fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/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         __be32 *start, *p;
 279
 280         if (cookie > 2) {
 281                 readdir->cookie = cookie;
 282                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 283                 return;
 284         }
 285
 286         readdir->cookie = 0;
 287         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 288         if (cookie == 2)
 289                 return;
 290
 291         /*
 292          * NFSv4 servers do not return entries for '.' and '..'
 293          * Therefore, we fake these entries here.  We let '.'
 294          * have cookie 0 and '..' have cookie 1.  Note that
 295          * when talking to the server, we always send cookie 0
 296          * instead of 1 or 2.
 297          */
 298         start = p = kmap_atomic(*readdir->pages);
 299
 300         if (cookie == 0) {
 301                 *p++ = xdr_one;                                  /* next */
 302                 *p++ = xdr_zero;                   /* cookie, first word */
 303                 *p++ = xdr_one;                   /* cookie, second word */
 304                 *p++ = xdr_one;                             /* entry len */
 305                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 306                 p++;
 307                 *p++ = xdr_one;                         /* bitmap length */
 308                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309                 *p++ = htonl(8);              /* attribute buffer length */
 310                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 311         }
 312
 313         *p++ = xdr_one;                                  /* next */
 314         *p++ = xdr_zero;                   /* cookie, first word */
 315         *p++ = xdr_two;                   /* cookie, second word */
 316         *p++ = xdr_two;                             /* entry len */
 317         memcpy(p, "..\0\0", 4);                         /* entry */
 318         p++;
 319         *p++ = xdr_one;                         /* bitmap length */
 320         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 321         *p++ = htonl(8);              /* attribute buffer length */
 322         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 323
 324         readdir->pgbase = (char *)p - (char *)start;
 325         readdir->count -= readdir->pgbase;
 326         kunmap_atomic(start);
 327 }
 328
 329 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 330                 nfs4_stateid *stateid,
 331                 struct rpc_cred *cred)
 332 {
 333         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 334
 335         ops->test_and_free_expired(server, stateid, cred);
 336 }
 337
 338 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 339                 nfs4_stateid *stateid,
 340                 struct rpc_cred *cred)
 341 {
 342         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 343         nfs4_test_and_free_stateid(server, stateid, cred);
 344 }
 345
 346 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 347                 const nfs4_stateid *stateid,
 348                 struct rpc_cred *cred)
 349 {
 350         nfs4_stateid tmp;
 351
 352         nfs4_stateid_copy(&tmp, stateid);
 353         __nfs4_free_revoked_stateid(server, &tmp, cred);
 354 }
 355
 356 static long nfs4_update_delay(long *timeout)
 357 {
 358         long ret;
 359         if (!timeout)
 360                 return NFS4_POLL_RETRY_MAX;
 361         if (*timeout <= 0)
 362                 *timeout = NFS4_POLL_RETRY_MIN;
 363         if (*timeout > NFS4_POLL_RETRY_MAX)
 364                 *timeout = NFS4_POLL_RETRY_MAX;
 365         ret = *timeout;
 366         *timeout <<= 1;
 367         return ret;
 368 }
 369
 370 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 371 {
 372         int res = 0;
 373
 374         might_sleep();
 375
 376         freezable_schedule_timeout_killable_unsafe(
 377                 nfs4_update_delay(timeout));
 378         if (fatal_signal_pending(current))
 379                 res = -ERESTARTSYS;
 380         return res;
 381 }
 382
 383 /* This is the error handling routine for processes that are allowed
 384  * to sleep.
 385  */
 386 static int nfs4_do_handle_exception(struct nfs_server *server,
 387                 int errorcode, struct nfs4_exception *exception)
 388 {
 389         struct nfs_client *clp = server->nfs_client;
 390         struct nfs4_state *state = exception->state;
 391         const nfs4_stateid *stateid = exception->stateid;
 392         struct inode *inode = exception->inode;
 393         int ret = errorcode;
 394
 395         exception->delay = 0;
 396         exception->recovering = 0;
 397         exception->retry = 0;
 398
 399         if (stateid == NULL && state != NULL)
 400                 stateid = &state->stateid;
 401
 402         switch(errorcode) {
 403                 case 0:
 404                         return 0;
 405                 case -NFS4ERR_DELEG_REVOKED:
 406                 case -NFS4ERR_ADMIN_REVOKED:
 407                 case -NFS4ERR_EXPIRED:
 408                 case -NFS4ERR_BAD_STATEID:
 409                         if (inode != NULL && stateid != NULL) {
 410                                 nfs_inode_find_state_and_recover(inode,
 411                                                 stateid);
 412                                 goto wait_on_recovery;
 413                         }
 414                 case -NFS4ERR_OPENMODE:
 415                         if (inode) {
 416                                 int err;
 417
 418                                 err = nfs_async_inode_return_delegation(inode,
 419                                                 stateid);
 420                                 if (err == 0)
 421                                         goto wait_on_recovery;
 422                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 423                                         exception->retry = 1;
 424                                         break;
 425                                 }
 426                         }
 427                         if (state == NULL)
 428                                 break;
 429                         ret = nfs4_schedule_stateid_recovery(server, state);
 430                         if (ret < 0)
 431                                 break;
 432                         goto wait_on_recovery;
 433                 case -NFS4ERR_STALE_STATEID:
 434                 case -NFS4ERR_STALE_CLIENTID:
 435                         nfs4_schedule_lease_recovery(clp);
 436                         goto wait_on_recovery;
 437                 case -NFS4ERR_MOVED:
 438                         ret = nfs4_schedule_migration_recovery(server);
 439                         if (ret < 0)
 440                                 break;
 441                         goto wait_on_recovery;
 442                 case -NFS4ERR_LEASE_MOVED:
 443                         nfs4_schedule_lease_moved_recovery(clp);
 444                         goto wait_on_recovery;
 445 #if defined(CONFIG_NFS_V4_1)
 446                 case -NFS4ERR_BADSESSION:
 447                 case -NFS4ERR_BADSLOT:
 448                 case -NFS4ERR_BAD_HIGH_SLOT:
 449                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 450                 case -NFS4ERR_DEADSESSION:
 451                 case -NFS4ERR_SEQ_FALSE_RETRY:
 452                 case -NFS4ERR_SEQ_MISORDERED:
 453                         dprintk("%s ERROR: %d Reset session\n", __func__,
 454                                 errorcode);
 455                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 456                         goto wait_on_recovery;
 457 #endif /* defined(CONFIG_NFS_V4_1) */
 458                 case -NFS4ERR_FILE_OPEN:
 459                         if (exception->timeout > HZ) {
 460                                 /* We have retried a decent amount, time to
 461                                  * fail
 462                                  */
 463                                 ret = -EBUSY;
 464                                 break;
 465                         }
 466                 case -NFS4ERR_DELAY:
 467                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 468                 case -NFS4ERR_GRACE:
 469                 case -NFS4ERR_LAYOUTTRYLATER:
 470                 case -NFS4ERR_RECALLCONFLICT:
 471                         exception->delay = 1;
 472                         return 0;
 473
 474                 case -NFS4ERR_RETRY_UNCACHED_REP:
 475                 case -NFS4ERR_OLD_STATEID:
 476                         exception->retry = 1;
 477                         break;
 478                 case -NFS4ERR_BADOWNER:
 479                         /* The following works around a Linux server bug! */
 480                 case -NFS4ERR_BADNAME:
 481                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 482                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 483                                 exception->retry = 1;
 484                                 printk(KERN_WARNING "NFS: v4 server %s "
 485                                                 "does not accept raw "
 486                                                 "uid/gids. "
 487                                                 "Reenabling the idmapper.\n",
 488                                                 server->nfs_client->cl_hostname);
 489                         }
 490         }
 491         /* We failed to handle the error */
 492         return nfs4_map_errors(ret);
 493 wait_on_recovery:
 494         exception->recovering = 1;
 495         return 0;
 496 }
 497
 498 /* This is the error handling routine for processes that are allowed
 499  * to sleep.
