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 <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  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/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57
  58 #include "nfs4_fs.h"
  59 #include "delegation.h"
  60 #include "internal.h"
  61 #include "iostat.h"
  62 #include "callback.h"
  63 #include "pnfs.h"
  64 #include "netns.h"
  65 #include "nfs4idmap.h"
  66 #include "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #include "nfs4trace.h"
  70
  71 #define NFSDBG_FACILITY         NFSDBG_PROC
  72
  73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  75
  76 /* file attributes which can be mapped to nfs attributes */
  77 #define NFS4_VALID_ATTRS (ATTR_MODE \
  78         | ATTR_UID \
  79         | ATTR_GID \
  80         | ATTR_SIZE \
  81         | ATTR_ATIME \
  82         | ATTR_MTIME \
  83         | ATTR_CTIME \
  84         | ATTR_ATIME_SET \
  85         | ATTR_MTIME_SET)
  86
  87 struct nfs4_opendata;
  88 static int _nfs4_proc_open(struct nfs4_opendata *data);
  89 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  90 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  91 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  92 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  93 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  94 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  95                             struct nfs_fattr *fattr, struct iattr *sattr,
  96                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  97                             struct nfs4_label *olabel);
  98 #ifdef CONFIG_NFS_V4_1
  99 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 100                 struct rpc_cred *);
 101 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 102                 struct rpc_cred *, bool);
 103 #endif
 104
 105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 106 static inline struct nfs4_label *
 107 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 108         struct iattr *sattr, struct nfs4_label *label)
 109 {
 110         int err;
 111
 112         if (label == NULL)
 113                 return NULL;
 114
 115         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 116                 return NULL;
 117
 118         err = security_dentry_init_security(dentry, sattr->ia_mode,
 119                                 &dentry->d_name, (void **)&label->label, &label->len);
 120         if (err == 0)
 121                 return label;
 122
 123         return NULL;
 124 }
 125 static inline void
 126 nfs4_label_release_security(struct nfs4_label *label)
 127 {
 128         if (label)
 129                 security_release_secctx(label->label, label->len);
 130 }
 131 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 132 {
 133         if (label)
 134                 return server->attr_bitmask;
 135
 136         return server->attr_bitmask_nl;
 137 }
 138 #else
 139 static inline struct nfs4_label *
 140 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 141         struct iattr *sattr, struct nfs4_label *l)
 142 { return NULL; }
 143 static inline void
 144 nfs4_label_release_security(struct nfs4_label *label)
 145 { return; }
 146 static inline u32 *
 147 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 148 { return server->attr_bitmask; }
 149 #endif
 150
 151 /* Prevent leaks of NFSv4 errors into userland */
 152 static int nfs4_map_errors(int err)
 153 {
 154         if (err >= -1000)
 155                 return err;
 156         switch (err) {
 157         case -NFS4ERR_RESOURCE:
 158         case -NFS4ERR_LAYOUTTRYLATER:
 159         case -NFS4ERR_RECALLCONFLICT:
 160                 return -EREMOTEIO;
 161         case -NFS4ERR_WRONGSEC:
 162         case -NFS4ERR_WRONG_CRED:
 163                 return -EPERM;
 164         case -NFS4ERR_BADOWNER:
 165         case -NFS4ERR_BADNAME:
 166                 return -EINVAL;
 167         case -NFS4ERR_SHARE_DENIED:
 168                 return -EACCES;
 169         case -NFS4ERR_MINOR_VERS_MISMATCH:
 170                 return -EPROTONOSUPPORT;
 171         case -NFS4ERR_FILE_OPEN:
 172                 return -EBUSY;
 173         default:
 174                 dprintk("%s could not handle NFSv4 error %d\n",
 175                                 __func__, -err);
 176                 break;
 177         }
 178         return -EIO;
 179 }
 180
 181 /*
 182  * This is our standard bitmap for GETATTR requests.
 183  */
 184 const u32 nfs4_fattr_bitmap[3] = {
 185         FATTR4_WORD0_TYPE
 186         | FATTR4_WORD0_CHANGE
 187         | FATTR4_WORD0_SIZE
 188         | FATTR4_WORD0_FSID
 189         | FATTR4_WORD0_FILEID,
 190         FATTR4_WORD1_MODE
 191         | FATTR4_WORD1_NUMLINKS
 192         | FATTR4_WORD1_OWNER
 193         | FATTR4_WORD1_OWNER_GROUP
 194         | FATTR4_WORD1_RAWDEV
 195         | FATTR4_WORD1_SPACE_USED
 196         | FATTR4_WORD1_TIME_ACCESS
 197         | FATTR4_WORD1_TIME_METADATA
 198         | FATTR4_WORD1_TIME_MODIFY
 199         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 200 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 201         FATTR4_WORD2_SECURITY_LABEL
 202 #endif
 203 };
 204
 205 static const u32 nfs4_pnfs_open_bitmap[3] = {
 206         FATTR4_WORD0_TYPE
 207         | FATTR4_WORD0_CHANGE
 208         | FATTR4_WORD0_SIZE
 209         | FATTR4_WORD0_FSID
 210         | FATTR4_WORD0_FILEID,
 211         FATTR4_WORD1_MODE
 212         | FATTR4_WORD1_NUMLINKS
 213         | FATTR4_WORD1_OWNER
 214         | FATTR4_WORD1_OWNER_GROUP
 215         | FATTR4_WORD1_RAWDEV
 216         | FATTR4_WORD1_SPACE_USED
 217         | FATTR4_WORD1_TIME_ACCESS
 218         | FATTR4_WORD1_TIME_METADATA
 219         | FATTR4_WORD1_TIME_MODIFY,
 220         FATTR4_WORD2_MDSTHRESHOLD
 221 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 222         | FATTR4_WORD2_SECURITY_LABEL
 223 #endif
 224 };
 225
 226 static const u32 nfs4_open_noattr_bitmap[3] = {
 227         FATTR4_WORD0_TYPE
 228         | FATTR4_WORD0_FILEID,
 229 };
 230
 231 const u32 nfs4_statfs_bitmap[3] = {
 232         FATTR4_WORD0_FILES_AVAIL
 233         | FATTR4_WORD0_FILES_FREE
 234         | FATTR4_WORD0_FILES_TOTAL,
 235         FATTR4_WORD1_SPACE_AVAIL
 236         | FATTR4_WORD1_SPACE_FREE
 237         | FATTR4_WORD1_SPACE_TOTAL
 238 };
 239
 240 const u32 nfs4_pathconf_bitmap[3] = {
 241         FATTR4_WORD0_MAXLINK
 242         | FATTR4_WORD0_MAXNAME,
 243         0
 244 };
 245
 246 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 247                         | FATTR4_WORD0_MAXREAD
 248                         | FATTR4_WORD0_MAXWRITE
 249                         | FATTR4_WORD0_LEASE_TIME,
 250                         FATTR4_WORD1_TIME_DELTA
 251                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 252                         FATTR4_WORD2_LAYOUT_BLKSIZE
 253                         | FATTR4_WORD2_CLONE_BLKSIZE
 254 };
 255
 256 const u32 nfs4_fs_locations_bitmap[3] = {
 257         FATTR4_WORD0_TYPE
 258         | FATTR4_WORD0_CHANGE
 259         | FATTR4_WORD0_SIZE
 260         | FATTR4_WORD0_FSID
 261         | FATTR4_WORD0_FILEID
 262         | FATTR4_WORD0_FS_LOCATIONS,
 263         FATTR4_WORD1_MODE
 264         | FATTR4_WORD1_NUMLINKS
 265         | FATTR4_WORD1_OWNER
 266         | FATTR4_WORD1_OWNER_GROUP
 267         | FATTR4_WORD1_RAWDEV
 268         | FATTR4_WORD1_SPACE_USED
 269         | FATTR4_WORD1_TIME_ACCESS
 270         | FATTR4_WORD1_TIME_METADATA
 271         | FATTR4_WORD1_TIME_MODIFY
 272         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 273 };
 274
 275 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 276                 struct nfs4_readdir_arg *readdir)
 277 {
 278         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 279         __be32 *start, *p;
 280
 281         if (cookie > 2) {
 282                 readdir->cookie = cookie;
 283                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 284                 return;
 285         }
 286
 287         readdir->cookie = 0;
 288         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 289         if (cookie == 2)
 290                 return;
 291
 292         /*
 293          * NFSv4 servers do not return entries for '.' and '..'
 294          * Therefore, we fake these entries here.  We let '.'
 295          * have cookie 0 and '..' have cookie 1.  Note that
 296          * when talking to the server, we always send cookie 0
 297          * instead of 1 or 2.
 298          */
 299         start = p = kmap_atomic(*readdir->pages);
 300
 301         if (cookie == 0) {
 302                 *p++ = xdr_one;                                  /* next */
 303                 *p++ = xdr_zero;                   /* cookie, first word */
 304                 *p++ = xdr_one;                   /* cookie, second word */
 305                 *p++ = xdr_one;                             /* entry len */
 306                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 307                 p++;
 308                 *p++ = xdr_one;                         /* bitmap length */
 309                 *p++ = htonl(attrs);                           /* bitmap */
 310                 *p++ = htonl(12);             /* attribute buffer length */
 311                 *p++ = htonl(NF4DIR);
 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(attrs);                           /* bitmap */
 323         *p++ = htonl(12);             /* attribute buffer length */
 324         *p++ = htonl(NF4DIR);
 325         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 326
 327         readdir->pgbase = (char *)p - (char *)start;
 328         readdir->count -= readdir->pgbase;
 329         kunmap_atomic(start);
 330 }
 331
 332 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 333                 nfs4_stateid *stateid,
 334                 struct rpc_cred *cred)
 335 {
 336         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 337
 338         ops->test_and_free_expired(server, stateid, cred);
 339 }
 340
 341 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 342                 nfs4_stateid *stateid,
 343                 struct rpc_cred *cred)
 344 {
 345         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 346         nfs4_test_and_free_stateid(server, stateid, cred);
 347 }
 348
 349 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 350                 const nfs4_stateid *stateid,
 351                 struct rpc_cred *cred)
 352 {
 353         nfs4_stateid tmp;
 354
 355         nfs4_stateid_copy(&tmp, stateid);
 356         __nfs4_free_revoked_stateid(server, &tmp, cred);
 357 }
 358
 359 static long nfs4_update_delay(long *timeout)
 360 {
 361         long ret;
 362         if (!timeout)
 363                 return NFS4_POLL_RETRY_MAX;
 364         if (*timeout <= 0)
 365                 *timeout = NFS4_POLL_RETRY_MIN;
 366         if (*timeout > NFS4_POLL_RETRY_MAX)
 367                 *timeout = NFS4_POLL_RETRY_MAX;
 368         ret = *timeout;
 369         *timeout <<= 1;
 370         return ret;
 371 }
 372
 373 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 374 {
 375         int res = 0;
 376
 377         might_sleep();
 378
 379         freezable_schedule_timeout_killable_unsafe(
 380                 nfs4_update_delay(timeout));
 381         if (fatal_signal_pending(current))
 382                 res = -ERESTARTSYS;
 383         return res;
 384 }
 385
 386 /* This is the error handling routine for processes that are allowed
 387  * to sleep.
 388  */
 389 static int nfs4_do_handle_exception(struct nfs_server *server,
 390                 int errorcode, struct nfs4_exception *exception)
 391 {
 392         struct nfs_client *clp = server->nfs_client;
 393         struct nfs4_state *state = exception->state;
 394         const nfs4_stateid *stateid = exception->stateid;
 395         struct inode *inode = exception->inode;
 396         int ret = errorcode;
 397
 398         exception->delay = 0;
 399         exception->recovering = 0;
 400         exception->retry = 0;
 401
 402         if (stateid == NULL && state != NULL)
 403                 stateid = &state->stateid;
 404
 405         switch(errorcode) {
 406                 case 0:
 407                         return 0;
 408                 case -NFS4ERR_DELEG_REVOKED:
 409                 case -NFS4ERR_ADMIN_REVOKED:
 410                 case -NFS4ERR_EXPIRED:
 411                 case -NFS4ERR_BAD_STATEID:
 412                         if (inode != NULL && stateid != NULL) {
 413                                 nfs_inode_find_state_and_recover(inode,
 414                                                 stateid);
 415                                 goto wait_on_recovery;
 416                         }
 417                 case -NFS4ERR_OPENMODE:
 418                         if (inode) {
 419                                 int err;
 420
 421                                 err = nfs_async_inode_return_delegation(inode,
 422                                                 stateid);
 423                                 if (err == 0)
 424                                         goto wait_on_recovery;
 425                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 426                                         exception->retry = 1;
 427                                         break;
 428                                 }
 429                         }
 430                         if (state == NULL)
 431                                 break;
 432                         ret = nfs4_schedule_stateid_recovery(server, state);
 433                         if (ret < 0)
 434                                 break;
 435                         goto wait_on_recovery;
 436                 case -NFS4ERR_STALE_STATEID:
 437                 case -NFS4ERR_STALE_CLIENTID:
 438                         nfs4_schedule_lease_recovery(clp);
 439                         goto wait_on_recovery;
 440                 case -NFS4ERR_MOVED:
 441                         ret = nfs4_schedule_migration_recovery(server);
 442                         if (ret < 0)
 443                                 break;
 444                         goto wait_on_recovery;
 445                 case -NFS4ERR_LEASE_MOVED:
 446                         nfs4_schedule_lease_moved_recovery(clp);
 447                         goto wait_on_recovery;
 448 #if defined(CONFIG_NFS_V4_1)
 449                 case -NFS4ERR_BADSESSION:
 450                 case -NFS4ERR_BADSLOT:
 451                 case -NFS4ERR_BAD_HIGH_SLOT:
 452                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 453                 case -NFS4ERR_DEADSESSION:
 454                 case -NFS4ERR_SEQ_FALSE_RETRY:
 455                 case -NFS4ERR_SEQ_MISORDERED:
 456                         dprintk("%s ERROR: %d Reset session\n", __func__,
 457                                 errorcode);
 458                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 459                         goto wait_on_recovery;
 460 #endif /* defined(CONFIG_NFS_V4_1) */
 461                 case -NFS4ERR_FILE_OPEN:
 462                         if (exception->timeout > HZ) {
 463                                 /* We have retried a decent amount, time to
 464                                  * fail
 465                                  */
 466                                 ret = -EBUSY;
 467                                 break;
 468                         }
 469                 case -NFS4ERR_DELAY:
 470                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 471                 case -NFS4ERR_GRACE:
 472                 case -NFS4ERR_LAYOUTTRYLATER:
 473                 case -NFS4ERR_RECALLCONFLICT:
 474                         exception->delay = 1;
 475                         return 0;
 476
 477                 case -NFS4ERR_RETRY_UNCACHED_REP:
 478                 case -NFS4ERR_OLD_STATEID:
 479                         exception->retry = 1;
 480                         break;
 481                 case -NFS4ERR_BADOWNER:
 482                         /* The following works around a Linux server bug! */
 483                 case -NFS4ERR_BADNAME:
 484                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 485                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 486                                 exception->retry = 1;
 487                                 printk(KERN_WARNING "NFS: v4 server %s "
 488                                                 "does not accept raw "
 489                                                 "uid/gids. "
 490                                                 "Reenabling the idmapper.\n",
 491                                                 server->nfs_client->cl_hostname);
 492                         }
 493         }
 494         /* We failed to handle the error */
 495         return nfs4_map_errors(ret);
 496 wait_on_recovery:
 497         exception->recovering = 1;
 498         return 0;
 499 }
 500
 501 /* This is the error handling routine for processes that are allowed
 502  * to sleep.
 503  */
 504 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 505 {
 506         struct nfs_client *clp = server->nfs_client;
 507         int ret;
 508
 509         ret = nfs4_do_handle_exception(server, errorcode, exception);
 510         if (exception->delay) {
 511                 ret = nfs4_delay(server->client, &exception->timeout);
 512                 goto out_retry;
 513         }
 514         if (exception->recovering) {
 515                 ret = nfs4_wait_clnt_recover(clp);
 516                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 517                         return -EIO;
 518                 goto out_retry;
 519         }
 520         return ret;
 521 out_retry:
 522         if (ret == 0)
 523                 exception->retry = 1;
 524         return ret;
 525 }
 526
 527 static int
 528 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 529                 int errorcode, struct nfs4_exception *exception)
 530 {
 531         struct nfs_client *clp = server->nfs_client;
 532         int ret;
 533
 534         ret = nfs4_do_handle_exception(server, errorcode, exception);
 535         if (exception->delay) {
 536                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 537                 goto out_retry;
 538         }
 539         if (exception->recovering) {
 540                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 541                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 542                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 543                 goto out_retry;
 544         }
 545         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 546                 ret = -EIO;
 547         return ret;
 548 out_retry:
 549         if (ret == 0)
 550                 exception->retry = 1;
 551         return ret;
 552 }
 553
 554 static int
 555 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 556                         struct nfs4_state *state, long *timeout)
 557 {
 558         struct nfs4_exception exception = {
 559                 .state = state,
 560         };
 561
 562         if (task->tk_status >= 0)
 563                 return 0;
 564         if (timeout)
 565                 exception.timeout = *timeout;
 566         task->tk_status = nfs4_async_handle_exception(task, server,
 567                         task->tk_status,
 568                         &exception);
 569         if (exception.delay && timeout)
 570                 *timeout = exception.timeout;
 571         if (exception.retry)
 572                 return -EAGAIN;
 573         return 0;
 574 }
 575
 576 /*
 577  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 578  * or 'false' otherwise.
 579  */
 580 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 581 {
 582         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 583         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 584 }
 585
 586 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 587 {
 588         spin_lock(&clp->cl_lock);
 589         if (time_before(clp->cl_last_renewal,timestamp))
 590                 clp->cl_last_renewal = timestamp;
 591         spin_unlock(&clp->cl_lock);
 592 }
 593
 594 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 595 {
 596         struct nfs_client *clp = server->nfs_client;
 597
 598         if (!nfs4_has_session(clp))
 599                 do_renew_lease(clp, timestamp);
 600 }
 601
 602 struct nfs4_call_sync_data {
 603         const struct nfs_server *seq_server;
 604         struct nfs4_sequence_args *seq_args;
 605         struct nfs4_sequence_res *seq_res;
 606 };
 607
 608 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 609                         struct nfs4_sequence_res *res, int cache_reply)
 610 {
 611         args->sa_slot = NULL;
 612         args->sa_cache_this = cache_reply;
 613         args->sa_privileged = 0;
 614
 615         res->sr_slot = NULL;
 616 }
 617
 618 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 619 {
 620         args->sa_privileged = 1;
 621 }
 622
 623 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 624 {
 625         struct nfs4_slot *slot = res->sr_slot;
 626         struct nfs4_slot_table *tbl;
 627
 628         tbl = slot->table;
 629         spin_lock(&tbl->slot_tbl_lock);
 630         if (!nfs41_wake_and_assign_slot(tbl, slot))
 631                 nfs4_free_slot(tbl, slot);
 632         spin_unlock(&tbl->slot_tbl_lock);
 633
 634         res->sr_slot = NULL;
 635 }
 636
 637 static int nfs40_sequence_done(struct rpc_task *task,
 638                                struct nfs4_sequence_res *res)
 639 {
 640         if (res->sr_slot != NULL)
 641                 nfs40_sequence_free_slot(res);
 642         return 1;
 643 }
 644
 645 #if defined(CONFIG_NFS_V4_1)
 646
 647 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 648 {
 649         struct nfs4_session *session;
 650         struct nfs4_slot_table *tbl;
 651         struct nfs4_slot *slot = res->sr_slot;
 652         bool send_new_highest_used_slotid = false;
 653
 654         tbl = slot->table;
 655         session = tbl->session;
 656
 657         /* Bump the slot sequence number */
 658         if (slot->seq_done)
 659                 slot->seq_nr++;
 660         slot->seq_done = 0;
 661
 662         spin_lock(&tbl->slot_tbl_lock);
 663         /* Be nice to the server: try to ensure that the last transmitted
 664          * value for highest_user_slotid <= target_highest_slotid
 665          */
 666         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 667                 send_new_highest_used_slotid = true;
 668
 669         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 670                 send_new_highest_used_slotid = false;
 671                 goto out_unlock;
 672         }
 673         nfs4_free_slot(tbl, slot);
 674
 675         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 676                 send_new_highest_used_slotid = false;
 677 out_unlock:
 678         spin_unlock(&tbl->slot_tbl_lock);
 679         res->sr_slot = NULL;
 680         if (send_new_highest_used_slotid)
 681                 nfs41_notify_server(session->clp);
 682         if (waitqueue_active(&tbl->slot_waitq))
 683                 wake_up_all(&tbl->slot_waitq);
 684 }
 685
 686 static int nfs41_sequence_process(struct rpc_task *task,
 687                 struct nfs4_sequence_res *res)
 688 {
 689         struct nfs4_session *session;
 690         struct nfs4_slot *slot = res->sr_slot;
 691         struct nfs_client *clp;
 692         bool interrupted = false;
 693         int ret = 1;
 694
 695         if (slot == NULL)
 696                 goto out_noaction;
 697         /* don't increment the sequence number if the task wasn't sent */
 698         if (!RPC_WAS_SENT(task))
 699                 goto out;
 700
 701         session = slot->table->session;
 702
 703         if (slot->interrupted) {
 704                 if (res->sr_status != -NFS4ERR_DELAY)
 705                         slot->interrupted = 0;
 706                 interrupted = true;
 707         }
 708
 709         trace_nfs4_sequence_done(session, res);
 710         /* Check the SEQUENCE operation status */
 711         switch (res->sr_status) {
 712         case 0:
 713                 /* If previous op on slot was interrupted and we reused
 714                  * the seq# and got a reply from the cache, then retry
 715                  */
 716                 if (task->tk_status == -EREMOTEIO && interrupted) {
 717                         ++slot->seq_nr;
 718                         goto retry_nowait;
 719                 }
 720                 /* Update the slot's sequence and clientid lease timer */
 721                 slot->seq_done = 1;
 722                 clp = session->clp;
 723                 do_renew_lease(clp, res->sr_timestamp);
 724                 /* Check sequence flags */
 725                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 726                                 !!slot->privileged);
 727                 nfs41_update_target_slotid(slot->table, slot, res);
 728                 break;
 729         case 1:
 730                 /*
 731                  * sr_status remains 1 if an RPC level error occurred.
 732                  * The server may or may not have processed the sequence
 733                  * operation..
 734                  * Mark the slot as having hosted an interrupted RPC call.
 735                  */
 736                 slot->interrupted = 1;
 737                 goto out;
 738         case -NFS4ERR_DELAY:
 739                 /* The server detected a resend of the RPC call and
 740                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 741                  * of RFC5661.
 742                  */
 743                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 744                         __func__,
 745                         slot->slot_nr,
 746                         slot->seq_nr);
 747                 goto out_retry;
 748         case -NFS4ERR_BADSLOT:
 749                 /*
 750                  * The slot id we used was probably retired. Try again
 751                  * using a different slot id.
 752                  */
 753                 goto retry_nowait;
 754         case -NFS4ERR_SEQ_MISORDERED:
 755                 /*
 756                  * Was the last operation on this sequence interrupted?
 757                  * If so, retry after bumping the sequence number.
 758                  */
 759                 if (interrupted) {
 760                         ++slot->seq_nr;
 761                         goto retry_nowait;
 762                 }
 763                 /*
 764                  * Could this slot have been previously retired?
 765                  * If so, then the server may be expecting seq_nr = 1!
 766                  */
 767                 if (slot->seq_nr != 1) {
 768                         slot->seq_nr = 1;
 769                         goto retry_nowait;
 770                 }
 771                 break;
 772         case -NFS4ERR_SEQ_FALSE_RETRY:
 773                 ++slot->seq_nr;
 774                 goto retry_nowait;
 775         default:
 776                 /* Just update the slot sequence no. */
 777                 slot->seq_done = 1;
 778         }
 779 out:
 780         /* The session may be reset by one of the error handlers. */
 781         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 782 out_noaction:
 783         return ret;
 784 retry_nowait:
 785         if (rpc_restart_call_prepare(task)) {
 786                 nfs41_sequence_free_slot(res);
 787                 task->tk_status = 0;
 788                 ret = 0;
 789         }
 790         goto out;
 791 out_retry:
 792         if (!rpc_restart_call(task))
 793                 goto out;
 794         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 795         return 0;
 796 }
 797
 798 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 799 {
 800         if (!nfs41_sequence_process(task, res))
 801                 return 0;
 802         if (res->sr_slot != NULL)
 803                 nfs41_sequence_free_slot(res);
 804         return 1;
 805
 806 }
 807 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 808
 809 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 810 {
 811         if (res->sr_slot == NULL)
 812                 return 1;
 813         if (res->sr_slot->table->session != NULL)
 814                 return nfs41_sequence_process(task, res);
 815         return nfs40_sequence_done(task, res);
 816 }
 817
 818 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 819 {
 820         if (res->sr_slot != NULL) {
 821                 if (res->sr_slot->table->session != NULL)
 822                         nfs41_sequence_free_slot(res);
 823                 else
 824                         nfs40_sequence_free_slot(res);
 825         }
 826 }
 827
 828 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 829 {
 830         if (res->sr_slot == NULL)
 831                 return 1;
 832         if (!res->sr_slot->table->session)
 833                 return nfs40_sequence_done(task, res);
 834         return nfs41_sequence_done(task, res);
 835 }
 836 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 837
 838 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 839 {
 840         struct nfs4_call_sync_data *data = calldata;
 841
 842         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 843
 844         nfs4_setup_sequence(data->seq_server->nfs_client,
 845                             data->seq_args, data->seq_res, task);
 846 }
 847
 848 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 849 {
 850         struct nfs4_call_sync_data *data = calldata;
 851
 852         nfs41_sequence_done(task, data->seq_res);
 853 }
 854
 855 static const struct rpc_call_ops nfs41_call_sync_ops = {
 856         .rpc_call_prepare = nfs41_call_sync_prepare,
 857         .rpc_call_done = nfs41_call_sync_done,
 858 };
 859
 860 #else   /* !CONFIG_NFS_V4_1 */
 861
 862 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 863 {
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL)
 870                 nfs40_sequence_free_slot(res);
 871 }
 872
 873 int nfs4_sequence_done(struct rpc_task *task,
 874                        struct nfs4_sequence_res *res)
 875 {
 876         return nfs40_sequence_done(task, res);
 877 }
 878 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 879
 880 #endif  /* !CONFIG_NFS_V4_1 */
 881
 882 int nfs4_setup_sequence(const struct nfs_client *client,
 883                         struct nfs4_sequence_args *args,
 884                         struct nfs4_sequence_res *res,
 885                         struct rpc_task *task)
 886 {
 887         struct nfs4_session *session = nfs4_get_session(client);
 888         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 889         struct nfs4_slot *slot;
 890
 891         /* slot already allocated? */
 892         if (res->sr_slot != NULL)
 893                 goto out_start;
 894
 895         if (session) {
 896                 tbl = &session->fc_slot_table;
 897                 task->tk_timeout = 0;
 898         }
 899
 900         spin_lock(&tbl->slot_tbl_lock);
 901         /* The state manager will wait until the slot table is empty */
 902         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 903                 goto out_sleep;
 904
 905         slot = nfs4_alloc_slot(tbl);
 906         if (IS_ERR(slot)) {
 907                 /* Try again in 1/4 second */
 908                 if (slot == ERR_PTR(-ENOMEM))
 909                         task->tk_timeout = HZ >> 2;
 910                 goto out_sleep;
 911         }
 912         spin_unlock(&tbl->slot_tbl_lock);
 913
 914         slot->privileged = args->sa_privileged ? 1 : 0;
 915         args->sa_slot = slot;
 916
 917         res->sr_slot = slot;
 918         if (session) {
 919                 res->sr_timestamp = jiffies;
 920                 res->sr_status_flags = 0;
 921                 res->sr_status = 1;
 922         }
 923
 924         trace_nfs4_setup_sequence(session, args);
 925 out_start:
 926         rpc_call_start(task);
 927         return 0;
 928
 929 out_sleep:
 930         if (args->sa_privileged)
 931                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 932                                 NULL, RPC_PRIORITY_PRIVILEGED);
 933         else
 934                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 935         spin_unlock(&tbl->slot_tbl_lock);
 936         return -EAGAIN;
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
 939
 940 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 941 {
 942         struct nfs4_call_sync_data *data = calldata;
 943         nfs4_setup_sequence(data->seq_server->nfs_client,
 944                                 data->seq_args, data->seq_res, task);
 945 }
 946
 947 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950         nfs4_sequence_done(task, data->seq_res);
 951 }
 952
 953 static const struct rpc_call_ops nfs40_call_sync_ops = {
 954         .rpc_call_prepare = nfs40_call_sync_prepare,
 955         .rpc_call_done = nfs40_call_sync_done,
 956 };
 957
 958 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 959                                    struct nfs_server *server,
 960                                    struct rpc_message *msg,
 961                                    struct nfs4_sequence_args *args,
 962                                    struct nfs4_sequence_res *res)
 963 {
 964         int ret;
 965         struct rpc_task *task;
 966         struct nfs_client *clp = server->nfs_client;
 967         struct nfs4_call_sync_data data = {
 968                 .seq_server = server,
 969                 .seq_args = args,
 970                 .seq_res = res,
 971         };
 972         struct rpc_task_setup task_setup = {
 973                 .rpc_client = clnt,
 974                 .rpc_message = msg,
 975                 .callback_ops = clp->cl_mvops->call_sync_ops,
 976                 .callback_data = &data
 977         };
 978
 979         task = rpc_run_task(&task_setup);
 980         if (IS_ERR(task))
 981                 ret = PTR_ERR(task);
 982         else {
 983                 ret = task->tk_status;
 984                 rpc_put_task(task);
 985         }
 986         return ret;
 987 }
 988
 989 int nfs4_call_sync(struct rpc_clnt *clnt,
 990                    struct nfs_server *server,
 991                    struct rpc_message *msg,
 992                    struct nfs4_sequence_args *args,
 993                    struct nfs4_sequence_res *res,
 994                    int cache_reply)
 995 {
 996         nfs4_init_sequence(args, res, cache_reply);
 997         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 998 }
 999
1000 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1001                 unsigned long timestamp)
1002 {
1003         struct nfs_inode *nfsi = NFS_I(dir);
1004
1005         spin_lock(&dir->i_lock);
1006         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1007         if (cinfo->atomic && cinfo->before == dir->i_version) {
1008                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1009                 nfsi->attrtimeo_timestamp = jiffies;
1010         } else {
1011                 nfs_force_lookup_revalidate(dir);
1012                 if (cinfo->before != dir->i_version)
1013                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1014                                 NFS_INO_INVALID_ACL;
1015         }
1016         dir->i_version = cinfo->after;
1017         nfsi->read_cache_jiffies = timestamp;
1018         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1019         nfs_fscache_invalidate(dir);
1020         spin_unlock(&dir->i_lock);
1021 }
1022
1023 struct nfs4_opendata {
1024         struct kref kref;
1025         struct nfs_openargs o_arg;
1026         struct nfs_openres o_res;
1027         struct nfs_open_confirmargs c_arg;
1028         struct nfs_open_confirmres c_res;
1029         struct nfs4_string owner_name;
1030         struct nfs4_string group_name;
1031         struct nfs4_label *a_label;
1032         struct nfs_fattr f_attr;
1033         struct nfs4_label *f_label;
1034         struct dentry *dir;
1035         struct dentry *dentry;
1036         struct nfs4_state_owner *owner;
1037         struct nfs4_state *state;
1038         struct iattr attrs;
1039         unsigned long timestamp;
1040         bool rpc_done;
1041         bool file_created;
1042         bool is_recover;
1043         bool cancelled;
1044         int rpc_status;
1045 };
1046
1047 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1048                 int err, struct nfs4_exception *exception)
1049 {
1050         if (err != -EINVAL)
1051                 return false;
1052         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1053                 return false;
1054         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1055         exception->retry = 1;
1056         return true;
1057 }
1058
1059 static u32
1060 nfs4_map_atomic_open_share(struct nfs_server *server,
1061                 fmode_t fmode, int openflags)
1062 {
1063         u32 res = 0;
1064
1065         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1066         case FMODE_READ:
1067                 res = NFS4_SHARE_ACCESS_READ;
1068                 break;
1069         case FMODE_WRITE:
1070                 res = NFS4_SHARE_ACCESS_WRITE;
1071                 break;
1072         case FMODE_READ|FMODE_WRITE:
1073                 res = NFS4_SHARE_ACCESS_BOTH;
1074         }
1075         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1076                 goto out;
1077         /* Want no delegation if we're using O_DIRECT */
1078         if (openflags & O_DIRECT)
1079                 res |= NFS4_SHARE_WANT_NO_DELEG;
1080 out:
1081         return res;
1082 }
1083
1084 static enum open_claim_type4
1085 nfs4_map_atomic_open_claim(struct nfs_server *server,
1086                 enum open_claim_type4 claim)
1087 {
1088         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1089                 return claim;
1090         switch (claim) {
1091         default:
1092                 return claim;
1093         case NFS4_OPEN_CLAIM_FH:
1094                 return NFS4_OPEN_CLAIM_NULL;
1095         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1096                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1097         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1098                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1099         }
1100 }
1101
1102 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1103 {
1104         p->o_res.f_attr = &p->f_attr;
1105         p->o_res.f_label = p->f_label;
1106         p->o_res.seqid = p->o_arg.seqid;
1107         p->c_res.seqid = p->c_arg.seqid;
1108         p->o_res.server = p->o_arg.server;
1109         p->o_res.access_request = p->o_arg.access;
1110         nfs_fattr_init(&p->f_attr);
1111         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1112 }
1113
1114 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1115                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1116                 const struct iattr *attrs,
1117                 struct nfs4_label *label,
1118                 enum open_claim_type4 claim,
1119                 gfp_t gfp_mask)
1120 {
1121         struct dentry *parent = dget_parent(dentry);
1122         struct inode *dir = d_inode(parent);
1123         struct nfs_server *server = NFS_SERVER(dir);
1124         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1125         struct nfs4_opendata *p;
1126
1127         p = kzalloc(sizeof(*p), gfp_mask);
1128         if (p == NULL)
1129                 goto err;
1130
1131         p->f_label = nfs4_label_alloc(server, gfp_mask);
1132         if (IS_ERR(p->f_label))
1133                 goto err_free_p;
1134
1135         p->a_label = nfs4_label_alloc(server, gfp_mask);
1136         if (IS_ERR(p->a_label))
1137                 goto err_free_f;
1138
1139         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1140         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1141         if (IS_ERR(p->o_arg.seqid))
1142                 goto err_free_label;
1143         nfs_sb_active(dentry->d_sb);
1144         p->dentry = dget(dentry);
1145         p->dir = parent;
1146         p->owner = sp;
1147         atomic_inc(&sp->so_count);
1148         p->o_arg.open_flags = flags;
1149         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1150         p->o_arg.umask = current_umask();
1151         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1152         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1153                         fmode, flags);
1154         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1155          * will return permission denied for all bits until close */
1156         if (!(flags & O_EXCL)) {
1157                 /* ask server to check for all possible rights as results
1158                  * are cached */
1159                 switch (p->o_arg.claim) {
1160                 default:
1161                         break;
1162                 case NFS4_OPEN_CLAIM_NULL:
1163                 case NFS4_OPEN_CLAIM_FH:
1164                         p->o_arg.access = NFS4_ACCESS_READ |
1165                                 NFS4_ACCESS_MODIFY |
1166                                 NFS4_ACCESS_EXTEND |
1167                                 NFS4_ACCESS_EXECUTE;
1168                 }
1169         }
1170         p->o_arg.clientid = server->nfs_client->cl_clientid;
1171         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1172         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1173         p->o_arg.name = &dentry->d_name;
1174         p->o_arg.server = server;
1175         p->o_arg.bitmask = nfs4_bitmask(server, label);
1176         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1177         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1178         switch (p->o_arg.claim) {
1179         case NFS4_OPEN_CLAIM_NULL:
1180         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1181         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1182                 p->o_arg.fh = NFS_FH(dir);
1183                 break;
1184         case NFS4_OPEN_CLAIM_PREVIOUS:
1185         case NFS4_OPEN_CLAIM_FH:
1186         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1187         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1188                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1189         }
1190         if (attrs != NULL && attrs->ia_valid != 0) {
1191                 __u32 verf[2];
1192
1193                 p->o_arg.u.attrs = &p->attrs;
1194                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1195
1196                 verf[0] = jiffies;
1197                 verf[1] = current->pid;
1198                 memcpy(p->o_arg.u.verifier.data, verf,
1199                                 sizeof(p->o_arg.u.verifier.data));
1200         }
1201         p->c_arg.fh = &p->o_res.fh;
1202         p->c_arg.stateid = &p->o_res.stateid;
1203         p->c_arg.seqid = p->o_arg.seqid;
1204         nfs4_init_opendata_res(p);
1205         kref_init(&p->kref);
1206         return p;
1207
1208 err_free_label:
1209         nfs4_label_free(p->a_label);
1210 err_free_f:
1211         nfs4_label_free(p->f_label);
1212 err_free_p:
1213         kfree(p);
1214 err:
1215         dput(parent);
1216         return NULL;
1217 }
1218
1219 static void nfs4_opendata_free(struct kref *kref)
1220 {
1221         struct nfs4_opendata *p = container_of(kref,
1222                         struct nfs4_opendata, kref);
1223         struct super_block *sb = p->dentry->d_sb;
1224
1225         nfs_free_seqid(p->o_arg.seqid);
1226         nfs4_sequence_free_slot(&p->o_res.seq_res);
1227         if (p->state != NULL)
1228                 nfs4_put_open_state(p->state);
1229         nfs4_put_state_owner(p->owner);
1230
1231         nfs4_label_free(p->a_label);
1232         nfs4_label_free(p->f_label);
1233
1234         dput(p->dir);
1235         dput(p->dentry);
1236         nfs_sb_deactive(sb);
