7686 struct nfs4_get_lease_time_data {
7687         struct nfs4_get_lease_time_args *args;
7688         struct nfs4_get_lease_time_res *res;
7689         struct nfs_client *clp;
7690 };
7691
7692 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7693                                         void *calldata)
7694 {
7695         struct nfs4_get_lease_time_data *data =
7696                         (struct nfs4_get_lease_time_data *)calldata;
7697
7698         dprintk("--> %s\n", __func__);
7699         /* just setup sequence, do not trigger session recovery
7700            since we're invoked within one */
7701         nfs41_setup_sequence(data->clp->cl_session,
7702                         &data->args->la_seq_args,
7703                         &data->res->lr_seq_res,
7704                         task);
7705         dprintk("<-- %s\n", __func__);
7706 }
7707
7708 /*
7709  * Called from nfs4_state_manager thread for session setup, so don't recover
7710  * from sequence operation or clientid errors.
7711  */
7712 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7713 {
7714         struct nfs4_get_lease_time_data *data =
7715                         (struct nfs4_get_lease_time_data *)calldata;
7716
7717         dprintk("--> %s\n", __func__);
7718         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7719                 return;
7720         switch (task->tk_status) {
7721         case -NFS4ERR_DELAY:
7722         case -NFS4ERR_GRACE:
7723                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7724                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7725                 task->tk_status = 0;
7726                 /* fall through */
7727         case -NFS4ERR_RETRY_UNCACHED_REP:
7728                 rpc_restart_call_prepare(task);
7729                 return;
7730         }
7731         dprintk("<-- %s\n", __func__);
7732 }
7733
7734 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7735         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7736         .rpc_call_done = nfs4_get_lease_time_done,
7737 };
7738
7739 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7740 {
7741         struct rpc_task *task;
7742         struct nfs4_get_lease_time_args args;
7743         struct nfs4_get_lease_time_res res = {
7744                 .lr_fsinfo = fsinfo,
7745         };
7746         struct nfs4_get_lease_time_data data = {
7747                 .args = &args,
7748                 .res = &res,
7749                 .clp = clp,
7750         };
7751         struct rpc_message msg = {
7752                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7753                 .rpc_argp = &args,
7754                 .rpc_resp = &res,
7755         };
7756         struct rpc_task_setup task_setup = {
7757                 .rpc_client = clp->cl_rpcclient,
7758                 .rpc_message = &msg,
7759                 .callback_ops = &nfs4_get_lease_time_ops,
7760                 .callback_data = &data,
7761                 .flags = RPC_TASK_TIMEOUT,
7762         };
7763         int status;
7764
7765         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7766         nfs4_set_sequence_privileged(&args.la_seq_args);
7767         dprintk("--> %s\n", __func__);
7768         task = rpc_run_task(&task_setup);
7769
7770         if (IS_ERR(task))
7771                 status = PTR_ERR(task);
7772         else {
7773                 status = task->tk_status;
7774                 rpc_put_task(task);
7775         }
7776         dprintk("<-- %s return %d\n", __func__, status);
7777
7778         return status;
7779 }
7780
7781 /*
7782  * Initialize the values to be used by the client in CREATE_SESSION
7783  * If nfs4_init_session set the fore channel request and response sizes,
7784  * use them.
7785  *
7786  * Set the back channel max_resp_sz_cached to zero to force the client to
7787  * always set csa_cachethis to FALSE because the current implementation
7788  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7789  */
7790 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7791                                     struct rpc_clnt *clnt)
7792 {
7793         unsigned int max_rqst_sz, max_resp_sz;
7794         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7795
7796         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7797         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7798
7799         /* Fore channel attributes */
7800         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7801         args->fc_attrs.max_resp_sz = max_resp_sz;
7802         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7803         args->fc_attrs.max_reqs = max_session_slots;
7804
7805         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7806                 "max_ops=%u max_reqs=%u\n",
7807                 __func__,
7808                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7809                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7810
7811         /* Back channel attributes */
7812         args->bc_attrs.max_rqst_sz = max_bc_payload;
7813         args->bc_attrs.max_resp_sz = max_bc_payload;
7814         args->bc_attrs.max_resp_sz_cached = 0;
7815         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7816         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7817
7818         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7819                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7820                 __func__,
7821                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7822                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7823                 args->bc_attrs.max_reqs);
7824 }
7825
7826 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7827                 struct nfs41_create_session_res *res)
7828 {
7829         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7830         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7831
7832         if (rcvd->max_resp_sz > sent->max_resp_sz)
7833                 return -EINVAL;
7834         /*
7835          * Our requested max_ops is the minimum we need; we're not
7836          * prepared to break up compounds into smaller pieces than that.
7837          * So, no point even trying to continue if the server won't
7838          * cooperate:
7839          */
7840         if (rcvd->max_ops < sent->max_ops)
7841                 return -EINVAL;
7842         if (rcvd->max_reqs == 0)
7843                 return -EINVAL;
7844         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7845                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7846         return 0;
7847 }
7848
7849 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7850                 struct nfs41_create_session_res *res)
7851 {
7852         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7853         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7854
7855         if (!(res->flags & SESSION4_BACK_CHAN))
7856                 goto out;
7857         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7858                 return -EINVAL;
7859         if (rcvd->max_resp_sz < sent->max_resp_sz)
7860                 return -EINVAL;
7861         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7862                 return -EINVAL;
7863         if (rcvd->max_ops > sent->max_ops)
7864                 return -EINVAL;
7865         if (rcvd->max_reqs > sent->max_reqs)
7866                 return -EINVAL;
7867 out:
7868         return 0;
7869 }
7870
7871 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7872                                      struct nfs41_create_session_res *res)
7873 {
7874         int ret;
7875
7876         ret = nfs4_verify_fore_channel_attrs(args, res);
7877         if (ret)
7878                 return ret;
7879         return nfs4_verify_back_channel_attrs(args, res);
7880 }
7881
7882 static void nfs4_update_session(struct nfs4_session *session,
7883                 struct nfs41_create_session_res *res)
7884 {
7885         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7886         /* Mark client id and session as being confirmed */
7887         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7888         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7889         session->flags = res->flags;
7890         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7891         if (res->flags & SESSION4_BACK_CHAN)
7892                 memcpy(&session->bc_attrs, &res->bc_attrs,
7893                                 sizeof(session->bc_attrs));
7894 }
7895
7896 static int _nfs4_proc_create_session(struct nfs_client *clp,
7897                 struct rpc_cred *cred)
7898 {
7899         struct nfs4_session *session = clp->cl_session;
7900         struct nfs41_create_session_args args = {
7901                 .client = clp,
7902                 .clientid = clp->cl_clientid,
7903                 .seqid = clp->cl_seqid,
7904                 .cb_program = NFS4_CALLBACK,
7905         };
7906         struct nfs41_create_session_res res;
7907
7908         struct rpc_message msg = {
7909                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7910                 .rpc_argp = &args,
7911                 .rpc_resp = &res,
7912                 .rpc_cred = cred,
7913         };
7914         int status;
7915
7916         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7917         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7918
7919         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7920         trace_nfs4_create_session(clp, status);
7921
7922         switch (status) {
7923         case -NFS4ERR_STALE_CLIENTID:
7924         case -NFS4ERR_DELAY:
7925         case -ETIMEDOUT:
7926         case -EACCES:
7927         case -EAGAIN:
7928                 goto out;
7929         };
7930
7931         clp->cl_seqid++;
7932         if (!status) {
7933                 /* Verify the session's negotiated channel_attrs values */
7934                 status = nfs4_verify_channel_attrs(&args, &res);
7935                 /* Increment the clientid slot sequence id */
7936                 if (status)
7937                         goto out;
7938                 nfs4_update_session(session, &res);
7939         }
7940 out:
7941         return status;
7942 }
7943
7944 /*
7945  * Issues a CREATE_SESSION operation to the server.