 500  */
 501 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 502 {
 503         struct nfs_client *clp = server->nfs_client;
 504         int ret;
 505
 506         ret = nfs4_do_handle_exception(server, errorcode, exception);
 507         if (exception->delay) {
 508                 ret = nfs4_delay(server->client, &exception->timeout);
 509                 goto out_retry;
 510         }
 511         if (exception->recovering) {
 512                 ret = nfs4_wait_clnt_recover(clp);
 513                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 514                         return -EIO;
 515                 goto out_retry;
 516         }
 517         return ret;
 518 out_retry:
 519         if (ret == 0)
 520                 exception->retry = 1;
 521         return ret;
 522 }
 523
 524 static int
 525 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 526                 int errorcode, struct nfs4_exception *exception)
 527 {
 528         struct nfs_client *clp = server->nfs_client;
 529         int ret;
 530
 531         ret = nfs4_do_handle_exception(server, errorcode, exception);
 532         if (exception->delay) {
 533                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 534                 goto out_retry;
 535         }
 536         if (exception->recovering) {
 537                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 538                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 539                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 540                 goto out_retry;
 541         }
 542         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 543                 ret = -EIO;
 544         return ret;
 545 out_retry:
 546         if (ret == 0)
 547                 exception->retry = 1;
 548         return ret;
 549 }
 550
 551 static int
 552 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 553                         struct nfs4_state *state, long *timeout)
 554 {
 555         struct nfs4_exception exception = {
 556                 .state = state,
 557         };
 558
 559         if (task->tk_status >= 0)
 560                 return 0;
 561         if (timeout)
 562                 exception.timeout = *timeout;
 563         task->tk_status = nfs4_async_handle_exception(task, server,
 564                         task->tk_status,
 565                         &exception);
 566         if (exception.delay && timeout)
 567                 *timeout = exception.timeout;
 568         if (exception.retry)
 569                 return -EAGAIN;
 570         return 0;
 571 }
 572
 573 /*
 574  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 575  * or 'false' otherwise.
 576  */
 577 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 578 {
 579         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 580         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 581 }
 582
 583 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 584 {
 585         spin_lock(&clp->cl_lock);
 586         if (time_before(clp->cl_last_renewal,timestamp))
 587                 clp->cl_last_renewal = timestamp;
 588         spin_unlock(&clp->cl_lock);
 589 }
 590
 591 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 592 {
 593         struct nfs_client *clp = server->nfs_client;
 594
 595         if (!nfs4_has_session(clp))
 596                 do_renew_lease(clp, timestamp);
 597 }
 598
 599 struct nfs4_call_sync_data {
 600         const struct nfs_server *seq_server;
 601         struct nfs4_sequence_args *seq_args;
 602         struct nfs4_sequence_res *seq_res;
 603 };
 604
 605 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 606                         struct nfs4_sequence_res *res, int cache_reply)
 607 {
 608         args->sa_slot = NULL;
 609         args->sa_cache_this = cache_reply;
 610         args->sa_privileged = 0;
 611
 612         res->sr_slot = NULL;
 613 }
 614
 615 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 616 {
 617         args->sa_privileged = 1;
 618 }
 619
 620 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 621 {
 622         struct nfs4_slot *slot = res->sr_slot;
 623         struct nfs4_slot_table *tbl;
 624
 625         tbl = slot->table;
 626         spin_lock(&tbl->slot_tbl_lock);
 627         if (!nfs41_wake_and_assign_slot(tbl, slot))
 628                 nfs4_free_slot(tbl, slot);
 629         spin_unlock(&tbl->slot_tbl_lock);
 630
 631         res->sr_slot = NULL;
 632 }
 633
 634 static int nfs40_sequence_done(struct rpc_task *task,
 635                                struct nfs4_sequence_res *res)
 636 {
 637         if (res->sr_slot != NULL)
 638                 nfs40_sequence_free_slot(res);
 639         return 1;
 640 }
 641
 642 #if defined(CONFIG_NFS_V4_1)
 643
 644 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 645 {
 646         struct nfs4_session *session;
 647         struct nfs4_slot_table *tbl;
 648         struct nfs4_slot *slot = res->sr_slot;
 649         bool send_new_highest_used_slotid = false;
 650
 651         tbl = slot->table;
 652         session = tbl->session;
 653
 654         /* Bump the slot sequence number */
 655         if (slot->seq_done)
 656                 slot->seq_nr++;
 657         slot->seq_done = 0;
 658
 659         spin_lock(&tbl->slot_tbl_lock);
 660         /* Be nice to the server: try to ensure that the last transmitted
 661          * value for highest_user_slotid <= target_highest_slotid
 662          */
 663         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 664                 send_new_highest_used_slotid = true;
 665
 666         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 667                 send_new_highest_used_slotid = false;
 668                 goto out_unlock;
 669         }
 670         nfs4_free_slot(tbl, slot);
 671
 672         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 673                 send_new_highest_used_slotid = false;
 674 out_unlock:
 675         spin_unlock(&tbl->slot_tbl_lock);
 676         res->sr_slot = NULL;
 677         if (send_new_highest_used_slotid)
 678                 nfs41_notify_server(session->clp);
 679         if (waitqueue_active(&tbl->slot_waitq))
 680                 wake_up_all(&tbl->slot_waitq);
 681 }
 682
 683 static int nfs41_sequence_process(struct rpc_task *task,
 684                 struct nfs4_sequence_res *res)
 685 {
 686         struct nfs4_session *session;
 687         struct nfs4_slot *slot = res->sr_slot;
 688         struct nfs_client *clp;
 689         bool interrupted = false;
 690         int ret = 1;
 691
 692         if (slot == NULL)
 693                 goto out_noaction;
 694         /* don't increment the sequence number if the task wasn't sent */
 695         if (!RPC_WAS_SENT(task))
 696                 goto out;
 697
 698         session = slot->table->session;
 699
 700         if (slot->interrupted) {
 701                 if (res->sr_status != -NFS4ERR_DELAY)
 702                         slot->interrupted = 0;
 703                 interrupted = true;
 704         }
 705
 706         trace_nfs4_sequence_done(session, res);
 707         /* Check the SEQUENCE operation status */
 708         switch (res->sr_status) {
 709         case 0:
 710                 /* If previous op on slot was interrupted and we reused
 711                  * the seq# and got a reply from the cache, then retry
 712                  */
 713                 if (task->tk_status == -EREMOTEIO && interrupted) {
 714                         ++slot->seq_nr;
 715                         goto retry_nowait;
 716                 }
 717                 /* Update the slot's sequence and clientid lease timer */
 718                 slot->seq_done = 1;
 719                 clp = session->clp;
 720                 do_renew_lease(clp, res->sr_timestamp);
 721                 /* Check sequence flags */
 722                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 723                                 !!slot->privileged);
 724                 nfs41_update_target_slotid(slot->table, slot, res);
 725                 break;
 726         case 1:
 727                 /*
 728                  * sr_status remains 1 if an RPC level error occurred.