1237         nfs_fattr_free_names(&p->f_attr);
1238         kfree(p->f_attr.mdsthreshold);
1239         kfree(p);
1240 }
1241
1242 static void nfs4_opendata_put(struct nfs4_opendata *p)
1243 {
1244         if (p != NULL)
1245                 kref_put(&p->kref, nfs4_opendata_free);
1246 }
1247
1248 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1249                 fmode_t fmode)
1250 {
1251         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1252         case FMODE_READ|FMODE_WRITE:
1253                 return state->n_rdwr != 0;
1254         case FMODE_WRITE:
1255                 return state->n_wronly != 0;
1256         case FMODE_READ:
1257                 return state->n_rdonly != 0;
1258         }
1259         WARN_ON_ONCE(1);
1260         return false;
1261 }
1262
1263 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1264 {
1265         int ret = 0;
1266
1267         if (open_mode & (O_EXCL|O_TRUNC))
1268                 goto out;
1269         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1270                 case FMODE_READ:
1271                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1272                                 && state->n_rdonly != 0;
1273                         break;
1274                 case FMODE_WRITE:
1275                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1276                                 && state->n_wronly != 0;
1277                         break;
1278                 case FMODE_READ|FMODE_WRITE:
1279                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1280                                 && state->n_rdwr != 0;
1281         }
1282 out:
1283         return ret;
1284 }
1285
1286 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1287                 enum open_claim_type4 claim)
1288 {
1289         if (delegation == NULL)
1290                 return 0;
1291         if ((delegation->type & fmode) != fmode)
1292                 return 0;
1293         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1294                 return 0;
1295         switch (claim) {
1296         case NFS4_OPEN_CLAIM_NULL:
1297         case NFS4_OPEN_CLAIM_FH:
1298                 break;
1299         case NFS4_OPEN_CLAIM_PREVIOUS:
1300                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1301                         break;
1302         default:
1303                 return 0;
1304         }
1305         nfs_mark_delegation_referenced(delegation);
1306         return 1;
1307 }
1308
1309 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1310 {
1311         switch (fmode) {
1312                 case FMODE_WRITE:
1313                         state->n_wronly++;
1314                         break;
1315                 case FMODE_READ:
1316                         state->n_rdonly++;
1317                         break;
1318                 case FMODE_READ|FMODE_WRITE:
1319                         state->n_rdwr++;
1320         }
1321         nfs4_state_set_mode_locked(state, state->state | fmode);
1322 }
1323
1324 #ifdef CONFIG_NFS_V4_1
1325 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1326 {
1327         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1328                 return true;
1329         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1330                 return true;
1331         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1332                 return true;
1333         return false;
1334 }
1335 #endif /* CONFIG_NFS_V4_1 */
1336
1337 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1338 {
1339         struct nfs_client *clp = state->owner->so_server->nfs_client;
1340         bool need_recover = false;
1341
1342         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1343                 need_recover = true;
1344         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1345                 need_recover = true;
1346         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1347                 need_recover = true;
1348         if (need_recover)
1349                 nfs4_state_mark_reclaim_nograce(clp, state);
1350 }
1351
1352 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1353                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1354 {
1355         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1356                 return true;
1357         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1358                 nfs4_stateid_copy(freeme, &state->open_stateid);
1359                 nfs_test_and_clear_all_open_stateid(state);
1360                 return true;
1361         }
1362         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1363                 return true;
1364         return false;
1365 }
1366
1367 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1368 {
1369         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1370                 return;
1371         if (state->n_wronly)
1372                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1373         if (state->n_rdonly)
1374                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1375         if (state->n_rdwr)
1376                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1377         set_bit(NFS_OPEN_STATE, &state->flags);
1378 }
1379
1380 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1381                 nfs4_stateid *stateid, fmode_t fmode)
1382 {
1383         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1384         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1385         case FMODE_WRITE:
1386                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1387                 break;
1388         case FMODE_READ:
1389                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1390                 break;
1391         case 0:
1392                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1393                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1394                 clear_bit(NFS_OPEN_STATE, &state->flags);
1395         }
1396         if (stateid == NULL)
1397                 return;
1398         /* Handle OPEN+OPEN_DOWNGRADE races */
1399         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1400             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1401                 nfs_resync_open_stateid_locked(state);
1402                 return;
1403         }
1404         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1405                 nfs4_stateid_copy(&state->stateid, stateid);
1406         nfs4_stateid_copy(&state->open_stateid, stateid);
1407 }
1408
1409 static void nfs_clear_open_stateid(struct nfs4_state *state,
1410         nfs4_stateid *arg_stateid,
1411         nfs4_stateid *stateid, fmode_t fmode)
1412 {
1413         write_seqlock(&state->seqlock);
1414         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1415         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1416                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1417         write_sequnlock(&state->seqlock);
1418         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1419                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1420 }
1421
1422 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1423                 const nfs4_stateid *stateid, fmode_t fmode,
1424                 nfs4_stateid *freeme)
1425 {
1426         switch (fmode) {
1427                 case FMODE_READ:
1428                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1429                         break;
1430                 case FMODE_WRITE:
1431                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1432                         break;
1433                 case FMODE_READ|FMODE_WRITE:
1434                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1435         }
1436         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1437                 return;
1438         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1439                 nfs4_stateid_copy(&state->stateid, stateid);
1440         nfs4_stateid_copy(&state->open_stateid, stateid);
1441 }
1442
1443 static void __update_open_stateid(struct nfs4_state *state,
1444                 const nfs4_stateid *open_stateid,
1445                 const nfs4_stateid *deleg_stateid,
1446                 fmode_t fmode,
1447                 nfs4_stateid *freeme)
1448 {
1449         /*
1450          * Protect the call to nfs4_state_set_mode_locked and
1451          * serialise the stateid update
1452          */
1453         spin_lock(&state->owner->so_lock);
1454         write_seqlock(&state->seqlock);
1455         if (deleg_stateid != NULL) {
1456                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1457                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1458         }
1459         if (open_stateid != NULL)
1460                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1461         write_sequnlock(&state->seqlock);
1462         update_open_stateflags(state, fmode);
1463         spin_unlock(&state->owner->so_lock);
1464 }
1465
1466 static int update_open_stateid(struct nfs4_state *state,
1467                 const nfs4_stateid *open_stateid,
1468                 const nfs4_stateid *delegation,
1469                 fmode_t fmode)
1470 {
1471         struct nfs_server *server = NFS_SERVER(state->inode);
1472         struct nfs_client *clp = server->nfs_client;
1473         struct nfs_inode *nfsi = NFS_I(state->inode);
1474         struct nfs_delegation *deleg_cur;
1475         nfs4_stateid freeme = { };
1476         int ret = 0;
1477
1478         fmode &= (FMODE_READ|FMODE_WRITE);
1479
1480         rcu_read_lock();
1481         deleg_cur = rcu_dereference(nfsi->delegation);
1482         if (deleg_cur == NULL)
1483                 goto no_delegation;
1484
1485         spin_lock(&deleg_cur->lock);
1486         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1487            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1488             (deleg_cur->type & fmode) != fmode)
1489                 goto no_delegation_unlock;
1490
1491         if (delegation == NULL)
1492                 delegation = &deleg_cur->stateid;
1493         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1494                 goto no_delegation_unlock;
1495
1496         nfs_mark_delegation_referenced(deleg_cur);
1497         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1498                         fmode, &freeme);
1499         ret = 1;
1500 no_delegation_unlock:
1501         spin_unlock(&deleg_cur->lock);
1502 no_delegation:
1503         rcu_read_unlock();
1504
1505         if (!ret && open_stateid != NULL) {
1506                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1507                 ret = 1;
1508         }
1509         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1510                 nfs4_schedule_state_manager(clp);
1511         if (freeme.type != 0)
1512                 nfs4_test_and_free_stateid(server, &freeme,
1513                                 state->owner->so_cred);
1514
1515         return ret;
1516 }
1517
1518 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1519                 const nfs4_stateid *stateid)
1520 {
1521         struct nfs4_state *state = lsp->ls_state;
1522         bool ret = false;
1523
1524         spin_lock(&state->state_lock);
1525         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1526                 goto out_noupdate;
1527         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1528                 goto out_noupdate;
1529         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1530         ret = true;
1531 out_noupdate:
1532         spin_unlock(&state->state_lock);
1533         return ret;
1534 }
1535
1536 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1537 {
1538         struct nfs_delegation *delegation;
1539
1540         rcu_read_lock();
1541         delegation = rcu_dereference(NFS_I(inode)->delegation);
1542         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1543                 rcu_read_unlock();
1544                 return;
1545         }
1546         rcu_read_unlock();
1547         nfs4_inode_return_delegation(inode);
1548 }
1549
1550 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1551 {
1552         struct nfs4_state *state = opendata->state;
1553         struct nfs_inode *nfsi = NFS_I(state->inode);
1554         struct nfs_delegation *delegation;
1555         int open_mode = opendata->o_arg.open_flags;
1556         fmode_t fmode = opendata->o_arg.fmode;
1557         enum open_claim_type4 claim = opendata->o_arg.claim;
1558         nfs4_stateid stateid;
1559         int ret = -EAGAIN;
1560
1561         for (;;) {
1562                 spin_lock(&state->owner->so_lock);
1563                 if (can_open_cached(state, fmode, open_mode)) {
1564                         update_open_stateflags(state, fmode);
1565                         spin_unlock(&state->owner->so_lock);
1566                         goto out_return_state;
1567                 }
1568                 spin_unlock(&state->owner->so_lock);
1569                 rcu_read_lock();
1570                 delegation = rcu_dereference(nfsi->delegation);
1571                 if (!can_open_delegated(delegation, fmode, claim)) {
1572                         rcu_read_unlock();
1573                         break;
1574                 }
1575                 /* Save the delegation */
1576                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1577                 rcu_read_unlock();
1578                 nfs_release_seqid(opendata->o_arg.seqid);
1579                 if (!opendata->is_recover) {
1580                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1581                         if (ret != 0)
1582                                 goto out;
1583                 }
1584                 ret = -EAGAIN;
1585
1586                 /* Try to update the stateid using the delegation */
1587                 if (update_open_stateid(state, NULL, &stateid, fmode))
1588                         goto out_return_state;
1589         }
1590 out:
1591         return ERR_PTR(ret);
1592 out_return_state:
1593         atomic_inc(&state->count);
1594         return state;
1595 }
1596
1597 static void
1598 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1599 {
1600         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1601         struct nfs_delegation *delegation;
1602         int delegation_flags = 0;
1603
1604         rcu_read_lock();
1605         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1606         if (delegation)
1607                 delegation_flags = delegation->flags;
1608         rcu_read_unlock();
1609         switch (data->o_arg.claim) {
1610         default:
1611                 break;
1612         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1613         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1614                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1615                                    "returning a delegation for "
1616                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1617                                    clp->cl_hostname);
1618                 return;
1619         }
1620         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1621                 nfs_inode_set_delegation(state->inode,
1622                                          data->owner->so_cred,
1623                                          &data->o_res);
1624         else
1625                 nfs_inode_reclaim_delegation(state->inode,
1626                                              data->owner->so_cred,
1627                                              &data->o_res);
1628 }
1629
1630 /*
1631  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1632  * and update the nfs4_state.
1633  */
1634 static struct nfs4_state *
1635 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1636 {
1637         struct inode *inode = data->state->inode;
1638         struct nfs4_state *state = data->state;
1639         int ret;
1640
1641         if (!data->rpc_done) {
1642                 if (data->rpc_status)
1643                         return ERR_PTR(data->rpc_status);
1644                 /* cached opens have already been processed */
1645                 goto update;
1646         }
1647
1648         ret = nfs_refresh_inode(inode, &data->f_attr);
1649         if (ret)
1650                 return ERR_PTR(ret);
1651
1652         if (data->o_res.delegation_type != 0)
1653                 nfs4_opendata_check_deleg(data, state);
1654 update:
1655         update_open_stateid(state, &data->o_res.stateid, NULL,
1656                             data->o_arg.fmode);
1657         atomic_inc(&state->count);
1658
1659         return state;
1660 }
1661
1662 static struct inode *
1663 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1664 {
1665         struct inode *inode;
1666
1667         switch (data->o_arg.claim) {
1668         case NFS4_OPEN_CLAIM_NULL:
1669         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1670         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1671                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1672                         return ERR_PTR(-EAGAIN);
1673                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1674                                 &data->f_attr, data->f_label);
1675                 break;
1676         default:
1677                 inode = d_inode(data->dentry);
1678                 ihold(inode);
1679                 nfs_refresh_inode(inode, &data->f_attr);
1680         }
1681         return inode;
1682 }
1683
1684 static struct nfs4_state *
1685 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1686 {
1687         struct nfs4_state *state;
1688         struct inode *inode;
1689
1690         inode = nfs4_opendata_get_inode(data);
1691         if (IS_ERR(inode))
1692                 return ERR_CAST(inode);
1693         if (data->state != NULL && data->state->inode == inode) {
1694                 state = data->state;
1695                 atomic_inc(&state->count);
1696         } else
1697                 state = nfs4_get_open_state(inode, data->owner);
1698         iput(inode);
1699         if (state == NULL)
1700                 state = ERR_PTR(-ENOMEM);
1701         return state;
1702 }
1703
1704 static struct nfs4_state *
1705 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1706 {
1707         struct nfs4_state *state;
1708
1709         if (!data->rpc_done) {
1710                 state = nfs4_try_open_cached(data);
1711                 trace_nfs4_cached_open(data->state);
1712                 goto out;
1713         }
1714
1715         state = nfs4_opendata_find_nfs4_state(data);
1716         if (IS_ERR(state))
1717                 goto out;
1718
1719         if (data->o_res.delegation_type != 0)
1720                 nfs4_opendata_check_deleg(data, state);
1721         update_open_stateid(state, &data->o_res.stateid, NULL,
1722                         data->o_arg.fmode);
1723 out:
1724         nfs_release_seqid(data->o_arg.seqid);
1725         return state;
1726 }
1727
1728 static struct nfs4_state *
1729 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1730 {
1731         struct nfs4_state *ret;
1732
1733         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1734                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1735         else
1736                 ret = _nfs4_opendata_to_nfs4_state(data);
1737         nfs4_sequence_free_slot(&data->o_res.seq_res);
1738         return ret;
1739 }
1740
1741 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1742 {
1743         struct nfs_inode *nfsi = NFS_I(state->inode);
1744         struct nfs_open_context *ctx;
1745
1746         spin_lock(&state->inode->i_lock);
1747         list_for_each_entry(ctx, &nfsi->open_files, list) {
1748                 if (ctx->state != state)
1749                         continue;
1750                 get_nfs_open_context(ctx);
1751                 spin_unlock(&state->inode->i_lock);
1752                 return ctx;
1753         }
1754         spin_unlock(&state->inode->i_lock);
1755         return ERR_PTR(-ENOENT);
1756 }
1757
1758 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1759                 struct nfs4_state *state, enum open_claim_type4 claim)
1760 {
1761         struct nfs4_opendata *opendata;
1762
1763         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1764                         NULL, NULL, claim, GFP_NOFS);
1765         if (opendata == NULL)
1766                 return ERR_PTR(-ENOMEM);
1767         opendata->state = state;
1768         atomic_inc(&state->count);
1769         return opendata;
1770 }
1771
1772 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1773                 fmode_t fmode)
1774 {
1775         struct nfs4_state *newstate;
1776         int ret;
1777
1778         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1779                 return 0;
1780         opendata->o_arg.open_flags = 0;
1781         opendata->o_arg.fmode = fmode;
1782         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1783                         NFS_SB(opendata->dentry->d_sb),
1784                         fmode, 0);
1785         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1786         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1787         nfs4_init_opendata_res(opendata);
1788         ret = _nfs4_recover_proc_open(opendata);
1789         if (ret != 0)
1790                 return ret;
1791         newstate = nfs4_opendata_to_nfs4_state(opendata);
1792         if (IS_ERR(newstate))
1793                 return PTR_ERR(newstate);
1794         if (newstate != opendata->state)
1795                 ret = -ESTALE;
1796         nfs4_close_state(newstate, fmode);
1797         return ret;
1798 }
1799
1800 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1801 {
1802         int ret;
1803
1804         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1805         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1806         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1807         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1808         /* memory barrier prior to reading state->n_* */
1809         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1810         clear_bit(NFS_OPEN_STATE, &state->flags);
1811         smp_rmb();
1812         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1813         if (ret != 0)
1814                 return ret;
1815         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1816         if (ret != 0)
1817                 return ret;
1818         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1819         if (ret != 0)
1820                 return ret;
1821         /*
1822          * We may have performed cached opens for all three recoveries.
1823          * Check if we need to update the current stateid.
1824          */
1825         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1826             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1827                 write_seqlock(&state->seqlock);
1828                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1829                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1830                 write_sequnlock(&state->seqlock);
1831         }
1832         return 0;
1833 }
1834
1835 /*
1836  * OPEN_RECLAIM:
1837  *      reclaim state on the server after a reboot.
1838  */
1839 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1840 {
1841         struct nfs_delegation *delegation;
1842         struct nfs4_opendata *opendata;
1843         fmode_t delegation_type = 0;
1844         int status;
1845
1846         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1847                         NFS4_OPEN_CLAIM_PREVIOUS);
1848         if (IS_ERR(opendata))
1849                 return PTR_ERR(opendata);
1850         rcu_read_lock();
1851         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1852         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1853                 delegation_type = delegation->type;
1854         rcu_read_unlock();
1855         opendata->o_arg.u.delegation_type = delegation_type;
1856         status = nfs4_open_recover(opendata, state);
1857         nfs4_opendata_put(opendata);
1858         return status;
1859 }
1860
1861 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1862 {
1863         struct nfs_server *server = NFS_SERVER(state->inode);
1864         struct nfs4_exception exception = { };
1865         int err;
1866         do {
1867                 err = _nfs4_do_open_reclaim(ctx, state);
1868                 trace_nfs4_open_reclaim(ctx, 0, err);
1869                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1870                         continue;
1871                 if (err != -NFS4ERR_DELAY)
1872                         break;
1873                 nfs4_handle_exception(server, err, &exception);
1874         } while (exception.retry);
1875         return err;
1876 }
1877
1878 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1879 {
1880         struct nfs_open_context *ctx;
1881         int ret;
1882
1883         ctx = nfs4_state_find_open_context(state);
1884         if (IS_ERR(ctx))
1885                 return -EAGAIN;
1886         ret = nfs4_do_open_reclaim(ctx, state);
1887         put_nfs_open_context(ctx);
1888         return ret;
1889 }
1890
1891 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1892 {
1893         switch (err) {
1894                 default:
1895                         printk(KERN_ERR "NFS: %s: unhandled error "
1896                                         "%d.\n", __func__, err);
1897                 case 0:
1898                 case -ENOENT:
1899                 case -EAGAIN:
1900                 case -ESTALE:
1901                         break;
1902                 case -NFS4ERR_BADSESSION:
1903                 case -NFS4ERR_BADSLOT:
1904                 case -NFS4ERR_BAD_HIGH_SLOT:
1905                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1906                 case -NFS4ERR_DEADSESSION:
1907                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1908                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1909                         return -EAGAIN;
1910                 case -NFS4ERR_STALE_CLIENTID:
1911                 case -NFS4ERR_STALE_STATEID:
1912                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1913                         /* Don't recall a delegation if it was lost */
1914                         nfs4_schedule_lease_recovery(server->nfs_client);
1915                         return -EAGAIN;
1916                 case -NFS4ERR_MOVED:
1917                         nfs4_schedule_migration_recovery(server);
1918                         return -EAGAIN;
1919                 case -NFS4ERR_LEASE_MOVED:
1920                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1921                         return -EAGAIN;
1922                 case -NFS4ERR_DELEG_REVOKED:
1923                 case -NFS4ERR_ADMIN_REVOKED:
1924                 case -NFS4ERR_EXPIRED:
1925                 case -NFS4ERR_BAD_STATEID:
1926                 case -NFS4ERR_OPENMODE:
1927                         nfs_inode_find_state_and_recover(state->inode,
1928                                         stateid);
1929                         nfs4_schedule_stateid_recovery(server, state);
1930                         return -EAGAIN;
1931                 case -NFS4ERR_DELAY:
1932                 case -NFS4ERR_GRACE:
1933                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1934                         ssleep(1);
1935                         return -EAGAIN;
1936                 case -ENOMEM:
1937                 case -NFS4ERR_DENIED:
1938                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1939                         return 0;
1940         }
1941         return err;
1942 }
1943
1944 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1945                 struct nfs4_state *state, const nfs4_stateid *stateid,
1946                 fmode_t type)
1947 {
1948         struct nfs_server *server = NFS_SERVER(state->inode);
1949         struct nfs4_opendata *opendata;
1950         int err = 0;
1951
1952         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1953                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1954         if (IS_ERR(opendata))
1955                 return PTR_ERR(opendata);
1956         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1957         write_seqlock(&state->seqlock);
1958         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1959         write_sequnlock(&state->seqlock);
1960         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1961         switch (type & (FMODE_READ|FMODE_WRITE)) {
1962         case FMODE_READ|FMODE_WRITE:
1963         case FMODE_WRITE:
1964                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1965                 if (err)
1966                         break;
1967                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1968                 if (err)
1969                         break;
1970         case FMODE_READ:
1971                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1972         }
1973         nfs4_opendata_put(opendata);
1974         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1975 }
1976
1977 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1978 {
1979         struct nfs4_opendata *data = calldata;
1980
1981         nfs4_setup_sequence(data->o_arg.server->nfs_client,
1982                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
1983 }
1984
1985 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1986 {
1987         struct nfs4_opendata *data = calldata;
1988
1989         nfs40_sequence_done(task, &data->c_res.seq_res);
1990
1991         data->rpc_status = task->tk_status;
1992         if (data->rpc_status == 0) {
1993                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1994                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1995                 renew_lease(data->o_res.server, data->timestamp);
1996                 data->rpc_done = true;
1997         }
1998 }
1999
2000 static void nfs4_open_confirm_release(void *calldata)
2001 {
2002         struct nfs4_opendata *data = calldata;
2003         struct nfs4_state *state = NULL;
2004
2005         /* If this request hasn't been cancelled, do nothing */
2006         if (!data->cancelled)
2007                 goto out_free;
2008         /* In case of error, no cleanup! */
2009         if (!data->rpc_done)
2010                 goto out_free;
2011         state = nfs4_opendata_to_nfs4_state(data);
2012         if (!IS_ERR(state))
2013                 nfs4_close_state(state, data->o_arg.fmode);
2014 out_free:
2015         nfs4_opendata_put(data);
2016 }
2017
2018 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2019         .rpc_call_prepare = nfs4_open_confirm_prepare,
2020         .rpc_call_done = nfs4_open_confirm_done,
2021         .rpc_release = nfs4_open_confirm_release,
2022 };
2023
2024 /*
2025  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2026  */
2027 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2028 {
2029         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2030         struct rpc_task *task;
2031         struct  rpc_message msg = {
2032                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2033                 .rpc_argp = &data->c_arg,
2034                 .rpc_resp = &data->c_res,
2035                 .rpc_cred = data->owner->so_cred,
2036         };
2037         struct rpc_task_setup task_setup_data = {
2038                 .rpc_client = server->client,
2039                 .rpc_message = &msg,
2040                 .callback_ops = &nfs4_open_confirm_ops,
2041                 .callback_data = data,
2042                 .workqueue = nfsiod_workqueue,
2043                 .flags = RPC_TASK_ASYNC,
2044         };
2045         int status;
2046
2047         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2048         kref_get(&data->kref);
2049         data->rpc_done = false;
2050         data->rpc_status = 0;
2051         data->timestamp = jiffies;
2052         if (data->is_recover)
2053                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2054         task = rpc_run_task(&task_setup_data);
2055         if (IS_ERR(task))
2056                 return PTR_ERR(task);
2057         status = rpc_wait_for_completion_task(task);
2058         if (status != 0) {
2059                 data->cancelled = true;
2060                 smp_wmb();
2061         } else
2062                 status = data->rpc_status;
2063         rpc_put_task(task);
2064         return status;
2065 }
2066
2067 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2068 {
2069         struct nfs4_opendata *data = calldata;
2070         struct nfs4_state_owner *sp = data->owner;
2071         struct nfs_client *clp = sp->so_server->nfs_client;
2072         enum open_claim_type4 claim = data->o_arg.claim;
2073
2074         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2075                 goto out_wait;
2076         /*
2077          * Check if we still need to send an OPEN call, or if we can use
2078          * a delegation instead.
2079          */
2080         if (data->state != NULL) {
2081                 struct nfs_delegation *delegation;
2082
2083                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2084                         goto out_no_action;
2085                 rcu_read_lock();
2086                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2087                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2088                         goto unlock_no_action;
2089                 rcu_read_unlock();
2090         }
2091         /* Update client id. */
2092         data->o_arg.clientid = clp->cl_clientid;
2093         switch (claim) {
2094         default:
2095                 break;
2096         case NFS4_OPEN_CLAIM_PREVIOUS:
2097         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2098         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2099                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2100         case NFS4_OPEN_CLAIM_FH:
2101                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2102         }
2103         data->timestamp = jiffies;
2104         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2105                                 &data->o_arg.seq_args,
2106                                 &data->o_res.seq_res,
2107                                 task) != 0)
2108                 nfs_release_seqid(data->o_arg.seqid);
2109
2110         /* Set the create mode (note dependency on the session type) */
2111         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2112         if (data->o_arg.open_flags & O_EXCL) {
2113                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2114                 if (nfs4_has_persistent_session(clp))
2115                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2116                 else if (clp->cl_mvops->minor_version > 0)
2117                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2118         }
2119         return;
2120 unlock_no_action:
2121         trace_nfs4_cached_open(data->state);
2122         rcu_read_unlock();
2123 out_no_action:
2124         task->tk_action = NULL;
2125 out_wait:
2126         nfs4_sequence_done(task, &data->o_res.seq_res);
2127 }
2128
2129 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2130 {
2131         struct nfs4_opendata *data = calldata;
2132
2133         data->rpc_status = task->tk_status;
2134
2135         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2136                 return;
2137
2138         if (task->tk_status == 0) {
2139                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2140                         switch (data->o_res.f_attr->mode & S_IFMT) {
2141                         case S_IFREG:
2142                                 break;
2143                         case S_IFLNK:
2144                                 data->rpc_status = -ELOOP;
2145                                 break;
2146                         case S_IFDIR:
2147                                 data->rpc_status = -EISDIR;
2148                                 break;
2149                         default:
2150                                 data->rpc_status = -ENOTDIR;
2151                         }
2152                 }
2153                 renew_lease(data->o_res.server, data->timestamp);
2154                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2155                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2156         }
2157         data->rpc_done = true;
2158 }
2159
2160 static void nfs4_open_release(void *calldata)
2161 {
2162         struct nfs4_opendata *data = calldata;
2163         struct nfs4_state *state = NULL;
2164
2165         /* If this request hasn't been cancelled, do nothing */
2166         if (!data->cancelled)
2167                 goto out_free;
2168         /* In case of error, no cleanup! */
2169         if (data->rpc_status != 0 || !data->rpc_done)
2170                 goto out_free;
2171         /* In case we need an open_confirm, no cleanup! */
2172         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2173                 goto out_free;
2174         state = nfs4_opendata_to_nfs4_state(data);
2175         if (!IS_ERR(state))
2176                 nfs4_close_state(state, data->o_arg.fmode);
2177 out_free:
2178         nfs4_opendata_put(data);
2179 }
2180
2181 static const struct rpc_call_ops nfs4_open_ops = {
2182         .rpc_call_prepare = nfs4_open_prepare,
2183         .rpc_call_done = nfs4_open_done,
2184         .rpc_release = nfs4_open_release,
2185 };
2186
2187 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2188 {
2189         struct inode *dir = d_inode(data->dir);
2190         struct nfs_server *server = NFS_SERVER(dir);
2191         struct nfs_openargs *o_arg = &data->o_arg;
2192         struct nfs_openres *o_res = &data->o_res;
2193         struct rpc_task *task;
2194         struct rpc_message msg = {
2195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2196                 .rpc_argp = o_arg,
2197                 .rpc_resp = o_res,
2198                 .rpc_cred = data->owner->so_cred,
2199         };
2200         struct rpc_task_setup task_setup_data = {
2201                 .rpc_client = server->client,
2202                 .rpc_message = &msg,
2203                 .callback_ops = &nfs4_open_ops,
2204                 .callback_data = data,
2205                 .workqueue = nfsiod_workqueue,
2206                 .flags = RPC_TASK_ASYNC,
2207         };
2208         int status;
2209
2210         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2211         kref_get(&data->kref);
2212         data->rpc_done = false;
2213         data->rpc_status = 0;
2214         data->cancelled = false;
2215         data->is_recover = false;
2216         if (isrecover) {
2217                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2218                 data->is_recover = true;
2219         }
2220         task = rpc_run_task(&task_setup_data);
2221         if (IS_ERR(task))
2222                 return PTR_ERR(task);
2223         status = rpc_wait_for_completion_task(task);
2224         if (status != 0) {
2225                 data->cancelled = true;
2226                 smp_wmb();
2227         } else
2228                 status = data->rpc_status;
2229         rpc_put_task(task);
2230
2231         return status;
2232 }
2233
2234 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2235 {
2236         struct inode *dir = d_inode(data->dir);
2237         struct nfs_openres *o_res = &data->o_res;
2238         int status;
2239
2240         status = nfs4_run_open_task(data, 1);
2241         if (status != 0 || !data->rpc_done)
2242                 return status;
2243
2244         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2245
2246         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2247                 status = _nfs4_proc_open_confirm(data);
2248
2249         return status;
2250 }
2251
2252 /*
2253  * Additional permission checks in order to distinguish between an
2254  * open for read, and an open for execute. This works around the
2255  * fact that NFSv4 OPEN treats read and execute permissions as being
2256  * the same.
2257  * Note that in the non-execute case, we want to turn off permission
2258  * checking if we just created a new file (POSIX open() semantics).
2259  */
2260 static int nfs4_opendata_access(struct rpc_cred *cred,
2261                                 struct nfs4_opendata *opendata,
2262                                 struct nfs4_state *state, fmode_t fmode,
2263                                 int openflags)
2264 {
2265         struct nfs_access_entry cache;
2266         u32 mask, flags;
2267
2268         /* access call failed or for some reason the server doesn't
2269          * support any access modes -- defer access call until later */
2270         if (opendata->o_res.access_supported == 0)
2271                 return 0;
2272
2273         mask = 0;
2274         /*
2275          * Use openflags to check for exec, because fmode won't
2276          * always have FMODE_EXEC set when file open for exec.
2277          */
2278         if (openflags & __FMODE_EXEC) {
2279                 /* ONLY check for exec rights */
2280                 if (S_ISDIR(state->inode->i_mode))
2281                         mask = NFS4_ACCESS_LOOKUP;
2282                 else
2283                         mask = NFS4_ACCESS_EXECUTE;
2284         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2285                 mask = NFS4_ACCESS_READ;
2286
2287         cache.cred = cred;
2288         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2289         nfs_access_add_cache(state->inode, &cache);
2290
2291         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2292         if ((mask & ~cache.mask & flags) == 0)
2293                 return 0;
2294
2295         return -EACCES;
2296 }
2297
2298 /*
2299  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2300  */
2301 static int _nfs4_proc_open(struct nfs4_opendata *data)
2302 {
2303         struct inode *dir = d_inode(data->dir);
2304         struct nfs_server *server = NFS_SERVER(dir);
2305         struct nfs_openargs *o_arg = &data->o_arg;
2306         struct nfs_openres *o_res = &data->o_res;
2307         int status;
2308
2309         status = nfs4_run_open_task(data, 0);
2310         if (!data->rpc_done)
2311                 return status;
2312         if (status != 0) {
2313                 if (status == -NFS4ERR_BADNAME &&
2314                                 !(o_arg->open_flags & O_CREAT))
2315                         return -ENOENT;
2316                 return status;
2317         }
2318
2319         nfs_fattr_map_and_free_names(server, &data->f_attr);
2320
2321         if (o_arg->open_flags & O_CREAT) {
2322                 if (o_arg->open_flags & O_EXCL)
2323                         data->file_created = true;
2324                 else if (o_res->cinfo.before != o_res->cinfo.after)
2325                         data->file_created = true;
2326                 if (data->file_created || dir->i_version != o_res->cinfo.after)
2327                         update_changeattr(dir, &o_res->cinfo,
2328                                         o_res->f_attr->time_start);
2329         }
2330         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2331                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2332         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2333                 status = _nfs4_proc_open_confirm(data);
2334                 if (status != 0)
2335                         return status;
2336         }
2337         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2338                 nfs4_sequence_free_slot(&o_res->seq_res);
2339                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2340         }
2341         return 0;
2342 }
2343
2344 /*
2345  * OPEN_EXPIRED:
2346  *      reclaim state on the server after a network partition.