7946  * It is the responsibility of the caller to verify the session is
7947  * expired before calling this routine.
7948  */
7949 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7950 {
7951         int status;
7952         unsigned *ptr;
7953         struct nfs4_session *session = clp->cl_session;
7954
7955         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7956
7957         status = _nfs4_proc_create_session(clp, cred);
7958         if (status)
7959                 goto out;
7960
7961         /* Init or reset the session slot tables */
7962         status = nfs4_setup_session_slot_tables(session);
7963         dprintk("slot table setup returned %d\n", status);
7964         if (status)
7965                 goto out;
7966
7967         ptr = (unsigned *)&session->sess_id.data[0];
7968         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7969                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7970 out:
7971         dprintk("<-- %s\n", __func__);
7972         return status;
7973 }
7974
7975 /*
7976  * Issue the over-the-wire RPC DESTROY_SESSION.
7977  * The caller must serialize access to this routine.
7978  */
7979 int nfs4_proc_destroy_session(struct nfs4_session *session,
7980                 struct rpc_cred *cred)
7981 {
7982         struct rpc_message msg = {
7983                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7984                 .rpc_argp = session,
7985                 .rpc_cred = cred,
7986         };
7987         int status = 0;
7988
7989         dprintk("--> nfs4_proc_destroy_session\n");
7990
7991         /* session is still being setup */
7992         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7993                 return 0;
7994
7995         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7996         trace_nfs4_destroy_session(session->clp, status);
7997
7998         if (status)
7999                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8000                         "Session has been destroyed regardless...\n", status);
8001
8002         dprintk("<-- nfs4_proc_destroy_session\n");
8003         return status;
8004 }
8005
8006 /*
8007  * Renew the cl_session lease.
8008  */
8009 struct nfs4_sequence_data {
8010         struct nfs_client *clp;
8011         struct nfs4_sequence_args args;
8012         struct nfs4_sequence_res res;
8013 };
8014
8015 static void nfs41_sequence_release(void *data)
8016 {
8017         struct nfs4_sequence_data *calldata = data;
8018         struct nfs_client *clp = calldata->clp;
8019
8020         if (atomic_read(&clp->cl_count) > 1)
8021                 nfs4_schedule_state_renewal(clp);
8022         nfs_put_client(clp);
8023         kfree(calldata);
8024 }
8025
8026 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8027 {
8028         switch(task->tk_status) {
8029         case -NFS4ERR_DELAY:
8030                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8031                 return -EAGAIN;
8032         default:
8033                 nfs4_schedule_lease_recovery(clp);
8034         }
8035         return 0;
8036 }
8037
8038 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8039 {
8040         struct nfs4_sequence_data *calldata = data;
8041         struct nfs_client *clp = calldata->clp;
8042
8043         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8044                 return;
8045
8046         trace_nfs4_sequence(clp, task->tk_status);
8047         if (task->tk_status < 0) {
8048                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8049                 if (atomic_read(&clp->cl_count) == 1)
8050                         goto out;
8051
8052                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8053                         rpc_restart_call_prepare(task);
8054                         return;
8055                 }
8056         }
8057         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8058 out:
8059         dprintk("<-- %s\n", __func__);
8060 }
8061
8062 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8063 {
8064         struct nfs4_sequence_data *calldata = data;
8065         struct nfs_client *clp = calldata->clp;
8066         struct nfs4_sequence_args *args;
8067         struct nfs4_sequence_res *res;
8068
8069         args = task->tk_msg.rpc_argp;
8070         res = task->tk_msg.rpc_resp;
8071
8072         nfs41_setup_sequence(clp->cl_session, args, res, task);
8073 }
8074
8075 static const struct rpc_call_ops nfs41_sequence_ops = {
8076         .rpc_call_done = nfs41_sequence_call_done,
8077         .rpc_call_prepare = nfs41_sequence_prepare,
8078         .rpc_release = nfs41_sequence_release,
8079 };
8080
8081 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8082                 struct rpc_cred *cred,
8083                 bool is_privileged)
8084 {
8085         struct nfs4_sequence_data *calldata;
8086         struct rpc_message msg = {
8087                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8088                 .rpc_cred = cred,
8089         };
8090         struct rpc_task_setup task_setup_data = {
8091                 .rpc_client = clp->cl_rpcclient,
8092                 .rpc_message = &msg,
8093                 .callback_ops = &nfs41_sequence_ops,
8094                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8095         };
8096
8097         if (!atomic_inc_not_zero(&clp->cl_count))
8098                 return ERR_PTR(-EIO);
8099         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8100         if (calldata == NULL) {
8101                 nfs_put_client(clp);
8102                 return ERR_PTR(-ENOMEM);
8103         }
8104         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8105         if (is_privileged)
8106                 nfs4_set_sequence_privileged(&calldata->args);
8107         msg.rpc_argp = &calldata->args;
8108         msg.rpc_resp = &calldata->res;
8109         calldata->clp = clp;
8110         task_setup_data.callback_data = calldata;
8111
8112         return rpc_run_task(&task_setup_data);
8113 }
8114
8115 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8116 {
8117         struct rpc_task *task;
8118         int ret = 0;
8119
8120         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8121                 return -EAGAIN;
8122         task = _nfs41_proc_sequence(clp, cred, false);
8123         if (IS_ERR(task))
8124                 ret = PTR_ERR(task);
8125         else
8126                 rpc_put_task_async(task);
8127         dprintk("<-- %s status=%d\n", __func__, ret);
8128         return ret;
8129 }
8130
8131 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8132 {
8133         struct rpc_task *task;
8134         int ret;
8135
8136         task = _nfs41_proc_sequence(clp, cred, true);