 729                  * The server may or may not have processed the sequence
 730                  * operation..
 731                  * Mark the slot as having hosted an interrupted RPC call.
 732                  */
 733                 slot->interrupted = 1;
 734                 goto out;
 735         case -NFS4ERR_DELAY:
 736                 /* The server detected a resend of the RPC call and
 737                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 738                  * of RFC5661.
 739                  */
 740                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 741                         __func__,
 742                         slot->slot_nr,
 743                         slot->seq_nr);
 744                 goto out_retry;
 745         case -NFS4ERR_BADSLOT:
 746                 /*
 747                  * The slot id we used was probably retired. Try again
 748                  * using a different slot id.
 749                  */
 750                 goto retry_nowait;
 751         case -NFS4ERR_SEQ_MISORDERED:
 752                 /*
 753                  * Was the last operation on this sequence interrupted?
 754                  * If so, retry after bumping the sequence number.
 755                  */
 756                 if (interrupted) {
 757                         ++slot->seq_nr;
 758                         goto retry_nowait;
 759                 }
 760                 /*
 761                  * Could this slot have been previously retired?
 762                  * If so, then the server may be expecting seq_nr = 1!
 763                  */
 764                 if (slot->seq_nr != 1) {
 765                         slot->seq_nr = 1;
 766                         goto retry_nowait;
 767                 }
 768                 break;
 769         case -NFS4ERR_SEQ_FALSE_RETRY:
 770                 ++slot->seq_nr;
 771                 goto retry_nowait;
 772         default:
 773                 /* Just update the slot sequence no. */
 774                 slot->seq_done = 1;
 775         }
 776 out:
 777         /* The session may be reset by one of the error handlers. */
 778         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 779 out_noaction:
 780         return ret;
 781 retry_nowait:
 782         if (rpc_restart_call_prepare(task)) {
 783                 nfs41_sequence_free_slot(res);
 784                 task->tk_status = 0;
 785                 ret = 0;
 786         }
 787         goto out;
 788 out_retry:
 789         if (!rpc_restart_call(task))
 790                 goto out;
 791         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 792         return 0;
 793 }
 794
 795 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 796 {
 797         if (!nfs41_sequence_process(task, res))
 798                 return 0;
 799         if (res->sr_slot != NULL)
 800                 nfs41_sequence_free_slot(res);
 801         return 1;
 802
 803 }
 804 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 805
 806 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 807 {
 808         if (res->sr_slot == NULL)
 809                 return 1;
 810         if (res->sr_slot->table->session != NULL)
 811                 return nfs41_sequence_process(task, res);
 812         return nfs40_sequence_done(task, res);
 813 }
 814
 815 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 816 {
 817         if (res->sr_slot != NULL) {
 818                 if (res->sr_slot->table->session != NULL)
 819                         nfs41_sequence_free_slot(res);
 820                 else
 821                         nfs40_sequence_free_slot(res);
 822         }
 823 }
 824
 825 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 826 {
 827         if (res->sr_slot == NULL)
 828                 return 1;
 829         if (!res->sr_slot->table->session)
 830                 return nfs40_sequence_done(task, res);
 831         return nfs41_sequence_done(task, res);
 832 }
 833 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 834
 835 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 836 {
 837         struct nfs4_call_sync_data *data = calldata;
 838
 839         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 840
 841         nfs4_setup_sequence(data->seq_server->nfs_client,
 842                             data->seq_args, data->seq_res, task);
 843 }
 844
 845 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 846 {
 847         struct nfs4_call_sync_data *data = calldata;
 848
 849         nfs41_sequence_done(task, data->seq_res);
 850 }
 851
 852 static const struct rpc_call_ops nfs41_call_sync_ops = {
 853         .rpc_call_prepare = nfs41_call_sync_prepare,
 854         .rpc_call_done = nfs41_call_sync_done,
 855 };
 856
 857 #else   /* !CONFIG_NFS_V4_1 */
 858
 859 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 860 {
 861         return nfs40_sequence_done(task, res);
 862 }
 863
 864 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 865 {
 866         if (res->sr_slot != NULL)
 867                 nfs40_sequence_free_slot(res);
 868 }
 869
 870 int nfs4_sequence_done(struct rpc_task *task,
 871                        struct nfs4_sequence_res *res)
 872 {
 873         return nfs40_sequence_done(task, res);
 874 }
 875 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 876
 877 #endif  /* !CONFIG_NFS_V4_1 */
 878
 879 int nfs4_setup_sequence(const struct nfs_client *client,
 880                         struct nfs4_sequence_args *args,
 881                         struct nfs4_sequence_res *res,
 882                         struct rpc_task *task)
 883 {
 884         struct nfs4_session *session = nfs4_get_session(client);
 885         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 886         struct nfs4_slot *slot;
 887
 888         /* slot already allocated? */
 889         if (res->sr_slot != NULL)
 890                 goto out_start;
 891
 892         if (session) {
 893                 tbl = &session->fc_slot_table;
 894                 task->tk_timeout = 0;
 895         }
 896
 897         spin_lock(&tbl->slot_tbl_lock);
 898         /* The state manager will wait until the slot table is empty */
 899         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 900                 goto out_sleep;
 901
 902         slot = nfs4_alloc_slot(tbl);
 903         if (IS_ERR(slot)) {
 904                 /* Try again in 1/4 second */
 905                 if (slot == ERR_PTR(-ENOMEM))
 906                         task->tk_timeout = HZ >> 2;
 907                 goto out_sleep;
 908         }
 909         spin_unlock(&tbl->slot_tbl_lock);
 910
 911         slot->privileged = args->sa_privileged ? 1 : 0;
 912         args->sa_slot = slot;
 913
 914         res->sr_slot = slot;
 915         if (session) {
 916                 res->sr_timestamp = jiffies;
 917                 res->sr_status_flags = 0;
 918                 res->sr_status = 1;
 919         }
 920
 921         trace_nfs4_setup_sequence(session, args);
 922 out_start:
 923         rpc_call_start(task);
 924         return 0;
 925
 926 out_sleep:
 927         if (args->sa_privileged)
 928                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 929                                 NULL, RPC_PRIORITY_PRIVILEGED);
 930         else
 931                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 932         spin_unlock(&tbl->slot_tbl_lock);
 933         return -EAGAIN;
 934 }
 935 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
 936
 937 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 938 {
 939         struct nfs4_call_sync_data *data = calldata;
 940         nfs4_setup_sequence(data->seq_server->nfs_client,
 941                                 data->seq_args, data->seq_res, task);
 942 }
 943
 944 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 945 {
 946         struct nfs4_call_sync_data *data = calldata;
 947         nfs4_sequence_done(task, data->seq_res);
 948 }
 949
 950 static const struct rpc_call_ops nfs40_call_sync_ops = {
 951         .rpc_call_prepare = nfs40_call_sync_prepare,
 952         .rpc_call_done = nfs40_call_sync_done,
 953 };
 954
 955 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 956                                    struct nfs_server *server,
 957                                    struct rpc_message *msg,
 958                                    struct nfs4_sequence_args *args,
 959                                    struct nfs4_sequence_res *res)
 960 {
 961         int ret;
 962         struct rpc_task *task;
 963         struct nfs_client *clp = server->nfs_client;
 964         struct nfs4_call_sync_data data = {
 965                 .seq_server = server,
 966                 .seq_args = args,
 967                 .seq_res = res,
 968         };