2347  *      Assumes caller holds the appropriate lock
2348  */
2349 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2350 {
2351         struct nfs4_opendata *opendata;
2352         int ret;
2353
2354         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2355                         NFS4_OPEN_CLAIM_FH);
2356         if (IS_ERR(opendata))
2357                 return PTR_ERR(opendata);
2358         ret = nfs4_open_recover(opendata, state);
2359         if (ret == -ESTALE)
2360                 d_drop(ctx->dentry);
2361         nfs4_opendata_put(opendata);
2362         return ret;
2363 }
2364
2365 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2366 {
2367         struct nfs_server *server = NFS_SERVER(state->inode);
2368         struct nfs4_exception exception = { };
2369         int err;
2370
2371         do {
2372                 err = _nfs4_open_expired(ctx, state);
2373                 trace_nfs4_open_expired(ctx, 0, err);
2374                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2375                         continue;
2376                 switch (err) {
2377                 default:
2378                         goto out;
2379                 case -NFS4ERR_GRACE:
2380                 case -NFS4ERR_DELAY:
2381                         nfs4_handle_exception(server, err, &exception);
2382                         err = 0;
2383                 }
2384         } while (exception.retry);
2385 out:
2386         return err;
2387 }
2388
2389 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2390 {
2391         struct nfs_open_context *ctx;
2392         int ret;
2393
2394         ctx = nfs4_state_find_open_context(state);
2395         if (IS_ERR(ctx))
2396                 return -EAGAIN;
2397         ret = nfs4_do_open_expired(ctx, state);
2398         put_nfs_open_context(ctx);
2399         return ret;
2400 }
2401
2402 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2403                 const nfs4_stateid *stateid)
2404 {
2405         nfs_remove_bad_delegation(state->inode, stateid);
2406         write_seqlock(&state->seqlock);
2407         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2408         write_sequnlock(&state->seqlock);
2409         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2410 }
2411
2412 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2413 {
2414         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2415                 nfs_finish_clear_delegation_stateid(state, NULL);
2416 }
2417
2418 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2419 {
2420         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2421         nfs40_clear_delegation_stateid(state);
2422         return nfs4_open_expired(sp, state);
2423 }
2424
2425 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2426                 nfs4_stateid *stateid,
2427                 struct rpc_cred *cred)
2428 {
2429         return -NFS4ERR_BAD_STATEID;
2430 }
2431
2432 #if defined(CONFIG_NFS_V4_1)
2433 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2434                 nfs4_stateid *stateid,
2435                 struct rpc_cred *cred)
2436 {
2437         int status;
2438
2439         switch (stateid->type) {
2440         default:
2441                 break;
2442         case NFS4_INVALID_STATEID_TYPE:
2443         case NFS4_SPECIAL_STATEID_TYPE:
2444                 return -NFS4ERR_BAD_STATEID;
2445         case NFS4_REVOKED_STATEID_TYPE:
2446                 goto out_free;
2447         }
2448
2449         status = nfs41_test_stateid(server, stateid, cred);
2450         switch (status) {
2451         case -NFS4ERR_EXPIRED:
2452         case -NFS4ERR_ADMIN_REVOKED:
2453         case -NFS4ERR_DELEG_REVOKED:
2454                 break;
2455         default:
2456                 return status;
2457         }
2458 out_free:
2459         /* Ack the revoked state to the server */
2460         nfs41_free_stateid(server, stateid, cred, true);
2461         return -NFS4ERR_EXPIRED;
2462 }
2463
2464 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2465 {
2466         struct nfs_server *server = NFS_SERVER(state->inode);
2467         nfs4_stateid stateid;
2468         struct nfs_delegation *delegation;
2469         struct rpc_cred *cred;
2470         int status;
2471
2472         /* Get the delegation credential for use by test/free_stateid */
2473         rcu_read_lock();
2474         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2475         if (delegation == NULL) {
2476                 rcu_read_unlock();
2477                 return;
2478         }
2479
2480         nfs4_stateid_copy(&stateid, &delegation->stateid);
2481         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2482                 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2483                         &delegation->flags)) {
2484                 rcu_read_unlock();
2485                 nfs_finish_clear_delegation_stateid(state, &stateid);
2486                 return;
2487         }
2488
2489         cred = get_rpccred(delegation->cred);
2490         rcu_read_unlock();
2491         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2492         trace_nfs4_test_delegation_stateid(state, NULL, status);
2493         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2494                 nfs_finish_clear_delegation_stateid(state, &stateid);
2495
2496         put_rpccred(cred);
2497 }
2498
2499 /**
2500  * nfs41_check_expired_locks - possibly free a lock stateid
2501  *
2502  * @state: NFSv4 state for an inode
2503  *
2504  * Returns NFS_OK if recovery for this stateid is now finished.
2505  * Otherwise a negative NFS4ERR value is returned.
2506  */
2507 static int nfs41_check_expired_locks(struct nfs4_state *state)
2508 {
2509         int status, ret = NFS_OK;
2510         struct nfs4_lock_state *lsp, *prev = NULL;
2511         struct nfs_server *server = NFS_SERVER(state->inode);
2512
2513         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2514                 goto out;
2515
2516         spin_lock(&state->state_lock);
2517         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2518                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2519                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2520
2521                         atomic_inc(&lsp->ls_count);
2522                         spin_unlock(&state->state_lock);
2523
2524                         nfs4_put_lock_state(prev);
2525                         prev = lsp;
2526
2527                         status = nfs41_test_and_free_expired_stateid(server,
2528                                         &lsp->ls_stateid,
2529                                         cred);
2530                         trace_nfs4_test_lock_stateid(state, lsp, status);
2531                         if (status == -NFS4ERR_EXPIRED ||
2532                             status == -NFS4ERR_BAD_STATEID) {
2533                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2534                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2535                                 if (!recover_lost_locks)
2536                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2537                         } else if (status != NFS_OK) {
2538                                 ret = status;
2539                                 nfs4_put_lock_state(prev);
2540                                 goto out;
2541                         }
2542                         spin_lock(&state->state_lock);
2543                 }
2544         }
2545         spin_unlock(&state->state_lock);
2546         nfs4_put_lock_state(prev);
2547 out:
2548         return ret;
2549 }
2550
2551 /**
2552  * nfs41_check_open_stateid - possibly free an open stateid
2553  *
2554  * @state: NFSv4 state for an inode
2555  *
2556  * Returns NFS_OK if recovery for this stateid is now finished.
2557  * Otherwise a negative NFS4ERR value is returned.
2558  */
2559 static int nfs41_check_open_stateid(struct nfs4_state *state)
2560 {
2561         struct nfs_server *server = NFS_SERVER(state->inode);
2562         nfs4_stateid *stateid = &state->open_stateid;
2563         struct rpc_cred *cred = state->owner->so_cred;
2564         int status;
2565
2566         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2567                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2568                         if (nfs4_have_delegation(state->inode, state->state))
2569                                 return NFS_OK;
2570                         return -NFS4ERR_OPENMODE;
2571                 }
2572                 return -NFS4ERR_BAD_STATEID;
2573         }
2574         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2575         trace_nfs4_test_open_stateid(state, NULL, status);
2576         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2577                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2578                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2579                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2580                 clear_bit(NFS_OPEN_STATE, &state->flags);
2581                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2582                 return status;
2583         }
2584         if (nfs_open_stateid_recover_openmode(state))
2585                 return -NFS4ERR_OPENMODE;
2586         return NFS_OK;
2587 }
2588
2589 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2590 {
2591         int status;
2592
2593         nfs41_check_delegation_stateid(state);
2594         status = nfs41_check_expired_locks(state);
2595         if (status != NFS_OK)
2596                 return status;
2597         status = nfs41_check_open_stateid(state);
2598         if (status != NFS_OK)
2599                 status = nfs4_open_expired(sp, state);
2600         return status;
2601 }
2602 #endif
2603
2604 /*
2605  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2606  * fields corresponding to attributes that were used to store the verifier.
2607  * Make sure we clobber those fields in the later setattr call
2608  */
2609 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2610                                 struct iattr *sattr, struct nfs4_label **label)
2611 {
2612         const u32 *attrset = opendata->o_res.attrset;
2613
2614         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2615             !(sattr->ia_valid & ATTR_ATIME_SET))
2616                 sattr->ia_valid |= ATTR_ATIME;
2617
2618         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2619             !(sattr->ia_valid & ATTR_MTIME_SET))
2620                 sattr->ia_valid |= ATTR_MTIME;
2621
2622         /* Except MODE, it seems harmless of setting twice. */
2623         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2624                 (attrset[1] & FATTR4_WORD1_MODE ||
2625                  attrset[2] & FATTR4_WORD2_MODE_UMASK))
2626                 sattr->ia_valid &= ~ATTR_MODE;
2627
2628         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2629                 *label = NULL;
2630 }
2631
2632 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2633                 fmode_t fmode,
2634                 int flags,
2635                 struct nfs_open_context *ctx)
2636 {
2637         struct nfs4_state_owner *sp = opendata->owner;
2638         struct nfs_server *server = sp->so_server;
2639         struct dentry *dentry;
2640         struct nfs4_state *state;
2641         unsigned int seq;
2642         int ret;
2643
2644         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2645
2646         ret = _nfs4_proc_open(opendata);
2647         if (ret != 0)
2648                 goto out;
2649
2650         state = nfs4_opendata_to_nfs4_state(opendata);
2651         ret = PTR_ERR(state);
2652         if (IS_ERR(state))
2653                 goto out;
2654         ctx->state = state;
2655         if (server->caps & NFS_CAP_POSIX_LOCK)
2656                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2657         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2658                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2659
2660         dentry = opendata->dentry;
2661         if (d_really_is_negative(dentry)) {
2662                 struct dentry *alias;
2663                 d_drop(dentry);
2664                 alias = d_exact_alias(dentry, state->inode);
2665                 if (!alias)
2666                         alias = d_splice_alias(igrab(state->inode), dentry);
2667                 /* d_splice_alias() can't fail here - it's a non-directory */
2668                 if (alias) {
2669                         dput(ctx->dentry);
2670                         ctx->dentry = dentry = alias;
2671                 }
2672                 nfs_set_verifier(dentry,
2673                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2674         }
2675
2676         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2677         if (ret != 0)
2678                 goto out;
2679
2680         if (d_inode(dentry) == state->inode) {
2681                 nfs_inode_attach_open_context(ctx);
2682                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2683                         nfs4_schedule_stateid_recovery(server, state);
2684         }
2685 out:
2686         return ret;
2687 }
2688
2689 /*
2690  * Returns a referenced nfs4_state
2691  */
2692 static int _nfs4_do_open(struct inode *dir,
2693                         struct nfs_open_context *ctx,
2694                         int flags,
2695                         struct iattr *sattr,
2696                         struct nfs4_label *label,
2697                         int *opened)
2698 {
2699         struct nfs4_state_owner  *sp;
2700         struct nfs4_state     *state = NULL;
2701         struct nfs_server       *server = NFS_SERVER(dir);
2702         struct nfs4_opendata *opendata;
2703         struct dentry *dentry = ctx->dentry;
2704         struct rpc_cred *cred = ctx->cred;
2705         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2706         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2707         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2708         struct nfs4_label *olabel = NULL;
2709         int status;
2710
2711         /* Protect against reboot recovery conflicts */
2712         status = -ENOMEM;
2713         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2714         if (sp == NULL) {
2715                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2716                 goto out_err;
2717         }
2718         status = nfs4_client_recover_expired_lease(server->nfs_client);
2719         if (status != 0)
2720                 goto err_put_state_owner;
2721         if (d_really_is_positive(dentry))
2722                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2723         status = -ENOMEM;
2724         if (d_really_is_positive(dentry))
2725                 claim = NFS4_OPEN_CLAIM_FH;
2726         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2727                         label, claim, GFP_KERNEL);
2728         if (opendata == NULL)
2729                 goto err_put_state_owner;
2730
2731         if (label) {
2732                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2733                 if (IS_ERR(olabel)) {
2734                         status = PTR_ERR(olabel);
2735                         goto err_opendata_put;
2736                 }
2737         }
2738
2739         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2740                 if (!opendata->f_attr.mdsthreshold) {
2741                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2742                         if (!opendata->f_attr.mdsthreshold)
2743                                 goto err_free_label;
2744                 }
2745                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2746         }
2747         if (d_really_is_positive(dentry))
2748                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2749
2750         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2751         if (status != 0)
2752                 goto err_free_label;
2753         state = ctx->state;
2754
2755         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2756             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2757                 nfs4_exclusive_attrset(opendata, sattr, &label);
2758                 /*
2759                  * send create attributes which was not set by open
2760                  * with an extra setattr.
2761                  */
2762                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2763                         nfs_fattr_init(opendata->o_res.f_attr);
2764                         status = nfs4_do_setattr(state->inode, cred,
2765                                         opendata->o_res.f_attr, sattr,
2766                                         ctx, label, olabel);
2767                         if (status == 0) {
2768                                 nfs_setattr_update_inode(state->inode, sattr,
2769                                                 opendata->o_res.f_attr);
2770                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2771                         }
2772                 }
2773         }
2774         if (opened && opendata->file_created)
2775                 *opened |= FILE_CREATED;
2776
2777         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2778                 *ctx_th = opendata->f_attr.mdsthreshold;
2779                 opendata->f_attr.mdsthreshold = NULL;
2780         }
2781
2782         nfs4_label_free(olabel);
2783
2784         nfs4_opendata_put(opendata);
2785         nfs4_put_state_owner(sp);
2786         return 0;
2787 err_free_label:
2788         nfs4_label_free(olabel);
2789 err_opendata_put:
2790         nfs4_opendata_put(opendata);
2791 err_put_state_owner:
2792         nfs4_put_state_owner(sp);
2793 out_err:
2794         return status;
2795 }
2796
2797
2798 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2799                                         struct nfs_open_context *ctx,
2800                                         int flags,
2801                                         struct iattr *sattr,
2802                                         struct nfs4_label *label,
2803                                         int *opened)
2804 {
2805         struct nfs_server *server = NFS_SERVER(dir);
2806         struct nfs4_exception exception = { };
2807         struct nfs4_state *res;
2808         int status;
2809
2810         do {
2811                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2812                 res = ctx->state;
2813                 trace_nfs4_open_file(ctx, flags, status);
2814                 if (status == 0)
2815                         break;
2816                 /* NOTE: BAD_SEQID means the server and client disagree about the
2817                  * book-keeping w.r.t. state-changing operations
2818                  * (OPEN/CLOSE/LOCK/LOCKU...)
2819                  * It is actually a sign of a bug on the client or on the server.
2820                  *
2821                  * If we receive a BAD_SEQID error in the particular case of
2822                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2823                  * have unhashed the old state_owner for us, and that we can
2824                  * therefore safely retry using a new one. We should still warn
2825                  * the user though...
2826                  */
2827                 if (status == -NFS4ERR_BAD_SEQID) {
2828                         pr_warn_ratelimited("NFS: v4 server %s "
2829                                         " returned a bad sequence-id error!\n",
2830                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2831                         exception.retry = 1;
2832                         continue;
2833                 }
2834                 /*
2835                  * BAD_STATEID on OPEN means that the server cancelled our
2836                  * state before it received the OPEN_CONFIRM.
2837                  * Recover by retrying the request as per the discussion
2838                  * on Page 181 of RFC3530.
2839                  */
2840                 if (status == -NFS4ERR_BAD_STATEID) {
2841                         exception.retry = 1;
2842                         continue;
2843                 }
2844                 if (status == -EAGAIN) {
2845                         /* We must have found a delegation */
2846                         exception.retry = 1;
2847                         continue;
2848                 }
2849                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2850                         continue;
2851                 res = ERR_PTR(nfs4_handle_exception(server,
2852                                         status, &exception));
2853         } while (exception.retry);
2854         return res;
2855 }
2856
2857 static int _nfs4_do_setattr(struct inode *inode,
2858                             struct nfs_setattrargs *arg,
2859                             struct nfs_setattrres *res,
2860                             struct rpc_cred *cred,
2861                             struct nfs_open_context *ctx)
2862 {
2863         struct nfs_server *server = NFS_SERVER(inode);
2864         struct rpc_message msg = {
2865                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2866                 .rpc_argp       = arg,
2867                 .rpc_resp       = res,
2868                 .rpc_cred       = cred,
2869         };
2870         struct rpc_cred *delegation_cred = NULL;
2871         unsigned long timestamp = jiffies;
2872         fmode_t fmode;
2873         bool truncate;
2874         int status;
2875
2876         nfs_fattr_init(res->fattr);
2877
2878         /* Servers should only apply open mode checks for file size changes */
2879         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2880         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2881
2882         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2883                 /* Use that stateid */
2884         } else if (truncate && ctx != NULL) {
2885                 struct nfs_lock_context *l_ctx;
2886                 if (!nfs4_valid_open_stateid(ctx->state))
2887                         return -EBADF;
2888                 l_ctx = nfs_get_lock_context(ctx);
2889                 if (IS_ERR(l_ctx))
2890                         return PTR_ERR(l_ctx);
2891                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2892                                                 &arg->stateid, &delegation_cred);
2893                 nfs_put_lock_context(l_ctx);
2894                 if (status == -EIO)
2895                         return -EBADF;
2896         } else
2897                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2898         if (delegation_cred)
2899                 msg.rpc_cred = delegation_cred;
2900
2901         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2902
2903         put_rpccred(delegation_cred);
2904         if (status == 0 && ctx != NULL)
2905                 renew_lease(server, timestamp);
2906         trace_nfs4_setattr(inode, &arg->stateid, status);
2907         return status;
2908 }
2909
2910 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2911                            struct nfs_fattr *fattr, struct iattr *sattr,
2912                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2913                            struct nfs4_label *olabel)
2914 {
2915         struct nfs_server *server = NFS_SERVER(inode);
2916         struct nfs4_state *state = ctx ? ctx->state : NULL;
2917         struct nfs_setattrargs  arg = {
2918                 .fh             = NFS_FH(inode),
2919                 .iap            = sattr,
2920                 .server         = server,
2921                 .bitmask = server->attr_bitmask,
2922                 .label          = ilabel,
2923         };
2924         struct nfs_setattrres  res = {
2925                 .fattr          = fattr,
2926                 .label          = olabel,
2927                 .server         = server,
2928         };
2929         struct nfs4_exception exception = {
2930                 .state = state,
2931                 .inode = inode,
2932                 .stateid = &arg.stateid,
2933         };
2934         int err;
2935
2936         arg.bitmask = nfs4_bitmask(server, ilabel);
2937         if (ilabel)
2938                 arg.bitmask = nfs4_bitmask(server, olabel);
2939
2940         do {
2941                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
2942                 switch (err) {
2943                 case -NFS4ERR_OPENMODE:
2944                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2945                                 pr_warn_once("NFSv4: server %s is incorrectly "
2946                                                 "applying open mode checks to "
2947                                                 "a SETATTR that is not "
2948                                                 "changing file size.\n",
2949                                                 server->nfs_client->cl_hostname);
2950                         }
2951                         if (state && !(state->state & FMODE_WRITE)) {
2952                                 err = -EBADF;
2953                                 if (sattr->ia_valid & ATTR_OPEN)
2954                                         err = -EACCES;
2955                                 goto out;
2956                         }
2957                 }
2958                 err = nfs4_handle_exception(server, err, &exception);
2959         } while (exception.retry);
2960 out:
2961         return err;
2962 }
2963
2964 static bool
2965 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2966 {
2967         if (inode == NULL || !nfs_have_layout(inode))
2968                 return false;
2969
2970         return pnfs_wait_on_layoutreturn(inode, task);
2971 }
2972
2973 struct nfs4_closedata {
2974         struct inode *inode;
2975         struct nfs4_state *state;
2976         struct nfs_closeargs arg;
2977         struct nfs_closeres res;
2978         struct {
2979                 struct nfs4_layoutreturn_args arg;
2980                 struct nfs4_layoutreturn_res res;
2981                 struct nfs4_xdr_opaque_data ld_private;
2982                 u32 roc_barrier;
2983                 bool roc;
2984         } lr;
2985         struct nfs_fattr fattr;
2986         unsigned long timestamp;
2987 };
2988
2989 static void nfs4_free_closedata(void *data)
2990 {
2991         struct nfs4_closedata *calldata = data;
2992         struct nfs4_state_owner *sp = calldata->state->owner;
2993         struct super_block *sb = calldata->state->inode->i_sb;
2994
2995         if (calldata->lr.roc)
2996                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
2997                                 calldata->res.lr_ret);
2998         nfs4_put_open_state(calldata->state);
2999         nfs_free_seqid(calldata->arg.seqid);
3000         nfs4_put_state_owner(sp);
3001         nfs_sb_deactive(sb);
3002         kfree(calldata);
3003 }
3004
3005 static void nfs4_close_done(struct rpc_task *task, void *data)
3006 {
3007         struct nfs4_closedata *calldata = data;
3008         struct nfs4_state *state = calldata->state;
3009         struct nfs_server *server = NFS_SERVER(calldata->inode);
3010         nfs4_stateid *res_stateid = NULL;
3011
3012         dprintk("%s: begin!\n", __func__);
3013         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3014                 return;
3015         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3016
3017         /* Handle Layoutreturn errors */
3018         if (calldata->arg.lr_args && task->tk_status != 0) {
3019                 switch (calldata->res.lr_ret) {
3020                 default:
3021                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3022                         break;
3023                 case 0:
3024                         calldata->arg.lr_args = NULL;
3025                         calldata->res.lr_res = NULL;
3026                         break;
3027                 case -NFS4ERR_ADMIN_REVOKED:
3028                 case -NFS4ERR_DELEG_REVOKED:
3029                 case -NFS4ERR_EXPIRED:
3030                 case -NFS4ERR_BAD_STATEID:
3031                 case -NFS4ERR_OLD_STATEID:
3032                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3033                 case -NFS4ERR_WRONG_CRED:
3034                         calldata->arg.lr_args = NULL;
3035                         calldata->res.lr_res = NULL;
3036                         calldata->res.lr_ret = 0;
3037                         rpc_restart_call_prepare(task);
3038                         return;
3039                 }
3040         }
3041
3042         /* hmm. we are done with the inode, and in the process of freeing
3043          * the state_owner. we keep this around to process errors
3044          */
3045         switch (task->tk_status) {
3046                 case 0:
3047                         res_stateid = &calldata->res.stateid;
3048                         renew_lease(server, calldata->timestamp);
3049                         break;
3050                 case -NFS4ERR_ACCESS:
3051                         if (calldata->arg.bitmask != NULL) {
3052                                 calldata->arg.bitmask = NULL;
3053                                 calldata->res.fattr = NULL;
3054                                 task->tk_status = 0;
3055                                 rpc_restart_call_prepare(task);
3056                                 goto out_release;
3057
3058                         }
3059                         break;
3060                 case -NFS4ERR_ADMIN_REVOKED:
3061                 case -NFS4ERR_STALE_STATEID:
3062                 case -NFS4ERR_EXPIRED:
3063                         nfs4_free_revoked_stateid(server,
3064                                         &calldata->arg.stateid,
3065                                         task->tk_msg.rpc_cred);
3066                 case -NFS4ERR_OLD_STATEID:
3067                 case -NFS4ERR_BAD_STATEID:
3068                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3069                                                 &state->open_stateid)) {
3070                                 rpc_restart_call_prepare(task);
3071                                 goto out_release;
3072                         }
3073                         if (calldata->arg.fmode == 0)
3074                                 break;
3075                 default:
3076                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3077                                 rpc_restart_call_prepare(task);
3078                                 goto out_release;
3079                         }
3080         }
3081         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3082                         res_stateid, calldata->arg.fmode);
3083 out_release:
3084         nfs_release_seqid(calldata->arg.seqid);
3085         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3086         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3087 }
3088
3089 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3090 {
3091         struct nfs4_closedata *calldata = data;
3092         struct nfs4_state *state = calldata->state;
3093         struct inode *inode = calldata->inode;
3094         bool is_rdonly, is_wronly, is_rdwr;
3095         int call_close = 0;
3096
3097         dprintk("%s: begin!\n", __func__);
3098         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3099                 goto out_wait;
3100
3101         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3102         spin_lock(&state->owner->so_lock);
3103         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3104         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3105         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3106         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3107         /* Calculate the change in open mode */
3108         calldata->arg.fmode = 0;
3109         if (state->n_rdwr == 0) {
3110                 if (state->n_rdonly == 0)
3111                         call_close |= is_rdonly;
3112                 else if (is_rdonly)
3113                         calldata->arg.fmode |= FMODE_READ;
3114                 if (state->n_wronly == 0)
3115                         call_close |= is_wronly;
3116                 else if (is_wronly)
3117                         calldata->arg.fmode |= FMODE_WRITE;
3118                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3119                         call_close |= is_rdwr;
3120         } else if (is_rdwr)
3121                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3122
3123         if (!nfs4_valid_open_stateid(state) ||
3124             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3125                 call_close = 0;
3126         spin_unlock(&state->owner->so_lock);
3127
3128         if (!call_close) {
3129                 /* Note: exit _without_ calling nfs4_close_done */
3130                 goto out_no_action;
3131         }
3132
3133         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3134                 nfs_release_seqid(calldata->arg.seqid);
3135                 goto out_wait;
3136         }
3137
3138         if (calldata->arg.fmode == 0)
3139                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3140
3141         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3142                 /* Close-to-open cache consistency revalidation */
3143                 if (!nfs4_have_delegation(inode, FMODE_READ))
3144                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3145                 else
3146                         calldata->arg.bitmask = NULL;
3147         }
3148
3149         calldata->arg.share_access =
3150                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3151                                 calldata->arg.fmode, 0);
3152
3153         if (calldata->res.fattr == NULL)
3154                 calldata->arg.bitmask = NULL;
3155         else if (calldata->arg.bitmask == NULL)
3156                 calldata->res.fattr = NULL;
3157         calldata->timestamp = jiffies;
3158         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3159                                 &calldata->arg.seq_args,
3160                                 &calldata->res.seq_res,
3161                                 task) != 0)
3162                 nfs_release_seqid(calldata->arg.seqid);
3163         dprintk("%s: done!\n", __func__);
3164         return;
3165 out_no_action:
3166         task->tk_action = NULL;
3167 out_wait:
3168         nfs4_sequence_done(task, &calldata->res.seq_res);
3169 }
3170
3171 static const struct rpc_call_ops nfs4_close_ops = {
3172         .rpc_call_prepare = nfs4_close_prepare,
3173         .rpc_call_done = nfs4_close_done,
3174         .rpc_release = nfs4_free_closedata,
3175 };
3176
3177 /*
3178  * It is possible for data to be read/written from a mem-mapped file
3179  * after the sys_close call (which hits the vfs layer as a flush).
3180  * This means that we can't safely call nfsv4 close on a file until
3181  * the inode is cleared. This in turn means that we are not good
3182  * NFSv4 citizens - we do not indicate to the server to update the file's
3183  * share state even when we are done with one of the three share
3184  * stateid's in the inode.
3185  *
3186  * NOTE: Caller must be holding the sp->so_owner semaphore!
3187  */
3188 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3189 {
3190         struct nfs_server *server = NFS_SERVER(state->inode);
3191         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3192         struct nfs4_closedata *calldata;
3193         struct nfs4_state_owner *sp = state->owner;
3194         struct rpc_task *task;
3195         struct rpc_message msg = {
3196                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3197                 .rpc_cred = state->owner->so_cred,
3198         };
3199         struct rpc_task_setup task_setup_data = {
3200                 .rpc_client = server->client,
3201                 .rpc_message = &msg,
3202                 .callback_ops = &nfs4_close_ops,
3203                 .workqueue = nfsiod_workqueue,
3204                 .flags = RPC_TASK_ASYNC,
3205         };
3206         int status = -ENOMEM;
3207
3208         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3209                 &task_setup_data.rpc_client, &msg);
3210
3211         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3212         if (calldata == NULL)
3213                 goto out;
3214         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3215         calldata->inode = state->inode;
3216         calldata->state = state;
3217         calldata->arg.fh = NFS_FH(state->inode);
3218         /* Serialization for the sequence id */
3219         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3220         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3221         if (IS_ERR(calldata->arg.seqid))
3222                 goto out_free_calldata;
3223         nfs_fattr_init(&calldata->fattr);
3224         calldata->arg.fmode = 0;
3225         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3226         calldata->res.fattr = &calldata->fattr;
3227         calldata->res.seqid = calldata->arg.seqid;
3228         calldata->res.server = server;
3229         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3230         calldata->lr.roc = pnfs_roc(state->inode,
3231                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3232         if (calldata->lr.roc) {
3233                 calldata->arg.lr_args = &calldata->lr.arg;
3234                 calldata->res.lr_res = &calldata->lr.res;
3235         }
3236         nfs_sb_active(calldata->inode->i_sb);
3237
3238         msg.rpc_argp = &calldata->arg;
3239         msg.rpc_resp = &calldata->res;
3240         task_setup_data.callback_data = calldata;
3241         task = rpc_run_task(&task_setup_data);
3242         if (IS_ERR(task))
3243                 return PTR_ERR(task);
3244         status = 0;
3245         if (wait)
3246                 status = rpc_wait_for_completion_task(task);
3247         rpc_put_task(task);
3248         return status;
3249 out_free_calldata:
3250         kfree(calldata);
3251 out:
3252         nfs4_put_open_state(state);
3253         nfs4_put_state_owner(sp);
3254         return status;
3255 }
3256
3257 static struct inode *
3258 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3259                 int open_flags, struct iattr *attr, int *opened)
3260 {
3261         struct nfs4_state *state;
3262         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3263
3264         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3265
3266         /* Protect against concurrent sillydeletes */
3267         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3268
3269         nfs4_label_release_security(label);
3270
3271         if (IS_ERR(state))
3272                 return ERR_CAST(state);
3273         return state->inode;
3274 }
3275
3276 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3277 {
3278         if (ctx->state == NULL)
3279                 return;
3280         if (is_sync)
3281                 nfs4_close_sync(ctx->state, ctx->mode);
3282         else
3283                 nfs4_close_state(ctx->state, ctx->mode);
3284 }
3285
3286 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3287 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3288 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3289
3290 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3291 {
3292         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3293         struct nfs4_server_caps_arg args = {
3294                 .fhandle = fhandle,
3295                 .bitmask = bitmask,
3296         };
3297         struct nfs4_server_caps_res res = {};
3298         struct rpc_message msg = {
3299                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3300                 .rpc_argp = &args,
3301                 .rpc_resp = &res,
3302         };
3303         int status;
3304         int i;
3305
3306         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3307                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3308                      FATTR4_WORD0_LINK_SUPPORT |
3309                      FATTR4_WORD0_SYMLINK_SUPPORT |
3310                      FATTR4_WORD0_ACLSUPPORT;
3311         if (minorversion)
3312                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3313
3314         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3315         if (status == 0) {
3316                 /* Sanity check the server answers */
3317                 switch (minorversion) {
3318                 case 0:
3319                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3320                         res.attr_bitmask[2] = 0;
3321                         break;
3322                 case 1:
3323                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3324                         break;
3325                 case 2:
3326                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3327                 }
3328                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3329                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3330                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3331                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3332                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3333                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3334                                 NFS_CAP_SECURITY_LABEL);
3335                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3336                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3337                         server->caps |= NFS_CAP_ACLS;
3338                 if (res.has_links != 0)
3339                         server->caps |= NFS_CAP_HARDLINKS;
3340                 if (res.has_symlinks != 0)
3341                         server->caps |= NFS_CAP_SYMLINKS;
3342                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3343                         server->caps |= NFS_CAP_FILEID;
3344                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3345                         server->caps |= NFS_CAP_MODE;
3346                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3347                         server->caps |= NFS_CAP_NLINK;
3348                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3349                         server->caps |= NFS_CAP_OWNER;
3350                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3351                         server->caps |= NFS_CAP_OWNER_GROUP;
3352                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3353                         server->caps |= NFS_CAP_ATIME;
3354                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3355                         server->caps |= NFS_CAP_CTIME;
3356                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3357                         server->caps |= NFS_CAP_MTIME;
3358 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3359                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3360                         server->caps |= NFS_CAP_SECURITY_LABEL;
3361 #endif
3362                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3363                                 sizeof(server->attr_bitmask));
3364                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3365
3366                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3367                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3368                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3369                 server->cache_consistency_bitmask[2] = 0;
3370
3371                 /* Avoid a regression due to buggy server */
3372                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3373                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3374                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3375                         sizeof(server->exclcreat_bitmask));
3376
3377                 server->acl_bitmask = res.acl_bitmask;
3378                 server->fh_expire_type = res.fh_expire_type;
3379         }
3380
3381         return status;
3382 }
3383
3384 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3385 {
3386         struct nfs4_exception exception = { };
3387         int err;
3388         do {
3389                 err = nfs4_handle_exception(server,
3390                                 _nfs4_server_capabilities(server, fhandle),
3391                                 &exception);
3392         } while (exception.retry);
3393         return err;
3394 }
3395
3396 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3397                 struct nfs_fsinfo *info)
3398 {
3399         u32 bitmask[3];
3400         struct nfs4_lookup_root_arg args = {
3401                 .bitmask = bitmask,
3402         };
3403         struct nfs4_lookup_res res = {
3404                 .server = server,
3405                 .fattr = info->fattr,
3406                 .fh = fhandle,
3407         };
3408         struct rpc_message msg = {
3409                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3410                 .rpc_argp = &args,
3411                 .rpc_resp = &res,
3412         };
3413
3414         bitmask[0] = nfs4_fattr_bitmap[0];
3415         bitmask[1] = nfs4_fattr_bitmap[1];
3416         /*
3417          * Process the label in the upcoming getfattr
3418          */
3419         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3420
3421         nfs_fattr_init(info->fattr);
3422         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3423 }
3424
3425 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3426                 struct nfs_fsinfo *info)
3427 {
3428         struct nfs4_exception exception = { };
3429         int err;
3430         do {
3431                 err = _nfs4_lookup_root(server, fhandle, info);
3432                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3433                 switch (err) {
3434                 case 0:
3435                 case -NFS4ERR_WRONGSEC:
3436                         goto out;
3437                 default:
3438                         err = nfs4_handle_exception(server, err, &exception);
3439                 }
3440         } while (exception.retry);
3441 out:
3442         return err;
3443 }
3444
3445 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3446                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3447 {
3448         struct rpc_auth_create_args auth_args = {
3449                 .pseudoflavor = flavor,
3450         };
3451         struct rpc_auth *auth;
3452
3453         auth = rpcauth_create(&auth_args, server->client);
3454         if (IS_ERR(auth))
3455                 return -EACCES;
3456         return nfs4_lookup_root(server, fhandle, info);
3457 }
3458
3459 /*
3460  * Retry pseudoroot lookup with various security flavors.  We do this when:
3461  *
3462  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3463  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3464  *
3465  * Returns zero on success, or a negative NFS4ERR value, or a
3466  * negative errno value.