8137         if (IS_ERR(task)) {
8138                 ret = PTR_ERR(task);
8139                 goto out;
8140         }
8141         ret = rpc_wait_for_completion_task(task);
8142         if (!ret)
8143                 ret = task->tk_status;
8144         rpc_put_task(task);
8145 out:
8146         dprintk("<-- %s status=%d\n", __func__, ret);
8147         return ret;
8148 }
8149
8150 struct nfs4_reclaim_complete_data {
8151         struct nfs_client *clp;
8152         struct nfs41_reclaim_complete_args arg;
8153         struct nfs41_reclaim_complete_res res;
8154 };
8155
8156 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8157 {
8158         struct nfs4_reclaim_complete_data *calldata = data;
8159
8160         nfs41_setup_sequence(calldata->clp->cl_session,
8161                         &calldata->arg.seq_args,
8162                         &calldata->res.seq_res,
8163                         task);
8164 }
8165
8166 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8167 {
8168         switch(task->tk_status) {
8169         case 0:
8170         case -NFS4ERR_COMPLETE_ALREADY:
8171         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8172                 break;
8173         case -NFS4ERR_DELAY:
8174                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8175                 /* fall through */
8176         case -NFS4ERR_RETRY_UNCACHED_REP:
8177                 return -EAGAIN;
8178         default:
8179                 nfs4_schedule_lease_recovery(clp);
8180         }
8181         return 0;
8182 }
8183
8184 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8185 {
8186         struct nfs4_reclaim_complete_data *calldata = data;
8187         struct nfs_client *clp = calldata->clp;
8188         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8189
8190         dprintk("--> %s\n", __func__);
8191         if (!nfs41_sequence_done(task, res))
8192                 return;
8193
8194         trace_nfs4_reclaim_complete(clp, task->tk_status);
8195         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8196                 rpc_restart_call_prepare(task);
8197                 return;
8198         }
8199         dprintk("<-- %s\n", __func__);
8200 }
8201
8202 static void nfs4_free_reclaim_complete_data(void *data)
8203 {
8204         struct nfs4_reclaim_complete_data *calldata = data;
8205
8206         kfree(calldata);
8207 }
8208
8209 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8210         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8211         .rpc_call_done = nfs4_reclaim_complete_done,
8212         .rpc_release = nfs4_free_reclaim_complete_data,
8213 };
8214
8215 /*
8216  * Issue a global reclaim complete.
8217  */
8218 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8219                 struct rpc_cred *cred)
8220 {
8221         struct nfs4_reclaim_complete_data *calldata;
8222         struct rpc_task *task;
8223         struct rpc_message msg = {
8224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8225                 .rpc_cred = cred,
8226         };
8227         struct rpc_task_setup task_setup_data = {
8228                 .rpc_client = clp->cl_rpcclient,
8229                 .rpc_message = &msg,
8230                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8231                 .flags = RPC_TASK_ASYNC,
8232         };
8233         int status = -ENOMEM;
8234
8235         dprintk("--> %s\n", __func__);
8236         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8237         if (calldata == NULL)
8238                 goto out;
8239         calldata->clp = clp;
8240         calldata->arg.one_fs = 0;
8241
8242         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8243         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8244         msg.rpc_argp = &calldata->arg;
8245         msg.rpc_resp = &calldata->res;
8246         task_setup_data.callback_data = calldata;
8247         task = rpc_run_task(&task_setup_data);
8248         if (IS_ERR(task)) {
8249                 status = PTR_ERR(task);
8250                 goto out;
8251         }
8252         status = nfs4_wait_for_completion_rpc_task(task);
8253         if (status == 0)
8254                 status = task->tk_status;
8255         rpc_put_task(task);
8256         return 0;
8257 out:
8258         dprintk("<-- %s status=%d\n", __func__, status);
8259         return status;
8260 }
8261
8262 static void
8263 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8264 {
8265         struct nfs4_layoutget *lgp = calldata;
8266         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8267         struct nfs4_session *session = nfs4_get_session(server);
8268
8269         dprintk("--> %s\n", __func__);
8270         nfs41_setup_sequence(session, &lgp->args.seq_args,
8271                                 &lgp->res.seq_res, task);
8272         dprintk("<-- %s\n", __func__);
8273 }
8274
8275 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8276 {
8277         struct nfs4_layoutget *lgp = calldata;
8278
8279         dprintk("--> %s\n", __func__);
8280         nfs41_sequence_process(task, &lgp->res.seq_res);
8281         dprintk("<-- %s\n", __func__);
8282 }
8283
8284 static int
8285 nfs4_layoutget_handle_exception(struct rpc_task *task,
8286                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8287 {
8288         struct inode *inode = lgp->args.inode;
8289         struct nfs_server *server = NFS_SERVER(inode);
8290         struct pnfs_layout_hdr *lo;
8291         int nfs4err = task->tk_status;
8292         int err, status = 0;
8293         LIST_HEAD(head);
8294
8295         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8296
8297         switch (nfs4err) {
8298         case 0:
8299                 goto out;
8300
8301         /*
8302          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8303          * on the file. set tk_status to -ENODATA to tell upper layer to
8304          * retry go inband.
8305          */
8306         case -NFS4ERR_LAYOUTUNAVAILABLE:
8307                 status = -ENODATA;
8308                 goto out;
8309         /*
8310          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8311          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8312          */
8313         case -NFS4ERR_BADLAYOUT:
8314                 status = -EOVERFLOW;
8315                 goto out;
8316         /*
8317          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8318          * (or clients) writing to the same RAID stripe except when
8319          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8320          *
8321          * Treat it like we would RECALLCONFLICT -- we retry for a little
8322          * while, and then eventually give up.