 969         struct rpc_task_setup task_setup = {
 970                 .rpc_client = clnt,
 971                 .rpc_message = msg,
 972                 .callback_ops = clp->cl_mvops->call_sync_ops,
 973                 .callback_data = &data
 974         };
 975
 976         task = rpc_run_task(&task_setup);
 977         if (IS_ERR(task))
 978                 ret = PTR_ERR(task);
 979         else {
 980                 ret = task->tk_status;
 981                 rpc_put_task(task);
 982         }
 983         return ret;
 984 }
 985
 986 int nfs4_call_sync(struct rpc_clnt *clnt,
 987                    struct nfs_server *server,
 988                    struct rpc_message *msg,
 989                    struct nfs4_sequence_args *args,
 990                    struct nfs4_sequence_res *res,
 991                    int cache_reply)
 992 {
 993         nfs4_init_sequence(args, res, cache_reply);
 994         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 995 }
 996
 997 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
 998                 unsigned long timestamp)
 999 {
1000         struct nfs_inode *nfsi = NFS_I(dir);
1001
1002         spin_lock(&dir->i_lock);
1003         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1004         if (cinfo->atomic && cinfo->before == dir->i_version) {
1005                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1006                 nfsi->attrtimeo_timestamp = jiffies;
1007         } else {
1008                 nfs_force_lookup_revalidate(dir);
1009                 if (cinfo->before != dir->i_version)
1010                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1011                                 NFS_INO_INVALID_ACL;
1012         }
1013         dir->i_version = cinfo->after;
1014         nfsi->read_cache_jiffies = timestamp;
1015         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1016         nfs_fscache_invalidate(dir);
1017         spin_unlock(&dir->i_lock);
1018 }
1019
1020 struct nfs4_opendata {
1021         struct kref kref;
1022         struct nfs_openargs o_arg;
1023         struct nfs_openres o_res;
1024         struct nfs_open_confirmargs c_arg;
1025         struct nfs_open_confirmres c_res;
1026         struct nfs4_string owner_name;
1027         struct nfs4_string group_name;
1028         struct nfs4_label *a_label;
1029         struct nfs_fattr f_attr;
1030         struct nfs4_label *f_label;
1031         struct dentry *dir;
1032         struct dentry *dentry;
1033         struct nfs4_state_owner *owner;
1034         struct nfs4_state *state;
1035         struct iattr attrs;
1036         unsigned long timestamp;
1037         unsigned int rpc_done : 1;
1038         unsigned int file_created : 1;
1039         unsigned int is_recover : 1;
1040         int rpc_status;
1041         int cancelled;
1042 };
1043
1044 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1045                 int err, struct nfs4_exception *exception)
1046 {
1047         if (err != -EINVAL)
1048                 return false;
1049         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1050                 return false;
1051         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1052         exception->retry = 1;
1053         return true;
1054 }
1055
1056 static u32
1057 nfs4_map_atomic_open_share(struct nfs_server *server,
1058                 fmode_t fmode, int openflags)
1059 {
1060         u32 res = 0;
1061
1062         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1063         case FMODE_READ:
1064                 res = NFS4_SHARE_ACCESS_READ;
1065                 break;
1066         case FMODE_WRITE:
1067                 res = NFS4_SHARE_ACCESS_WRITE;
1068                 break;
1069         case FMODE_READ|FMODE_WRITE:
1070                 res = NFS4_SHARE_ACCESS_BOTH;
1071         }
1072         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1073                 goto out;
1074         /* Want no delegation if we're using O_DIRECT */
1075         if (openflags & O_DIRECT)
1076                 res |= NFS4_SHARE_WANT_NO_DELEG;
1077 out:
1078         return res;
1079 }
1080
1081 static enum open_claim_type4
1082 nfs4_map_atomic_open_claim(struct nfs_server *server,
1083                 enum open_claim_type4 claim)
1084 {
1085         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1086                 return claim;
1087         switch (claim) {
1088         default:
1089                 return claim;
1090         case NFS4_OPEN_CLAIM_FH:
1091                 return NFS4_OPEN_CLAIM_NULL;
1092         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1093                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1094         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1095                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1096         }
1097 }
1098
1099 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1100 {
1101         p->o_res.f_attr = &p->f_attr;
1102         p->o_res.f_label = p->f_label;
1103         p->o_res.seqid = p->o_arg.seqid;
1104         p->c_res.seqid = p->c_arg.seqid;
1105         p->o_res.server = p->o_arg.server;
1106         p->o_res.access_request = p->o_arg.access;
1107         nfs_fattr_init(&p->f_attr);
1108         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1109 }
1110
1111 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1112                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1113                 const struct iattr *attrs,
1114                 struct nfs4_label *label,
1115                 enum open_claim_type4 claim,
1116                 gfp_t gfp_mask)
1117 {
1118         struct dentry *parent = dget_parent(dentry);
1119         struct inode *dir = d_inode(parent);
1120         struct nfs_server *server = NFS_SERVER(dir);
1121         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1122         struct nfs4_opendata *p;
1123
1124         p = kzalloc(sizeof(*p), gfp_mask);
1125         if (p == NULL)
1126                 goto err;
1127
1128         p->f_label = nfs4_label_alloc(server, gfp_mask);
1129         if (IS_ERR(p->f_label))
1130                 goto err_free_p;
1131
1132         p->a_label = nfs4_label_alloc(server, gfp_mask);
1133         if (IS_ERR(p->a_label))
1134                 goto err_free_f;
1135
1136         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1137         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1138         if (IS_ERR(p->o_arg.seqid))
1139                 goto err_free_label;
1140         nfs_sb_active(dentry->d_sb);
1141         p->dentry = dget(dentry);
1142         p->dir = parent;
1143         p->owner = sp;
1144         atomic_inc(&sp->so_count);
1145         p->o_arg.open_flags = flags;
1146         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1147         p->o_arg.umask = current_umask();
1148         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1149         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1150                         fmode, flags);
1151         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1152          * will return permission denied for all bits until close */
1153         if (!(flags & O_EXCL)) {
1154                 /* ask server to check for all possible rights as results
1155                  * are cached */
1156                 switch (p->o_arg.claim) {
1157                 default:
1158                         break;
1159                 case NFS4_OPEN_CLAIM_NULL:
1160                 case NFS4_OPEN_CLAIM_FH:
1161                         p->o_arg.access = NFS4_ACCESS_READ |
1162                                 NFS4_ACCESS_MODIFY |
1163                                 NFS4_ACCESS_EXTEND |
1164                                 NFS4_ACCESS_EXECUTE;
1165                 }
1166         }
1167         p->o_arg.clientid = server->nfs_client->cl_clientid;
1168         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1169         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1170         p->o_arg.name = &dentry->d_name;
1171         p->o_arg.server = server;
1172         p->o_arg.bitmask = nfs4_bitmask(server, label);
1173         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1174         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1175         switch (p->o_arg.claim) {
1176         case NFS4_OPEN_CLAIM_NULL:
1177         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1178         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1179                 p->o_arg.fh = NFS_FH(dir);
1180                 break;
1181         case NFS4_OPEN_CLAIM_PREVIOUS:
1182         case NFS4_OPEN_CLAIM_FH:
1183         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1184         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1185                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1186         }