3467  */
3468 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3469                               struct nfs_fsinfo *info)
3470 {
3471         /* Per 3530bis 15.33.5 */
3472         static const rpc_authflavor_t flav_array[] = {
3473                 RPC_AUTH_GSS_KRB5P,
3474                 RPC_AUTH_GSS_KRB5I,
3475                 RPC_AUTH_GSS_KRB5,
3476                 RPC_AUTH_UNIX,                  /* courtesy */
3477                 RPC_AUTH_NULL,
3478         };
3479         int status = -EPERM;
3480         size_t i;
3481
3482         if (server->auth_info.flavor_len > 0) {
3483                 /* try each flavor specified by user */
3484                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3485                         status = nfs4_lookup_root_sec(server, fhandle, info,
3486                                                 server->auth_info.flavors[i]);
3487                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3488                                 continue;
3489                         break;
3490                 }
3491         } else {
3492                 /* no flavors specified by user, try default list */
3493                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3494                         status = nfs4_lookup_root_sec(server, fhandle, info,
3495                                                       flav_array[i]);
3496                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3497                                 continue;
3498                         break;
3499                 }
3500         }
3501
3502         /*
3503          * -EACCESS could mean that the user doesn't have correct permissions
3504          * to access the mount.  It could also mean that we tried to mount
3505          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3506          * existing mount programs don't handle -EACCES very well so it should
3507          * be mapped to -EPERM instead.
3508          */
3509         if (status == -EACCES)
3510                 status = -EPERM;
3511         return status;
3512 }
3513
3514 /**
3515  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3516  * @server: initialized nfs_server handle
3517  * @fhandle: we fill in the pseudo-fs root file handle
3518  * @info: we fill in an FSINFO struct
3519  * @auth_probe: probe the auth flavours
3520  *
3521  * Returns zero on success, or a negative errno.
3522  */
3523 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3524                          struct nfs_fsinfo *info,
3525                          bool auth_probe)
3526 {
3527         int status = 0;
3528
3529         if (!auth_probe)
3530                 status = nfs4_lookup_root(server, fhandle, info);
3531
3532         if (auth_probe || status == NFS4ERR_WRONGSEC)
3533                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3534                                 fhandle, info);
3535
3536         if (status == 0)
3537                 status = nfs4_server_capabilities(server, fhandle);
3538         if (status == 0)
3539                 status = nfs4_do_fsinfo(server, fhandle, info);
3540
3541         return nfs4_map_errors(status);
3542 }
3543
3544 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3545                               struct nfs_fsinfo *info)
3546 {
3547         int error;
3548         struct nfs_fattr *fattr = info->fattr;
3549         struct nfs4_label *label = NULL;
3550
3551         error = nfs4_server_capabilities(server, mntfh);
3552         if (error < 0) {
3553                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3554                 return error;
3555         }
3556
3557         label = nfs4_label_alloc(server, GFP_KERNEL);
3558         if (IS_ERR(label))
3559                 return PTR_ERR(label);
3560
3561         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3562         if (error < 0) {
3563                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3564                 goto err_free_label;
3565         }
3566
3567         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3568             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3569                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3570
3571 err_free_label:
3572         nfs4_label_free(label);
3573
3574         return error;
3575 }
3576
3577 /*
3578  * Get locations and (maybe) other attributes of a referral.
3579  * Note that we'll actually follow the referral later when
3580  * we detect fsid mismatch in inode revalidation
3581  */
3582 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3583                              const struct qstr *name, struct nfs_fattr *fattr,
3584                              struct nfs_fh *fhandle)
3585 {
3586         int status = -ENOMEM;
3587         struct page *page = NULL;
3588         struct nfs4_fs_locations *locations = NULL;
3589
3590         page = alloc_page(GFP_KERNEL);
3591         if (page == NULL)
3592                 goto out;
3593         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3594         if (locations == NULL)
3595                 goto out;
3596
3597         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3598         if (status != 0)
3599                 goto out;
3600
3601         /*
3602          * If the fsid didn't change, this is a migration event, not a
3603          * referral.  Cause us to drop into the exception handler, which
3604          * will kick off migration recovery.
3605          */
3606         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3607                 dprintk("%s: server did not return a different fsid for"
3608                         " a referral at %s\n", __func__, name->name);
3609                 status = -NFS4ERR_MOVED;
3610                 goto out;
3611         }
3612         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3613         nfs_fixup_referral_attributes(&locations->fattr);
3614
3615         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3616         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3617         memset(fhandle, 0, sizeof(struct nfs_fh));
3618 out:
3619         if (page)
3620                 __free_page(page);
3621         kfree(locations);
3622         return status;
3623 }
3624
3625 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3626                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3627 {
3628         struct nfs4_getattr_arg args = {
3629                 .fh = fhandle,
3630                 .bitmask = server->attr_bitmask,
3631         };
3632         struct nfs4_getattr_res res = {
3633                 .fattr = fattr,
3634                 .label = label,
3635                 .server = server,
3636         };
3637         struct rpc_message msg = {
3638                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3639                 .rpc_argp = &args,
3640                 .rpc_resp = &res,
3641         };
3642
3643         args.bitmask = nfs4_bitmask(server, label);
3644
3645         nfs_fattr_init(fattr);
3646         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3647 }
3648
3649 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3650                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3651 {
3652         struct nfs4_exception exception = { };
3653         int err;
3654         do {
3655                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3656                 trace_nfs4_getattr(server, fhandle, fattr, err);
3657                 err = nfs4_handle_exception(server, err,
3658                                 &exception);
3659         } while (exception.retry);
3660         return err;
3661 }
3662
3663 /*
3664  * The file is not closed if it is opened due to the a request to change
3665  * the size of the file. The open call will not be needed once the
3666  * VFS layer lookup-intents are implemented.
3667  *
3668  * Close is called when the inode is destroyed.
3669  * If we haven't opened the file for O_WRONLY, we
3670  * need to in the size_change case to obtain a stateid.
3671  *
3672  * Got race?
3673  * Because OPEN is always done by name in nfsv4, it is
3674  * possible that we opened a different file by the same
3675  * name.  We can recognize this race condition, but we
3676  * can't do anything about it besides returning an error.
3677  *
3678  * This will be fixed with VFS changes (lookup-intent).
3679  */
3680 static int
3681 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3682                   struct iattr *sattr)
3683 {
3684         struct inode *inode = d_inode(dentry);
3685         struct rpc_cred *cred = NULL;
3686         struct nfs_open_context *ctx = NULL;
3687         struct nfs4_label *label = NULL;
3688         int status;
3689
3690         if (pnfs_ld_layoutret_on_setattr(inode) &&
3691             sattr->ia_valid & ATTR_SIZE &&
3692             sattr->ia_size < i_size_read(inode))
3693                 pnfs_commit_and_return_layout(inode);
3694
3695         nfs_fattr_init(fattr);
3696
3697         /* Deal with open(O_TRUNC) */
3698         if (sattr->ia_valid & ATTR_OPEN)
3699                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3700
3701         /* Optimization: if the end result is no change, don't RPC */
3702         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3703                 return 0;
3704
3705         /* Search for an existing open(O_WRITE) file */
3706         if (sattr->ia_valid & ATTR_FILE) {
3707
3708                 ctx = nfs_file_open_context(sattr->ia_file);
3709                 if (ctx)
3710                         cred = ctx->cred;
3711         }
3712
3713         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3714         if (IS_ERR(label))
3715                 return PTR_ERR(label);
3716
3717         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3718         if (status == 0) {
3719                 nfs_setattr_update_inode(inode, sattr, fattr);
3720                 nfs_setsecurity(inode, fattr, label);
3721         }
3722         nfs4_label_free(label);
3723         return status;
3724 }
3725
3726 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3727                 const struct qstr *name, struct nfs_fh *fhandle,
3728                 struct nfs_fattr *fattr, struct nfs4_label *label)
3729 {
3730         struct nfs_server *server = NFS_SERVER(dir);
3731         int                    status;
3732         struct nfs4_lookup_arg args = {
3733                 .bitmask = server->attr_bitmask,
3734                 .dir_fh = NFS_FH(dir),
3735                 .name = name,
3736         };
3737         struct nfs4_lookup_res res = {
3738                 .server = server,
3739                 .fattr = fattr,
3740                 .label = label,
3741                 .fh = fhandle,
3742         };
3743         struct rpc_message msg = {
3744                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3745                 .rpc_argp = &args,
3746                 .rpc_resp = &res,
3747         };
3748
3749         args.bitmask = nfs4_bitmask(server, label);
3750
3751         nfs_fattr_init(fattr);
3752
3753         dprintk("NFS call  lookup %s\n", name->name);
3754         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3755         dprintk("NFS reply lookup: %d\n", status);
3756         return status;
3757 }
3758
3759 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3760 {
3761         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3762                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3763         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3764         fattr->nlink = 2;
3765 }
3766
3767 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3768                                    const struct qstr *name, struct nfs_fh *fhandle,
3769                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3770 {
3771         struct nfs4_exception exception = { };
3772         struct rpc_clnt *client = *clnt;
3773         int err;
3774         do {
3775                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3776                 trace_nfs4_lookup(dir, name, err);
3777                 switch (err) {
3778                 case -NFS4ERR_BADNAME:
3779                         err = -ENOENT;
3780                         goto out;
3781                 case -NFS4ERR_MOVED:
3782                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3783                         if (err == -NFS4ERR_MOVED)
3784                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3785                         goto out;
3786                 case -NFS4ERR_WRONGSEC:
3787                         err = -EPERM;
3788                         if (client != *clnt)
3789                                 goto out;
3790                         client = nfs4_negotiate_security(client, dir, name);
3791                         if (IS_ERR(client))
3792                                 return PTR_ERR(client);
3793
3794                         exception.retry = 1;
3795                         break;
3796                 default:
3797                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3798                 }
3799         } while (exception.retry);
3800
3801 out:
3802         if (err == 0)
3803                 *clnt = client;
3804         else if (client != *clnt)
3805                 rpc_shutdown_client(client);
3806
3807         return err;
3808 }
3809
3810 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3811                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3812                             struct nfs4_label *label)
3813 {
3814         int status;
3815         struct rpc_clnt *client = NFS_CLIENT(dir);
3816
3817         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3818         if (client != NFS_CLIENT(dir)) {
3819                 rpc_shutdown_client(client);
3820                 nfs_fixup_secinfo_attributes(fattr);
3821         }
3822         return status;
3823 }
3824
3825 struct rpc_clnt *
3826 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3827                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3828 {
3829         struct rpc_clnt *client = NFS_CLIENT(dir);
3830         int status;
3831
3832         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3833         if (status < 0)
3834                 return ERR_PTR(status);
3835         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3836 }
3837
3838 static int _nfs4_proc_lookupp(struct inode *inode,
3839                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3840                 struct nfs4_label *label)
3841 {
3842         struct rpc_clnt *clnt = NFS_CLIENT(inode);
3843         struct nfs_server *server = NFS_SERVER(inode);
3844         int                    status;
3845         struct nfs4_lookupp_arg args = {
3846                 .bitmask = server->attr_bitmask,
3847                 .fh = NFS_FH(inode),
3848         };
3849         struct nfs4_lookupp_res res = {
3850                 .server = server,
3851                 .fattr = fattr,
3852                 .label = label,
3853                 .fh = fhandle,
3854         };
3855         struct rpc_message msg = {
3856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
3857                 .rpc_argp = &args,
3858                 .rpc_resp = &res,
3859         };
3860
3861         args.bitmask = nfs4_bitmask(server, label);
3862
3863         nfs_fattr_init(fattr);
3864
3865         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
3866         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
3867                                 &res.seq_res, 0);
3868         dprintk("NFS reply lookupp: %d\n", status);
3869         return status;
3870 }
3871
3872 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
3873                              struct nfs_fattr *fattr, struct nfs4_label *label)
3874 {
3875         struct nfs4_exception exception = { };
3876         int err;
3877         do {
3878                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
3879                 trace_nfs4_lookupp(inode, err);
3880                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3881                                 &exception);
3882         } while (exception.retry);
3883         return err;
3884 }
3885
3886 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3887 {
3888         struct nfs_server *server = NFS_SERVER(inode);
3889         struct nfs4_accessargs args = {
3890                 .fh = NFS_FH(inode),
3891                 .bitmask = server->cache_consistency_bitmask,
3892         };
3893         struct nfs4_accessres res = {
3894                 .server = server,
3895         };
3896         struct rpc_message msg = {
3897                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3898                 .rpc_argp = &args,
3899                 .rpc_resp = &res,
3900                 .rpc_cred = entry->cred,
3901         };
3902         int mode = entry->mask;
3903         int status = 0;
3904
3905         /*
3906          * Determine which access bits we want to ask for...
3907          */
3908         if (mode & MAY_READ)
3909                 args.access |= NFS4_ACCESS_READ;
3910         if (S_ISDIR(inode->i_mode)) {
3911                 if (mode & MAY_WRITE)
3912                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3913                 if (mode & MAY_EXEC)
3914                         args.access |= NFS4_ACCESS_LOOKUP;
3915         } else {
3916                 if (mode & MAY_WRITE)
3917                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3918                 if (mode & MAY_EXEC)
3919                         args.access |= NFS4_ACCESS_EXECUTE;
3920         }
3921
3922         res.fattr = nfs_alloc_fattr();
3923         if (res.fattr == NULL)
3924                 return -ENOMEM;
3925
3926         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3927         if (!status) {
3928                 nfs_access_set_mask(entry, res.access);
3929                 nfs_refresh_inode(inode, res.fattr);
3930         }
3931         nfs_free_fattr(res.fattr);
3932         return status;
3933 }
3934
3935 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3936 {
3937         struct nfs4_exception exception = { };
3938         int err;
3939         do {
3940                 err = _nfs4_proc_access(inode, entry);
3941                 trace_nfs4_access(inode, err);
3942                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3943                                 &exception);
3944         } while (exception.retry);
3945         return err;
3946 }
3947
3948 /*
3949  * TODO: For the time being, we don't try to get any attributes
3950  * along with any of the zero-copy operations READ, READDIR,
3951  * READLINK, WRITE.
3952  *
3953  * In the case of the first three, we want to put the GETATTR
3954  * after the read-type operation -- this is because it is hard
3955  * to predict the length of a GETATTR response in v4, and thus
3956  * align the READ data correctly.  This means that the GETATTR
3957  * may end up partially falling into the page cache, and we should
3958  * shift it into the 'tail' of the xdr_buf before processing.
3959  * To do this efficiently, we need to know the total length
3960  * of data received, which doesn't seem to be available outside
3961  * of the RPC layer.
3962  *
3963  * In the case of WRITE, we also want to put the GETATTR after
3964  * the operation -- in this case because we want to make sure
3965  * we get the post-operation mtime and size.
3966  *
3967  * Both of these changes to the XDR layer would in fact be quite
3968  * minor, but I decided to leave them for a subsequent patch.
3969  */
3970 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3971                 unsigned int pgbase, unsigned int pglen)
3972 {
3973         struct nfs4_readlink args = {
3974                 .fh       = NFS_FH(inode),
3975                 .pgbase   = pgbase,
3976                 .pglen    = pglen,
3977                 .pages    = &page,
3978         };
3979         struct nfs4_readlink_res res;
3980         struct rpc_message msg = {
3981                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3982                 .rpc_argp = &args,
3983                 .rpc_resp = &res,
3984         };
3985
3986         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3987 }
3988
3989 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3990                 unsigned int pgbase, unsigned int pglen)
3991 {
3992         struct nfs4_exception exception = { };
3993         int err;
3994         do {
3995                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3996                 trace_nfs4_readlink(inode, err);
3997                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3998                                 &exception);
3999         } while (exception.retry);
4000         return err;
4001 }
4002
4003 /*
4004  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4005  */
4006 static int
4007 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4008                  int flags)
4009 {
4010         struct nfs_server *server = NFS_SERVER(dir);
4011         struct nfs4_label l, *ilabel = NULL;
4012         struct nfs_open_context *ctx;
4013         struct nfs4_state *state;
4014         int status = 0;
4015
4016         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4017         if (IS_ERR(ctx))
4018                 return PTR_ERR(ctx);
4019
4020         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4021
4022         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4023                 sattr->ia_mode &= ~current_umask();
4024         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4025         if (IS_ERR(state)) {
4026                 status = PTR_ERR(state);
4027                 goto out;
4028         }
4029 out:
4030         nfs4_label_release_security(ilabel);
4031         put_nfs_open_context(ctx);
4032         return status;
4033 }
4034
4035 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4036 {
4037         struct nfs_server *server = NFS_SERVER(dir);
4038         struct nfs_removeargs args = {
4039                 .fh = NFS_FH(dir),
4040                 .name = *name,
4041         };
4042         struct nfs_removeres res = {
4043                 .server = server,
4044         };
4045         struct rpc_message msg = {
4046                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4047                 .rpc_argp = &args,
4048                 .rpc_resp = &res,
4049         };
4050         unsigned long timestamp = jiffies;
4051         int status;
4052
4053         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4054         if (status == 0)
4055                 update_changeattr(dir, &res.cinfo, timestamp);
4056         return status;
4057 }
4058
4059 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4060 {
4061         struct nfs4_exception exception = { };
4062         int err;
4063         do {
4064                 err = _nfs4_proc_remove(dir, name);
4065                 trace_nfs4_remove(dir, name, err);
4066                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4067                                 &exception);
4068         } while (exception.retry);
4069         return err;
4070 }
4071
4072 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4073 {
4074         struct nfs_server *server = NFS_SERVER(dir);
4075         struct nfs_removeargs *args = msg->rpc_argp;
4076         struct nfs_removeres *res = msg->rpc_resp;
4077
4078         res->server = server;
4079         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4080         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4081
4082         nfs_fattr_init(res->dir_attr);
4083 }
4084
4085 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4086 {
4087         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4088                         &data->args.seq_args,
4089                         &data->res.seq_res,
4090                         task);
4091 }
4092
4093 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4094 {
4095         struct nfs_unlinkdata *data = task->tk_calldata;
4096         struct nfs_removeres *res = &data->res;
4097
4098         if (!nfs4_sequence_done(task, &res->seq_res))
4099                 return 0;
4100         if (nfs4_async_handle_error(task, res->server, NULL,
4101                                     &data->timeout) == -EAGAIN)
4102                 return 0;
4103         if (task->tk_status == 0)
4104                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4105         return 1;
4106 }
4107
4108 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4109 {
4110         struct nfs_server *server = NFS_SERVER(dir);
4111         struct nfs_renameargs *arg = msg->rpc_argp;
4112         struct nfs_renameres *res = msg->rpc_resp;
4113
4114         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4115         res->server = server;
4116         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4117 }
4118
4119 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4120 {
4121         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4122                         &data->args.seq_args,
4123                         &data->res.seq_res,
4124                         task);
4125 }
4126
4127 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4128                                  struct inode *new_dir)
4129 {
4130         struct nfs_renamedata *data = task->tk_calldata;
4131         struct nfs_renameres *res = &data->res;
4132
4133         if (!nfs4_sequence_done(task, &res->seq_res))
4134                 return 0;
4135         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4136                 return 0;
4137
4138         if (task->tk_status == 0) {
4139                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4140                 if (new_dir != old_dir)
4141                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4142         }
4143         return 1;
4144 }
4145
4146 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4147 {
4148         struct nfs_server *server = NFS_SERVER(inode);
4149         struct nfs4_link_arg arg = {
4150                 .fh     = NFS_FH(inode),
4151                 .dir_fh = NFS_FH(dir),
4152                 .name   = name,
4153                 .bitmask = server->attr_bitmask,
4154         };
4155         struct nfs4_link_res res = {
4156                 .server = server,
4157                 .label = NULL,
4158         };
4159         struct rpc_message msg = {
4160                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4161                 .rpc_argp = &arg,
4162                 .rpc_resp = &res,
4163         };
4164         int status = -ENOMEM;
4165
4166         res.fattr = nfs_alloc_fattr();
4167         if (res.fattr == NULL)
4168                 goto out;
4169
4170         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4171         if (IS_ERR(res.label)) {
4172                 status = PTR_ERR(res.label);
4173                 goto out;
4174         }
4175         arg.bitmask = nfs4_bitmask(server, res.label);
4176
4177         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4178         if (!status) {
4179                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4180                 status = nfs_post_op_update_inode(inode, res.fattr);
4181                 if (!status)
4182                         nfs_setsecurity(inode, res.fattr, res.label);
4183         }
4184
4185
4186         nfs4_label_free(res.label);
4187
4188 out:
4189         nfs_free_fattr(res.fattr);
4190         return status;
4191 }
4192
4193 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4194 {
4195         struct nfs4_exception exception = { };
4196         int err;
4197         do {
4198                 err = nfs4_handle_exception(NFS_SERVER(inode),
4199                                 _nfs4_proc_link(inode, dir, name),
4200                                 &exception);
4201         } while (exception.retry);
4202         return err;
4203 }
4204
4205 struct nfs4_createdata {
4206         struct rpc_message msg;
4207         struct nfs4_create_arg arg;
4208         struct nfs4_create_res res;
4209         struct nfs_fh fh;
4210         struct nfs_fattr fattr;
4211         struct nfs4_label *label;
4212 };
4213
4214 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4215                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4216 {
4217         struct nfs4_createdata *data;
4218
4219         data = kzalloc(sizeof(*data), GFP_KERNEL);
4220         if (data != NULL) {
4221                 struct nfs_server *server = NFS_SERVER(dir);
4222
4223                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4224                 if (IS_ERR(data->label))
4225                         goto out_free;
4226
4227                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4228                 data->msg.rpc_argp = &data->arg;
4229                 data->msg.rpc_resp = &data->res;
4230                 data->arg.dir_fh = NFS_FH(dir);
4231                 data->arg.server = server;
4232                 data->arg.name = name;
4233                 data->arg.attrs = sattr;
4234                 data->arg.ftype = ftype;
4235                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4236                 data->arg.umask = current_umask();
4237                 data->res.server = server;
4238                 data->res.fh = &data->fh;
4239                 data->res.fattr = &data->fattr;
4240                 data->res.label = data->label;
4241                 nfs_fattr_init(data->res.fattr);
4242         }
4243         return data;
4244 out_free:
4245         kfree(data);
4246         return NULL;
4247 }
4248
4249 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4250 {
4251         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4252                                     &data->arg.seq_args, &data->res.seq_res, 1);
4253         if (status == 0) {
4254                 update_changeattr(dir, &data->res.dir_cinfo,
4255                                 data->res.fattr->time_start);
4256                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4257         }
4258         return status;
4259 }
4260
4261 static void nfs4_free_createdata(struct nfs4_createdata *data)
4262 {
4263         nfs4_label_free(data->label);
4264         kfree(data);
4265 }
4266
4267 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4268                 struct page *page, unsigned int len, struct iattr *sattr,
4269                 struct nfs4_label *label)
4270 {
4271         struct nfs4_createdata *data;
4272         int status = -ENAMETOOLONG;
4273
4274         if (len > NFS4_MAXPATHLEN)
4275                 goto out;
4276
4277         status = -ENOMEM;
4278         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4279         if (data == NULL)
4280                 goto out;
4281
4282         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4283         data->arg.u.symlink.pages = &page;
4284         data->arg.u.symlink.len = len;
4285         data->arg.label = label;
4286
4287         status = nfs4_do_create(dir, dentry, data);
4288
4289         nfs4_free_createdata(data);
4290 out:
4291         return status;
4292 }
4293
4294 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4295                 struct page *page, unsigned int len, struct iattr *sattr)
4296 {
4297         struct nfs4_exception exception = { };
4298         struct nfs4_label l, *label = NULL;
4299         int err;
4300
4301         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4302
4303         do {
4304                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4305                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4306                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4307                                 &exception);
4308         } while (exception.retry);
4309
4310         nfs4_label_release_security(label);
4311         return err;
4312 }
4313
4314 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4315                 struct iattr *sattr, struct nfs4_label *label)
4316 {
4317         struct nfs4_createdata *data;
4318         int status = -ENOMEM;
4319
4320         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4321         if (data == NULL)
4322                 goto out;
4323
4324         data->arg.label = label;
4325         status = nfs4_do_create(dir, dentry, data);
4326
4327         nfs4_free_createdata(data);
4328 out:
4329         return status;
4330 }
4331
4332 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4333                 struct iattr *sattr)
4334 {
4335         struct nfs_server *server = NFS_SERVER(dir);
4336         struct nfs4_exception exception = { };
4337         struct nfs4_label l, *label = NULL;
4338         int err;
4339
4340         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4341
4342         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4343                 sattr->ia_mode &= ~current_umask();
4344         do {
4345                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4346                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4347                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4348                                 &exception);
4349         } while (exception.retry);
4350         nfs4_label_release_security(label);
4351
4352         return err;
4353 }
4354
4355 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4356                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4357 {
4358         struct inode            *dir = d_inode(dentry);
4359         struct nfs4_readdir_arg args = {
4360                 .fh = NFS_FH(dir),
4361                 .pages = pages,
4362                 .pgbase = 0,
4363                 .count = count,
4364                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4365                 .plus = plus,
4366         };
4367         struct nfs4_readdir_res res;
4368         struct rpc_message msg = {
4369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4370                 .rpc_argp = &args,
4371                 .rpc_resp = &res,
4372                 .rpc_cred = cred,
4373         };
4374         int                     status;
4375
4376         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4377                         dentry,
4378                         (unsigned long long)cookie);
4379         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4380         res.pgbase = args.pgbase;
4381         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4382         if (status >= 0) {
4383                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4384                 status += args.pgbase;
4385         }
4386
4387         nfs_invalidate_atime(dir);
4388
4389         dprintk("%s: returns %d\n", __func__, status);
4390         return status;
4391 }
4392
4393 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4394                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4395 {
4396         struct nfs4_exception exception = { };
4397         int err;
4398         do {
4399                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4400                                 pages, count, plus);
4401                 trace_nfs4_readdir(d_inode(dentry), err);
4402                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4403                                 &exception);
4404         } while (exception.retry);
4405         return err;
4406 }
4407
4408 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4409                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4410 {
4411         struct nfs4_createdata *data;
4412         int mode = sattr->ia_mode;
4413         int status = -ENOMEM;
4414
4415         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4416         if (data == NULL)
4417                 goto out;
4418
4419         if (S_ISFIFO(mode))
4420                 data->arg.ftype = NF4FIFO;
4421         else if (S_ISBLK(mode)) {
4422                 data->arg.ftype = NF4BLK;
4423                 data->arg.u.device.specdata1 = MAJOR(rdev);
4424                 data->arg.u.device.specdata2 = MINOR(rdev);
4425         }
4426         else if (S_ISCHR(mode)) {
4427                 data->arg.ftype = NF4CHR;
4428                 data->arg.u.device.specdata1 = MAJOR(rdev);
4429                 data->arg.u.device.specdata2 = MINOR(rdev);
4430         } else if (!S_ISSOCK(mode)) {
4431                 status = -EINVAL;
4432                 goto out_free;
4433         }
4434
4435         data->arg.label = label;
4436         status = nfs4_do_create(dir, dentry, data);
4437 out_free:
4438         nfs4_free_createdata(data);
4439 out:
4440         return status;
4441 }
4442
4443 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4444                 struct iattr *sattr, dev_t rdev)
4445 {
4446         struct nfs_server *server = NFS_SERVER(dir);
4447         struct nfs4_exception exception = { };
4448         struct nfs4_label l, *label = NULL;
4449         int err;
4450
4451         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4452
4453         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4454                 sattr->ia_mode &= ~current_umask();
4455         do {
4456                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4457                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4458                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4459                                 &exception);
4460         } while (exception.retry);
4461
4462         nfs4_label_release_security(label);
4463
4464         return err;
4465 }
4466
4467 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4468                  struct nfs_fsstat *fsstat)
4469 {
4470         struct nfs4_statfs_arg args = {
4471                 .fh = fhandle,
4472                 .bitmask = server->attr_bitmask,
4473         };
4474         struct nfs4_statfs_res res = {
4475                 .fsstat = fsstat,
4476         };
4477         struct rpc_message msg = {
4478                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4479                 .rpc_argp = &args,
4480                 .rpc_resp = &res,
4481         };
4482
4483         nfs_fattr_init(fsstat->fattr);
4484         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4485 }
4486
4487 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4488 {
4489         struct nfs4_exception exception = { };
4490         int err;
4491         do {
4492                 err = nfs4_handle_exception(server,
4493                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4494                                 &exception);
4495         } while (exception.retry);
4496         return err;
4497 }
4498
4499 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4500                 struct nfs_fsinfo *fsinfo)
4501 {
4502         struct nfs4_fsinfo_arg args = {
4503                 .fh = fhandle,
4504                 .bitmask = server->attr_bitmask,
4505         };
4506         struct nfs4_fsinfo_res res = {
4507                 .fsinfo = fsinfo,
4508         };
4509         struct rpc_message msg = {
4510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4511                 .rpc_argp = &args,
4512                 .rpc_resp = &res,
4513         };
4514
4515         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4516 }
4517
4518 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4519 {
4520         struct nfs4_exception exception = { };
4521         unsigned long now = jiffies;
4522         int err;
4523
4524         do {
4525                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4526                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4527                 if (err == 0) {
4528                         nfs4_set_lease_period(server->nfs_client,
4529                                         fsinfo->lease_time * HZ,
4530                                         now);
4531                         break;
4532                 }
4533                 err = nfs4_handle_exception(server, err, &exception);
4534         } while (exception.retry);
4535         return err;
4536 }
4537
4538 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4539 {
4540         int error;
4541
4542         nfs_fattr_init(fsinfo->fattr);
4543         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4544         if (error == 0) {
4545                 /* block layout checks this! */
4546                 server->pnfs_blksize = fsinfo->blksize;
4547                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4548         }
4549
4550         return error;
4551 }
4552
4553 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4554                 struct nfs_pathconf *pathconf)
4555 {
4556         struct nfs4_pathconf_arg args = {
4557                 .fh = fhandle,
4558                 .bitmask = server->attr_bitmask,
4559         };
4560         struct nfs4_pathconf_res res = {
4561                 .pathconf = pathconf,
4562         };
4563         struct rpc_message msg = {
4564                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4565                 .rpc_argp = &args,
4566                 .rpc_resp = &res,
4567         };
4568
4569         /* None of the pathconf attributes are mandatory to implement */
4570         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4571                 memset(pathconf, 0, sizeof(*pathconf));
4572                 return 0;
4573         }
4574
4575         nfs_fattr_init(pathconf->fattr);
4576         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4577 }
4578
4579 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4580                 struct nfs_pathconf *pathconf)
4581 {
4582         struct nfs4_exception exception = { };
4583         int err;
4584
4585         do {
4586                 err = nfs4_handle_exception(server,
4587                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4588                                 &exception);
4589         } while (exception.retry);
4590         return err;
4591 }
4592
4593 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4594                 const struct nfs_open_context *ctx,
4595                 const struct nfs_lock_context *l_ctx,
4596                 fmode_t fmode)
4597 {
4598         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4599 }
4600 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4601
4602 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4603                 const struct nfs_open_context *ctx,
4604                 const struct nfs_lock_context *l_ctx,
4605                 fmode_t fmode)
4606 {
4607         nfs4_stateid current_stateid;
4608
4609         /* If the current stateid represents a lost lock, then exit */
4610         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4611                 return true;
4612         return nfs4_stateid_match(stateid, &current_stateid);
4613 }
4614
4615 static bool nfs4_error_stateid_expired(int err)
4616 {
4617         switch (err) {
4618         case -NFS4ERR_DELEG_REVOKED:
4619         case -NFS4ERR_ADMIN_REVOKED:
4620         case -NFS4ERR_BAD_STATEID:
4621         case -NFS4ERR_STALE_STATEID:
4622         case -NFS4ERR_OLD_STATEID:
4623         case -NFS4ERR_OPENMODE:
4624         case -NFS4ERR_EXPIRED:
4625                 return true;
4626         }
4627         return false;
4628 }
4629
4630 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4631 {
4632         struct nfs_server *server = NFS_SERVER(hdr->inode);
4633
4634         trace_nfs4_read(hdr, task->tk_status);
4635         if (task->tk_status < 0) {
4636                 struct nfs4_exception exception = {
4637                         .inode = hdr->inode,
4638                         .state = hdr->args.context->state,
4639                         .stateid = &hdr->args.stateid,
4640                 };
4641                 task->tk_status = nfs4_async_handle_exception(task,
4642                                 server, task->tk_status, &exception);
4643                 if (exception.retry) {
4644                         rpc_restart_call_prepare(task);
4645                         return -EAGAIN;
4646                 }
4647         }
4648
4649         if (task->tk_status > 0)
4650                 renew_lease(server, hdr->timestamp);
4651         return 0;
4652 }
4653
4654 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4655                 struct nfs_pgio_args *args)
4656 {
4657
4658         if (!nfs4_error_stateid_expired(task->tk_status) ||
4659                 nfs4_stateid_is_current(&args->stateid,
4660                                 args->context,
4661                                 args->lock_context,
4662                                 FMODE_READ))
4663                 return false;
4664         rpc_restart_call_prepare(task);
4665         return true;
4666 }
4667
4668 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4669 {
4670
4671         dprintk("--> %s\n", __func__);
4672
4673         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4674                 return -EAGAIN;
4675         if (nfs4_read_stateid_changed(task, &hdr->args))
4676                 return -EAGAIN;
4677         if (task->tk_status > 0)
4678                 nfs_invalidate_atime(hdr->inode);
4679         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4680                                     nfs4_read_done_cb(task, hdr);
4681 }
4682
4683 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4684                                  struct rpc_message *msg)
4685 {
4686         hdr->timestamp   = jiffies;
4687         if (!hdr->pgio_done_cb)
4688                 hdr->pgio_done_cb = nfs4_read_done_cb;
4689         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4690         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4691 }
4692
4693 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4694                                       struct nfs_pgio_header *hdr)
4695 {
4696         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4697                         &hdr->args.seq_args,
4698                         &hdr->res.seq_res,
4699                         task))
4700                 return 0;
4701         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4702                                 hdr->args.lock_context,
4703                                 hdr->rw_mode) == -EIO)
4704                 return -EIO;
4705         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4706                 return -EIO;
4707         return 0;
4708 }
4709
4710 static int nfs4_write_done_cb(struct rpc_task *task,
4711                               struct nfs_pgio_header *hdr)
4712 {
4713         struct inode *inode = hdr->inode;
4714
4715         trace_nfs4_write(hdr, task->tk_status);
4716         if (task->tk_status < 0) {
4717                 struct nfs4_exception exception = {
4718                         .inode = hdr->inode,
4719                         .state = hdr->args.context->state,
4720                         .stateid = &hdr->args.stateid,
4721                 };
4722                 task->tk_status = nfs4_async_handle_exception(task,
4723                                 NFS_SERVER(inode), task->tk_status,
4724                                 &exception);
4725                 if (exception.retry) {
4726                         rpc_restart_call_prepare(task);
4727                         return -EAGAIN;
4728                 }
4729         }
4730         if (task->tk_status >= 0) {
4731                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4732                 nfs_writeback_update_inode(hdr);
4733         }
4734         return 0;
4735 }
4736
4737 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4738                 struct nfs_pgio_args *args)
4739 {
4740
4741         if (!nfs4_error_stateid_expired(task->tk_status) ||
4742                 nfs4_stateid_is_current(&args->stateid,
4743                                 args->context,
4744                                 args->lock_context,
4745                                 FMODE_WRITE))
4746                 return false;
4747         rpc_restart_call_prepare(task);
4748         return true;
4749 }
4750
4751 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4752 {
4753         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4754                 return -EAGAIN;
4755         if (nfs4_write_stateid_changed(task, &hdr->args))
4756                 return -EAGAIN;
4757         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4758                 nfs4_write_done_cb(task, hdr);
4759 }
4760
4761 static
4762 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4763 {
4764         /* Don't request attributes for pNFS or O_DIRECT writes */
4765         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4766                 return false;
4767         /* Otherwise, request attributes if and only if we don't hold
4768          * a delegation
4769          */
4770         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4771 }
4772
4773 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4774                                   struct rpc_message *msg)
4775 {
4776         struct nfs_server *server = NFS_SERVER(hdr->inode);
4777
4778         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4779                 hdr->args.bitmask = NULL;
4780                 hdr->res.fattr = NULL;
4781         } else
4782                 hdr->args.bitmask = server->cache_consistency_bitmask;
4783
4784         if (!hdr->pgio_done_cb)
4785                 hdr->pgio_done_cb = nfs4_write_done_cb;
4786         hdr->res.server = server;
4787         hdr->timestamp   = jiffies;
4788
4789         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4790         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4791 }
4792
4793 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4794 {
4795         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
4796                         &data->args.seq_args,
4797                         &data->res.seq_res,
4798                         task);
4799 }
4800
4801 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4802 {
4803         struct inode *inode = data->inode;
4804
4805         trace_nfs4_commit(data, task->tk_status);
4806         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4807                                     NULL, NULL) == -EAGAIN) {
4808                 rpc_restart_call_prepare(task);
4809                 return -EAGAIN;
4810         }
4811         return 0;
4812 }
4813
4814 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4815 {
4816         if (!nfs4_sequence_done(task, &data->res.seq_res))
4817                 return -EAGAIN;
4818         return data->commit_done_cb(task, data);
4819 }
4820
4821 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4822 {
4823         struct nfs_server *server = NFS_SERVER(data->inode);
4824
4825         if (data->commit_done_cb == NULL)
4826                 data->commit_done_cb = nfs4_commit_done_cb;
4827         data->res.server = server;
4828         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4829         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4830 }
4831
4832 struct nfs4_renewdata {
4833         struct nfs_client       *client;
4834         unsigned long           timestamp;
4835 };
4836
4837 /*
4838  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4839  * standalone procedure for queueing an asynchronous RENEW.