8323          */
8324         case -NFS4ERR_LAYOUTTRYLATER:
8325                 if (lgp->args.minlength == 0) {
8326                         status = -EOVERFLOW;
8327                         goto out;
8328                 }
8329                 status = -EBUSY;
8330                 break;
8331         case -NFS4ERR_RECALLCONFLICT:
8332                 status = -ERECALLCONFLICT;
8333                 break;
8334         case -NFS4ERR_DELEG_REVOKED:
8335         case -NFS4ERR_ADMIN_REVOKED:
8336         case -NFS4ERR_EXPIRED:
8337         case -NFS4ERR_BAD_STATEID:
8338                 exception->timeout = 0;
8339                 spin_lock(&inode->i_lock);
8340                 lo = NFS_I(inode)->layout;
8341                 /* If the open stateid was bad, then recover it. */
8342                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8343                     nfs4_stateid_match_other(&lgp->args.stateid,
8344                                         &lgp->args.ctx->state->stateid)) {
8345                         spin_unlock(&inode->i_lock);
8346                         exception->state = lgp->args.ctx->state;
8347                         exception->stateid = &lgp->args.stateid;
8348                         break;
8349                 }
8350
8351                 /*
8352                  * Mark the bad layout state as invalid, then retry
8353                  */
8354                 pnfs_mark_layout_stateid_invalid(lo, &head);
8355                 spin_unlock(&inode->i_lock);
8356                 pnfs_free_lseg_list(&head);
8357                 status = -EAGAIN;
8358                 goto out;
8359         }
8360
8361         err = nfs4_handle_exception(server, nfs4err, exception);
8362         if (!status) {
8363                 if (exception->retry)
8364                         status = -EAGAIN;
8365                 else
8366                         status = err;
8367         }
8368 out:
8369         dprintk("<-- %s\n", __func__);
8370         return status;
8371 }
8372
8373 static size_t max_response_pages(struct nfs_server *server)
8374 {
8375         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8376         return nfs_page_array_len(0, max_resp_sz);
8377 }
8378
8379 static void nfs4_free_pages(struct page **pages, size_t size)
8380 {
8381         int i;
8382
8383         if (!pages)
8384                 return;
8385
8386         for (i = 0; i < size; i++) {
8387                 if (!pages[i])
8388                         break;
8389                 __free_page(pages[i]);
8390         }
8391         kfree(pages);
8392 }
8393
8394 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8395 {
8396         struct page **pages;
8397         int i;
8398
8399         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8400         if (!pages) {
8401                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8402                 return NULL;
8403         }
8404
8405         for (i = 0; i < size; i++) {
8406                 pages[i] = alloc_page(gfp_flags);
8407                 if (!pages[i]) {
8408                         dprintk("%s: failed to allocate page\n", __func__);
8409                         nfs4_free_pages(pages, size);
8410                         return NULL;
8411                 }
8412         }
8413
8414         return pages;
8415 }
8416
8417 static void nfs4_layoutget_release(void *calldata)
8418 {
8419         struct nfs4_layoutget *lgp = calldata;
8420         struct inode *inode = lgp->args.inode;
8421         struct nfs_server *server = NFS_SERVER(inode);
8422         size_t max_pages = max_response_pages(server);
8423
8424         dprintk("--> %s\n", __func__);
8425         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8426         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8427         put_nfs_open_context(lgp->args.ctx);
8428         kfree(calldata);
8429         dprintk("<-- %s\n", __func__);
8430 }
8431
8432 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8433         .rpc_call_prepare = nfs4_layoutget_prepare,
8434         .rpc_call_done = nfs4_layoutget_done,
8435         .rpc_release = nfs4_layoutget_release,
8436 };
8437
8438 struct pnfs_layout_segment *
8439 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8440 {
8441         struct inode *inode = lgp->args.inode;
8442         struct nfs_server *server = NFS_SERVER(inode);
8443         size_t max_pages = max_response_pages(server);
8444         struct rpc_task *task;
8445         struct rpc_message msg = {
8446                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8447                 .rpc_argp = &lgp->args,
8448                 .rpc_resp = &lgp->res,
8449                 .rpc_cred = lgp->cred,
8450         };
8451         struct rpc_task_setup task_setup_data = {
8452                 .rpc_client = server->client,
8453                 .rpc_message = &msg,
8454                 .callback_ops = &nfs4_layoutget_call_ops,
8455                 .callback_data = lgp,
8456                 .flags = RPC_TASK_ASYNC,
8457         };
8458         struct pnfs_layout_segment *lseg = NULL;
8459         struct nfs4_exception exception = {
8460                 .inode = inode,
8461                 .timeout = *timeout,
8462         };
8463         int status = 0;
8464
8465         dprintk("--> %s\n", __func__);
8466
8467         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8468         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8469
8470         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8471         if (!lgp->args.layout.pages) {
8472                 nfs4_layoutget_release(lgp);
8473                 return ERR_PTR(-ENOMEM);
8474         }
8475         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8476
8477         lgp->res.layoutp = &lgp->args.layout;
8478         lgp->res.seq_res.sr_slot = NULL;
8479         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8480
8481         task = rpc_run_task(&task_setup_data);
8482         if (IS_ERR(task))
8483                 return ERR_CAST(task);
8484         status = nfs4_wait_for_completion_rpc_task(task);
8485         if (status == 0) {
8486                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8487                 *timeout = exception.timeout;
8488         }
8489
8490         trace_nfs4_layoutget(lgp->args.ctx,
8491                         &lgp->args.range,
8492                         &lgp->res.range,
8493                         &lgp->res.stateid,
8494                         status);
8495
8496         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8497         if (status == 0 && lgp->res.layoutp->len)
8498                 lseg = pnfs_layout_process(lgp);
8499         nfs4_sequence_free_slot(&lgp->res.seq_res);
8500         rpc_put_task(task);
8501         dprintk("<-- %s status=%d\n", __func__, status);
8502         if (status)
8503                 return ERR_PTR(status);
8504         return lseg;
8505 }
8506
8507 static void
8508 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8509 {
8510         struct nfs4_layoutreturn *lrp = calldata;
8511
8512         dprintk("--> %s\n", __func__);
8513         nfs41_setup_sequence(lrp->clp->cl_session,
8514                         &lrp->args.seq_args,
8515                         &lrp->res.seq_res,
8516                         task);
8517 }
8518
8519 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8520 {
8521         struct nfs4_layoutreturn *lrp = calldata;
8522         struct nfs_server *server;
8523
8524         dprintk("--> %s\n", __func__);
8525
8526         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8527                 return;
8528
8529         server = NFS_SERVER(lrp->args.inode);
8530         switch (task->tk_status) {
8531         default:
8532                 task->tk_status = 0;
8533         case 0:
8534                 break;
8535         case -NFS4ERR_DELAY:
8536                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8537                         break;
8538                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8539                 rpc_restart_call_prepare(task);
8540                 return;
8541         }
8542         dprintk("<-- %s\n", __func__);
8543 }
8544
8545 static void nfs4_layoutreturn_release(void *calldata)
8546 {
8547         struct nfs4_layoutreturn *lrp = calldata;
8548         struct pnfs_layout_hdr *lo = lrp->args.layout;
8549         LIST_HEAD(freeme);
8550
8551         dprintk("--> %s\n", __func__);
8552         spin_lock(&lo->plh_inode->i_lock);
8553         if (lrp->res.lrs_present) {
8554                 pnfs_mark_matching_lsegs_invalid(lo, &freeme,
8555                                 &lrp->args.range,
8556                                 be32_to_cpu(lrp->args.stateid.seqid));
8557                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8558         } else
8559                 pnfs_mark_layout_stateid_invalid(lo, &freeme);
8560         pnfs_clear_layoutreturn_waitbit(lo);
8561         spin_unlock(&lo->plh_inode->i_lock);
8562         nfs4_sequence_free_slot(&lrp->res.seq_res);
8563         pnfs_free_lseg_list(&freeme);
8564         pnfs_put_layout_hdr(lrp->args.layout);
8565         nfs_iput_and_deactive(lrp->inode);
8566         kfree(calldata);
8567         dprintk("<-- %s\n", __func__);
8568 }
8569
8570 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8571         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8572         .rpc_call_done = nfs4_layoutreturn_done,
8573         .rpc_release = nfs4_layoutreturn_release,
8574 };
8575
8576 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8577 {
8578         struct rpc_task *task;
8579         struct rpc_message msg = {