1187         if (attrs != NULL && attrs->ia_valid != 0) {
1188                 __u32 verf[2];
1189
1190                 p->o_arg.u.attrs = &p->attrs;
1191                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1192
1193                 verf[0] = jiffies;
1194                 verf[1] = current->pid;
1195                 memcpy(p->o_arg.u.verifier.data, verf,
1196                                 sizeof(p->o_arg.u.verifier.data));
1197         }
1198         p->c_arg.fh = &p->o_res.fh;
1199         p->c_arg.stateid = &p->o_res.stateid;
1200         p->c_arg.seqid = p->o_arg.seqid;
1201         nfs4_init_opendata_res(p);
1202         kref_init(&p->kref);
1203         return p;
1204
1205 err_free_label:
1206         nfs4_label_free(p->a_label);
1207 err_free_f:
1208         nfs4_label_free(p->f_label);
1209 err_free_p:
1210         kfree(p);
1211 err:
1212         dput(parent);
1213         return NULL;
1214 }
1215
1216 static void nfs4_opendata_free(struct kref *kref)
1217 {
1218         struct nfs4_opendata *p = container_of(kref,
1219                         struct nfs4_opendata, kref);
1220         struct super_block *sb = p->dentry->d_sb;
1221
1222         nfs_free_seqid(p->o_arg.seqid);
1223         nfs4_sequence_free_slot(&p->o_res.seq_res);
1224         if (p->state != NULL)
1225                 nfs4_put_open_state(p->state);
1226         nfs4_put_state_owner(p->owner);
1227
1228         nfs4_label_free(p->a_label);
1229         nfs4_label_free(p->f_label);
1230
1231         dput(p->dir);
1232         dput(p->dentry);
1233         nfs_sb_deactive(sb);
1234         nfs_fattr_free_names(&p->f_attr);
1235         kfree(p->f_attr.mdsthreshold);
1236         kfree(p);
1237 }
1238
1239 static void nfs4_opendata_put(struct nfs4_opendata *p)
1240 {
1241         if (p != NULL)
1242                 kref_put(&p->kref, nfs4_opendata_free);
1243 }
1244
1245 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1246                 fmode_t fmode)
1247 {
1248         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1249         case FMODE_READ|FMODE_WRITE:
1250                 return state->n_rdwr != 0;
1251         case FMODE_WRITE:
1252                 return state->n_wronly != 0;
1253         case FMODE_READ:
1254                 return state->n_rdonly != 0;
1255         }
1256         WARN_ON_ONCE(1);
1257         return false;
1258 }
1259
1260 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1261 {
1262         int ret = 0;
1263
1264         if (open_mode & (O_EXCL|O_TRUNC))
1265                 goto out;
1266         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1267                 case FMODE_READ:
1268                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1269                                 && state->n_rdonly != 0;
1270                         break;
1271                 case FMODE_WRITE:
1272                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1273                                 && state->n_wronly != 0;
1274                         break;
1275                 case FMODE_READ|FMODE_WRITE:
1276                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1277                                 && state->n_rdwr != 0;
1278         }
1279 out:
1280         return ret;
1281 }
1282
1283 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1284                 enum open_claim_type4 claim)
1285 {
1286         if (delegation == NULL)
1287                 return 0;
1288         if ((delegation->type & fmode) != fmode)
1289                 return 0;
1290         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1291                 return 0;
1292         switch (claim) {
1293         case NFS4_OPEN_CLAIM_NULL:
1294         case NFS4_OPEN_CLAIM_FH:
1295                 break;
1296         case NFS4_OPEN_CLAIM_PREVIOUS:
1297                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1298                         break;
1299         default:
1300                 return 0;
1301         }
1302         nfs_mark_delegation_referenced(delegation);
1303         return 1;
1304 }
1305
1306 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1307 {
1308         switch (fmode) {
1309                 case FMODE_WRITE:
1310                         state->n_wronly++;
1311                         break;
1312                 case FMODE_READ:
1313                         state->n_rdonly++;
1314                         break;
1315                 case FMODE_READ|FMODE_WRITE:
1316                         state->n_rdwr++;
1317         }
1318         nfs4_state_set_mode_locked(state, state->state | fmode);
1319 }
1320
1321 #ifdef CONFIG_NFS_V4_1
1322 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1323 {
1324         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1325                 return true;
1326         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1327                 return true;
1328         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1329                 return true;
1330         return false;
1331 }
1332 #endif /* CONFIG_NFS_V4_1 */
1333
1334 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1335 {
1336         struct nfs_client *clp = state->owner->so_server->nfs_client;
1337         bool need_recover = false;
1338
1339         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1340                 need_recover = true;
1341         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1342                 need_recover = true;
1343         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1344                 need_recover = true;
1345         if (need_recover)
1346                 nfs4_state_mark_reclaim_nograce(clp, state);
1347 }
1348
1349 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1350                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1351 {
1352         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1353                 return true;
1354         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1355                 nfs4_stateid_copy(freeme, &state->open_stateid);
1356                 nfs_test_and_clear_all_open_stateid(state);
1357                 return true;
1358         }
1359         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1360                 return true;
1361         return false;
1362 }
1363
1364 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1365 {
1366         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1367                 return;
1368         if (state->n_wronly)
1369                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1370         if (state->n_rdonly)
1371                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1372         if (state->n_rdwr)
1373                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1374         set_bit(NFS_OPEN_STATE, &state->flags);
1375 }
1376
1377 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1378                 nfs4_stateid *stateid, fmode_t fmode)
1379 {
1380         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1381         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1382         case FMODE_WRITE:
1383                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1384                 break;
1385         case FMODE_READ:
1386                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1387                 break;
1388         case 0:
1389                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1390                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1391                 clear_bit(NFS_OPEN_STATE, &state->flags);
1392         }
1393         if (stateid == NULL)
1394                 return;
1395         /* Handle OPEN+OPEN_DOWNGRADE races */
1396         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1397             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1398                 nfs_resync_open_stateid_locked(state);
1399                 return;
1400         }
1401         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1402                 nfs4_stateid_copy(&state->stateid, stateid);
1403         nfs4_stateid_copy(&state->open_stateid, stateid);
1404 }
1405
1406 static void nfs_clear_open_stateid(struct nfs4_state *state,
1407         nfs4_stateid *arg_stateid,
1408         nfs4_stateid *stateid, fmode_t fmode)
1409 {
1410         write_seqlock(&state->seqlock);
1411         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1412         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1413                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1414         write_sequnlock(&state->seqlock);