4840  */
4841 static void nfs4_renew_release(void *calldata)
4842 {
4843         struct nfs4_renewdata *data = calldata;
4844         struct nfs_client *clp = data->client;
4845
4846         if (atomic_read(&clp->cl_count) > 1)
4847                 nfs4_schedule_state_renewal(clp);
4848         nfs_put_client(clp);
4849         kfree(data);
4850 }
4851
4852 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4853 {
4854         struct nfs4_renewdata *data = calldata;
4855         struct nfs_client *clp = data->client;
4856         unsigned long timestamp = data->timestamp;
4857
4858         trace_nfs4_renew_async(clp, task->tk_status);
4859         switch (task->tk_status) {
4860         case 0:
4861                 break;
4862         case -NFS4ERR_LEASE_MOVED:
4863                 nfs4_schedule_lease_moved_recovery(clp);
4864                 break;
4865         default:
4866                 /* Unless we're shutting down, schedule state recovery! */
4867                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4868                         return;
4869                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4870                         nfs4_schedule_lease_recovery(clp);
4871                         return;
4872                 }
4873                 nfs4_schedule_path_down_recovery(clp);
4874         }
4875         do_renew_lease(clp, timestamp);
4876 }
4877
4878 static const struct rpc_call_ops nfs4_renew_ops = {
4879         .rpc_call_done = nfs4_renew_done,
4880         .rpc_release = nfs4_renew_release,
4881 };
4882
4883 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4884 {
4885         struct rpc_message msg = {
4886                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4887                 .rpc_argp       = clp,
4888                 .rpc_cred       = cred,
4889         };
4890         struct nfs4_renewdata *data;
4891
4892         if (renew_flags == 0)
4893                 return 0;
4894         if (!atomic_inc_not_zero(&clp->cl_count))
4895                 return -EIO;
4896         data = kmalloc(sizeof(*data), GFP_NOFS);
4897         if (data == NULL) {
4898                 nfs_put_client(clp);
4899                 return -ENOMEM;
4900         }
4901         data->client = clp;
4902         data->timestamp = jiffies;
4903         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4904                         &nfs4_renew_ops, data);
4905 }
4906
4907 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4908 {
4909         struct rpc_message msg = {
4910                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4911                 .rpc_argp       = clp,
4912                 .rpc_cred       = cred,
4913         };
4914         unsigned long now = jiffies;
4915         int status;
4916
4917         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4918         if (status < 0)
4919                 return status;
4920         do_renew_lease(clp, now);
4921         return 0;
4922 }
4923
4924 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4925 {
4926         return server->caps & NFS_CAP_ACLS;
4927 }
4928
4929 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4930  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4931  * the stack.
4932  */
4933 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4934
4935 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4936                 struct page **pages)
4937 {
4938         struct page *newpage, **spages;
4939         int rc = 0;
4940         size_t len;
4941         spages = pages;
4942
4943         do {
4944                 len = min_t(size_t, PAGE_SIZE, buflen);
4945                 newpage = alloc_page(GFP_KERNEL);
4946
4947                 if (newpage == NULL)
4948                         goto unwind;
4949                 memcpy(page_address(newpage), buf, len);
4950                 buf += len;
4951                 buflen -= len;
4952                 *pages++ = newpage;
4953                 rc++;
4954         } while (buflen != 0);
4955
4956         return rc;
4957
4958 unwind:
4959         for(; rc > 0; rc--)
4960                 __free_page(spages[rc-1]);
4961         return -ENOMEM;
4962 }
4963
4964 struct nfs4_cached_acl {
4965         int cached;
4966         size_t len;
4967         char data[0];
4968 };
4969
4970 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4971 {
4972         struct nfs_inode *nfsi = NFS_I(inode);
4973
4974         spin_lock(&inode->i_lock);
4975         kfree(nfsi->nfs4_acl);
4976         nfsi->nfs4_acl = acl;
4977         spin_unlock(&inode->i_lock);
4978 }
4979
4980 static void nfs4_zap_acl_attr(struct inode *inode)
4981 {
4982         nfs4_set_cached_acl(inode, NULL);
4983 }
4984
4985 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4986 {
4987         struct nfs_inode *nfsi = NFS_I(inode);
4988         struct nfs4_cached_acl *acl;
4989         int ret = -ENOENT;
4990
4991         spin_lock(&inode->i_lock);
4992         acl = nfsi->nfs4_acl;
4993         if (acl == NULL)
4994                 goto out;
4995         if (buf == NULL) /* user is just asking for length */
4996                 goto out_len;
4997         if (acl->cached == 0)
4998                 goto out;
4999         ret = -ERANGE; /* see getxattr(2) man page */
5000         if (acl->len > buflen)
5001                 goto out;
5002         memcpy(buf, acl->data, acl->len);
5003 out_len:
5004         ret = acl->len;
5005 out:
5006         spin_unlock(&inode->i_lock);
5007         return ret;
5008 }
5009
5010 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5011 {
5012         struct nfs4_cached_acl *acl;
5013         size_t buflen = sizeof(*acl) + acl_len;
5014
5015         if (buflen <= PAGE_SIZE) {
5016                 acl = kmalloc(buflen, GFP_KERNEL);
5017                 if (acl == NULL)
5018                         goto out;
5019                 acl->cached = 1;
5020                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5021         } else {
5022                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5023                 if (acl == NULL)
5024                         goto out;
5025                 acl->cached = 0;
5026         }
5027         acl->len = acl_len;
5028 out:
5029         nfs4_set_cached_acl(inode, acl);
5030 }
5031
5032 /*
5033  * The getxattr API returns the required buffer length when called with a
5034  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5035  * the required buf.  On a NULL buf, we send a page of data to the server
5036  * guessing that the ACL request can be serviced by a page. If so, we cache
5037  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5038  * the cache. If not so, we throw away the page, and cache the required
5039  * length. The next getxattr call will then produce another round trip to
5040  * the server, this time with the input buf of the required size.
5041  */
5042 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5043 {
5044         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5045         struct nfs_getaclargs args = {
5046                 .fh = NFS_FH(inode),
5047                 .acl_pages = pages,
5048                 .acl_len = buflen,
5049         };
5050         struct nfs_getaclres res = {
5051                 .acl_len = buflen,
5052         };
5053         struct rpc_message msg = {
5054                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5055                 .rpc_argp = &args,
5056                 .rpc_resp = &res,
5057         };
5058         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5059         int ret = -ENOMEM, i;
5060
5061         if (npages > ARRAY_SIZE(pages))
5062                 return -ERANGE;
5063
5064         for (i = 0; i < npages; i++) {
5065                 pages[i] = alloc_page(GFP_KERNEL);
5066                 if (!pages[i])
5067                         goto out_free;
5068         }
5069
5070         /* for decoding across pages */
5071         res.acl_scratch = alloc_page(GFP_KERNEL);
5072         if (!res.acl_scratch)
5073                 goto out_free;
5074
5075         args.acl_len = npages * PAGE_SIZE;
5076
5077         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5078                 __func__, buf, buflen, npages, args.acl_len);
5079         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5080                              &msg, &args.seq_args, &res.seq_res, 0);
5081         if (ret)
5082                 goto out_free;
5083
5084         /* Handle the case where the passed-in buffer is too short */
5085         if (res.acl_flags & NFS4_ACL_TRUNC) {
5086                 /* Did the user only issue a request for the acl length? */
5087                 if (buf == NULL)
5088                         goto out_ok;
5089                 ret = -ERANGE;
5090                 goto out_free;
5091         }
5092         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5093         if (buf) {
5094                 if (res.acl_len > buflen) {
5095                         ret = -ERANGE;
5096                         goto out_free;
5097                 }
5098                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5099         }
5100 out_ok:
5101         ret = res.acl_len;
5102 out_free:
5103         for (i = 0; i < npages; i++)
5104                 if (pages[i])
5105                         __free_page(pages[i]);
5106         if (res.acl_scratch)
5107                 __free_page(res.acl_scratch);
5108         return ret;
5109 }
5110
5111 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5112 {
5113         struct nfs4_exception exception = { };
5114         ssize_t ret;
5115         do {
5116                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5117                 trace_nfs4_get_acl(inode, ret);
5118                 if (ret >= 0)
5119                         break;
5120                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5121         } while (exception.retry);
5122         return ret;
5123 }
5124
5125 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5126 {
5127         struct nfs_server *server = NFS_SERVER(inode);
5128         int ret;
5129
5130         if (!nfs4_server_supports_acls(server))
5131                 return -EOPNOTSUPP;
5132         ret = nfs_revalidate_inode(server, inode);
5133         if (ret < 0)
5134                 return ret;
5135         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5136                 nfs_zap_acl_cache(inode);
5137         ret = nfs4_read_cached_acl(inode, buf, buflen);
5138         if (ret != -ENOENT)
5139                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5140                  * but no cached acl data, just the acl length */
5141                 return ret;
5142         return nfs4_get_acl_uncached(inode, buf, buflen);
5143 }
5144
5145 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5146 {
5147         struct nfs_server *server = NFS_SERVER(inode);
5148         struct page *pages[NFS4ACL_MAXPAGES];
5149         struct nfs_setaclargs arg = {
5150                 .fh             = NFS_FH(inode),
5151                 .acl_pages      = pages,
5152                 .acl_len        = buflen,
5153         };
5154         struct nfs_setaclres res;
5155         struct rpc_message msg = {
5156                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5157                 .rpc_argp       = &arg,
5158                 .rpc_resp       = &res,
5159         };
5160         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5161         int ret, i;
5162
5163         if (!nfs4_server_supports_acls(server))
5164                 return -EOPNOTSUPP;
5165         if (npages > ARRAY_SIZE(pages))
5166                 return -ERANGE;
5167         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5168         if (i < 0)
5169                 return i;
5170         nfs4_inode_return_delegation(inode);
5171         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5172
5173         /*
5174          * Free each page after tx, so the only ref left is
5175          * held by the network stack
5176          */
5177         for (; i > 0; i--)
5178                 put_page(pages[i-1]);
5179
5180         /*
5181          * Acl update can result in inode attribute update.
5182          * so mark the attribute cache invalid.
5183          */
5184         spin_lock(&inode->i_lock);
5185         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5186         spin_unlock(&inode->i_lock);
5187         nfs_access_zap_cache(inode);
5188         nfs_zap_acl_cache(inode);
5189         return ret;
5190 }
5191
5192 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5193 {
5194         struct nfs4_exception exception = { };
5195         int err;
5196         do {
5197                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5198                 trace_nfs4_set_acl(inode, err);
5199                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5200                                 &exception);
5201         } while (exception.retry);
5202         return err;
5203 }
5204
5205 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5206 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5207                                         size_t buflen)
5208 {
5209         struct nfs_server *server = NFS_SERVER(inode);
5210         struct nfs_fattr fattr;
5211         struct nfs4_label label = {0, 0, buflen, buf};
5212
5213         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5214         struct nfs4_getattr_arg arg = {
5215                 .fh             = NFS_FH(inode),
5216                 .bitmask        = bitmask,
5217         };
5218         struct nfs4_getattr_res res = {
5219                 .fattr          = &fattr,
5220                 .label          = &label,
5221                 .server         = server,
5222         };
5223         struct rpc_message msg = {
5224                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5225                 .rpc_argp       = &arg,
5226                 .rpc_resp       = &res,
5227         };
5228         int ret;
5229
5230         nfs_fattr_init(&fattr);
5231
5232         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5233         if (ret)
5234                 return ret;
5235         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5236                 return -ENOENT;
5237         if (buflen < label.len)
5238                 return -ERANGE;
5239         return 0;
5240 }
5241
5242 static int nfs4_get_security_label(struct inode *inode, void *buf,
5243                                         size_t buflen)
5244 {
5245         struct nfs4_exception exception = { };
5246         int err;
5247
5248         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5249                 return -EOPNOTSUPP;
5250
5251         do {
5252                 err = _nfs4_get_security_label(inode, buf, buflen);
5253                 trace_nfs4_get_security_label(inode, err);
5254                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5255                                 &exception);
5256         } while (exception.retry);
5257         return err;
5258 }
5259
5260 static int _nfs4_do_set_security_label(struct inode *inode,
5261                 struct nfs4_label *ilabel,
5262                 struct nfs_fattr *fattr,
5263                 struct nfs4_label *olabel)
5264 {
5265
5266         struct iattr sattr = {0};
5267         struct nfs_server *server = NFS_SERVER(inode);
5268         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5269         struct nfs_setattrargs arg = {
5270                 .fh             = NFS_FH(inode),
5271                 .iap            = &sattr,
5272                 .server         = server,
5273                 .bitmask        = bitmask,
5274                 .label          = ilabel,
5275         };
5276         struct nfs_setattrres res = {
5277                 .fattr          = fattr,
5278                 .label          = olabel,
5279                 .server         = server,
5280         };
5281         struct rpc_message msg = {
5282                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5283                 .rpc_argp       = &arg,
5284                 .rpc_resp       = &res,
5285         };
5286         int status;
5287
5288         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5289
5290         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5291         if (status)
5292                 dprintk("%s failed: %d\n", __func__, status);
5293
5294         return status;
5295 }
5296
5297 static int nfs4_do_set_security_label(struct inode *inode,
5298                 struct nfs4_label *ilabel,
5299                 struct nfs_fattr *fattr,
5300                 struct nfs4_label *olabel)
5301 {
5302         struct nfs4_exception exception = { };
5303         int err;
5304
5305         do {
5306                 err = _nfs4_do_set_security_label(inode, ilabel,
5307                                 fattr, olabel);
5308                 trace_nfs4_set_security_label(inode, err);
5309                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5310                                 &exception);
5311         } while (exception.retry);
5312         return err;
5313 }
5314
5315 static int
5316 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5317 {
5318         struct nfs4_label ilabel, *olabel = NULL;
5319         struct nfs_fattr fattr;
5320         struct rpc_cred *cred;
5321         int status;
5322
5323         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5324                 return -EOPNOTSUPP;
5325
5326         nfs_fattr_init(&fattr);
5327
5328         ilabel.pi = 0;
5329         ilabel.lfs = 0;
5330         ilabel.label = (char *)buf;
5331         ilabel.len = buflen;
5332
5333         cred = rpc_lookup_cred();
5334         if (IS_ERR(cred))
5335                 return PTR_ERR(cred);
5336
5337         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5338         if (IS_ERR(olabel)) {
5339                 status = -PTR_ERR(olabel);
5340                 goto out;
5341         }
5342
5343         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5344         if (status == 0)
5345                 nfs_setsecurity(inode, &fattr, olabel);
5346
5347         nfs4_label_free(olabel);
5348 out:
5349         put_rpccred(cred);
5350         return status;
5351 }
5352 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5353
5354
5355 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5356                                     nfs4_verifier *bootverf)
5357 {
5358         __be32 verf[2];
5359
5360         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5361                 /* An impossible timestamp guarantees this value
5362                  * will never match a generated boot time. */
5363                 verf[0] = cpu_to_be32(U32_MAX);
5364                 verf[1] = cpu_to_be32(U32_MAX);
5365         } else {
5366                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5367                 u64 ns = ktime_to_ns(nn->boot_time);
5368
5369                 verf[0] = cpu_to_be32(ns >> 32);
5370                 verf[1] = cpu_to_be32(ns);
5371         }
5372         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5373 }
5374
5375 static int
5376 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5377 {
5378         size_t len;
5379         char *str;
5380
5381         if (clp->cl_owner_id != NULL)
5382                 return 0;
5383
5384         rcu_read_lock();
5385         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5386                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5387                 1 +
5388                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5389                 1;
5390         rcu_read_unlock();
5391
5392         if (len > NFS4_OPAQUE_LIMIT + 1)
5393                 return -EINVAL;
5394
5395         /*
5396          * Since this string is allocated at mount time, and held until the
5397          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5398          * about a memory-reclaim deadlock.
5399          */
5400         str = kmalloc(len, GFP_KERNEL);
5401         if (!str)
5402                 return -ENOMEM;
5403
5404         rcu_read_lock();
5405         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5406                         clp->cl_ipaddr,
5407                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5408                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5409         rcu_read_unlock();
5410
5411         clp->cl_owner_id = str;
5412         return 0;
5413 }
5414
5415 static int
5416 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5417 {
5418         size_t len;
5419         char *str;
5420
5421         len = 10 + 10 + 1 + 10 + 1 +
5422                 strlen(nfs4_client_id_uniquifier) + 1 +
5423                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5424
5425         if (len > NFS4_OPAQUE_LIMIT + 1)
5426                 return -EINVAL;
5427
5428         /*
5429          * Since this string is allocated at mount time, and held until the
5430          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5431          * about a memory-reclaim deadlock.
5432          */
5433         str = kmalloc(len, GFP_KERNEL);
5434         if (!str)
5435                 return -ENOMEM;
5436
5437         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5438                         clp->rpc_ops->version, clp->cl_minorversion,
5439                         nfs4_client_id_uniquifier,
5440                         clp->cl_rpcclient->cl_nodename);
5441         clp->cl_owner_id = str;
5442         return 0;
5443 }
5444
5445 static int
5446 nfs4_init_uniform_client_string(struct nfs_client *clp)
5447 {
5448         size_t len;
5449         char *str;
5450
5451         if (clp->cl_owner_id != NULL)
5452                 return 0;
5453
5454         if (nfs4_client_id_uniquifier[0] != '\0')
5455                 return nfs4_init_uniquifier_client_string(clp);
5456
5457         len = 10 + 10 + 1 + 10 + 1 +
5458                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5459
5460         if (len > NFS4_OPAQUE_LIMIT + 1)
5461                 return -EINVAL;
5462
5463         /*
5464          * Since this string is allocated at mount time, and held until the
5465          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5466          * about a memory-reclaim deadlock.
5467          */
5468         str = kmalloc(len, GFP_KERNEL);
5469         if (!str)
5470                 return -ENOMEM;
5471
5472         scnprintf(str, len, "Linux NFSv%u.%u %s",
5473                         clp->rpc_ops->version, clp->cl_minorversion,
5474                         clp->cl_rpcclient->cl_nodename);
5475         clp->cl_owner_id = str;
5476         return 0;
5477 }
5478
5479 /*
5480  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5481  * services.  Advertise one based on the address family of the
5482  * clientaddr.
5483  */
5484 static unsigned int
5485 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5486 {
5487         if (strchr(clp->cl_ipaddr, ':') != NULL)
5488                 return scnprintf(buf, len, "tcp6");
5489         else
5490                 return scnprintf(buf, len, "tcp");
5491 }
5492
5493 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5494 {
5495         struct nfs4_setclientid *sc = calldata;
5496
5497         if (task->tk_status == 0)
5498                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5499 }
5500
5501 static const struct rpc_call_ops nfs4_setclientid_ops = {
5502         .rpc_call_done = nfs4_setclientid_done,
5503 };
5504
5505 /**
5506  * nfs4_proc_setclientid - Negotiate client ID
5507  * @clp: state data structure
5508  * @program: RPC program for NFSv4 callback service
5509  * @port: IP port number for NFS4 callback service
5510  * @cred: RPC credential to use for this call
5511  * @res: where to place the result
5512  *
5513  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5514  */
5515 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5516                 unsigned short port, struct rpc_cred *cred,
5517                 struct nfs4_setclientid_res *res)
5518 {
5519         nfs4_verifier sc_verifier;
5520         struct nfs4_setclientid setclientid = {
5521                 .sc_verifier = &sc_verifier,
5522                 .sc_prog = program,
5523                 .sc_clnt = clp,
5524         };
5525         struct rpc_message msg = {
5526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5527                 .rpc_argp = &setclientid,
5528                 .rpc_resp = res,
5529                 .rpc_cred = cred,
5530         };
5531         struct rpc_task *task;
5532         struct rpc_task_setup task_setup_data = {
5533                 .rpc_client = clp->cl_rpcclient,
5534                 .rpc_message = &msg,
5535                 .callback_ops = &nfs4_setclientid_ops,
5536                 .callback_data = &setclientid,
5537                 .flags = RPC_TASK_TIMEOUT,
5538         };
5539         int status;
5540
5541         /* nfs_client_id4 */
5542         nfs4_init_boot_verifier(clp, &sc_verifier);
5543
5544         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5545                 status = nfs4_init_uniform_client_string(clp);
5546         else
5547                 status = nfs4_init_nonuniform_client_string(clp);
5548
5549         if (status)
5550                 goto out;
5551
5552         /* cb_client4 */
5553         setclientid.sc_netid_len =
5554                                 nfs4_init_callback_netid(clp,
5555                                                 setclientid.sc_netid,
5556                                                 sizeof(setclientid.sc_netid));
5557         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5558                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5559                                 clp->cl_ipaddr, port >> 8, port & 255);
5560
5561         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5562                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5563                 clp->cl_owner_id);
5564         task = rpc_run_task(&task_setup_data);
5565         if (IS_ERR(task)) {
5566                 status = PTR_ERR(task);
5567                 goto out;
5568         }
5569         status = task->tk_status;
5570         if (setclientid.sc_cred) {
5571                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5572                 put_rpccred(setclientid.sc_cred);
5573         }
5574         rpc_put_task(task);
5575 out:
5576         trace_nfs4_setclientid(clp, status);
5577         dprintk("NFS reply setclientid: %d\n", status);
5578         return status;
5579 }
5580
5581 /**
5582  * nfs4_proc_setclientid_confirm - Confirm client ID
5583  * @clp: state data structure
5584  * @res: result of a previous SETCLIENTID
5585  * @cred: RPC credential to use for this call
5586  *
5587  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5588  */
5589 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5590                 struct nfs4_setclientid_res *arg,
5591                 struct rpc_cred *cred)
5592 {
5593         struct rpc_message msg = {
5594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5595                 .rpc_argp = arg,
5596                 .rpc_cred = cred,
5597         };
5598         int status;
5599
5600         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5601                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5602                 clp->cl_clientid);
5603         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5604         trace_nfs4_setclientid_confirm(clp, status);
5605         dprintk("NFS reply setclientid_confirm: %d\n", status);
5606         return status;
5607 }
5608
5609 struct nfs4_delegreturndata {
5610         struct nfs4_delegreturnargs args;
5611         struct nfs4_delegreturnres res;
5612         struct nfs_fh fh;
5613         nfs4_stateid stateid;
5614         unsigned long timestamp;
5615         struct {
5616                 struct nfs4_layoutreturn_args arg;
5617                 struct nfs4_layoutreturn_res res;
5618                 struct nfs4_xdr_opaque_data ld_private;
5619                 u32 roc_barrier;
5620                 bool roc;
5621         } lr;
5622         struct nfs_fattr fattr;
5623         int rpc_status;
5624         struct inode *inode;
5625 };
5626
5627 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5628 {
5629         struct nfs4_delegreturndata *data = calldata;
5630
5631         if (!nfs4_sequence_done(task, &data->res.seq_res))
5632                 return;
5633
5634         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5635
5636         /* Handle Layoutreturn errors */
5637         if (data->args.lr_args && task->tk_status != 0) {
5638                 switch(data->res.lr_ret) {
5639                 default:
5640                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5641                         break;
5642                 case 0:
5643                         data->args.lr_args = NULL;
5644                         data->res.lr_res = NULL;
5645                         break;
5646                 case -NFS4ERR_ADMIN_REVOKED:
5647                 case -NFS4ERR_DELEG_REVOKED:
5648                 case -NFS4ERR_EXPIRED:
5649                 case -NFS4ERR_BAD_STATEID:
5650                 case -NFS4ERR_OLD_STATEID:
5651                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5652                 case -NFS4ERR_WRONG_CRED:
5653                         data->args.lr_args = NULL;
5654                         data->res.lr_res = NULL;
5655                         data->res.lr_ret = 0;
5656                         rpc_restart_call_prepare(task);
5657                         return;
5658                 }
5659         }
5660
5661         switch (task->tk_status) {
5662         case 0:
5663                 renew_lease(data->res.server, data->timestamp);
5664                 break;
5665         case -NFS4ERR_ADMIN_REVOKED:
5666         case -NFS4ERR_DELEG_REVOKED:
5667         case -NFS4ERR_EXPIRED:
5668                 nfs4_free_revoked_stateid(data->res.server,
5669                                 data->args.stateid,
5670                                 task->tk_msg.rpc_cred);
5671         case -NFS4ERR_BAD_STATEID:
5672         case -NFS4ERR_OLD_STATEID:
5673         case -NFS4ERR_STALE_STATEID:
5674                 task->tk_status = 0;
5675                 break;
5676         case -NFS4ERR_ACCESS:
5677                 if (data->args.bitmask) {
5678                         data->args.bitmask = NULL;
5679                         data->res.fattr = NULL;
5680                         task->tk_status = 0;
5681                         rpc_restart_call_prepare(task);
5682                         return;
5683                 }
5684         default:
5685                 if (nfs4_async_handle_error(task, data->res.server,
5686                                             NULL, NULL) == -EAGAIN) {
5687                         rpc_restart_call_prepare(task);
5688                         return;
5689                 }
5690         }
5691         data->rpc_status = task->tk_status;
5692 }
5693
5694 static void nfs4_delegreturn_release(void *calldata)
5695 {
5696         struct nfs4_delegreturndata *data = calldata;
5697         struct inode *inode = data->inode;
5698
5699         if (inode) {
5700                 if (data->lr.roc)
5701                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5702                                         data->res.lr_ret);
5703                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5704                 nfs_iput_and_deactive(inode);
5705         }
5706         kfree(calldata);
5707 }
5708
5709 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5710 {
5711         struct nfs4_delegreturndata *d_data;
5712
5713         d_data = (struct nfs4_delegreturndata *)data;
5714
5715         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5716                 return;
5717
5718         nfs4_setup_sequence(d_data->res.server->nfs_client,
5719                         &d_data->args.seq_args,
5720                         &d_data->res.seq_res,
5721                         task);
5722 }
5723
5724 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5725         .rpc_call_prepare = nfs4_delegreturn_prepare,
5726         .rpc_call_done = nfs4_delegreturn_done,
5727         .rpc_release = nfs4_delegreturn_release,
5728 };
5729
5730 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5731 {
5732         struct nfs4_delegreturndata *data;
5733         struct nfs_server *server = NFS_SERVER(inode);
5734         struct rpc_task *task;
5735         struct rpc_message msg = {
5736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5737                 .rpc_cred = cred,
5738         };
5739         struct rpc_task_setup task_setup_data = {
5740                 .rpc_client = server->client,
5741                 .rpc_message = &msg,
5742                 .callback_ops = &nfs4_delegreturn_ops,
5743                 .flags = RPC_TASK_ASYNC,
5744         };
5745         int status = 0;
5746
5747         data = kzalloc(sizeof(*data), GFP_NOFS);
5748         if (data == NULL)
5749                 return -ENOMEM;
5750         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5751
5752         nfs4_state_protect(server->nfs_client,
5753                         NFS_SP4_MACH_CRED_CLEANUP,
5754                         &task_setup_data.rpc_client, &msg);
5755
5756         data->args.fhandle = &data->fh;
5757         data->args.stateid = &data->stateid;
5758         data->args.bitmask = server->cache_consistency_bitmask;
5759         nfs_copy_fh(&data->fh, NFS_FH(inode));
5760         nfs4_stateid_copy(&data->stateid, stateid);
5761         data->res.fattr = &data->fattr;
5762         data->res.server = server;
5763         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5764         data->lr.arg.ld_private = &data->lr.ld_private;
5765         nfs_fattr_init(data->res.fattr);
5766         data->timestamp = jiffies;
5767         data->rpc_status = 0;
5768         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5769         data->inode = nfs_igrab_and_active(inode);
5770         if (data->inode) {
5771                 if (data->lr.roc) {
5772                         data->args.lr_args = &data->lr.arg;
5773                         data->res.lr_res = &data->lr.res;
5774                 }
5775         } else if (data->lr.roc) {
5776                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5777                 data->lr.roc = false;
5778         }
5779
5780         task_setup_data.callback_data = data;
5781         msg.rpc_argp = &data->args;
5782         msg.rpc_resp = &data->res;
5783         task = rpc_run_task(&task_setup_data);
5784         if (IS_ERR(task))
5785                 return PTR_ERR(task);
5786         if (!issync)
5787                 goto out;
5788         status = rpc_wait_for_completion_task(task);
5789         if (status != 0)
5790                 goto out;
5791         status = data->rpc_status;
5792 out:
5793         rpc_put_task(task);
5794         return status;
5795 }
5796
5797 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5798 {
5799         struct nfs_server *server = NFS_SERVER(inode);
5800         struct nfs4_exception exception = { };
5801         int err;
5802         do {
5803                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5804                 trace_nfs4_delegreturn(inode, stateid, err);
5805                 switch (err) {
5806                         case -NFS4ERR_STALE_STATEID:
5807                         case -NFS4ERR_EXPIRED:
5808                         case 0:
5809                                 return 0;
5810                 }
5811                 err = nfs4_handle_exception(server, err, &exception);
5812         } while (exception.retry);
5813         return err;
5814 }
5815
5816 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5817 {
5818         struct inode *inode = state->inode;
5819         struct nfs_server *server = NFS_SERVER(inode);
5820         struct nfs_client *clp = server->nfs_client;
5821         struct nfs_lockt_args arg = {
5822                 .fh = NFS_FH(inode),
5823                 .fl = request,
5824         };
5825         struct nfs_lockt_res res = {
5826                 .denied = request,
5827         };
5828         struct rpc_message msg = {
5829                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5830                 .rpc_argp       = &arg,
5831                 .rpc_resp       = &res,
5832                 .rpc_cred       = state->owner->so_cred,
5833         };
5834         struct nfs4_lock_state *lsp;
5835         int status;
5836
5837         arg.lock_owner.clientid = clp->cl_clientid;
5838         status = nfs4_set_lock_state(state, request);
5839         if (status != 0)
5840                 goto out;
5841         lsp = request->fl_u.nfs4_fl.owner;
5842         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5843         arg.lock_owner.s_dev = server->s_dev;
5844         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5845         switch (status) {
5846                 case 0:
5847                         request->fl_type = F_UNLCK;
5848                         break;
5849                 case -NFS4ERR_DENIED:
5850                         status = 0;
5851         }
5852         request->fl_ops->fl_release_private(request);
5853         request->fl_ops = NULL;
5854 out:
5855         return status;
5856 }
5857
5858 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5859 {
5860         struct nfs4_exception exception = { };
5861         int err;
5862
5863         do {
5864                 err = _nfs4_proc_getlk(state, cmd, request);
5865                 trace_nfs4_get_lock(request, state, cmd, err);
5866                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5867                                 &exception);
5868         } while (exception.retry);
5869         return err;
5870 }
5871
5872 struct nfs4_unlockdata {
5873         struct nfs_locku_args arg;
5874         struct nfs_locku_res res;
5875         struct nfs4_lock_state *lsp;
5876         struct nfs_open_context *ctx;
5877         struct nfs_lock_context *l_ctx;
5878         struct file_lock fl;
5879         struct nfs_server *server;
5880         unsigned long timestamp;
5881 };
5882
5883 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5884                 struct nfs_open_context *ctx,
5885                 struct nfs4_lock_state *lsp,
5886                 struct nfs_seqid *seqid)
5887 {
5888         struct nfs4_unlockdata *p;
5889         struct inode *inode = lsp->ls_state->inode;
5890
5891         p = kzalloc(sizeof(*p), GFP_NOFS);
5892         if (p == NULL)
5893                 return NULL;
5894         p->arg.fh = NFS_FH(inode);
5895         p->arg.fl = &p->fl;
5896         p->arg.seqid = seqid;
5897         p->res.seqid = seqid;
5898         p->lsp = lsp;
5899         atomic_inc(&lsp->ls_count);
5900         /* Ensure we don't close file until we're done freeing locks! */
5901         p->ctx = get_nfs_open_context(ctx);
5902         p->l_ctx = nfs_get_lock_context(ctx);
5903         memcpy(&p->fl, fl, sizeof(p->fl));
5904         p->server = NFS_SERVER(inode);
5905         return p;
5906 }
5907
5908 static void nfs4_locku_release_calldata(void *data)
5909 {
5910         struct nfs4_unlockdata *calldata = data;
5911         nfs_free_seqid(calldata->arg.seqid);
5912         nfs4_put_lock_state(calldata->lsp);
5913         nfs_put_lock_context(calldata->l_ctx);
5914         put_nfs_open_context(calldata->ctx);
5915         kfree(calldata);
5916 }
5917
5918 static void nfs4_locku_done(struct rpc_task *task, void *data)
5919 {
5920         struct nfs4_unlockdata *calldata = data;
5921
5922         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5923                 return;
5924         switch (task->tk_status) {
5925                 case 0:
5926                         renew_lease(calldata->server, calldata->timestamp);
5927                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5928                         if (nfs4_update_lock_stateid(calldata->lsp,
5929                                         &calldata->res.stateid))
5930                                 break;
5931                 case -NFS4ERR_ADMIN_REVOKED:
5932                 case -NFS4ERR_EXPIRED:
5933                         nfs4_free_revoked_stateid(calldata->server,
5934                                         &calldata->arg.stateid,
5935                                         task->tk_msg.rpc_cred);
5936                 case -NFS4ERR_BAD_STATEID:
5937                 case -NFS4ERR_OLD_STATEID:
5938                 case -NFS4ERR_STALE_STATEID:
5939                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5940                                                 &calldata->lsp->ls_stateid))
5941                                 rpc_restart_call_prepare(task);
5942                         break;
5943                 default:
5944                         if (nfs4_async_handle_error(task, calldata->server,
5945                                                     NULL, NULL) == -EAGAIN)
5946                                 rpc_restart_call_prepare(task);
5947         }
5948         nfs_release_seqid(calldata->arg.seqid);
5949 }
5950
5951 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5952 {
5953         struct nfs4_unlockdata *calldata = data;
5954
5955         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
5956                 nfs_async_iocounter_wait(task, calldata->l_ctx))
5957                 return;
5958
5959         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5960                 goto out_wait;
5961         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5962         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5963                 /* Note: exit _without_ running nfs4_locku_done */
5964                 goto out_no_action;
5965         }
5966         calldata->timestamp = jiffies;
5967         if (nfs4_setup_sequence(calldata->server->nfs_client,
5968                                 &calldata->arg.seq_args,
5969                                 &calldata->res.seq_res,
5970                                 task) != 0)
5971                 nfs_release_seqid(calldata->arg.seqid);
5972         return;
5973 out_no_action:
5974         task->tk_action = NULL;
5975 out_wait:
5976         nfs4_sequence_done(task, &calldata->res.seq_res);
5977 }
5978
5979 static const struct rpc_call_ops nfs4_locku_ops = {
5980         .rpc_call_prepare = nfs4_locku_prepare,
5981         .rpc_call_done = nfs4_locku_done,
5982         .rpc_release = nfs4_locku_release_calldata,
5983 };
5984
5985 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5986                 struct nfs_open_context *ctx,
5987                 struct nfs4_lock_state *lsp,
5988                 struct nfs_seqid *seqid)
5989 {
5990         struct nfs4_unlockdata *data;
5991         struct rpc_message msg = {
5992                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5993                 .rpc_cred = ctx->cred,
5994         };
5995         struct rpc_task_setup task_setup_data = {
5996                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5997                 .rpc_message = &msg,
5998                 .callback_ops = &nfs4_locku_ops,
5999                 .workqueue = nfsiod_workqueue,
6000                 .flags = RPC_TASK_ASYNC,
6001         };
6002
6003         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6004                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6005
6006         /* Ensure this is an unlock - when canceling a lock, the
6007          * canceled lock is passed in, and it won't be an unlock.