8580                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8581                 .rpc_argp = &lrp->args,
8582                 .rpc_resp = &lrp->res,
8583                 .rpc_cred = lrp->cred,
8584         };
8585         struct rpc_task_setup task_setup_data = {
8586                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8587                 .rpc_message = &msg,
8588                 .callback_ops = &nfs4_layoutreturn_call_ops,
8589                 .callback_data = lrp,
8590         };
8591         int status = 0;
8592
8593         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8594                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8595                         &task_setup_data.rpc_client, &msg);
8596
8597         dprintk("--> %s\n", __func__);
8598         if (!sync) {
8599                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8600                 if (!lrp->inode) {
8601                         nfs4_layoutreturn_release(lrp);
8602                         return -EAGAIN;
8603                 }
8604                 task_setup_data.flags |= RPC_TASK_ASYNC;
8605         }
8606         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8607         task = rpc_run_task(&task_setup_data);
8608         if (IS_ERR(task))
8609                 return PTR_ERR(task);
8610         if (sync)
8611                 status = task->tk_status;
8612         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8613         dprintk("<-- %s status=%d\n", __func__, status);
8614         rpc_put_task(task);
8615         return status;
8616 }
8617
8618 static int
8619 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8620                 struct pnfs_device *pdev,
8621                 struct rpc_cred *cred)
8622 {
8623         struct nfs4_getdeviceinfo_args args = {
8624                 .pdev = pdev,
8625                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8626                         NOTIFY_DEVICEID4_DELETE,
8627         };
8628         struct nfs4_getdeviceinfo_res res = {
8629                 .pdev = pdev,
8630         };
8631         struct rpc_message msg = {
8632                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8633                 .rpc_argp = &args,
8634                 .rpc_resp = &res,
8635                 .rpc_cred = cred,
8636         };
8637         int status;
8638
8639         dprintk("--> %s\n", __func__);
8640         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8641         if (res.notification & ~args.notify_types)
8642                 dprintk("%s: unsupported notification\n", __func__);
8643         if (res.notification != args.notify_types)
8644                 pdev->nocache = 1;
8645
8646         dprintk("<-- %s status=%d\n", __func__, status);
8647
8648         return status;
8649 }
8650
8651 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8652                 struct pnfs_device *pdev,
8653                 struct rpc_cred *cred)
8654 {
8655         struct nfs4_exception exception = { };
8656         int err;
8657
8658         do {
8659                 err = nfs4_handle_exception(server,
8660                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8661                                         &exception);
8662         } while (exception.retry);
8663         return err;
8664 }
8665 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8666
8667 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8668 {
8669         struct nfs4_layoutcommit_data *data = calldata;
8670         struct nfs_server *server = NFS_SERVER(data->args.inode);
8671         struct nfs4_session *session = nfs4_get_session(server);
8672
8673         nfs41_setup_sequence(session,
8674                         &data->args.seq_args,
8675                         &data->res.seq_res,
8676                         task);
8677 }
8678
8679 static void
8680 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8681 {
8682         struct nfs4_layoutcommit_data *data = calldata;
8683         struct nfs_server *server = NFS_SERVER(data->args.inode);
8684
8685         if (!nfs41_sequence_done(task, &data->res.seq_res))
8686                 return;
8687
8688         switch (task->tk_status) { /* Just ignore these failures */
8689         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8690         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8691         case -NFS4ERR_BADLAYOUT:     /* no layout */
8692         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8693                 task->tk_status = 0;
8694         case 0:
8695                 break;
8696         default:
8697                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8698                         rpc_restart_call_prepare(task);
8699                         return;
8700                 }
8701         }
8702 }
8703
8704 static void nfs4_layoutcommit_release(void *calldata)
8705 {
8706         struct nfs4_layoutcommit_data *data = calldata;
8707
8708         pnfs_cleanup_layoutcommit(data);
8709         nfs_post_op_update_inode_force_wcc(data->args.inode,
8710                                            data->res.fattr);
8711         put_rpccred(data->cred);
8712         nfs_iput_and_deactive(data->inode);
8713         kfree(data);
8714 }
8715
8716 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8717         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8718         .rpc_call_done = nfs4_layoutcommit_done,
8719         .rpc_release = nfs4_layoutcommit_release,
8720 };
8721
8722 int
8723 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8724 {
8725         struct rpc_message msg = {
8726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8727                 .rpc_argp = &data->args,
8728                 .rpc_resp = &data->res,
8729                 .rpc_cred = data->cred,
8730         };
8731         struct rpc_task_setup task_setup_data = {
8732                 .task = &data->task,
8733                 .rpc_client = NFS_CLIENT(data->args.inode),
8734                 .rpc_message = &msg,
8735                 .callback_ops = &nfs4_layoutcommit_ops,
8736                 .callback_data = data,
8737         };
8738         struct rpc_task *task;
8739         int status = 0;
8740
8741         dprintk("NFS: initiating layoutcommit call. sync %d "
8742                 "lbw: %llu inode %lu\n", sync,
8743                 data->args.lastbytewritten,
8744                 data->args.inode->i_ino);
8745
8746         if (!sync) {
8747                 data->inode = nfs_igrab_and_active(data->args.inode);
8748                 if (data->inode == NULL) {
8749                         nfs4_layoutcommit_release(data);
8750                         return -EAGAIN;
8751                 }
8752                 task_setup_data.flags = RPC_TASK_ASYNC;
8753         }
8754         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8755         task = rpc_run_task(&task_setup_data);
8756         if (IS_ERR(task))
8757                 return PTR_ERR(task);
8758         if (sync)
8759                 status = task->tk_status;
8760         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8761         dprintk("%s: status %d\n", __func__, status);
8762         rpc_put_task(task);
8763         return status;
8764 }
8765
8766 /**
8767  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8768  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8769  */
8770 static int
8771 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8772                     struct nfs_fsinfo *info,
8773                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8774 {
8775         struct nfs41_secinfo_no_name_args args = {
8776                 .style = SECINFO_STYLE_CURRENT_FH,
8777         };
8778         struct nfs4_secinfo_res res = {
8779                 .flavors = flavors,
8780         };
8781         struct rpc_message msg = {
8782                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8783                 .rpc_argp = &args,
8784                 .rpc_resp = &res,
8785         };
8786         struct rpc_clnt *clnt = server->client;
8787         struct rpc_cred *cred = NULL;
8788         int status;
8789
8790         if (use_integrity) {
8791                 clnt = server->nfs_client->cl_rpcclient;
8792                 cred = nfs4_get_clid_cred(server->nfs_client);
8793                 msg.rpc_cred = cred;
8794         }
8795
8796         dprintk("--> %s\n", __func__);
8797         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8798                                 &res.seq_res, 0);
8799         dprintk("<-- %s status=%d\n", __func__, status);
8800
8801         if (cred)
8802                 put_rpccred(cred);
8803
8804         return status;
8805 }
8806
8807 static int
8808 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8809                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8810 {
8811         struct nfs4_exception exception = { };
8812         int err;
8813         do {
8814                 /* first try using integrity protection */
8815                 err = -NFS4ERR_WRONGSEC;
8816
8817                 /* try to use integrity protection with machine cred */
8818                 if (_nfs4_is_integrity_protected(server->nfs_client))
8819                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8820                                                           flavors, true);
8821
8822                 /*
8823                  * if unable to use integrity protection, or SECINFO with
8824                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8825                  * disallowed by spec, but exists in deployed servers) use
8826                  * the current filesystem's rpc_client and the user cred.