1415         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1416                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1417 }
1418
1419 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1420                 const nfs4_stateid *stateid, fmode_t fmode,
1421                 nfs4_stateid *freeme)
1422 {
1423         switch (fmode) {
1424                 case FMODE_READ:
1425                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1426                         break;
1427                 case FMODE_WRITE:
1428                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1429                         break;
1430                 case FMODE_READ|FMODE_WRITE:
1431                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1432         }
1433         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1434                 return;
1435         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1436                 nfs4_stateid_copy(&state->stateid, stateid);
1437         nfs4_stateid_copy(&state->open_stateid, stateid);
1438 }
1439
1440 static void __update_open_stateid(struct nfs4_state *state,
1441                 const nfs4_stateid *open_stateid,
1442                 const nfs4_stateid *deleg_stateid,
1443                 fmode_t fmode,
1444                 nfs4_stateid *freeme)
1445 {
1446         /*
1447          * Protect the call to nfs4_state_set_mode_locked and
1448          * serialise the stateid update
1449          */
1450         spin_lock(&state->owner->so_lock);
1451         write_seqlock(&state->seqlock);
1452         if (deleg_stateid != NULL) {
1453                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1454                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1455         }
1456         if (open_stateid != NULL)
1457                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1458         write_sequnlock(&state->seqlock);
1459         update_open_stateflags(state, fmode);
1460         spin_unlock(&state->owner->so_lock);
1461 }
1462
1463 static int update_open_stateid(struct nfs4_state *state,
1464                 const nfs4_stateid *open_stateid,
1465                 const nfs4_stateid *delegation,
1466                 fmode_t fmode)
1467 {
1468         struct nfs_server *server = NFS_SERVER(state->inode);
1469         struct nfs_client *clp = server->nfs_client;
1470         struct nfs_inode *nfsi = NFS_I(state->inode);
1471         struct nfs_delegation *deleg_cur;
1472         nfs4_stateid freeme = { };
1473         int ret = 0;
1474
1475         fmode &= (FMODE_READ|FMODE_WRITE);
1476
1477         rcu_read_lock();
1478         deleg_cur = rcu_dereference(nfsi->delegation);
1479         if (deleg_cur == NULL)
1480                 goto no_delegation;
1481
1482         spin_lock(&deleg_cur->lock);
1483         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1484            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1485             (deleg_cur->type & fmode) != fmode)
1486                 goto no_delegation_unlock;
1487
1488         if (delegation == NULL)
1489                 delegation = &deleg_cur->stateid;
1490         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1491                 goto no_delegation_unlock;
1492
1493         nfs_mark_delegation_referenced(deleg_cur);
1494         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1495                         fmode, &freeme);
1496         ret = 1;
1497 no_delegation_unlock:
1498         spin_unlock(&deleg_cur->lock);
1499 no_delegation:
1500         rcu_read_unlock();
1501
1502         if (!ret && open_stateid != NULL) {
1503                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1504                 ret = 1;
1505         }
1506         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1507                 nfs4_schedule_state_manager(clp);
1508         if (freeme.type != 0)
1509                 nfs4_test_and_free_stateid(server, &freeme,
1510                                 state->owner->so_cred);
1511
1512         return ret;
1513 }
1514
1515 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1516                 const nfs4_stateid *stateid)
1517 {
1518         struct nfs4_state *state = lsp->ls_state;
1519         bool ret = false;
1520
1521         spin_lock(&state->state_lock);
1522         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1523                 goto out_noupdate;
1524         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1525                 goto out_noupdate;
1526         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1527         ret = true;
1528 out_noupdate:
1529         spin_unlock(&state->state_lock);
1530         return ret;
1531 }
1532
1533 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1534 {
1535         struct nfs_delegation *delegation;
1536
1537         rcu_read_lock();
1538         delegation = rcu_dereference(NFS_I(inode)->delegation);
1539         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1540                 rcu_read_unlock();
1541                 return;
1542         }
1543         rcu_read_unlock();
1544         nfs4_inode_return_delegation(inode);
1545 }
1546
1547 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1548 {
1549         struct nfs4_state *state = opendata->state;
1550         struct nfs_inode *nfsi = NFS_I(state->inode);
1551         struct nfs_delegation *delegation;
1552         int open_mode = opendata->o_arg.open_flags;
1553         fmode_t fmode = opendata->o_arg.fmode;
1554         enum open_claim_type4 claim = opendata->o_arg.claim;
1555         nfs4_stateid stateid;
1556         int ret = -EAGAIN;
1557
1558         for (;;) {
1559                 spin_lock(&state->owner->so_lock);
1560                 if (can_open_cached(state, fmode, open_mode)) {
1561                         update_open_stateflags(state, fmode);
1562                         spin_unlock(&state->owner->so_lock);
1563                         goto out_return_state;
1564                 }
1565                 spin_unlock(&state->owner->so_lock);
1566                 rcu_read_lock();
1567                 delegation = rcu_dereference(nfsi->delegation);
1568                 if (!can_open_delegated(delegation, fmode, claim)) {
1569                         rcu_read_unlock();
1570                         break;
1571                 }
1572                 /* Save the delegation */
1573                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1574                 rcu_read_unlock();
1575                 nfs_release_seqid(opendata->o_arg.seqid);
1576                 if (!opendata->is_recover) {
1577                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1578                         if (ret != 0)
1579                                 goto out;
1580                 }
1581                 ret = -EAGAIN;
1582
1583                 /* Try to update the stateid using the delegation */
1584                 if (update_open_stateid(state, NULL, &stateid, fmode))
1585                         goto out_return_state;
1586         }
1587 out:
1588         return ERR_PTR(ret);
1589 out_return_state:
1590         atomic_inc(&state->count);
1591         return state;
1592 }
1593
1594 static void
1595 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1596 {
1597         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1598         struct nfs_delegation *delegation;
1599         int delegation_flags = 0;
1600
1601         rcu_read_lock();
1602         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1603         if (delegation)
1604                 delegation_flags = delegation->flags;
1605         rcu_read_unlock();
1606         switch (data->o_arg.claim) {
1607         default:
1608                 break;
1609         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1610         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1611                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1612                                    "returning a delegation for "
1613                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1614                                    clp->cl_hostname);
1615                 return;
1616         }
1617         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1618                 nfs_inode_set_delegation(state->inode,
1619                                          data->owner->so_cred,
1620                                          &data->o_res);
1621         else
1622                 nfs_inode_reclaim_delegation(state->inode,
1623                                              data->owner->so_cred,
1624                                              &data->o_res);
1625 }
1626
1627 /*
1628  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1629  * and update the nfs4_state.