6008          */
6009         fl->fl_type = F_UNLCK;
6010         if (fl->fl_flags & FL_CLOSE)
6011                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6012
6013         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6014         if (data == NULL) {
6015                 nfs_free_seqid(seqid);
6016                 return ERR_PTR(-ENOMEM);
6017         }
6018
6019         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6020         msg.rpc_argp = &data->arg;
6021         msg.rpc_resp = &data->res;
6022         task_setup_data.callback_data = data;
6023         return rpc_run_task(&task_setup_data);
6024 }
6025
6026 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6027 {
6028         struct inode *inode = state->inode;
6029         struct nfs4_state_owner *sp = state->owner;
6030         struct nfs_inode *nfsi = NFS_I(inode);
6031         struct nfs_seqid *seqid;
6032         struct nfs4_lock_state *lsp;
6033         struct rpc_task *task;
6034         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6035         int status = 0;
6036         unsigned char fl_flags = request->fl_flags;
6037
6038         status = nfs4_set_lock_state(state, request);
6039         /* Unlock _before_ we do the RPC call */
6040         request->fl_flags |= FL_EXISTS;
6041         /* Exclude nfs_delegation_claim_locks() */
6042         mutex_lock(&sp->so_delegreturn_mutex);
6043         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6044         down_read(&nfsi->rwsem);
6045         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6046                 up_read(&nfsi->rwsem);
6047                 mutex_unlock(&sp->so_delegreturn_mutex);
6048                 goto out;
6049         }
6050         up_read(&nfsi->rwsem);
6051         mutex_unlock(&sp->so_delegreturn_mutex);
6052         if (status != 0)
6053                 goto out;
6054         /* Is this a delegated lock? */
6055         lsp = request->fl_u.nfs4_fl.owner;
6056         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6057                 goto out;
6058         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6059         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6060         status = -ENOMEM;
6061         if (IS_ERR(seqid))
6062                 goto out;
6063         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6064         status = PTR_ERR(task);
6065         if (IS_ERR(task))
6066                 goto out;
6067         status = rpc_wait_for_completion_task(task);
6068         rpc_put_task(task);
6069 out:
6070         request->fl_flags = fl_flags;
6071         trace_nfs4_unlock(request, state, F_SETLK, status);
6072         return status;
6073 }
6074
6075 struct nfs4_lockdata {
6076         struct nfs_lock_args arg;
6077         struct nfs_lock_res res;
6078         struct nfs4_lock_state *lsp;
6079         struct nfs_open_context *ctx;
6080         struct file_lock fl;
6081         unsigned long timestamp;
6082         int rpc_status;
6083         int cancelled;
6084         struct nfs_server *server;
6085 };
6086
6087 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6088                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6089                 gfp_t gfp_mask)
6090 {
6091         struct nfs4_lockdata *p;
6092         struct inode *inode = lsp->ls_state->inode;
6093         struct nfs_server *server = NFS_SERVER(inode);
6094         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6095
6096         p = kzalloc(sizeof(*p), gfp_mask);
6097         if (p == NULL)
6098                 return NULL;
6099
6100         p->arg.fh = NFS_FH(inode);
6101         p->arg.fl = &p->fl;
6102         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6103         if (IS_ERR(p->arg.open_seqid))
6104                 goto out_free;
6105         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6106         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6107         if (IS_ERR(p->arg.lock_seqid))
6108                 goto out_free_seqid;
6109         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6110         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6111         p->arg.lock_owner.s_dev = server->s_dev;
6112         p->res.lock_seqid = p->arg.lock_seqid;
6113         p->lsp = lsp;
6114         p->server = server;
6115         atomic_inc(&lsp->ls_count);
6116         p->ctx = get_nfs_open_context(ctx);
6117         memcpy(&p->fl, fl, sizeof(p->fl));
6118         return p;
6119 out_free_seqid:
6120         nfs_free_seqid(p->arg.open_seqid);
6121 out_free:
6122         kfree(p);
6123         return NULL;
6124 }
6125
6126 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6127 {
6128         struct nfs4_lockdata *data = calldata;
6129         struct nfs4_state *state = data->lsp->ls_state;
6130
6131         dprintk("%s: begin!\n", __func__);
6132         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6133                 goto out_wait;
6134         /* Do we need to do an open_to_lock_owner? */
6135         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6136                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6137                         goto out_release_lock_seqid;
6138                 }
6139                 nfs4_stateid_copy(&data->arg.open_stateid,
6140                                 &state->open_stateid);
6141                 data->arg.new_lock_owner = 1;
6142                 data->res.open_seqid = data->arg.open_seqid;
6143         } else {
6144                 data->arg.new_lock_owner = 0;
6145                 nfs4_stateid_copy(&data->arg.lock_stateid,
6146                                 &data->lsp->ls_stateid);
6147         }
6148         if (!nfs4_valid_open_stateid(state)) {
6149                 data->rpc_status = -EBADF;
6150                 task->tk_action = NULL;
6151                 goto out_release_open_seqid;
6152         }
6153         data->timestamp = jiffies;
6154         if (nfs4_setup_sequence(data->server->nfs_client,
6155                                 &data->arg.seq_args,
6156                                 &data->res.seq_res,
6157                                 task) == 0)
6158                 return;
6159 out_release_open_seqid:
6160         nfs_release_seqid(data->arg.open_seqid);
6161 out_release_lock_seqid:
6162         nfs_release_seqid(data->arg.lock_seqid);
6163 out_wait:
6164         nfs4_sequence_done(task, &data->res.seq_res);
6165         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6166 }
6167
6168 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6169 {
6170         struct nfs4_lockdata *data = calldata;
6171         struct nfs4_lock_state *lsp = data->lsp;
6172
6173         dprintk("%s: begin!\n", __func__);
6174
6175         if (!nfs4_sequence_done(task, &data->res.seq_res))
6176                 return;
6177
6178         data->rpc_status = task->tk_status;
6179         switch (task->tk_status) {
6180         case 0:
6181                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6182                                 data->timestamp);
6183                 if (data->arg.new_lock) {
6184                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6185                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6186                                 rpc_restart_call_prepare(task);
6187                                 break;
6188                         }
6189                 }
6190                 if (data->arg.new_lock_owner != 0) {
6191                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6192                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6193                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6194                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6195                         rpc_restart_call_prepare(task);
6196                 break;
6197         case -NFS4ERR_BAD_STATEID:
6198         case -NFS4ERR_OLD_STATEID:
6199         case -NFS4ERR_STALE_STATEID:
6200         case -NFS4ERR_EXPIRED:
6201                 if (data->arg.new_lock_owner != 0) {
6202                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6203                                                 &lsp->ls_state->open_stateid))
6204                                 rpc_restart_call_prepare(task);
6205                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6206                                                 &lsp->ls_stateid))
6207                                 rpc_restart_call_prepare(task);
6208         }
6209         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6210 }
6211
6212 static void nfs4_lock_release(void *calldata)
6213 {
6214         struct nfs4_lockdata *data = calldata;
6215
6216         dprintk("%s: begin!\n", __func__);
6217         nfs_free_seqid(data->arg.open_seqid);
6218         if (data->cancelled) {
6219                 struct rpc_task *task;
6220                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6221                                 data->arg.lock_seqid);
6222                 if (!IS_ERR(task))
6223                         rpc_put_task_async(task);
6224                 dprintk("%s: cancelling lock!\n", __func__);
6225         } else
6226                 nfs_free_seqid(data->arg.lock_seqid);
6227         nfs4_put_lock_state(data->lsp);
6228         put_nfs_open_context(data->ctx);
6229         kfree(data);
6230         dprintk("%s: done!\n", __func__);
6231 }
6232
6233 static const struct rpc_call_ops nfs4_lock_ops = {
6234         .rpc_call_prepare = nfs4_lock_prepare,
6235         .rpc_call_done = nfs4_lock_done,
6236         .rpc_release = nfs4_lock_release,
6237 };
6238
6239 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6240 {
6241         switch (error) {
6242         case -NFS4ERR_ADMIN_REVOKED:
6243         case -NFS4ERR_EXPIRED:
6244         case -NFS4ERR_BAD_STATEID:
6245                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6246                 if (new_lock_owner != 0 ||
6247                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6248                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6249                 break;
6250         case -NFS4ERR_STALE_STATEID:
6251                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6252                 nfs4_schedule_lease_recovery(server->nfs_client);
6253         };
6254 }
6255
6256 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6257 {
6258         struct nfs4_lockdata *data;
6259         struct rpc_task *task;
6260         struct rpc_message msg = {
6261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6262                 .rpc_cred = state->owner->so_cred,
6263         };
6264         struct rpc_task_setup task_setup_data = {
6265                 .rpc_client = NFS_CLIENT(state->inode),
6266                 .rpc_message = &msg,
6267                 .callback_ops = &nfs4_lock_ops,
6268                 .workqueue = nfsiod_workqueue,
6269                 .flags = RPC_TASK_ASYNC,
6270         };
6271         int ret;
6272
6273         dprintk("%s: begin!\n", __func__);
6274         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6275                         fl->fl_u.nfs4_fl.owner,
6276                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6277         if (data == NULL)
6278                 return -ENOMEM;
6279         if (IS_SETLKW(cmd))
6280                 data->arg.block = 1;
6281         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6282         msg.rpc_argp = &data->arg;
6283         msg.rpc_resp = &data->res;
6284         task_setup_data.callback_data = data;
6285         if (recovery_type > NFS_LOCK_NEW) {
6286                 if (recovery_type == NFS_LOCK_RECLAIM)
6287                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6288                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6289         } else
6290                 data->arg.new_lock = 1;
6291         task = rpc_run_task(&task_setup_data);
6292         if (IS_ERR(task))
6293                 return PTR_ERR(task);
6294         ret = rpc_wait_for_completion_task(task);
6295         if (ret == 0) {
6296                 ret = data->rpc_status;
6297                 if (ret)
6298                         nfs4_handle_setlk_error(data->server, data->lsp,
6299                                         data->arg.new_lock_owner, ret);
6300         } else
6301                 data->cancelled = true;
6302         rpc_put_task(task);
6303         dprintk("%s: done, ret = %d!\n", __func__, ret);
6304         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6305         return ret;
6306 }
6307
6308 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6309 {
6310         struct nfs_server *server = NFS_SERVER(state->inode);
6311         struct nfs4_exception exception = {
6312                 .inode = state->inode,
6313         };
6314         int err;
6315
6316         do {
6317                 /* Cache the lock if possible... */
6318                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6319                         return 0;
6320                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6321                 if (err != -NFS4ERR_DELAY)
6322                         break;
6323                 nfs4_handle_exception(server, err, &exception);
6324         } while (exception.retry);
6325         return err;
6326 }
6327
6328 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6329 {
6330         struct nfs_server *server = NFS_SERVER(state->inode);
6331         struct nfs4_exception exception = {
6332                 .inode = state->inode,
6333         };
6334         int err;
6335
6336         err = nfs4_set_lock_state(state, request);
6337         if (err != 0)
6338                 return err;
6339         if (!recover_lost_locks) {
6340                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6341                 return 0;
6342         }
6343         do {
6344                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6345                         return 0;
6346                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6347                 switch (err) {
6348                 default:
6349                         goto out;
6350                 case -NFS4ERR_GRACE:
6351                 case -NFS4ERR_DELAY:
6352                         nfs4_handle_exception(server, err, &exception);
6353                         err = 0;
6354                 }
6355         } while (exception.retry);
6356 out:
6357         return err;
6358 }
6359
6360 #if defined(CONFIG_NFS_V4_1)
6361 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6362 {
6363         struct nfs4_lock_state *lsp;
6364         int status;
6365
6366         status = nfs4_set_lock_state(state, request);
6367         if (status != 0)
6368                 return status;
6369         lsp = request->fl_u.nfs4_fl.owner;
6370         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6371             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6372                 return 0;
6373         return nfs4_lock_expired(state, request);
6374 }
6375 #endif
6376
6377 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6378 {
6379         struct nfs_inode *nfsi = NFS_I(state->inode);
6380         struct nfs4_state_owner *sp = state->owner;
6381         unsigned char fl_flags = request->fl_flags;
6382         int status;
6383
6384         request->fl_flags |= FL_ACCESS;
6385         status = locks_lock_inode_wait(state->inode, request);
6386         if (status < 0)
6387                 goto out;
6388         mutex_lock(&sp->so_delegreturn_mutex);
6389         down_read(&nfsi->rwsem);
6390         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6391                 /* Yes: cache locks! */
6392                 /* ...but avoid races with delegation recall... */
6393                 request->fl_flags = fl_flags & ~FL_SLEEP;
6394                 status = locks_lock_inode_wait(state->inode, request);
6395                 up_read(&nfsi->rwsem);
6396                 mutex_unlock(&sp->so_delegreturn_mutex);
6397                 goto out;
6398         }
6399         up_read(&nfsi->rwsem);
6400         mutex_unlock(&sp->so_delegreturn_mutex);
6401         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6402 out:
6403         request->fl_flags = fl_flags;
6404         return status;
6405 }
6406
6407 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6408 {
6409         struct nfs4_exception exception = {
6410                 .state = state,
6411                 .inode = state->inode,
6412         };
6413         int err;
6414
6415         do {
6416                 err = _nfs4_proc_setlk(state, cmd, request);
6417                 if (err == -NFS4ERR_DENIED)
6418                         err = -EAGAIN;
6419                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6420                                 err, &exception);
6421         } while (exception.retry);
6422         return err;
6423 }
6424
6425 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6426 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6427
6428 static int
6429 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6430                         struct file_lock *request)
6431 {
6432         int             status = -ERESTARTSYS;
6433         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6434
6435         while(!signalled()) {
6436                 status = nfs4_proc_setlk(state, cmd, request);
6437                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6438                         break;
6439                 freezable_schedule_timeout_interruptible(timeout);
6440                 timeout *= 2;
6441                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6442                 status = -ERESTARTSYS;
6443         }
6444         return status;
6445 }
6446
6447 #ifdef CONFIG_NFS_V4_1
6448 struct nfs4_lock_waiter {
6449         struct task_struct      *task;
6450         struct inode            *inode;
6451         struct nfs_lowner       *owner;
6452         bool                    notified;
6453 };
6454
6455 static int
6456 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6457 {
6458         int ret;
6459         struct cb_notify_lock_args *cbnl = key;
6460         struct nfs4_lock_waiter *waiter = wait->private;
6461         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6462                                 *wowner = waiter->owner;
6463
6464         /* Only wake if the callback was for the same owner */
6465         if (lowner->clientid != wowner->clientid ||
6466             lowner->id != wowner->id             ||
6467             lowner->s_dev != wowner->s_dev)
6468                 return 0;
6469
6470         /* Make sure it's for the right inode */
6471         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6472                 return 0;
6473
6474         waiter->notified = true;
6475
6476         /* override "private" so we can use default_wake_function */
6477         wait->private = waiter->task;
6478         ret = autoremove_wake_function(wait, mode, flags, key);
6479         wait->private = waiter;
6480         return ret;
6481 }
6482
6483 static int
6484 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6485 {
6486         int status = -ERESTARTSYS;
6487         unsigned long flags;
6488         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6489         struct nfs_server *server = NFS_SERVER(state->inode);
6490         struct nfs_client *clp = server->nfs_client;
6491         wait_queue_head_t *q = &clp->cl_lock_waitq;
6492         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6493                                     .id = lsp->ls_seqid.owner_id,
6494                                     .s_dev = server->s_dev };
6495         struct nfs4_lock_waiter waiter = { .task  = current,
6496                                            .inode = state->inode,
6497                                            .owner = &owner,
6498                                            .notified = false };
6499         wait_queue_entry_t wait;
6500
6501         /* Don't bother with waitqueue if we don't expect a callback */
6502         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6503                 return nfs4_retry_setlk_simple(state, cmd, request);
6504
6505         init_wait(&wait);
6506         wait.private = &waiter;
6507         wait.func = nfs4_wake_lock_waiter;
6508         add_wait_queue(q, &wait);
6509
6510         while(!signalled()) {
6511                 status = nfs4_proc_setlk(state, cmd, request);
6512                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6513                         break;
6514
6515                 status = -ERESTARTSYS;
6516                 spin_lock_irqsave(&q->lock, flags);
6517                 if (waiter.notified) {
6518                         spin_unlock_irqrestore(&q->lock, flags);
6519                         continue;
6520                 }
6521                 set_current_state(TASK_INTERRUPTIBLE);
6522                 spin_unlock_irqrestore(&q->lock, flags);
6523
6524                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6525         }
6526
6527         finish_wait(q, &wait);
6528         return status;
6529 }
6530 #else /* !CONFIG_NFS_V4_1 */
6531 static inline int
6532 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6533 {
6534         return nfs4_retry_setlk_simple(state, cmd, request);
6535 }
6536 #endif
6537
6538 static int
6539 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6540 {
6541         struct nfs_open_context *ctx;
6542         struct nfs4_state *state;
6543         int status;
6544
6545         /* verify open state */
6546         ctx = nfs_file_open_context(filp);
6547         state = ctx->state;
6548
6549         if (IS_GETLK(cmd)) {
6550                 if (state != NULL)
6551                         return nfs4_proc_getlk(state, F_GETLK, request);
6552                 return 0;
6553         }
6554
6555         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6556                 return -EINVAL;
6557
6558         if (request->fl_type == F_UNLCK) {
6559                 if (state != NULL)
6560                         return nfs4_proc_unlck(state, cmd, request);
6561                 return 0;
6562         }
6563
6564         if (state == NULL)
6565                 return -ENOLCK;
6566
6567         if ((request->fl_flags & FL_POSIX) &&
6568             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6569                 return -ENOLCK;
6570
6571         status = nfs4_set_lock_state(state, request);
6572         if (status != 0)
6573                 return status;
6574
6575         return nfs4_retry_setlk(state, cmd, request);
6576 }
6577
6578 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6579 {
6580         struct nfs_server *server = NFS_SERVER(state->inode);
6581         int err;
6582
6583         err = nfs4_set_lock_state(state, fl);
6584         if (err != 0)
6585                 return err;
6586         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6587         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6588 }
6589
6590 struct nfs_release_lockowner_data {
6591         struct nfs4_lock_state *lsp;
6592         struct nfs_server *server;
6593         struct nfs_release_lockowner_args args;
6594         struct nfs_release_lockowner_res res;
6595         unsigned long timestamp;
6596 };
6597
6598 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6599 {
6600         struct nfs_release_lockowner_data *data = calldata;
6601         struct nfs_server *server = data->server;
6602         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
6603                            &data->res.seq_res, task);
6604         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6605         data->timestamp = jiffies;
6606 }
6607
6608 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6609 {
6610         struct nfs_release_lockowner_data *data = calldata;
6611         struct nfs_server *server = data->server;
6612
6613         nfs40_sequence_done(task, &data->res.seq_res);
6614
6615         switch (task->tk_status) {
6616         case 0:
6617                 renew_lease(server, data->timestamp);
6618                 break;
6619         case -NFS4ERR_STALE_CLIENTID:
6620         case -NFS4ERR_EXPIRED:
6621                 nfs4_schedule_lease_recovery(server->nfs_client);
6622                 break;
6623         case -NFS4ERR_LEASE_MOVED:
6624         case -NFS4ERR_DELAY:
6625                 if (nfs4_async_handle_error(task, server,
6626                                             NULL, NULL) == -EAGAIN)
6627                         rpc_restart_call_prepare(task);
6628         }
6629 }
6630
6631 static void nfs4_release_lockowner_release(void *calldata)
6632 {
6633         struct nfs_release_lockowner_data *data = calldata;
6634         nfs4_free_lock_state(data->server, data->lsp);
6635         kfree(calldata);
6636 }
6637
6638 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6639         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6640         .rpc_call_done = nfs4_release_lockowner_done,
6641         .rpc_release = nfs4_release_lockowner_release,
6642 };
6643
6644 static void
6645 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6646 {
6647         struct nfs_release_lockowner_data *data;
6648         struct rpc_message msg = {
6649                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6650         };
6651
6652         if (server->nfs_client->cl_mvops->minor_version != 0)
6653                 return;
6654
6655         data = kmalloc(sizeof(*data), GFP_NOFS);
6656         if (!data)
6657                 return;
6658         data->lsp = lsp;
6659         data->server = server;
6660         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6661         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6662         data->args.lock_owner.s_dev = server->s_dev;
6663
6664         msg.rpc_argp = &data->args;
6665         msg.rpc_resp = &data->res;
6666         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6667         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6668 }
6669
6670 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6671
6672 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6673                                    struct dentry *unused, struct inode *inode,
6674                                    const char *key, const void *buf,
6675                                    size_t buflen, int flags)
6676 {
6677         return nfs4_proc_set_acl(inode, buf, buflen);
6678 }
6679
6680 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6681                                    struct dentry *unused, struct inode *inode,
6682                                    const char *key, void *buf, size_t buflen)
6683 {
6684         return nfs4_proc_get_acl(inode, buf, buflen);
6685 }
6686
6687 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6688 {
6689         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6690 }
6691
6692 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6693
6694 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6695                                      struct dentry *unused, struct inode *inode,
6696                                      const char *key, const void *buf,
6697                                      size_t buflen, int flags)
6698 {
6699         if (security_ismaclabel(key))
6700                 return nfs4_set_security_label(inode, buf, buflen);
6701
6702         return -EOPNOTSUPP;
6703 }
6704
6705 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6706                                      struct dentry *unused, struct inode *inode,
6707                                      const char *key, void *buf, size_t buflen)
6708 {
6709         if (security_ismaclabel(key))
6710                 return nfs4_get_security_label(inode, buf, buflen);
6711         return -EOPNOTSUPP;
6712 }
6713
6714 static ssize_t
6715 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6716 {
6717         int len = 0;
6718
6719         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6720                 len = security_inode_listsecurity(inode, list, list_len);
6721                 if (list_len && len > list_len)
6722                         return -ERANGE;
6723         }
6724         return len;
6725 }
6726
6727 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6728         .prefix = XATTR_SECURITY_PREFIX,
6729         .get    = nfs4_xattr_get_nfs4_label,
6730         .set    = nfs4_xattr_set_nfs4_label,
6731 };
6732
6733 #else
6734
6735 static ssize_t
6736 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6737 {
6738         return 0;
6739 }
6740
6741 #endif
6742
6743 /*
6744  * nfs_fhget will use either the mounted_on_fileid or the fileid
6745  */
6746 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6747 {
6748         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6749                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6750               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6751               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6752                 return;
6753
6754         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6755                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6756         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6757         fattr->nlink = 2;
6758 }
6759
6760 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6761                                    const struct qstr *name,
6762                                    struct nfs4_fs_locations *fs_locations,
6763                                    struct page *page)
6764 {
6765         struct nfs_server *server = NFS_SERVER(dir);
6766         u32 bitmask[3] = {
6767                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6768         };
6769         struct nfs4_fs_locations_arg args = {
6770                 .dir_fh = NFS_FH(dir),
6771                 .name = name,
6772                 .page = page,
6773                 .bitmask = bitmask,
6774         };
6775         struct nfs4_fs_locations_res res = {
6776                 .fs_locations = fs_locations,
6777         };
6778         struct rpc_message msg = {
6779                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6780                 .rpc_argp = &args,
6781                 .rpc_resp = &res,
6782         };
6783         int status;
6784
6785         dprintk("%s: start\n", __func__);
6786
6787         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6788          * is not supported */
6789         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6790                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6791         else
6792                 bitmask[0] |= FATTR4_WORD0_FILEID;
6793
6794         nfs_fattr_init(&fs_locations->fattr);
6795         fs_locations->server = server;
6796         fs_locations->nlocations = 0;
6797         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6798         dprintk("%s: returned status = %d\n", __func__, status);
6799         return status;
6800 }
6801
6802 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6803                            const struct qstr *name,
6804                            struct nfs4_fs_locations *fs_locations,
6805                            struct page *page)
6806 {
6807         struct nfs4_exception exception = { };
6808         int err;
6809         do {
6810                 err = _nfs4_proc_fs_locations(client, dir, name,
6811                                 fs_locations, page);
6812                 trace_nfs4_get_fs_locations(dir, name, err);
6813                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6814                                 &exception);
6815         } while (exception.retry);
6816         return err;
6817 }
6818
6819 /*
6820  * This operation also signals the server that this client is
6821  * performing migration recovery.  The server can stop returning
6822  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6823  * appended to this compound to identify the client ID which is
6824  * performing recovery.
6825  */
6826 static int _nfs40_proc_get_locations(struct inode *inode,
6827                                      struct nfs4_fs_locations *locations,
6828                                      struct page *page, struct rpc_cred *cred)
6829 {
6830         struct nfs_server *server = NFS_SERVER(inode);
6831         struct rpc_clnt *clnt = server->client;
6832         u32 bitmask[2] = {
6833                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6834         };
6835         struct nfs4_fs_locations_arg args = {
6836                 .clientid       = server->nfs_client->cl_clientid,
6837                 .fh             = NFS_FH(inode),
6838                 .page           = page,
6839                 .bitmask        = bitmask,
6840                 .migration      = 1,            /* skip LOOKUP */
6841                 .renew          = 1,            /* append RENEW */
6842         };
6843         struct nfs4_fs_locations_res res = {
6844                 .fs_locations   = locations,
6845                 .migration      = 1,
6846                 .renew          = 1,
6847         };
6848         struct rpc_message msg = {
6849                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6850                 .rpc_argp       = &args,
6851                 .rpc_resp       = &res,
6852                 .rpc_cred       = cred,
6853         };
6854         unsigned long now = jiffies;
6855         int status;
6856
6857         nfs_fattr_init(&locations->fattr);
6858         locations->server = server;
6859         locations->nlocations = 0;
6860
6861         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6862         nfs4_set_sequence_privileged(&args.seq_args);
6863         status = nfs4_call_sync_sequence(clnt, server, &msg,
6864                                         &args.seq_args, &res.seq_res);
6865         if (status)
6866                 return status;
6867
6868         renew_lease(server, now);
6869         return 0;
6870 }
6871
6872 #ifdef CONFIG_NFS_V4_1
6873
6874 /*
6875  * This operation also signals the server that this client is
6876  * performing migration recovery.  The server can stop asserting
6877  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6878  * performing this operation is identified in the SEQUENCE
6879  * operation in this compound.
6880  *
6881  * When the client supports GETATTR(fs_locations_info), it can
6882  * be plumbed in here.
6883  */
6884 static int _nfs41_proc_get_locations(struct inode *inode,
6885                                      struct nfs4_fs_locations *locations,
6886                                      struct page *page, struct rpc_cred *cred)
6887 {
6888         struct nfs_server *server = NFS_SERVER(inode);
6889         struct rpc_clnt *clnt = server->client;
6890         u32 bitmask[2] = {
6891                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6892         };
6893         struct nfs4_fs_locations_arg args = {
6894                 .fh             = NFS_FH(inode),
6895                 .page           = page,
6896                 .bitmask        = bitmask,
6897                 .migration      = 1,            /* skip LOOKUP */
6898         };
6899         struct nfs4_fs_locations_res res = {
6900                 .fs_locations   = locations,
6901                 .migration      = 1,
6902         };
6903         struct rpc_message msg = {
6904                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6905                 .rpc_argp       = &args,
6906                 .rpc_resp       = &res,
6907                 .rpc_cred       = cred,
6908         };
6909         int status;
6910
6911         nfs_fattr_init(&locations->fattr);
6912         locations->server = server;
6913         locations->nlocations = 0;
6914
6915         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6916         nfs4_set_sequence_privileged(&args.seq_args);
6917         status = nfs4_call_sync_sequence(clnt, server, &msg,
6918                                         &args.seq_args, &res.seq_res);
6919         if (status == NFS4_OK &&
6920             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6921                 status = -NFS4ERR_LEASE_MOVED;
6922         return status;
6923 }
6924
6925 #endif  /* CONFIG_NFS_V4_1 */
6926
6927 /**
6928  * nfs4_proc_get_locations - discover locations for a migrated FSID
6929  * @inode: inode on FSID that is migrating
6930  * @locations: result of query
6931  * @page: buffer
6932  * @cred: credential to use for this operation
6933  *
6934  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6935  * operation failed, or a negative errno if a local error occurred.
6936  *
6937  * On success, "locations" is filled in, but if the server has
6938  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6939  * asserted.
6940  *
6941  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6942  * from this client that require migration recovery.
6943  */
6944 int nfs4_proc_get_locations(struct inode *inode,
6945                             struct nfs4_fs_locations *locations,
6946                             struct page *page, struct rpc_cred *cred)
6947 {
6948         struct nfs_server *server = NFS_SERVER(inode);
6949         struct nfs_client *clp = server->nfs_client;
6950         const struct nfs4_mig_recovery_ops *ops =
6951                                         clp->cl_mvops->mig_recovery_ops;
6952         struct nfs4_exception exception = { };
6953         int status;
6954
6955         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6956                 (unsigned long long)server->fsid.major,
6957                 (unsigned long long)server->fsid.minor,
6958                 clp->cl_hostname);
6959         nfs_display_fhandle(NFS_FH(inode), __func__);
6960
6961         do {
6962                 status = ops->get_locations(inode, locations, page, cred);
6963                 if (status != -NFS4ERR_DELAY)
6964                         break;
6965                 nfs4_handle_exception(server, status, &exception);
6966         } while (exception.retry);
6967         return status;
6968 }
6969
6970 /*
6971  * This operation also signals the server that this client is
6972  * performing "lease moved" recovery.  The server can stop
6973  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6974  * is appended to this compound to identify the client ID which is
6975  * performing recovery.
6976  */
6977 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6978 {
6979         struct nfs_server *server = NFS_SERVER(inode);
6980         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6981         struct rpc_clnt *clnt = server->client;
6982         struct nfs4_fsid_present_arg args = {
6983                 .fh             = NFS_FH(inode),
6984                 .clientid       = clp->cl_clientid,
6985                 .renew          = 1,            /* append RENEW */
6986         };
6987         struct nfs4_fsid_present_res res = {
6988                 .renew          = 1,
6989         };
6990         struct rpc_message msg = {
6991                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6992                 .rpc_argp       = &args,
6993                 .rpc_resp       = &res,
6994                 .rpc_cred       = cred,
6995         };
6996         unsigned long now = jiffies;
6997         int status;
6998
6999         res.fh = nfs_alloc_fhandle();
7000         if (res.fh == NULL)
7001                 return -ENOMEM;
7002
7003         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7004         nfs4_set_sequence_privileged(&args.seq_args);
7005         status = nfs4_call_sync_sequence(clnt, server, &msg,
7006                                                 &args.seq_args, &res.seq_res);
7007         nfs_free_fhandle(res.fh);
7008         if (status)
7009                 return status;
7010
7011         do_renew_lease(clp, now);
7012         return 0;
7013 }
7014
7015 #ifdef CONFIG_NFS_V4_1
7016
7017 /*
7018  * This operation also signals the server that this client is
7019  * performing "lease moved" recovery.  The server can stop asserting
7020  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7021  * this operation is identified in the SEQUENCE operation in this
7022  * compound.
7023  */
7024 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7025 {
7026         struct nfs_server *server = NFS_SERVER(inode);
7027         struct rpc_clnt *clnt = server->client;
7028         struct nfs4_fsid_present_arg args = {
7029                 .fh             = NFS_FH(inode),
7030         };
7031         struct nfs4_fsid_present_res res = {
7032         };
7033         struct rpc_message msg = {
7034                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7035                 .rpc_argp       = &args,
7036                 .rpc_resp       = &res,
7037                 .rpc_cred       = cred,
7038         };
7039         int status;
7040
7041         res.fh = nfs_alloc_fhandle();
7042         if (res.fh == NULL)
7043                 return -ENOMEM;
7044
7045         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7046         nfs4_set_sequence_privileged(&args.seq_args);
7047         status = nfs4_call_sync_sequence(clnt, server, &msg,
7048                                                 &args.seq_args, &res.seq_res);
7049         nfs_free_fhandle(res.fh);
7050         if (status == NFS4_OK &&
7051             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7052                 status = -NFS4ERR_LEASE_MOVED;
7053         return status;
7054 }
7055
7056 #endif  /* CONFIG_NFS_V4_1 */
7057
7058 /**
7059  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7060  * @inode: inode on FSID to check
7061  * @cred: credential to use for this operation
7062  *
7063  * Server indicates whether the FSID is present, moved, or not
7064  * recognized.  This operation is necessary to clear a LEASE_MOVED
7065  * condition for this client ID.
7066  *
7067  * Returns NFS4_OK if the FSID is present on this server,
7068  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7069  *  NFS4ERR code if some error occurred on the server, or a
7070  *  negative errno if a local failure occurred.
7071  */
7072 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7073 {
7074         struct nfs_server *server = NFS_SERVER(inode);
7075         struct nfs_client *clp = server->nfs_client;
7076         const struct nfs4_mig_recovery_ops *ops =
7077                                         clp->cl_mvops->mig_recovery_ops;
7078         struct nfs4_exception exception = { };
7079         int status;
7080
7081         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7082                 (unsigned long long)server->fsid.major,
7083                 (unsigned long long)server->fsid.minor,
7084                 clp->cl_hostname);
7085         nfs_display_fhandle(NFS_FH(inode), __func__);
7086
7087         do {
7088                 status = ops->fsid_present(inode, cred);
7089                 if (status != -NFS4ERR_DELAY)
7090                         break;
7091                 nfs4_handle_exception(server, status, &exception);
7092         } while (exception.retry);
7093         return status;
7094 }
7095
7096 /**
7097  * If 'use_integrity' is true and the state managment nfs_client
7098  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7099  * and the machine credential as per RFC3530bis and RFC5661 Security
7100  * Considerations sections. Otherwise, just use the user cred with the
7101  * filesystem's rpc_client.