8827                  */
8828                 if (err == -NFS4ERR_WRONGSEC)
8829                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8830                                                           flavors, false);
8831
8832                 switch (err) {
8833                 case 0:
8834                 case -NFS4ERR_WRONGSEC:
8835                 case -ENOTSUPP:
8836                         goto out;
8837                 default:
8838                         err = nfs4_handle_exception(server, err, &exception);
8839                 }
8840         } while (exception.retry);
8841 out:
8842         return err;
8843 }
8844
8845 static int
8846 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8847                     struct nfs_fsinfo *info)
8848 {
8849         int err;
8850         struct page *page;
8851         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8852         struct nfs4_secinfo_flavors *flavors;
8853         struct nfs4_secinfo4 *secinfo;
8854         int i;
8855
8856         page = alloc_page(GFP_KERNEL);
8857         if (!page) {
8858                 err = -ENOMEM;
8859                 goto out;
8860         }
8861
8862         flavors = page_address(page);
8863         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8864
8865         /*
8866          * Fall back on "guess and check" method if
8867          * the server doesn't support SECINFO_NO_NAME
8868          */
8869         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8870                 err = nfs4_find_root_sec(server, fhandle, info);
8871                 goto out_freepage;
8872         }
8873         if (err)
8874                 goto out_freepage;
8875
8876         for (i = 0; i < flavors->num_flavors; i++) {
8877                 secinfo = &flavors->flavors[i];
8878
8879                 switch (secinfo->flavor) {
8880                 case RPC_AUTH_NULL:
8881                 case RPC_AUTH_UNIX:
8882                 case RPC_AUTH_GSS:
8883                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8884                                         &secinfo->flavor_info);
8885                         break;
8886                 default:
8887                         flavor = RPC_AUTH_MAXFLAVOR;
8888                         break;
8889                 }
8890
8891                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8892                         flavor = RPC_AUTH_MAXFLAVOR;
8893
8894                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8895                         err = nfs4_lookup_root_sec(server, fhandle,
8896                                                    info, flavor);
8897                         if (!err)
8898                                 break;
8899                 }
8900         }
8901
8902         if (flavor == RPC_AUTH_MAXFLAVOR)
8903                 err = -EPERM;
8904
8905 out_freepage:
8906         put_page(page);
8907         if (err == -EACCES)
8908                 return -EPERM;
8909 out:
8910         return err;
8911 }
8912
8913 static int _nfs41_test_stateid(struct nfs_server *server,
8914                 nfs4_stateid *stateid,
8915                 struct rpc_cred *cred)
8916 {
8917         int status;
8918         struct nfs41_test_stateid_args args = {
8919                 .stateid = stateid,
8920         };
8921         struct nfs41_test_stateid_res res;
8922         struct rpc_message msg = {
8923                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8924                 .rpc_argp = &args,
8925                 .rpc_resp = &res,
8926                 .rpc_cred = cred,
8927         };
8928         struct rpc_clnt *rpc_client = server->client;
8929
8930         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8931                 &rpc_client, &msg);
8932
8933         dprintk("NFS call  test_stateid %p\n", stateid);
8934         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8935         nfs4_set_sequence_privileged(&args.seq_args);
8936         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8937                         &args.seq_args, &res.seq_res);
8938         if (status != NFS_OK) {
8939                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8940                 return status;
8941         }
8942         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8943         return -res.status;
8944 }
8945
8946 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8947                 int err, struct nfs4_exception *exception)
8948 {
8949         exception->retry = 0;
8950         switch(err) {
8951         case -NFS4ERR_DELAY:
8952         case -NFS4ERR_RETRY_UNCACHED_REP:
8953                 nfs4_handle_exception(server, err, exception);
8954                 break;
8955         case -NFS4ERR_BADSESSION:
8956         case -NFS4ERR_BADSLOT:
8957         case -NFS4ERR_BAD_HIGH_SLOT:
8958         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8959         case -NFS4ERR_DEADSESSION:
8960                 nfs4_do_handle_exception(server, err, exception);
8961         }
8962 }
8963
8964 /**
8965  * nfs41_test_stateid - perform a TEST_STATEID operation
8966  *
8967  * @server: server / transport on which to perform the operation
8968  * @stateid: state ID to test
8969  * @cred: credential
8970  *
8971  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8972  * Otherwise a negative NFS4ERR value is returned if the operation
8973  * failed or the state ID is not currently valid.