1630  */
1631 static struct nfs4_state *
1632 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1633 {
1634         struct inode *inode = data->state->inode;
1635         struct nfs4_state *state = data->state;
1636         int ret;
1637
1638         if (!data->rpc_done) {
1639                 if (data->rpc_status)
1640                         return ERR_PTR(data->rpc_status);
1641                 /* cached opens have already been processed */
1642                 goto update;
1643         }
1644
1645         ret = nfs_refresh_inode(inode, &data->f_attr);
1646         if (ret)
1647                 return ERR_PTR(ret);
1648
1649         if (data->o_res.delegation_type != 0)
1650                 nfs4_opendata_check_deleg(data, state);
1651 update:
1652         update_open_stateid(state, &data->o_res.stateid, NULL,
1653                             data->o_arg.fmode);
1654         atomic_inc(&state->count);
1655
1656         return state;
1657 }
1658
1659 static struct nfs4_state *
1660 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1661 {
1662         struct inode *inode;
1663         struct nfs4_state *state = NULL;
1664         int ret;
1665
1666         if (!data->rpc_done) {
1667                 state = nfs4_try_open_cached(data);
1668                 trace_nfs4_cached_open(data->state);
1669                 goto out;
1670         }
1671
1672         ret = -EAGAIN;
1673         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1674                 goto err;
1675         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1676         ret = PTR_ERR(inode);
1677         if (IS_ERR(inode))
1678                 goto err;
1679         ret = -ENOMEM;
1680         state = nfs4_get_open_state(inode, data->owner);
1681         if (state == NULL)
1682                 goto err_put_inode;
1683         if (data->o_res.delegation_type != 0)
1684                 nfs4_opendata_check_deleg(data, state);
1685         update_open_stateid(state, &data->o_res.stateid, NULL,
1686                         data->o_arg.fmode);
1687         iput(inode);
1688 out:
1689         nfs_release_seqid(data->o_arg.seqid);
1690         return state;
1691 err_put_inode:
1692         iput(inode);
1693 err:
1694         return ERR_PTR(ret);
1695 }
1696
1697 static struct nfs4_state *
1698 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1699 {
1700         struct nfs4_state *ret;
1701
1702         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1703                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1704         else
1705                 ret = _nfs4_opendata_to_nfs4_state(data);
1706         nfs4_sequence_free_slot(&data->o_res.seq_res);
1707         return ret;
1708 }
1709
1710 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1711 {
1712         struct nfs_inode *nfsi = NFS_I(state->inode);
1713         struct nfs_open_context *ctx;
1714
1715         spin_lock(&state->inode->i_lock);
1716         list_for_each_entry(ctx, &nfsi->open_files, list) {
1717                 if (ctx->state != state)
1718                         continue;
1719                 get_nfs_open_context(ctx);
1720                 spin_unlock(&state->inode->i_lock);
1721                 return ctx;
1722         }
1723         spin_unlock(&state->inode->i_lock);
1724         return ERR_PTR(-ENOENT);
1725 }
1726
1727 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1728                 struct nfs4_state *state, enum open_claim_type4 claim)
1729 {
1730         struct nfs4_opendata *opendata;
1731
1732         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1733                         NULL, NULL, claim, GFP_NOFS);
1734         if (opendata == NULL)
1735                 return ERR_PTR(-ENOMEM);
1736         opendata->state = state;
1737         atomic_inc(&state->count);
1738         return opendata;
1739 }
1740
1741 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1742                 fmode_t fmode)
1743 {
1744         struct nfs4_state *newstate;
1745         int ret;
1746
1747         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1748                 return 0;
1749         opendata->o_arg.open_flags = 0;
1750         opendata->o_arg.fmode = fmode;
1751         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1752                         NFS_SB(opendata->dentry->d_sb),
1753                         fmode, 0);
1754         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1755         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1756         nfs4_init_opendata_res(opendata);
1757         ret = _nfs4_recover_proc_open(opendata);
1758         if (ret != 0)
1759                 return ret;
1760         newstate = nfs4_opendata_to_nfs4_state(opendata);
1761         if (IS_ERR(newstate))
1762                 return PTR_ERR(newstate);
1763         if (newstate != opendata->state)
1764                 ret = -ESTALE;
1765         nfs4_close_state(newstate, fmode);
1766         return ret;
1767 }
1768
1769 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1770 {
1771         int ret;
1772
1773         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1774         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1775         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1776         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1777         /* memory barrier prior to reading state->n_* */
1778         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1779         clear_bit(NFS_OPEN_STATE, &state->flags);
1780         smp_rmb();
1781         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1782         if (ret != 0)
1783                 return ret;
1784         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1785         if (ret != 0)
1786                 return ret;
1787         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1788         if (ret != 0)
1789                 return ret;
1790         /*
1791          * We may have performed cached opens for all three recoveries.
1792          * Check if we need to update the current stateid.
1793          */
1794         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1795             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1796                 write_seqlock(&state->seqlock);
1797                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1798                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1799                 write_sequnlock(&state->seqlock);
1800         }
1801         return 0;
1802 }
1803
1804 /*
1805  * OPEN_RECLAIM:
1806  *      reclaim state on the server after a reboot.
1807  */
1808 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1809 {
1810         struct nfs_delegation *delegation;
1811         struct nfs4_opendata *opendata;
1812         fmode_t delegation_type = 0;
1813         int status;
1814
1815         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1816                         NFS4_OPEN_CLAIM_PREVIOUS);
1817         if (IS_ERR(opendata))
1818                 return PTR_ERR(opendata);
1819         rcu_read_lock();
1820         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1821         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1822                 delegation_type = delegation->type;
1823         rcu_read_unlock();
1824         opendata->o_arg.u.delegation_type = delegation_type;
1825         status = nfs4_open_recover(opendata, state);
1826         nfs4_opendata_put(opendata);
1827         return status;
1828 }
1829
1830 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1831 {
1832         struct nfs_server *server = NFS_SERVER(state->inode);
1833         struct nfs4_exception exception = { };
1834         int err;
1835         do {
1836                 err = _nfs4_do_open_reclaim(ctx, state);
1837                 trace_nfs4_open_reclaim(ctx, 0, err);
1838                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1839                         continue;
1840                 if (err != -NFS4ERR_DELAY)
1841                         break;
1842                 nfs4_handle_exception(server, err, &exception);
1843         } while (exception.retry);
1844         return err;
1845 }
1846
1847 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1848 {
1849         struct nfs_open_context *ctx;
1850         int ret;
1851
1852         ctx = nfs4_state_find_open_context(state);
1853         if (IS_ERR(ctx))
1854                 return -EAGAIN;
1855         ret = nfs4_do_open_reclaim(ctx, state);
1856         put_nfs_open_context(ctx);
1857         return ret;
1858 }
1859
1860 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1861 {
1862         switch (err) {
1863                 default:
1864                         printk(KERN_ERR "NFS: %s: unhandled error "
1865                                         "%d.\n", __func__, err);
1866                 case 0:
1867                 case -ENOENT:
1868                 case -EAGAIN:
1869                 case -ESTALE:
1870                         break;
1871                 case -NFS4ERR_BADSESSION:
1872                 case -NFS4ERR_BADSLOT:
1873                 case -NFS4ERR_BAD_HIGH_SLOT:
1874                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1875                 case -NFS4ERR_DEADSESSION:
1876                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1877                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1878                         return -EAGAIN;
1879                 case -NFS4ERR_STALE_CLIENTID:
1880                 case -NFS4ERR_STALE_STATEID:
1881                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1882                         /* Don't recall a delegation if it was lost */
1883                         nfs4_schedule_lease_recovery(server->nfs_client);
1884                         return -EAGAIN;
1885                 case -NFS4ERR_MOVED:
1886                         nfs4_schedule_migration_recovery(server);