7102  */
7103 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7104 {
7105         int status;
7106         struct nfs4_secinfo_arg args = {
7107                 .dir_fh = NFS_FH(dir),
7108                 .name   = name,
7109         };
7110         struct nfs4_secinfo_res res = {
7111                 .flavors     = flavors,
7112         };
7113         struct rpc_message msg = {
7114                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7115                 .rpc_argp = &args,
7116                 .rpc_resp = &res,
7117         };
7118         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7119         struct rpc_cred *cred = NULL;
7120
7121         if (use_integrity) {
7122                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7123                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7124                 msg.rpc_cred = cred;
7125         }
7126
7127         dprintk("NFS call  secinfo %s\n", name->name);
7128
7129         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7130                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7131
7132         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7133                                 &res.seq_res, 0);
7134         dprintk("NFS reply  secinfo: %d\n", status);
7135
7136         if (cred)
7137                 put_rpccred(cred);
7138
7139         return status;
7140 }
7141
7142 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7143                       struct nfs4_secinfo_flavors *flavors)
7144 {
7145         struct nfs4_exception exception = { };
7146         int err;
7147         do {
7148                 err = -NFS4ERR_WRONGSEC;
7149
7150                 /* try to use integrity protection with machine cred */
7151                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7152                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7153
7154                 /*
7155                  * if unable to use integrity protection, or SECINFO with
7156                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7157                  * disallowed by spec, but exists in deployed servers) use
7158                  * the current filesystem's rpc_client and the user cred.
7159                  */
7160                 if (err == -NFS4ERR_WRONGSEC)
7161                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7162
7163                 trace_nfs4_secinfo(dir, name, err);
7164                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7165                                 &exception);
7166         } while (exception.retry);
7167         return err;
7168 }
7169
7170 #ifdef CONFIG_NFS_V4_1
7171 /*
7172  * Check the exchange flags returned by the server for invalid flags, having
7173  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7174  * DS flags set.
7175  */
7176 static int nfs4_check_cl_exchange_flags(u32 flags)
7177 {
7178         if (flags & ~EXCHGID4_FLAG_MASK_R)
7179                 goto out_inval;
7180         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7181             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7182                 goto out_inval;
7183         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7184                 goto out_inval;
7185         return NFS_OK;
7186 out_inval:
7187         return -NFS4ERR_INVAL;
7188 }
7189
7190 static bool
7191 nfs41_same_server_scope(struct nfs41_server_scope *a,
7192                         struct nfs41_server_scope *b)
7193 {
7194         if (a->server_scope_sz != b->server_scope_sz)
7195                 return false;
7196         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7197 }
7198
7199 static void
7200 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7201 {
7202 }
7203
7204 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7205         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7206 };
7207
7208 /*
7209  * nfs4_proc_bind_one_conn_to_session()
7210  *
7211  * The 4.1 client currently uses the same TCP connection for the
7212  * fore and backchannel.
7213  */
7214 static
7215 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7216                 struct rpc_xprt *xprt,
7217                 struct nfs_client *clp,
7218                 struct rpc_cred *cred)
7219 {
7220         int status;
7221         struct nfs41_bind_conn_to_session_args args = {
7222                 .client = clp,
7223                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7224         };
7225         struct nfs41_bind_conn_to_session_res res;
7226         struct rpc_message msg = {
7227                 .rpc_proc =
7228                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7229                 .rpc_argp = &args,
7230                 .rpc_resp = &res,
7231                 .rpc_cred = cred,
7232         };
7233         struct rpc_task_setup task_setup_data = {
7234                 .rpc_client = clnt,
7235                 .rpc_xprt = xprt,
7236                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7237                 .rpc_message = &msg,
7238                 .flags = RPC_TASK_TIMEOUT,
7239         };
7240         struct rpc_task *task;
7241
7242         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7243         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7244                 args.dir = NFS4_CDFC4_FORE;
7245
7246         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7247         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7248                 args.dir = NFS4_CDFC4_FORE;
7249
7250         task = rpc_run_task(&task_setup_data);
7251         if (!IS_ERR(task)) {
7252                 status = task->tk_status;
7253                 rpc_put_task(task);
7254         } else
7255                 status = PTR_ERR(task);
7256         trace_nfs4_bind_conn_to_session(clp, status);
7257         if (status == 0) {
7258                 if (memcmp(res.sessionid.data,
7259                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7260                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7261                         return -EIO;
7262                 }
7263                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7264                         dprintk("NFS: %s: Unexpected direction from server\n",
7265                                 __func__);
7266                         return -EIO;
7267                 }
7268                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7269                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7270                                 __func__);
7271                         return -EIO;
7272                 }
7273         }
7274
7275         return status;
7276 }
7277
7278 struct rpc_bind_conn_calldata {
7279         struct nfs_client *clp;
7280         struct rpc_cred *cred;
7281 };
7282
7283 static int
7284 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7285                 struct rpc_xprt *xprt,
7286                 void *calldata)
7287 {
7288         struct rpc_bind_conn_calldata *p = calldata;
7289
7290         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7291 }
7292
7293 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7294 {
7295         struct rpc_bind_conn_calldata data = {
7296                 .clp = clp,
7297                 .cred = cred,
7298         };
7299         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7300                         nfs4_proc_bind_conn_to_session_callback, &data);
7301 }
7302
7303 /*
7304  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7305  * and operations we'd like to see to enable certain features in the allow map
7306  */
7307 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7308         .how = SP4_MACH_CRED,
7309         .enforce.u.words = {
7310                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7311                       1 << (OP_EXCHANGE_ID - 32) |
7312                       1 << (OP_CREATE_SESSION - 32) |
7313                       1 << (OP_DESTROY_SESSION - 32) |
7314                       1 << (OP_DESTROY_CLIENTID - 32)
7315         },
7316         .allow.u.words = {
7317                 [0] = 1 << (OP_CLOSE) |
7318                       1 << (OP_OPEN_DOWNGRADE) |
7319                       1 << (OP_LOCKU) |
7320                       1 << (OP_DELEGRETURN) |
7321                       1 << (OP_COMMIT),
7322                 [1] = 1 << (OP_SECINFO - 32) |
7323                       1 << (OP_SECINFO_NO_NAME - 32) |
7324                       1 << (OP_LAYOUTRETURN - 32) |
7325                       1 << (OP_TEST_STATEID - 32) |
7326                       1 << (OP_FREE_STATEID - 32) |
7327                       1 << (OP_WRITE - 32)
7328         }
7329 };
7330
7331 /*
7332  * Select the state protection mode for client `clp' given the server results
7333  * from exchange_id in `sp'.
7334  *
7335  * Returns 0 on success, negative errno otherwise.
7336  */
7337 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7338                                  struct nfs41_state_protection *sp)
7339 {
7340         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7341                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7342                       1 << (OP_EXCHANGE_ID - 32) |
7343                       1 << (OP_CREATE_SESSION - 32) |
7344                       1 << (OP_DESTROY_SESSION - 32) |
7345                       1 << (OP_DESTROY_CLIENTID - 32)
7346         };
7347         unsigned long flags = 0;
7348         unsigned int i;
7349         int ret = 0;
7350
7351         if (sp->how == SP4_MACH_CRED) {
7352                 /* Print state protect result */
7353                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7354                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7355                         if (test_bit(i, sp->enforce.u.longs))
7356                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7357                         if (test_bit(i, sp->allow.u.longs))
7358                                 dfprintk(MOUNT, "  allow op %d\n", i);
7359                 }
7360
7361                 /* make sure nothing is on enforce list that isn't supported */
7362                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7363                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7364                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7365                                 ret = -EINVAL;
7366                                 goto out;
7367                         }
7368                 }
7369
7370                 /*
7371                  * Minimal mode - state operations are allowed to use machine
7372                  * credential.  Note this already happens by default, so the
7373                  * client doesn't have to do anything more than the negotiation.
7374                  *
7375                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7376                  *       we're already using the machine cred for exchange_id
7377                  *       and will never use a different cred.
7378                  */
7379                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7380                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7381                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7382                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7383                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7384                         dfprintk(MOUNT, "  minimal mode enabled\n");
7385                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7386                 } else {
7387                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7388                         ret = -EINVAL;
7389                         goto out;
7390                 }
7391
7392                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7393                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7394                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7395                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7396                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7397                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7398                 }
7399
7400                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7401                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7402                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7403                 }
7404
7405                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7406                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7407                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7408                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7409                 }
7410
7411                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7412                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7413                         dfprintk(MOUNT, "  stateid mode enabled\n");
7414                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7415                 }
7416
7417                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7418                         dfprintk(MOUNT, "  write mode enabled\n");
7419                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7420                 }
7421
7422                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7423                         dfprintk(MOUNT, "  commit mode enabled\n");
7424                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7425                 }
7426         }
7427 out:
7428         clp->cl_sp4_flags = flags;
7429         return 0;
7430 }
7431
7432 struct nfs41_exchange_id_data {
7433         struct nfs41_exchange_id_res res;
7434         struct nfs41_exchange_id_args args;
7435 };
7436
7437 static void nfs4_exchange_id_release(void *data)
7438 {
7439         struct nfs41_exchange_id_data *cdata =
7440                                         (struct nfs41_exchange_id_data *)data;
7441
7442         nfs_put_client(cdata->args.client);
7443         kfree(cdata->res.impl_id);
7444         kfree(cdata->res.server_scope);
7445         kfree(cdata->res.server_owner);
7446         kfree(cdata);
7447 }
7448
7449 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7450         .rpc_release = nfs4_exchange_id_release,
7451 };
7452
7453 /*
7454  * _nfs4_proc_exchange_id()
7455  *
7456  * Wrapper for EXCHANGE_ID operation.
7457  */
7458 static struct rpc_task *
7459 nfs4_run_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7460                         u32 sp4_how, struct rpc_xprt *xprt)
7461 {
7462         struct rpc_message msg = {
7463                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7464                 .rpc_cred = cred,
7465         };
7466         struct rpc_task_setup task_setup_data = {
7467                 .rpc_client = clp->cl_rpcclient,
7468                 .callback_ops = &nfs4_exchange_id_call_ops,
7469                 .rpc_message = &msg,
7470                 .flags = RPC_TASK_TIMEOUT,
7471         };
7472         struct nfs41_exchange_id_data *calldata;
7473         int status;
7474
7475         if (!atomic_inc_not_zero(&clp->cl_count))
7476                 return ERR_PTR(-EIO);
7477
7478         status = -ENOMEM;
7479         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7480         if (!calldata)
7481                 goto out;
7482
7483         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7484
7485         status = nfs4_init_uniform_client_string(clp);
7486         if (status)
7487                 goto out_calldata;
7488
7489         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7490                                                 GFP_NOFS);
7491         status = -ENOMEM;
7492         if (unlikely(calldata->res.server_owner == NULL))
7493                 goto out_calldata;
7494
7495         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7496                                         GFP_NOFS);
7497         if (unlikely(calldata->res.server_scope == NULL))
7498                 goto out_server_owner;
7499
7500         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7501         if (unlikely(calldata->res.impl_id == NULL))
7502                 goto out_server_scope;
7503
7504         switch (sp4_how) {
7505         case SP4_NONE:
7506                 calldata->args.state_protect.how = SP4_NONE;
7507                 break;
7508
7509         case SP4_MACH_CRED:
7510                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7511                 break;
7512
7513         default:
7514                 /* unsupported! */
7515                 WARN_ON_ONCE(1);
7516                 status = -EINVAL;
7517                 goto out_impl_id;
7518         }
7519         if (xprt) {
7520                 task_setup_data.rpc_xprt = xprt;
7521                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7522                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7523                                 sizeof(calldata->args.verifier.data));
7524         }
7525         calldata->args.client = clp;
7526         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7527         EXCHGID4_FLAG_BIND_PRINC_STATEID;
7528 #ifdef CONFIG_NFS_V4_1_MIGRATION
7529         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7530 #endif
7531         msg.rpc_argp = &calldata->args;
7532         msg.rpc_resp = &calldata->res;
7533         task_setup_data.callback_data = calldata;
7534
7535         return rpc_run_task(&task_setup_data);
7536
7537 out_impl_id:
7538         kfree(calldata->res.impl_id);
7539 out_server_scope:
7540         kfree(calldata->res.server_scope);
7541 out_server_owner:
7542         kfree(calldata->res.server_owner);
7543 out_calldata:
7544         kfree(calldata);
7545 out:
7546         nfs_put_client(clp);
7547         return ERR_PTR(status);
7548 }
7549
7550 /*
7551  * _nfs4_proc_exchange_id()
7552  *
7553  * Wrapper for EXCHANGE_ID operation.
7554  */
7555 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7556                         u32 sp4_how)
7557 {
7558         struct rpc_task *task;
7559         struct nfs41_exchange_id_args *argp;
7560         struct nfs41_exchange_id_res *resp;
7561         int status;
7562
7563         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
7564         if (IS_ERR(task))
7565                 return PTR_ERR(task);
7566
7567         argp = task->tk_msg.rpc_argp;
7568         resp = task->tk_msg.rpc_resp;
7569         status = task->tk_status;
7570         if (status  != 0)
7571                 goto out;
7572
7573         status = nfs4_check_cl_exchange_flags(resp->flags);
7574         if (status  != 0)
7575                 goto out;
7576
7577         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
7578         if (status != 0)
7579                 goto out;
7580
7581         clp->cl_clientid = resp->clientid;
7582         clp->cl_exchange_flags = resp->flags;
7583         clp->cl_seqid = resp->seqid;
7584         /* Client ID is not confirmed */
7585         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
7586                 clear_bit(NFS4_SESSION_ESTABLISHED,
7587                           &clp->cl_session->session_state);
7588
7589         if (clp->cl_serverscope != NULL &&
7590             !nfs41_same_server_scope(clp->cl_serverscope,
7591                                 resp->server_scope)) {
7592                 dprintk("%s: server_scope mismatch detected\n",
7593                         __func__);
7594                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7595         }
7596
7597         swap(clp->cl_serverowner, resp->server_owner);
7598         swap(clp->cl_serverscope, resp->server_scope);
7599         swap(clp->cl_implid, resp->impl_id);
7600
7601         /* Save the EXCHANGE_ID verifier session trunk tests */
7602         memcpy(clp->cl_confirm.data, argp->verifier.data,
7603                sizeof(clp->cl_confirm.data));
7604 out:
7605         trace_nfs4_exchange_id(clp, status);
7606         rpc_put_task(task);
7607         return status;
7608 }
7609
7610 /*
7611  * nfs4_proc_exchange_id()
7612  *
7613  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7614  *
7615  * Since the clientid has expired, all compounds using sessions
7616  * associated with the stale clientid will be returning
7617  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7618  * be in some phase of session reset.
7619  *
7620  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7621  */
7622 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7623 {
7624         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7625         int status;
7626
7627         /* try SP4_MACH_CRED if krb5i/p */
7628         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7629             authflavor == RPC_AUTH_GSS_KRB5P) {
7630                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7631                 if (!status)
7632                         return 0;
7633         }
7634
7635         /* try SP4_NONE */
7636         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7637 }
7638
7639 /**
7640  * nfs4_test_session_trunk
7641  *
7642  * This is an add_xprt_test() test function called from
7643  * rpc_clnt_setup_test_and_add_xprt.
7644  *
7645  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7646  * and is dereferrenced in nfs4_exchange_id_release
7647  *
7648  * Upon success, add the new transport to the rpc_clnt
7649  *
7650  * @clnt: struct rpc_clnt to get new transport
7651  * @xprt: the rpc_xprt to test
7652  * @data: call data for _nfs4_proc_exchange_id.
7653  */
7654 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7655                             void *data)
7656 {
7657         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7658         struct rpc_task *task;
7659         int status;
7660
7661         u32 sp4_how;
7662
7663         dprintk("--> %s try %s\n", __func__,
7664                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7665
7666         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7667
7668         /* Test connection for session trunking. Async exchange_id call */
7669         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7670         if (IS_ERR(task))
7671                 return PTR_ERR(task);
7672
7673         status = task->tk_status;
7674         if (status == 0)
7675                 status = nfs4_detect_session_trunking(adata->clp,
7676                                 task->tk_msg.rpc_resp, xprt);
7677
7678         rpc_put_task(task);
7679         return status;
7680 }
7681 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7682
7683 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7684                 struct rpc_cred *cred)
7685 {
7686         struct rpc_message msg = {
7687                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7688                 .rpc_argp = clp,
7689                 .rpc_cred = cred,
7690         };
7691         int status;
7692
7693         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7694         trace_nfs4_destroy_clientid(clp, status);
7695         if (status)
7696                 dprintk("NFS: Got error %d from the server %s on "
7697                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7698         return status;
7699 }
7700
7701 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7702                 struct rpc_cred *cred)
7703 {
7704         unsigned int loop;
7705         int ret;
7706
7707         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7708                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7709                 switch (ret) {
7710                 case -NFS4ERR_DELAY:
7711                 case -NFS4ERR_CLIENTID_BUSY:
7712                         ssleep(1);
7713                         break;
7714                 default:
7715                         return ret;
7716                 }
7717         }
7718         return 0;
7719 }
7720
7721 int nfs4_destroy_clientid(struct nfs_client *clp)
7722 {
7723         struct rpc_cred *cred;
7724         int ret = 0;
7725
7726         if (clp->cl_mvops->minor_version < 1)
7727                 goto out;
7728         if (clp->cl_exchange_flags == 0)
7729                 goto out;
7730         if (clp->cl_preserve_clid)
7731                 goto out;
7732         cred = nfs4_get_clid_cred(clp);
7733         ret = nfs4_proc_destroy_clientid(clp, cred);
7734         if (cred)
7735                 put_rpccred(cred);
7736         switch (ret) {
7737         case 0:
7738         case -NFS4ERR_STALE_CLIENTID:
7739                 clp->cl_exchange_flags = 0;
7740         }
7741 out:
7742         return ret;
7743 }
7744
7745 struct nfs4_get_lease_time_data {
7746         struct nfs4_get_lease_time_args *args;
7747         struct nfs4_get_lease_time_res *res;
7748         struct nfs_client *clp;
7749 };
7750
7751 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7752                                         void *calldata)
7753 {
7754         struct nfs4_get_lease_time_data *data =
7755                         (struct nfs4_get_lease_time_data *)calldata;
7756
7757         dprintk("--> %s\n", __func__);
7758         /* just setup sequence, do not trigger session recovery
7759            since we're invoked within one */
7760         nfs4_setup_sequence(data->clp,
7761                         &data->args->la_seq_args,
7762                         &data->res->lr_seq_res,
7763                         task);
7764         dprintk("<-- %s\n", __func__);
7765 }
7766
7767 /*
7768  * Called from nfs4_state_manager thread for session setup, so don't recover
7769  * from sequence operation or clientid errors.
7770  */
7771 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7772 {
7773         struct nfs4_get_lease_time_data *data =
7774                         (struct nfs4_get_lease_time_data *)calldata;
7775
7776         dprintk("--> %s\n", __func__);
7777         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7778                 return;
7779         switch (task->tk_status) {
7780         case -NFS4ERR_DELAY:
7781         case -NFS4ERR_GRACE:
7782                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7783                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7784                 task->tk_status = 0;
7785                 /* fall through */
7786         case -NFS4ERR_RETRY_UNCACHED_REP:
7787                 rpc_restart_call_prepare(task);
7788                 return;
7789         }
7790         dprintk("<-- %s\n", __func__);
7791 }
7792
7793 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7794         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7795         .rpc_call_done = nfs4_get_lease_time_done,
7796 };
7797
7798 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7799 {
7800         struct rpc_task *task;
7801         struct nfs4_get_lease_time_args args;
7802         struct nfs4_get_lease_time_res res = {
7803                 .lr_fsinfo = fsinfo,
7804         };
7805         struct nfs4_get_lease_time_data data = {
7806                 .args = &args,
7807                 .res = &res,
7808                 .clp = clp,
7809         };
7810         struct rpc_message msg = {
7811                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7812                 .rpc_argp = &args,
7813                 .rpc_resp = &res,
7814         };
7815         struct rpc_task_setup task_setup = {
7816                 .rpc_client = clp->cl_rpcclient,
7817                 .rpc_message = &msg,
7818                 .callback_ops = &nfs4_get_lease_time_ops,
7819                 .callback_data = &data,
7820                 .flags = RPC_TASK_TIMEOUT,
7821         };
7822         int status;
7823
7824         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7825         nfs4_set_sequence_privileged(&args.la_seq_args);
7826         task = rpc_run_task(&task_setup);
7827
7828         if (IS_ERR(task))
7829                 return PTR_ERR(task);
7830
7831         status = task->tk_status;
7832         rpc_put_task(task);
7833         return status;
7834 }
7835
7836 /*
7837  * Initialize the values to be used by the client in CREATE_SESSION
7838  * If nfs4_init_session set the fore channel request and response sizes,
7839  * use them.
7840  *
7841  * Set the back channel max_resp_sz_cached to zero to force the client to
7842  * always set csa_cachethis to FALSE because the current implementation
7843  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7844  */
7845 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7846                                     struct rpc_clnt *clnt)
7847 {
7848         unsigned int max_rqst_sz, max_resp_sz;
7849         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7850
7851         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7852         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7853
7854         /* Fore channel attributes */
7855         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7856         args->fc_attrs.max_resp_sz = max_resp_sz;
7857         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7858         args->fc_attrs.max_reqs = max_session_slots;
7859
7860         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7861                 "max_ops=%u max_reqs=%u\n",
7862                 __func__,
7863                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7864                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7865
7866         /* Back channel attributes */
7867         args->bc_attrs.max_rqst_sz = max_bc_payload;
7868         args->bc_attrs.max_resp_sz = max_bc_payload;
7869         args->bc_attrs.max_resp_sz_cached = 0;
7870         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7871         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7872
7873         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7874                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7875                 __func__,
7876                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7877                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7878                 args->bc_attrs.max_reqs);
7879 }
7880
7881 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7882                 struct nfs41_create_session_res *res)
7883 {
7884         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7885         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7886
7887         if (rcvd->max_resp_sz > sent->max_resp_sz)
7888                 return -EINVAL;
7889         /*
7890          * Our requested max_ops is the minimum we need; we're not
7891          * prepared to break up compounds into smaller pieces than that.
7892          * So, no point even trying to continue if the server won't
7893          * cooperate:
7894          */
7895         if (rcvd->max_ops < sent->max_ops)
7896                 return -EINVAL;
7897         if (rcvd->max_reqs == 0)
7898                 return -EINVAL;
7899         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7900                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7901         return 0;
7902 }
7903
7904 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7905                 struct nfs41_create_session_res *res)
7906 {
7907         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7908         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7909
7910         if (!(res->flags & SESSION4_BACK_CHAN))
7911                 goto out;
7912         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7913                 return -EINVAL;
7914         if (rcvd->max_resp_sz < sent->max_resp_sz)
7915                 return -EINVAL;
7916         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7917                 return -EINVAL;
7918         if (rcvd->max_ops > sent->max_ops)
7919                 return -EINVAL;
7920         if (rcvd->max_reqs > sent->max_reqs)
7921                 return -EINVAL;
7922 out:
7923         return 0;
7924 }
7925
7926 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7927                                      struct nfs41_create_session_res *res)
7928 {
7929         int ret;
7930
7931         ret = nfs4_verify_fore_channel_attrs(args, res);
7932         if (ret)
7933                 return ret;
7934         return nfs4_verify_back_channel_attrs(args, res);
7935 }
7936
7937 static void nfs4_update_session(struct nfs4_session *session,
7938                 struct nfs41_create_session_res *res)
7939 {
7940         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7941         /* Mark client id and session as being confirmed */
7942         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7943         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7944         session->flags = res->flags;
7945         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7946         if (res->flags & SESSION4_BACK_CHAN)
7947                 memcpy(&session->bc_attrs, &res->bc_attrs,
7948                                 sizeof(session->bc_attrs));
7949 }
7950
7951 static int _nfs4_proc_create_session(struct nfs_client *clp,
7952                 struct rpc_cred *cred)
7953 {
7954         struct nfs4_session *session = clp->cl_session;
7955         struct nfs41_create_session_args args = {
7956                 .client = clp,
7957                 .clientid = clp->cl_clientid,
7958                 .seqid = clp->cl_seqid,
7959                 .cb_program = NFS4_CALLBACK,
7960         };
7961         struct nfs41_create_session_res res;
7962
7963         struct rpc_message msg = {
7964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7965                 .rpc_argp = &args,
7966                 .rpc_resp = &res,
7967                 .rpc_cred = cred,
7968         };
7969         int status;
7970
7971         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7972         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7973
7974         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7975         trace_nfs4_create_session(clp, status);
7976
7977         switch (status) {
7978         case -NFS4ERR_STALE_CLIENTID:
7979         case -NFS4ERR_DELAY:
7980         case -ETIMEDOUT:
7981         case -EACCES:
7982         case -EAGAIN:
7983                 goto out;
7984         };
7985
7986         clp->cl_seqid++;
7987         if (!status) {
7988                 /* Verify the session's negotiated channel_attrs values */
7989                 status = nfs4_verify_channel_attrs(&args, &res);
7990                 /* Increment the clientid slot sequence id */
7991                 if (status)
7992                         goto out;
7993                 nfs4_update_session(session, &res);
7994         }
7995 out:
7996         return status;
7997 }
7998
7999 /*
8000  * Issues a CREATE_SESSION operation to the server.
8001  * It is the responsibility of the caller to verify the session is
8002  * expired before calling this routine.
8003  */
8004 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8005 {
8006         int status;
8007         unsigned *ptr;
8008         struct nfs4_session *session = clp->cl_session;
8009
8010         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8011
8012         status = _nfs4_proc_create_session(clp, cred);
8013         if (status)
8014                 goto out;
8015
8016         /* Init or reset the session slot tables */
8017         status = nfs4_setup_session_slot_tables(session);
8018         dprintk("slot table setup returned %d\n", status);
8019         if (status)
8020                 goto out;
8021
8022         ptr = (unsigned *)&session->sess_id.data[0];
8023         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8024                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8025 out:
8026         dprintk("<-- %s\n", __func__);
8027         return status;
8028 }
8029
8030 /*
8031  * Issue the over-the-wire RPC DESTROY_SESSION.
8032  * The caller must serialize access to this routine.
8033  */
8034 int nfs4_proc_destroy_session(struct nfs4_session *session,
8035                 struct rpc_cred *cred)
8036 {
8037         struct rpc_message msg = {
8038                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8039                 .rpc_argp = session,
8040                 .rpc_cred = cred,
8041         };
8042         int status = 0;
8043
8044         dprintk("--> nfs4_proc_destroy_session\n");
8045
8046         /* session is still being setup */
8047         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8048                 return 0;
8049
8050         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8051         trace_nfs4_destroy_session(session->clp, status);
8052
8053         if (status)
8054                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8055                         "Session has been destroyed regardless...\n", status);
8056
8057         dprintk("<-- nfs4_proc_destroy_session\n");
8058         return status;
8059 }
8060
8061 /*
8062  * Renew the cl_session lease.
8063  */
8064 struct nfs4_sequence_data {
8065         struct nfs_client *clp;
8066         struct nfs4_sequence_args args;
8067         struct nfs4_sequence_res res;
8068 };
8069
8070 static void nfs41_sequence_release(void *data)
8071 {
8072         struct nfs4_sequence_data *calldata = data;
8073         struct nfs_client *clp = calldata->clp;
8074
8075         if (atomic_read(&clp->cl_count) > 1)
8076                 nfs4_schedule_state_renewal(clp);
8077         nfs_put_client(clp);
8078         kfree(calldata);
8079 }
8080
8081 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8082 {
8083         switch(task->tk_status) {
8084         case -NFS4ERR_DELAY:
8085                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8086                 return -EAGAIN;
8087         default:
8088                 nfs4_schedule_lease_recovery(clp);
8089         }
8090         return 0;
8091 }
8092
8093 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8094 {
8095         struct nfs4_sequence_data *calldata = data;
8096         struct nfs_client *clp = calldata->clp;
8097
8098         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8099                 return;
8100
8101         trace_nfs4_sequence(clp, task->tk_status);
8102         if (task->tk_status < 0) {
8103                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8104                 if (atomic_read(&clp->cl_count) == 1)
8105                         goto out;
8106
8107                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8108                         rpc_restart_call_prepare(task);
8109                         return;
8110                 }
8111         }
8112         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8113 out:
8114         dprintk("<-- %s\n", __func__);
8115 }
8116
8117 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8118 {
8119         struct nfs4_sequence_data *calldata = data;
8120         struct nfs_client *clp = calldata->clp;
8121         struct nfs4_sequence_args *args;
8122         struct nfs4_sequence_res *res;
8123
8124         args = task->tk_msg.rpc_argp;
8125         res = task->tk_msg.rpc_resp;
8126
8127         nfs4_setup_sequence(clp, args, res, task);
8128 }
8129
8130 static const struct rpc_call_ops nfs41_sequence_ops = {
8131         .rpc_call_done = nfs41_sequence_call_done,
8132         .rpc_call_prepare = nfs41_sequence_prepare,
8133         .rpc_release = nfs41_sequence_release,
8134 };
8135
8136 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8137                 struct rpc_cred *cred,
8138                 bool is_privileged)
8139 {
8140         struct nfs4_sequence_data *calldata;
8141         struct rpc_message msg = {
8142                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8143                 .rpc_cred = cred,
8144         };
8145         struct rpc_task_setup task_setup_data = {
8146                 .rpc_client = clp->cl_rpcclient,
8147                 .rpc_message = &msg,
8148                 .callback_ops = &nfs41_sequence_ops,
8149                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8150         };
8151
8152         if (!atomic_inc_not_zero(&clp->cl_count))
8153                 return ERR_PTR(-EIO);
8154         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8155         if (calldata == NULL) {
8156                 nfs_put_client(clp);
8157                 return ERR_PTR(-ENOMEM);
8158         }
8159         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8160         if (is_privileged)
8161                 nfs4_set_sequence_privileged(&calldata->args);
8162         msg.rpc_argp = &calldata->args;
8163         msg.rpc_resp = &calldata->res;
8164         calldata->clp = clp;
8165         task_setup_data.callback_data = calldata;
8166
8167         return rpc_run_task(&task_setup_data);
8168 }
8169
8170 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8171 {
8172         struct rpc_task *task;
8173         int ret = 0;
8174
8175         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8176                 return -EAGAIN;
8177         task = _nfs41_proc_sequence(clp, cred, false);
8178         if (IS_ERR(task))
8179                 ret = PTR_ERR(task);
8180         else
8181                 rpc_put_task_async(task);
8182         dprintk("<-- %s status=%d\n", __func__, ret);
8183         return ret;
8184 }
8185
8186 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8187 {
8188         struct rpc_task *task;
8189         int ret;
8190
8191         task = _nfs41_proc_sequence(clp, cred, true);
8192         if (IS_ERR(task)) {
8193                 ret = PTR_ERR(task);
8194                 goto out;
8195         }
8196         ret = rpc_wait_for_completion_task(task);
8197         if (!ret)
8198                 ret = task->tk_status;
8199         rpc_put_task(task);
8200 out:
8201         dprintk("<-- %s status=%d\n", __func__, ret);
8202         return ret;
8203 }
8204
8205 struct nfs4_reclaim_complete_data {
8206         struct nfs_client *clp;
8207         struct nfs41_reclaim_complete_args arg;
8208         struct nfs41_reclaim_complete_res res;
8209 };
8210
8211 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8212 {
8213         struct nfs4_reclaim_complete_data *calldata = data;
8214
8215         nfs4_setup_sequence(calldata->clp,
8216                         &calldata->arg.seq_args,
8217                         &calldata->res.seq_res,
8218                         task);
8219 }
8220
8221 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8222 {
8223         switch(task->tk_status) {
8224         case 0:
8225         case -NFS4ERR_COMPLETE_ALREADY:
8226         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8227                 break;
8228         case -NFS4ERR_DELAY:
8229                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8230                 /* fall through */
8231         case -NFS4ERR_RETRY_UNCACHED_REP:
8232                 return -EAGAIN;
8233         case -NFS4ERR_BADSESSION:
8234         case -NFS4ERR_DEADSESSION:
8235         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8236                 nfs4_schedule_session_recovery(clp->cl_session,
8237                                 task->tk_status);
8238                 break;
8239         default:
8240                 nfs4_schedule_lease_recovery(clp);
8241         }
8242         return 0;
8243 }
8244
8245 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8246 {
8247         struct nfs4_reclaim_complete_data *calldata = data;
8248         struct nfs_client *clp = calldata->clp;
8249         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8250
8251         dprintk("--> %s\n", __func__);
8252         if (!nfs41_sequence_done(task, res))
8253                 return;
8254
8255         trace_nfs4_reclaim_complete(clp, task->tk_status);
8256         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8257                 rpc_restart_call_prepare(task);
8258                 return;
8259         }
8260         dprintk("<-- %s\n", __func__);
8261 }
8262
8263 static void nfs4_free_reclaim_complete_data(void *data)
8264 {
8265         struct nfs4_reclaim_complete_data *calldata = data;
8266
8267         kfree(calldata);
8268 }
8269
8270 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8271         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8272         .rpc_call_done = nfs4_reclaim_complete_done,
8273         .rpc_release = nfs4_free_reclaim_complete_data,
8274 };
8275
8276 /*
8277  * Issue a global reclaim complete.
8278  */
8279 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8280                 struct rpc_cred *cred)
8281 {
8282         struct nfs4_reclaim_complete_data *calldata;
8283         struct rpc_task *task;
8284         struct rpc_message msg = {
8285                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8286                 .rpc_cred = cred,
8287         };
8288         struct rpc_task_setup task_setup_data = {
8289                 .rpc_client = clp->cl_rpcclient,
8290                 .rpc_message = &msg,
8291                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8292                 .flags = RPC_TASK_ASYNC,
8293         };
8294         int status = -ENOMEM;
8295
8296         dprintk("--> %s\n", __func__);
8297         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8298         if (calldata == NULL)
8299                 goto out;
8300         calldata->clp = clp;
8301         calldata->arg.one_fs = 0;
8302
8303         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8304         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8305         msg.rpc_argp = &calldata->arg;
8306         msg.rpc_resp = &calldata->res;
8307         task_setup_data.callback_data = calldata;
8308         task = rpc_run_task(&task_setup_data);
8309         if (IS_ERR(task)) {
8310                 status = PTR_ERR(task);
8311                 goto out;
8312         }
8313         status = rpc_wait_for_completion_task(task);
8314         if (status == 0)
8315                 status = task->tk_status;
8316         rpc_put_task(task);
8317 out:
8318         dprintk("<-- %s status=%d\n", __func__, status);
8319         return status;
8320 }
8321
8322 static void
8323 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8324 {
8325         struct nfs4_layoutget *lgp = calldata;
8326         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8327
8328         dprintk("--> %s\n", __func__);
8329         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8330                                 &lgp->res.seq_res, task);
8331         dprintk("<-- %s\n", __func__);
8332 }
8333
8334 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8335 {
8336         struct nfs4_layoutget *lgp = calldata;
8337
8338         dprintk("--> %s\n", __func__);
8339         nfs41_sequence_process(task, &lgp->res.seq_res);
8340         dprintk("<-- %s\n", __func__);
8341 }
8342
8343 static int
8344 nfs4_layoutget_handle_exception(struct rpc_task *task,
8345                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8346 {
8347         struct inode *inode = lgp->args.inode;
8348         struct nfs_server *server = NFS_SERVER(inode);
8349         struct pnfs_layout_hdr *lo;
8350         int nfs4err = task->tk_status;
8351         int err, status = 0;
8352         LIST_HEAD(head);
8353
8354         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8355
8356         switch (nfs4err) {
8357         case 0:
8358                 goto out;
8359
8360         /*
8361          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8362          * on the file. set tk_status to -ENODATA to tell upper layer to
8363          * retry go inband.