8974  */
8975 static int nfs41_test_stateid(struct nfs_server *server,
8976                 nfs4_stateid *stateid,
8977                 struct rpc_cred *cred)
8978 {
8979         struct nfs4_exception exception = { };
8980         int err;
8981         do {
8982                 err = _nfs41_test_stateid(server, stateid, cred);
8983                 nfs4_handle_delay_or_session_error(server, err, &exception);
8984         } while (exception.retry);
8985         return err;
8986 }
8987
8988 struct nfs_free_stateid_data {
8989         struct nfs_server *server;
8990         struct nfs41_free_stateid_args args;
8991         struct nfs41_free_stateid_res res;
8992 };
8993
8994 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8995 {
8996         struct nfs_free_stateid_data *data = calldata;
8997         nfs41_setup_sequence(nfs4_get_session(data->server),
8998                         &data->args.seq_args,
8999                         &data->res.seq_res,
9000                         task);
9001 }
9002
9003 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9004 {
9005         struct nfs_free_stateid_data *data = calldata;
9006
9007         nfs41_sequence_done(task, &data->res.seq_res);
9008
9009         switch (task->tk_status) {
9010         case -NFS4ERR_DELAY:
9011                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9012                         rpc_restart_call_prepare(task);
9013         }
9014 }
9015
9016 static void nfs41_free_stateid_release(void *calldata)
9017 {
9018         kfree(calldata);
9019 }
9020
9021 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9022         .rpc_call_prepare = nfs41_free_stateid_prepare,
9023         .rpc_call_done = nfs41_free_stateid_done,
9024         .rpc_release = nfs41_free_stateid_release,
9025 };
9026
9027 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9028                 const nfs4_stateid *stateid,
9029                 struct rpc_cred *cred,
9030                 bool privileged)
9031 {
9032         struct rpc_message msg = {
9033                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9034                 .rpc_cred = cred,
9035         };
9036         struct rpc_task_setup task_setup = {
9037                 .rpc_client = server->client,
9038                 .rpc_message = &msg,
9039                 .callback_ops = &nfs41_free_stateid_ops,
9040                 .flags = RPC_TASK_ASYNC,
9041         };
9042         struct nfs_free_stateid_data *data;
9043
9044         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9045                 &task_setup.rpc_client, &msg);
9046
9047         dprintk("NFS call  free_stateid %p\n", stateid);
9048         data = kmalloc(sizeof(*data), GFP_NOFS);
9049         if (!data)
9050                 return ERR_PTR(-ENOMEM);
9051         data->server = server;
9052         nfs4_stateid_copy(&data->args.stateid, stateid);
9053
9054         task_setup.callback_data = data;
9055
9056         msg.rpc_argp = &data->args;
9057         msg.rpc_resp = &data->res;
9058         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9059         if (privileged)
9060                 nfs4_set_sequence_privileged(&data->args.seq_args);
9061
9062         return rpc_run_task(&task_setup);
9063 }
9064
9065 /**
9066  * nfs41_free_stateid - perform a FREE_STATEID operation
9067  *
9068  * @server: server / transport on which to perform the operation
9069  * @stateid: state ID to release
9070  * @cred: credential
9071  * @is_recovery: set to true if this call needs to be privileged
9072  *
9073  * Note: this function is always asynchronous.
9074  */
9075 static int nfs41_free_stateid(struct nfs_server *server,
9076                 const nfs4_stateid *stateid,
9077                 struct rpc_cred *cred,
9078                 bool is_recovery)
9079 {
9080         struct rpc_task *task;
9081
9082         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9083         if (IS_ERR(task))
9084                 return PTR_ERR(task);
9085         rpc_put_task(task);
9086         return 0;
9087 }
9088
9089 static void
9090 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9091 {
9092         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9093
9094         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9095         nfs4_free_lock_state(server, lsp);
9096 }
9097
9098 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9099                 const nfs4_stateid *s2)
9100 {
9101         if (s1->type != s2->type)
9102                 return false;
9103
9104         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9105                 return false;
9106
9107         if (s1->seqid == s2->seqid)
9108                 return true;
9109         if (s1->seqid == 0 || s2->seqid == 0)
9110                 return true;
9111
9112         return false;
9113 }
9114
9115 #endif /* CONFIG_NFS_V4_1 */
9116
9117 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9118                 const nfs4_stateid *s2)
9119 {
9120         return nfs4_stateid_match(s1, s2);
9121 }
9122
9123
9124 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9125         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9126         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9127         .recover_open   = nfs4_open_reclaim,
9128         .recover_lock   = nfs4_lock_reclaim,
9129         .establish_clid = nfs4_init_clientid,
9130         .detect_trunking = nfs40_discover_server_trunking,
9131 };
9132
9133 #if defined(CONFIG_NFS_V4_1)
9134 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9135         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9136         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9137         .recover_open   = nfs4_open_reclaim,
9138         .recover_lock   = nfs4_lock_reclaim,
9139         .establish_clid = nfs41_init_clientid,
9140         .reclaim_complete = nfs41_proc_reclaim_complete,
9141         .detect_trunking = nfs41_discover_server_trunking,
9142 };
9143 #endif /* CONFIG_NFS_V4_1 */
9144
9145 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9146         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9147         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9148         .recover_open   = nfs40_open_expired,
9149         .recover_lock   = nfs4_lock_expired,
9150         .establish_clid = nfs4_init_clientid,
9151 };
9152
9153 #if defined(CONFIG_NFS_V4_1)
9154 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9155         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9156         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9157         .recover_open   = nfs41_open_expired,
9158         .recover_lock   = nfs41_lock_expired,
9159         .establish_clid = nfs41_init_clientid,
9160 };
9161 #endif /* CONFIG_NFS_V4_1 */
9162
9163 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9164         .sched_state_renewal = nfs4_proc_async_renew,
9165         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9166         .renew_lease = nfs4_proc_renew,
9167 };
9168
9169 #if defined(CONFIG_NFS_V4_1)
9170 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9171         .sched_state_renewal = nfs41_proc_async_sequence,
9172         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9173         .renew_lease = nfs4_proc_sequence,
9174 };
9175 #endif
9176
9177 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9178         .get_locations = _nfs40_proc_get_locations,
9179         .fsid_present = _nfs40_proc_fsid_present,
9180 };
9181
9182 #if defined(CONFIG_NFS_V4_1)
9183 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9184         .get_locations = _nfs41_proc_get_locations,
9185         .fsid_present = _nfs41_proc_fsid_present,
9186 };
9187 #endif  /* CONFIG_NFS_V4_1 */
9188
9189 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9190         .minor_version = 0,
9191         .init_caps = NFS_CAP_READDIRPLUS
9192                 | NFS_CAP_ATOMIC_OPEN
9193                 | NFS_CAP_POSIX_LOCK,
9194         .init_client = nfs40_init_client,
9195         .shutdown_client = nfs40_shutdown_client,
9196         .match_stateid = nfs4_match_stateid,
9197         .find_root_sec = nfs4_find_root_sec,
9198         .free_lock_state = nfs4_release_lockowner,
9199         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9200         .alloc_seqid = nfs_alloc_seqid,