8364          */
8365         case -NFS4ERR_LAYOUTUNAVAILABLE:
8366                 status = -ENODATA;
8367                 goto out;
8368         /*
8369          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8370          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8371          */
8372         case -NFS4ERR_BADLAYOUT:
8373                 status = -EOVERFLOW;
8374                 goto out;
8375         /*
8376          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8377          * (or clients) writing to the same RAID stripe except when
8378          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8379          *
8380          * Treat it like we would RECALLCONFLICT -- we retry for a little
8381          * while, and then eventually give up.
8382          */
8383         case -NFS4ERR_LAYOUTTRYLATER:
8384                 if (lgp->args.minlength == 0) {
8385                         status = -EOVERFLOW;
8386                         goto out;
8387                 }
8388                 status = -EBUSY;
8389                 break;
8390         case -NFS4ERR_RECALLCONFLICT:
8391                 status = -ERECALLCONFLICT;
8392                 break;
8393         case -NFS4ERR_DELEG_REVOKED:
8394         case -NFS4ERR_ADMIN_REVOKED:
8395         case -NFS4ERR_EXPIRED:
8396         case -NFS4ERR_BAD_STATEID:
8397                 exception->timeout = 0;
8398                 spin_lock(&inode->i_lock);
8399                 lo = NFS_I(inode)->layout;
8400                 /* If the open stateid was bad, then recover it. */
8401                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8402                     nfs4_stateid_match_other(&lgp->args.stateid,
8403                                         &lgp->args.ctx->state->stateid)) {
8404                         spin_unlock(&inode->i_lock);
8405                         exception->state = lgp->args.ctx->state;
8406                         exception->stateid = &lgp->args.stateid;
8407                         break;
8408                 }
8409
8410                 /*
8411                  * Mark the bad layout state as invalid, then retry
8412                  */
8413                 pnfs_mark_layout_stateid_invalid(lo, &head);
8414                 spin_unlock(&inode->i_lock);
8415                 nfs_commit_inode(inode, 0);
8416                 pnfs_free_lseg_list(&head);
8417                 status = -EAGAIN;
8418                 goto out;
8419         }
8420
8421         nfs4_sequence_free_slot(&lgp->res.seq_res);
8422         err = nfs4_handle_exception(server, nfs4err, exception);
8423         if (!status) {
8424                 if (exception->retry)
8425                         status = -EAGAIN;
8426                 else
8427                         status = err;
8428         }
8429 out:
8430         dprintk("<-- %s\n", __func__);
8431         return status;
8432 }
8433
8434 static size_t max_response_pages(struct nfs_server *server)
8435 {
8436         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8437         return nfs_page_array_len(0, max_resp_sz);
8438 }
8439
8440 static void nfs4_free_pages(struct page **pages, size_t size)
8441 {
8442         int i;
8443
8444         if (!pages)
8445                 return;
8446
8447         for (i = 0; i < size; i++) {
8448                 if (!pages[i])
8449                         break;
8450                 __free_page(pages[i]);
8451         }
8452         kfree(pages);
8453 }
8454
8455 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8456 {
8457         struct page **pages;
8458         int i;
8459
8460         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8461         if (!pages) {
8462                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8463                 return NULL;
8464         }
8465
8466         for (i = 0; i < size; i++) {
8467                 pages[i] = alloc_page(gfp_flags);
8468                 if (!pages[i]) {
8469                         dprintk("%s: failed to allocate page\n", __func__);
8470                         nfs4_free_pages(pages, size);
8471                         return NULL;
8472                 }
8473         }
8474
8475         return pages;
8476 }
8477
8478 static void nfs4_layoutget_release(void *calldata)
8479 {
8480         struct nfs4_layoutget *lgp = calldata;
8481         struct inode *inode = lgp->args.inode;
8482         struct nfs_server *server = NFS_SERVER(inode);
8483         size_t max_pages = max_response_pages(server);
8484
8485         dprintk("--> %s\n", __func__);
8486         nfs4_sequence_free_slot(&lgp->res.seq_res);
8487         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8488         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8489         put_nfs_open_context(lgp->args.ctx);
8490         kfree(calldata);
8491         dprintk("<-- %s\n", __func__);
8492 }
8493
8494 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8495         .rpc_call_prepare = nfs4_layoutget_prepare,
8496         .rpc_call_done = nfs4_layoutget_done,
8497         .rpc_release = nfs4_layoutget_release,
8498 };
8499
8500 struct pnfs_layout_segment *
8501 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8502 {
8503         struct inode *inode = lgp->args.inode;
8504         struct nfs_server *server = NFS_SERVER(inode);
8505         size_t max_pages = max_response_pages(server);
8506         struct rpc_task *task;
8507         struct rpc_message msg = {
8508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8509                 .rpc_argp = &lgp->args,
8510                 .rpc_resp = &lgp->res,
8511                 .rpc_cred = lgp->cred,
8512         };
8513         struct rpc_task_setup task_setup_data = {
8514                 .rpc_client = server->client,
8515                 .rpc_message = &msg,
8516                 .callback_ops = &nfs4_layoutget_call_ops,
8517                 .callback_data = lgp,
8518                 .flags = RPC_TASK_ASYNC,
8519         };
8520         struct pnfs_layout_segment *lseg = NULL;
8521         struct nfs4_exception exception = {
8522                 .inode = inode,
8523                 .timeout = *timeout,
8524         };
8525         int status = 0;
8526
8527         dprintk("--> %s\n", __func__);
8528
8529         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8530         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8531
8532         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8533         if (!lgp->args.layout.pages) {
8534                 nfs4_layoutget_release(lgp);
8535                 return ERR_PTR(-ENOMEM);
8536         }
8537         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8538
8539         lgp->res.layoutp = &lgp->args.layout;
8540         lgp->res.seq_res.sr_slot = NULL;
8541         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8542
8543         task = rpc_run_task(&task_setup_data);
8544         if (IS_ERR(task))
8545                 return ERR_CAST(task);
8546         status = rpc_wait_for_completion_task(task);
8547         if (status == 0) {
8548                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8549                 *timeout = exception.timeout;
8550         }
8551
8552         trace_nfs4_layoutget(lgp->args.ctx,
8553                         &lgp->args.range,
8554                         &lgp->res.range,
8555                         &lgp->res.stateid,
8556                         status);
8557
8558         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8559         if (status == 0 && lgp->res.layoutp->len)
8560                 lseg = pnfs_layout_process(lgp);
8561         rpc_put_task(task);
8562         dprintk("<-- %s status=%d\n", __func__, status);
8563         if (status)
8564                 return ERR_PTR(status);
8565         return lseg;
8566 }
8567
8568 static void
8569 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8570 {
8571         struct nfs4_layoutreturn *lrp = calldata;
8572
8573         dprintk("--> %s\n", __func__);
8574         nfs4_setup_sequence(lrp->clp,
8575                         &lrp->args.seq_args,
8576                         &lrp->res.seq_res,
8577                         task);
8578 }
8579
8580 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8581 {
8582         struct nfs4_layoutreturn *lrp = calldata;
8583         struct nfs_server *server;
8584
8585         dprintk("--> %s\n", __func__);
8586
8587         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8588                 return;
8589
8590         server = NFS_SERVER(lrp->args.inode);
8591         switch (task->tk_status) {
8592         default:
8593                 task->tk_status = 0;
8594         case 0:
8595                 break;
8596         case -NFS4ERR_DELAY:
8597                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8598                         break;
8599                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8600                 rpc_restart_call_prepare(task);
8601                 return;
8602         }
8603         dprintk("<-- %s\n", __func__);
8604 }
8605
8606 static void nfs4_layoutreturn_release(void *calldata)
8607 {
8608         struct nfs4_layoutreturn *lrp = calldata;
8609         struct pnfs_layout_hdr *lo = lrp->args.layout;
8610
8611         dprintk("--> %s\n", __func__);
8612         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8613                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8614         nfs4_sequence_free_slot(&lrp->res.seq_res);
8615         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8616                 lrp->ld_private.ops->free(&lrp->ld_private);
8617         pnfs_put_layout_hdr(lrp->args.layout);
8618         nfs_iput_and_deactive(lrp->inode);
8619         kfree(calldata);
8620         dprintk("<-- %s\n", __func__);
8621 }
8622
8623 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8624         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8625         .rpc_call_done = nfs4_layoutreturn_done,
8626         .rpc_release = nfs4_layoutreturn_release,
8627 };
8628
8629 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8630 {
8631         struct rpc_task *task;
8632         struct rpc_message msg = {
8633                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8634                 .rpc_argp = &lrp->args,
8635                 .rpc_resp = &lrp->res,
8636                 .rpc_cred = lrp->cred,
8637         };
8638         struct rpc_task_setup task_setup_data = {
8639                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8640                 .rpc_message = &msg,
8641                 .callback_ops = &nfs4_layoutreturn_call_ops,
8642                 .callback_data = lrp,
8643         };
8644         int status = 0;
8645
8646         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8647                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8648                         &task_setup_data.rpc_client, &msg);
8649
8650         dprintk("--> %s\n", __func__);
8651         if (!sync) {
8652                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8653                 if (!lrp->inode) {
8654                         nfs4_layoutreturn_release(lrp);
8655                         return -EAGAIN;
8656                 }
8657                 task_setup_data.flags |= RPC_TASK_ASYNC;
8658         }
8659         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8660         task = rpc_run_task(&task_setup_data);
8661         if (IS_ERR(task))
8662                 return PTR_ERR(task);
8663         if (sync)
8664                 status = task->tk_status;
8665         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8666         dprintk("<-- %s status=%d\n", __func__, status);
8667         rpc_put_task(task);
8668         return status;
8669 }
8670
8671 static int
8672 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8673                 struct pnfs_device *pdev,
8674                 struct rpc_cred *cred)
8675 {
8676         struct nfs4_getdeviceinfo_args args = {
8677                 .pdev = pdev,
8678                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8679                         NOTIFY_DEVICEID4_DELETE,
8680         };
8681         struct nfs4_getdeviceinfo_res res = {
8682                 .pdev = pdev,
8683         };
8684         struct rpc_message msg = {
8685                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8686                 .rpc_argp = &args,
8687                 .rpc_resp = &res,
8688                 .rpc_cred = cred,
8689         };
8690         int status;
8691
8692         dprintk("--> %s\n", __func__);
8693         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8694         if (res.notification & ~args.notify_types)
8695                 dprintk("%s: unsupported notification\n", __func__);
8696         if (res.notification != args.notify_types)
8697                 pdev->nocache = 1;
8698
8699         dprintk("<-- %s status=%d\n", __func__, status);
8700
8701         return status;
8702 }
8703
8704 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8705                 struct pnfs_device *pdev,
8706                 struct rpc_cred *cred)
8707 {
8708         struct nfs4_exception exception = { };
8709         int err;
8710
8711         do {
8712                 err = nfs4_handle_exception(server,
8713                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8714                                         &exception);
8715         } while (exception.retry);
8716         return err;
8717 }
8718 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8719
8720 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8721 {
8722         struct nfs4_layoutcommit_data *data = calldata;
8723         struct nfs_server *server = NFS_SERVER(data->args.inode);
8724
8725         nfs4_setup_sequence(server->nfs_client,
8726                         &data->args.seq_args,
8727                         &data->res.seq_res,
8728                         task);
8729 }
8730
8731 static void
8732 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8733 {
8734         struct nfs4_layoutcommit_data *data = calldata;
8735         struct nfs_server *server = NFS_SERVER(data->args.inode);
8736
8737         if (!nfs41_sequence_done(task, &data->res.seq_res))
8738                 return;
8739
8740         switch (task->tk_status) { /* Just ignore these failures */
8741         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8742         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8743         case -NFS4ERR_BADLAYOUT:     /* no layout */
8744         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8745                 task->tk_status = 0;
8746         case 0:
8747                 break;
8748         default:
8749                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8750                         rpc_restart_call_prepare(task);
8751                         return;
8752                 }
8753         }
8754 }
8755
8756 static void nfs4_layoutcommit_release(void *calldata)
8757 {
8758         struct nfs4_layoutcommit_data *data = calldata;
8759
8760         pnfs_cleanup_layoutcommit(data);
8761         nfs_post_op_update_inode_force_wcc(data->args.inode,
8762                                            data->res.fattr);
8763         put_rpccred(data->cred);
8764         nfs_iput_and_deactive(data->inode);
8765         kfree(data);
8766 }
8767
8768 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8769         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8770         .rpc_call_done = nfs4_layoutcommit_done,
8771         .rpc_release = nfs4_layoutcommit_release,
8772 };
8773
8774 int
8775 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8776 {
8777         struct rpc_message msg = {
8778                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8779                 .rpc_argp = &data->args,
8780                 .rpc_resp = &data->res,
8781                 .rpc_cred = data->cred,
8782         };
8783         struct rpc_task_setup task_setup_data = {
8784                 .task = &data->task,
8785                 .rpc_client = NFS_CLIENT(data->args.inode),
8786                 .rpc_message = &msg,
8787                 .callback_ops = &nfs4_layoutcommit_ops,
8788                 .callback_data = data,
8789         };
8790         struct rpc_task *task;
8791         int status = 0;
8792
8793         dprintk("NFS: initiating layoutcommit call. sync %d "
8794                 "lbw: %llu inode %lu\n", sync,
8795                 data->args.lastbytewritten,
8796                 data->args.inode->i_ino);
8797
8798         if (!sync) {
8799                 data->inode = nfs_igrab_and_active(data->args.inode);
8800                 if (data->inode == NULL) {
8801                         nfs4_layoutcommit_release(data);
8802                         return -EAGAIN;
8803                 }
8804                 task_setup_data.flags = RPC_TASK_ASYNC;
8805         }
8806         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8807         task = rpc_run_task(&task_setup_data);
8808         if (IS_ERR(task))
8809                 return PTR_ERR(task);
8810         if (sync)
8811                 status = task->tk_status;
8812         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8813         dprintk("%s: status %d\n", __func__, status);
8814         rpc_put_task(task);
8815         return status;
8816 }
8817
8818 /**
8819  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8820  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8821  */
8822 static int
8823 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8824                     struct nfs_fsinfo *info,
8825                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8826 {
8827         struct nfs41_secinfo_no_name_args args = {
8828                 .style = SECINFO_STYLE_CURRENT_FH,
8829         };
8830         struct nfs4_secinfo_res res = {
8831                 .flavors = flavors,
8832         };
8833         struct rpc_message msg = {
8834                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8835                 .rpc_argp = &args,
8836                 .rpc_resp = &res,
8837         };
8838         struct rpc_clnt *clnt = server->client;
8839         struct rpc_cred *cred = NULL;
8840         int status;
8841
8842         if (use_integrity) {
8843                 clnt = server->nfs_client->cl_rpcclient;
8844                 cred = nfs4_get_clid_cred(server->nfs_client);
8845                 msg.rpc_cred = cred;
8846         }
8847
8848         dprintk("--> %s\n", __func__);
8849         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8850                                 &res.seq_res, 0);
8851         dprintk("<-- %s status=%d\n", __func__, status);
8852
8853         if (cred)
8854                 put_rpccred(cred);
8855
8856         return status;
8857 }
8858
8859 static int
8860 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8861                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8862 {
8863         struct nfs4_exception exception = { };
8864         int err;
8865         do {
8866                 /* first try using integrity protection */
8867                 err = -NFS4ERR_WRONGSEC;
8868
8869                 /* try to use integrity protection with machine cred */
8870                 if (_nfs4_is_integrity_protected(server->nfs_client))
8871                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8872                                                           flavors, true);
8873
8874                 /*
8875                  * if unable to use integrity protection, or SECINFO with
8876                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8877                  * disallowed by spec, but exists in deployed servers) use
8878                  * the current filesystem's rpc_client and the user cred.
8879                  */
8880                 if (err == -NFS4ERR_WRONGSEC)
8881                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8882                                                           flavors, false);
8883
8884                 switch (err) {
8885                 case 0:
8886                 case -NFS4ERR_WRONGSEC:
8887                 case -ENOTSUPP:
8888                         goto out;
8889                 default:
8890                         err = nfs4_handle_exception(server, err, &exception);
8891                 }
8892         } while (exception.retry);
8893 out:
8894         return err;
8895 }
8896
8897 static int
8898 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8899                     struct nfs_fsinfo *info)
8900 {
8901         int err;
8902         struct page *page;
8903         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8904         struct nfs4_secinfo_flavors *flavors;
8905         struct nfs4_secinfo4 *secinfo;
8906         int i;
8907
8908         page = alloc_page(GFP_KERNEL);
8909         if (!page) {
8910                 err = -ENOMEM;
8911                 goto out;
8912         }
8913
8914         flavors = page_address(page);
8915         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8916
8917         /*
8918          * Fall back on "guess and check" method if
8919          * the server doesn't support SECINFO_NO_NAME
8920          */
8921         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8922                 err = nfs4_find_root_sec(server, fhandle, info);
8923                 goto out_freepage;
8924         }
8925         if (err)
8926                 goto out_freepage;
8927
8928         for (i = 0; i < flavors->num_flavors; i++) {
8929                 secinfo = &flavors->flavors[i];
8930
8931                 switch (secinfo->flavor) {
8932                 case RPC_AUTH_NULL:
8933                 case RPC_AUTH_UNIX:
8934                 case RPC_AUTH_GSS:
8935                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8936                                         &secinfo->flavor_info);
8937                         break;
8938                 default:
8939                         flavor = RPC_AUTH_MAXFLAVOR;
8940                         break;
8941                 }
8942
8943                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8944                         flavor = RPC_AUTH_MAXFLAVOR;
8945
8946                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8947                         err = nfs4_lookup_root_sec(server, fhandle,
8948                                                    info, flavor);
8949                         if (!err)
8950                                 break;
8951                 }
8952         }
8953
8954         if (flavor == RPC_AUTH_MAXFLAVOR)
8955                 err = -EPERM;
8956
8957 out_freepage:
8958         put_page(page);
8959         if (err == -EACCES)
8960                 return -EPERM;
8961 out:
8962         return err;
8963 }
8964
8965 static int _nfs41_test_stateid(struct nfs_server *server,
8966                 nfs4_stateid *stateid,
8967                 struct rpc_cred *cred)
8968 {
8969         int status;
8970         struct nfs41_test_stateid_args args = {
8971                 .stateid = stateid,
8972         };
8973         struct nfs41_test_stateid_res res;
8974         struct rpc_message msg = {
8975                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8976                 .rpc_argp = &args,
8977                 .rpc_resp = &res,
8978                 .rpc_cred = cred,
8979         };
8980         struct rpc_clnt *rpc_client = server->client;
8981
8982         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8983                 &rpc_client, &msg);
8984
8985         dprintk("NFS call  test_stateid %p\n", stateid);
8986         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8987         nfs4_set_sequence_privileged(&args.seq_args);
8988         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8989                         &args.seq_args, &res.seq_res);
8990         if (status != NFS_OK) {
8991                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8992                 return status;
8993         }
8994         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8995         return -res.status;
8996 }
8997
8998 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8999                 int err, struct nfs4_exception *exception)
9000 {
9001         exception->retry = 0;
9002         switch(err) {
9003         case -NFS4ERR_DELAY:
9004         case -NFS4ERR_RETRY_UNCACHED_REP:
9005                 nfs4_handle_exception(server, err, exception);
9006                 break;
9007         case -NFS4ERR_BADSESSION:
9008         case -NFS4ERR_BADSLOT:
9009         case -NFS4ERR_BAD_HIGH_SLOT:
9010         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9011         case -NFS4ERR_DEADSESSION:
9012                 nfs4_do_handle_exception(server, err, exception);
9013         }
9014 }
9015
9016 /**
9017  * nfs41_test_stateid - perform a TEST_STATEID operation
9018  *
9019  * @server: server / transport on which to perform the operation
9020  * @stateid: state ID to test
9021  * @cred: credential
9022  *
9023  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9024  * Otherwise a negative NFS4ERR value is returned if the operation
9025  * failed or the state ID is not currently valid.
9026  */
9027 static int nfs41_test_stateid(struct nfs_server *server,
9028                 nfs4_stateid *stateid,
9029                 struct rpc_cred *cred)
9030 {
9031         struct nfs4_exception exception = { };
9032         int err;
9033         do {
9034                 err = _nfs41_test_stateid(server, stateid, cred);
9035                 nfs4_handle_delay_or_session_error(server, err, &exception);
9036         } while (exception.retry);
9037         return err;
9038 }
9039
9040 struct nfs_free_stateid_data {
9041         struct nfs_server *server;
9042         struct nfs41_free_stateid_args args;
9043         struct nfs41_free_stateid_res res;
9044 };
9045
9046 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9047 {
9048         struct nfs_free_stateid_data *data = calldata;
9049         nfs4_setup_sequence(data->server->nfs_client,
9050                         &data->args.seq_args,
9051                         &data->res.seq_res,
9052                         task);
9053 }
9054
9055 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9056 {
9057         struct nfs_free_stateid_data *data = calldata;
9058
9059         nfs41_sequence_done(task, &data->res.seq_res);
9060
9061         switch (task->tk_status) {
9062         case -NFS4ERR_DELAY:
9063                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9064                         rpc_restart_call_prepare(task);
9065         }
9066 }
9067
9068 static void nfs41_free_stateid_release(void *calldata)
9069 {
9070         kfree(calldata);
9071 }
9072
9073 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9074         .rpc_call_prepare = nfs41_free_stateid_prepare,
9075         .rpc_call_done = nfs41_free_stateid_done,
9076         .rpc_release = nfs41_free_stateid_release,
9077 };
9078
9079 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9080                 const nfs4_stateid *stateid,
9081                 struct rpc_cred *cred,
9082                 bool privileged)
9083 {
9084         struct rpc_message msg = {
9085                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9086                 .rpc_cred = cred,
9087         };
9088         struct rpc_task_setup task_setup = {
9089                 .rpc_client = server->client,
9090                 .rpc_message = &msg,
9091                 .callback_ops = &nfs41_free_stateid_ops,
9092                 .flags = RPC_TASK_ASYNC,
9093         };
9094         struct nfs_free_stateid_data *data;
9095
9096         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9097                 &task_setup.rpc_client, &msg);
9098
9099         dprintk("NFS call  free_stateid %p\n", stateid);
9100         data = kmalloc(sizeof(*data), GFP_NOFS);
9101         if (!data)
9102                 return ERR_PTR(-ENOMEM);
9103         data->server = server;
9104         nfs4_stateid_copy(&data->args.stateid, stateid);
9105
9106         task_setup.callback_data = data;
9107
9108         msg.rpc_argp = &data->args;
9109         msg.rpc_resp = &data->res;
9110         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9111         if (privileged)
9112                 nfs4_set_sequence_privileged(&data->args.seq_args);
9113
9114         return rpc_run_task(&task_setup);
9115 }
9116
9117 /**
9118  * nfs41_free_stateid - perform a FREE_STATEID operation
9119  *
9120  * @server: server / transport on which to perform the operation
9121  * @stateid: state ID to release
9122  * @cred: credential
9123  * @is_recovery: set to true if this call needs to be privileged
9124  *
9125  * Note: this function is always asynchronous.
9126  */
9127 static int nfs41_free_stateid(struct nfs_server *server,
9128                 const nfs4_stateid *stateid,
9129                 struct rpc_cred *cred,
9130                 bool is_recovery)
9131 {
9132         struct rpc_task *task;
9133
9134         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9135         if (IS_ERR(task))
9136                 return PTR_ERR(task);
9137         rpc_put_task(task);
9138         return 0;
9139 }
9140
9141 static void
9142 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9143 {
9144         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9145
9146         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9147         nfs4_free_lock_state(server, lsp);
9148 }
9149
9150 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9151                 const nfs4_stateid *s2)
9152 {
9153         if (s1->type != s2->type)
9154                 return false;
9155
9156         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9157                 return false;
9158
9159         if (s1->seqid == s2->seqid)
9160                 return true;
9161
9162         return s1->seqid == 0 || s2->seqid == 0;
9163 }
9164
9165 #endif /* CONFIG_NFS_V4_1 */
9166
9167 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9168                 const nfs4_stateid *s2)
9169 {
9170         return nfs4_stateid_match(s1, s2);
9171 }
9172
9173
9174 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9175         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9176         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9177         .recover_open   = nfs4_open_reclaim,
9178         .recover_lock   = nfs4_lock_reclaim,
9179         .establish_clid = nfs4_init_clientid,
9180         .detect_trunking = nfs40_discover_server_trunking,
9181 };
9182
9183 #if defined(CONFIG_NFS_V4_1)
9184 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9185         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9186         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9187         .recover_open   = nfs4_open_reclaim,
9188         .recover_lock   = nfs4_lock_reclaim,
9189         .establish_clid = nfs41_init_clientid,
9190         .reclaim_complete = nfs41_proc_reclaim_complete,
9191         .detect_trunking = nfs41_discover_server_trunking,
9192 };
9193 #endif /* CONFIG_NFS_V4_1 */
9194
9195 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9196         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9197         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9198         .recover_open   = nfs40_open_expired,
9199         .recover_lock   = nfs4_lock_expired,
9200         .establish_clid = nfs4_init_clientid,
9201 };
9202
9203 #if defined(CONFIG_NFS_V4_1)
9204 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9205         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9206         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9207         .recover_open   = nfs41_open_expired,
9208         .recover_lock   = nfs41_lock_expired,
9209         .establish_clid = nfs41_init_clientid,
9210 };
9211 #endif /* CONFIG_NFS_V4_1 */
9212
9213 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9214         .sched_state_renewal = nfs4_proc_async_renew,
9215         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9216         .renew_lease = nfs4_proc_renew,
9217 };
9218
9219 #if defined(CONFIG_NFS_V4_1)
9220 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9221         .sched_state_renewal = nfs41_proc_async_sequence,
9222         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9223         .renew_lease = nfs4_proc_sequence,
9224 };
9225 #endif
9226
9227 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9228         .get_locations = _nfs40_proc_get_locations,
9229         .fsid_present = _nfs40_proc_fsid_present,
9230 };
9231
9232 #if defined(CONFIG_NFS_V4_1)
9233 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9234         .get_locations = _nfs41_proc_get_locations,
9235         .fsid_present = _nfs41_proc_fsid_present,
9236 };
9237 #endif  /* CONFIG_NFS_V4_1 */
9238
9239 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9240         .minor_version = 0,
9241         .init_caps = NFS_CAP_READDIRPLUS
9242                 | NFS_CAP_ATOMIC_OPEN
9243                 | NFS_CAP_POSIX_LOCK,
9244         .init_client = nfs40_init_client,
9245         .shutdown_client = nfs40_shutdown_client,
9246         .match_stateid = nfs4_match_stateid,
9247         .find_root_sec = nfs4_find_root_sec,
9248         .free_lock_state = nfs4_release_lockowner,
9249         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9250         .alloc_seqid = nfs_alloc_seqid,
9251         .call_sync_ops = &nfs40_call_sync_ops,
9252         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9253         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9254         .state_renewal_ops = &nfs40_state_renewal_ops,
9255         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9256 };
9257
9258 #if defined(CONFIG_NFS_V4_1)
9259 static struct nfs_seqid *
9260 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9261 {
9262         return NULL;
9263 }
9264
9265 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9266         .minor_version = 1,
9267         .init_caps = NFS_CAP_READDIRPLUS
9268                 | NFS_CAP_ATOMIC_OPEN
9269                 | NFS_CAP_POSIX_LOCK
9270                 | NFS_CAP_STATEID_NFSV41
9271                 | NFS_CAP_ATOMIC_OPEN_V1,
9272         .init_client = nfs41_init_client,
9273         .shutdown_client = nfs41_shutdown_client,
9274         .match_stateid = nfs41_match_stateid,
9275         .find_root_sec = nfs41_find_root_sec,
9276         .free_lock_state = nfs41_free_lock_state,
9277         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9278         .alloc_seqid = nfs_alloc_no_seqid,
9279         .session_trunk = nfs4_test_session_trunk,
9280         .call_sync_ops = &nfs41_call_sync_ops,
9281         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9282         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9283         .state_renewal_ops = &nfs41_state_renewal_ops,
9284         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9285 };
9286 #endif
9287
9288 #if defined(CONFIG_NFS_V4_2)
9289 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9290         .minor_version = 2,
9291         .init_caps = NFS_CAP_READDIRPLUS
9292                 | NFS_CAP_ATOMIC_OPEN
9293                 | NFS_CAP_POSIX_LOCK
9294                 | NFS_CAP_STATEID_NFSV41
9295                 | NFS_CAP_ATOMIC_OPEN_V1
9296                 | NFS_CAP_ALLOCATE
9297                 | NFS_CAP_COPY
9298                 | NFS_CAP_DEALLOCATE
9299                 | NFS_CAP_SEEK
9300                 | NFS_CAP_LAYOUTSTATS
9301                 | NFS_CAP_CLONE,
9302         .init_client = nfs41_init_client,
9303         .shutdown_client = nfs41_shutdown_client,
9304         .match_stateid = nfs41_match_stateid,
9305         .find_root_sec = nfs41_find_root_sec,
9306         .free_lock_state = nfs41_free_lock_state,
9307         .call_sync_ops = &nfs41_call_sync_ops,
9308         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9309         .alloc_seqid = nfs_alloc_no_seqid,
9310         .session_trunk = nfs4_test_session_trunk,
9311         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9312         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9313         .state_renewal_ops = &nfs41_state_renewal_ops,
9314         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9315 };
9316 #endif
9317
9318 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9319         [0] = &nfs_v4_0_minor_ops,
9320 #if defined(CONFIG_NFS_V4_1)
9321         [1] = &nfs_v4_1_minor_ops,
9322 #endif
9323 #if defined(CONFIG_NFS_V4_2)
9324         [2] = &nfs_v4_2_minor_ops,
9325 #endif
9326 };
9327
9328 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9329 {
9330         ssize_t error, error2;
9331
9332         error = generic_listxattr(dentry, list, size);
9333         if (error < 0)
9334                 return error;
9335         if (list) {
9336                 list += error;
9337                 size -= error;
9338         }
9339
9340         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9341         if (error2 < 0)
9342                 return error2;
9343         return error + error2;
9344 }
9345
9346 static const struct inode_operations nfs4_dir_inode_operations = {
9347         .create         = nfs_create,
9348         .lookup         = nfs_lookup,
9349         .atomic_open    = nfs_atomic_open,
9350         .link           = nfs_link,
9351         .unlink         = nfs_unlink,
9352         .symlink        = nfs_symlink,
9353         .mkdir          = nfs_mkdir,
9354         .rmdir          = nfs_rmdir,
9355         .mknod          = nfs_mknod,
9356         .rename         = nfs_rename,
9357         .permission     = nfs_permission,
9358         .getattr        = nfs_getattr,
9359         .setattr        = nfs_setattr,
9360         .listxattr      = nfs4_listxattr,
9361 };
9362
9363 static const struct inode_operations nfs4_file_inode_operations = {
9364         .permission     = nfs_permission,
9365         .getattr        = nfs_getattr,
9366         .setattr        = nfs_setattr,
9367         .listxattr      = nfs4_listxattr,
9368 };
9369
9370 const struct nfs_rpc_ops nfs_v4_clientops = {
9371         .version        = 4,                    /* protocol version */
9372         .dentry_ops     = &nfs4_dentry_operations,
9373         .dir_inode_ops  = &nfs4_dir_inode_operations,
9374         .file_inode_ops = &nfs4_file_inode_operations,
9375         .file_ops       = &nfs4_file_operations,
9376         .getroot        = nfs4_proc_get_root,
9377         .submount       = nfs4_submount,
9378         .try_mount      = nfs4_try_mount,
9379         .getattr        = nfs4_proc_getattr,
9380         .setattr        = nfs4_proc_setattr,
9381         .lookup         = nfs4_proc_lookup,
9382         .lookupp        = nfs4_proc_lookupp,
9383         .access         = nfs4_proc_access,
9384         .readlink       = nfs4_proc_readlink,
9385         .create         = nfs4_proc_create,
9386         .remove         = nfs4_proc_remove,
9387         .unlink_setup   = nfs4_proc_unlink_setup,
9388         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9389         .unlink_done    = nfs4_proc_unlink_done,
9390         .rename_setup   = nfs4_proc_rename_setup,
9391         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9392         .rename_done    = nfs4_proc_rename_done,
9393         .link           = nfs4_proc_link,
9394         .symlink        = nfs4_proc_symlink,
9395         .mkdir          = nfs4_proc_mkdir,
9396         .rmdir          = nfs4_proc_remove,
9397         .readdir        = nfs4_proc_readdir,
9398         .mknod          = nfs4_proc_mknod,
9399         .statfs         = nfs4_proc_statfs,
9400         .fsinfo         = nfs4_proc_fsinfo,
9401         .pathconf       = nfs4_proc_pathconf,
9402         .set_capabilities = nfs4_server_capabilities,
9403         .decode_dirent  = nfs4_decode_dirent,
9404         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9405         .read_setup     = nfs4_proc_read_setup,
9406         .read_done      = nfs4_read_done,
9407         .write_setup    = nfs4_proc_write_setup,
9408         .write_done     = nfs4_write_done,
9409         .commit_setup   = nfs4_proc_commit_setup,
9410         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9411         .commit_done    = nfs4_commit_done,
9412         .lock           = nfs4_proc_lock,
9413         .clear_acl_cache = nfs4_zap_acl_attr,
9414         .close_context  = nfs4_close_context,
9415         .open_context   = nfs4_atomic_open,
9416         .have_delegation = nfs4_have_delegation,
9417         .return_delegation = nfs4_inode_return_delegation,
9418         .alloc_client   = nfs4_alloc_client,
9419         .init_client    = nfs4_init_client,
9420         .free_client    = nfs4_free_client,
9421         .create_server  = nfs4_create_server,
9422         .clone_server   = nfs_clone_server,
9423 };
9424
9425 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9426         .name   = XATTR_NAME_NFSV4_ACL,
9427         .list   = nfs4_xattr_list_nfs4_acl,
9428         .get    = nfs4_xattr_get_nfs4_acl,
9429         .set    = nfs4_xattr_set_nfs4_acl,
9430 };
9431
9432 const struct xattr_handler *nfs4_xattr_handlers[] = {
9433         &nfs4_xattr_nfs4_acl_handler,
9434 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9435         &nfs4_xattr_nfs4_label_handler,
9436 #endif
9437         NULL
9438 };
9439
9440 /*
9441  * Local variables:
9442  *  c-basic-offset: 8
9443  * End:
9444  */