9201         .call_sync_ops = &nfs40_call_sync_ops,
9202         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9203         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9204         .state_renewal_ops = &nfs40_state_renewal_ops,
9205         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9206 };
9207
9208 #if defined(CONFIG_NFS_V4_1)
9209 static struct nfs_seqid *
9210 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9211 {
9212         return NULL;
9213 }
9214
9215 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9216         .minor_version = 1,
9217         .init_caps = NFS_CAP_READDIRPLUS
9218                 | NFS_CAP_ATOMIC_OPEN
9219                 | NFS_CAP_POSIX_LOCK
9220                 | NFS_CAP_STATEID_NFSV41
9221                 | NFS_CAP_ATOMIC_OPEN_V1,
9222         .init_client = nfs41_init_client,
9223         .shutdown_client = nfs41_shutdown_client,
9224         .match_stateid = nfs41_match_stateid,
9225         .find_root_sec = nfs41_find_root_sec,
9226         .free_lock_state = nfs41_free_lock_state,
9227         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9228         .alloc_seqid = nfs_alloc_no_seqid,
9229         .session_trunk = nfs4_test_session_trunk,
9230         .call_sync_ops = &nfs41_call_sync_ops,
9231         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9232         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9233         .state_renewal_ops = &nfs41_state_renewal_ops,
9234         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9235 };
9236 #endif
9237
9238 #if defined(CONFIG_NFS_V4_2)
9239 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9240         .minor_version = 2,
9241         .init_caps = NFS_CAP_READDIRPLUS
9242                 | NFS_CAP_ATOMIC_OPEN
9243                 | NFS_CAP_POSIX_LOCK
9244                 | NFS_CAP_STATEID_NFSV41
9245                 | NFS_CAP_ATOMIC_OPEN_V1
9246                 | NFS_CAP_ALLOCATE
9247                 | NFS_CAP_COPY
9248                 | NFS_CAP_DEALLOCATE
9249                 | NFS_CAP_SEEK
9250                 | NFS_CAP_LAYOUTSTATS
9251                 | NFS_CAP_CLONE,
9252         .init_client = nfs41_init_client,
9253         .shutdown_client = nfs41_shutdown_client,
9254         .match_stateid = nfs41_match_stateid,
9255         .find_root_sec = nfs41_find_root_sec,
9256         .free_lock_state = nfs41_free_lock_state,
9257         .call_sync_ops = &nfs41_call_sync_ops,
9258         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9259         .alloc_seqid = nfs_alloc_no_seqid,
9260         .session_trunk = nfs4_test_session_trunk,
9261         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9262         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9263         .state_renewal_ops = &nfs41_state_renewal_ops,
9264         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9265 };
9266 #endif
9267
9268 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9269         [0] = &nfs_v4_0_minor_ops,
9270 #if defined(CONFIG_NFS_V4_1)
9271         [1] = &nfs_v4_1_minor_ops,
9272 #endif
9273 #if defined(CONFIG_NFS_V4_2)
9274         [2] = &nfs_v4_2_minor_ops,
9275 #endif
9276 };
9277
9278 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9279 {
9280         ssize_t error, error2;
9281
9282         error = generic_listxattr(dentry, list, size);
9283         if (error < 0)
9284                 return error;
9285         if (list) {
9286                 list += error;
9287                 size -= error;
9288         }
9289
9290         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9291         if (error2 < 0)
9292                 return error2;
9293         return error + error2;
9294 }
9295
9296 static const struct inode_operations nfs4_dir_inode_operations = {
9297         .create         = nfs_create,
9298         .lookup         = nfs_lookup,
9299         .atomic_open    = nfs_atomic_open,
9300         .link           = nfs_link,
9301         .unlink         = nfs_unlink,
9302         .symlink        = nfs_symlink,
9303         .mkdir          = nfs_mkdir,
9304         .rmdir          = nfs_rmdir,
9305         .mknod          = nfs_mknod,
9306         .rename         = nfs_rename,
9307         .permission     = nfs_permission,
9308         .getattr        = nfs_getattr,
9309         .setattr        = nfs_setattr,
9310         .getxattr       = generic_getxattr,
9311         .setxattr       = generic_setxattr,
9312         .listxattr      = nfs4_listxattr,
9313         .removexattr    = generic_removexattr,
9314 };
9315
9316 static const struct inode_operations nfs4_file_inode_operations = {
9317         .permission     = nfs_permission,
9318         .getattr        = nfs_getattr,
9319         .setattr        = nfs_setattr,
9320         .getxattr       = generic_getxattr,
9321         .setxattr       = generic_setxattr,
9322         .listxattr      = nfs4_listxattr,
9323         .removexattr    = generic_removexattr,
9324 };
9325
9326 const struct nfs_rpc_ops nfs_v4_clientops = {
9327         .version        = 4,                    /* protocol version */
9328         .dentry_ops     = &nfs4_dentry_operations,
9329         .dir_inode_ops  = &nfs4_dir_inode_operations,
9330         .file_inode_ops = &nfs4_file_inode_operations,
9331         .file_ops       = &nfs4_file_operations,
9332         .getroot        = nfs4_proc_get_root,
9333         .submount       = nfs4_submount,
9334         .try_mount      = nfs4_try_mount,
9335         .getattr        = nfs4_proc_getattr,
9336         .setattr        = nfs4_proc_setattr,
9337         .lookup         = nfs4_proc_lookup,
9338         .access         = nfs4_proc_access,
9339         .readlink       = nfs4_proc_readlink,
9340         .create         = nfs4_proc_create,
9341         .remove         = nfs4_proc_remove,
9342         .unlink_setup   = nfs4_proc_unlink_setup,
9343         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9344         .unlink_done    = nfs4_proc_unlink_done,
9345         .rename_setup   = nfs4_proc_rename_setup,
9346         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9347         .rename_done    = nfs4_proc_rename_done,
9348         .link           = nfs4_proc_link,
9349         .symlink        = nfs4_proc_symlink,
9350         .mkdir          = nfs4_proc_mkdir,
9351         .rmdir          = nfs4_proc_remove,
9352         .readdir        = nfs4_proc_readdir,
9353         .mknod          = nfs4_proc_mknod,
9354         .statfs         = nfs4_proc_statfs,
9355         .fsinfo         = nfs4_proc_fsinfo,
9356         .pathconf       = nfs4_proc_pathconf,
9357         .set_capabilities = nfs4_server_capabilities,
9358         .decode_dirent  = nfs4_decode_dirent,
9359         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9360         .read_setup     = nfs4_proc_read_setup,
9361         .read_done      = nfs4_read_done,
9362         .write_setup    = nfs4_proc_write_setup,
9363         .write_done     = nfs4_write_done,
9364         .commit_setup   = nfs4_proc_commit_setup,
9365         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9366         .commit_done    = nfs4_commit_done,
9367         .lock           = nfs4_proc_lock,
9368         .clear_acl_cache = nfs4_zap_acl_attr,
9369         .close_context  = nfs4_close_context,
9370         .open_context   = nfs4_atomic_open,
9371         .have_delegation = nfs4_have_delegation,
9372         .return_delegation = nfs4_inode_return_delegation,
9373         .alloc_client   = nfs4_alloc_client,
9374         .init_client    = nfs4_init_client,
9375         .free_client    = nfs4_free_client,
9376         .create_server  = nfs4_create_server,
9377         .clone_server   = nfs_clone_server,
9378 };
9379
9380 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9381         .name   = XATTR_NAME_NFSV4_ACL,
9382         .list   = nfs4_xattr_list_nfs4_acl,
9383         .get    = nfs4_xattr_get_nfs4_acl,
9384         .set    = nfs4_xattr_set_nfs4_acl,
9385 };
9386
9387 const struct xattr_handler *nfs4_xattr_handlers[] = {
9388         &nfs4_xattr_nfs4_acl_handler,
9389 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9390         &nfs4_xattr_nfs4_label_handler,
9391 #endif
9392         NULL
9393 };
9394
9395 /*
9396  * Local variables:
9397  *  c-basic-offset: 8
9398  * End:
9399  */