1887                         return -EAGAIN;
1888                 case -NFS4ERR_LEASE_MOVED:
1889                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1890                         return -EAGAIN;
1891                 case -NFS4ERR_DELEG_REVOKED:
1892                 case -NFS4ERR_ADMIN_REVOKED:
1893                 case -NFS4ERR_EXPIRED:
1894                 case -NFS4ERR_BAD_STATEID:
1895                 case -NFS4ERR_OPENMODE:
1896                         nfs_inode_find_state_and_recover(state->inode,
1897                                         stateid);
1898                         nfs4_schedule_stateid_recovery(server, state);
1899                         return -EAGAIN;
1900                 case -NFS4ERR_DELAY:
1901                 case -NFS4ERR_GRACE:
1902                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1903                         ssleep(1);
1904                         return -EAGAIN;
1905                 case -ENOMEM:
1906                 case -NFS4ERR_DENIED:
1907                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1908                         return 0;
1909         }
1910         return err;
1911 }
1912
1913 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1914                 struct nfs4_state *state, const nfs4_stateid *stateid,
1915                 fmode_t type)
1916 {
1917         struct nfs_server *server = NFS_SERVER(state->inode);
1918         struct nfs4_opendata *opendata;
1919         int err = 0;
1920
1921         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1922                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1923         if (IS_ERR(opendata))
1924                 return PTR_ERR(opendata);
1925         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1926         write_seqlock(&state->seqlock);
1927         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1928         write_sequnlock(&state->seqlock);
1929         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1930         switch (type & (FMODE_READ|FMODE_WRITE)) {
1931         case FMODE_READ|FMODE_WRITE:
1932         case FMODE_WRITE:
1933                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1934                 if (err)
1935                         break;
1936                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1937                 if (err)
1938                         break;
1939         case FMODE_READ:
1940                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1941         }
1942         nfs4_opendata_put(opendata);
1943         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1944 }
1945
1946 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1947 {
1948         struct nfs4_opendata *data = calldata;
1949
1950         nfs4_setup_sequence(data->o_arg.server->nfs_client,
1951                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
1952 }
1953
1954 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1955 {
1956         struct nfs4_opendata *data = calldata;
1957
1958         nfs40_sequence_done(task, &data->c_res.seq_res);
1959
1960         data->rpc_status = task->tk_status;
1961         if (data->rpc_status == 0) {
1962                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1963                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1964                 renew_lease(data->o_res.server, data->timestamp);
1965                 data->rpc_done = 1;
1966         }
1967 }
1968
1969 static void nfs4_open_confirm_release(void *calldata)
1970 {
1971         struct nfs4_opendata *data = calldata;
1972         struct nfs4_state *state = NULL;
1973
1974         /* If this request hasn't been cancelled, do nothing */
1975         if (data->cancelled == 0)
1976                 goto out_free;
1977         /* In case of error, no cleanup! */
1978         if (!data->rpc_done)
1979                 goto out_free;
1980         state = nfs4_opendata_to_nfs4_state(data);
1981         if (!IS_ERR(state))
1982                 nfs4_close_state(state, data->o_arg.fmode);
1983 out_free:
1984         nfs4_opendata_put(data);
1985 }
1986
1987 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1988         .rpc_call_prepare = nfs4_open_confirm_prepare,
1989         .rpc_call_done = nfs4_open_confirm_done,
1990         .rpc_release = nfs4_open_confirm_release,
1991 };
1992
1993 /*
1994  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1995  */
1996 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1997 {
1998         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1999         struct rpc_task *task;
2000         struct  rpc_message msg = {
2001                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2002                 .rpc_argp = &data->c_arg,
2003                 .rpc_resp = &data->c_res,
2004                 .rpc_cred = data->owner->so_cred,
2005         };
2006         struct rpc_task_setup task_setup_data = {
2007                 .rpc_client = server->client,
2008                 .rpc_message = &msg,
2009                 .callback_ops = &nfs4_open_confirm_ops,
2010                 .callback_data = data,
2011                 .workqueue = nfsiod_workqueue,
2012                 .flags = RPC_TASK_ASYNC,
2013         };
2014         int status;
2015
2016         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2017         kref_get(&data->kref);
2018         data->rpc_done = 0;
2019         data->rpc_status = 0;
2020         data->timestamp = jiffies;
2021         if (data->is_recover)
2022                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2023         task = rpc_run_task(&task_setup_data);
2024         if (IS_ERR(task))
2025                 return PTR_ERR(task);
2026         status = rpc_wait_for_completion_task(task);
2027         if (status != 0) {
2028                 data->cancelled = 1;
2029                 smp_wmb();
2030         } else
2031                 status = data->rpc_status;
2032         rpc_put_task(task);
2033         return status;
2034 }
2035
2036 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2037 {
2038         struct nfs4_opendata *data = calldata;
2039         struct nfs4_state_owner *sp = data->owner;
2040         struct nfs_client *clp = sp->so_server->nfs_client;
2041         enum open_claim_type4 claim = data->o_arg.claim;
2042
2043         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2044                 goto out_wait;
2045         /*
2046          * Check if we still need to send an OPEN call, or if we can use
2047          * a delegation instead.
2048          */
2049         if (data->state != NULL) {
2050                 struct nfs_delegation *delegation;
2051
2052                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2053                         goto out_no_action;
2054                 rcu_read_lock();
2055                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2056                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2057                         goto unlock_no_action;
2058                 rcu_read_unlock();
2059         }
2060         /* Update client id. */
2061         data->o_arg.clientid = clp->cl_clientid;
2062         switch (claim) {
2063         default:
2064                 break;
2065         case NFS4_OPEN_CLAIM_PREVIOUS:
2066         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2067         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2068                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2069         case NFS4_OPEN_CLAIM_FH:
2070                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2071                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2072         }
2073         data->timestamp = jiffies;
2074         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2075                                 &data->o_arg.seq_args,
2076                                 &data->o_res.seq_res,
2077                                 task) != 0)
2078                 nfs_release_seqid(data->o_arg.seqid);
2079
2080         /* Set the create mode (note dependency on the session type) */
2081         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2082         if (data->o_arg.open_flags & O_EXCL) {
2083                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2084                 if (nfs4_has_persistent_session(clp))
2085                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2086                 else if (clp->cl_mvops->minor_version > 0)
2087                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2088         }
2089         return;
2090 unlock_no_action:
2091         trace_nfs4_cached_open(data->state);
2092         rcu_read_unlock();
2093 out_no_action:
2094         task->tk_action = NULL;
2095 out_wait:
2096         nfs4_sequence_done(task, &data->o_res.seq_res);
2097 }
2098
2099 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2100 {
2101         struct nfs4_opendata *data = calldata;
2102
2103         data->rpc_status = task->tk_status;
2104
2105         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2106                 return;
2107
2108         if (task->tk_status == 0) {
2109                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2110                         switch (data->o_res.f_attr->mode & S_IFMT) {
2111                         case S_IFREG:
2112                                 break;
2113                         case S_IFLNK:
2114                                 data->rpc_status = -ELOOP;
2115                                 break;
2116                         case S_IFDIR:
2117                                 data->rpc_status = -EISDIR;
2118                                 break;
2119                         default:
2120                                 data->rpc_status = -ENOTDIR;
2121                         }
2122                 }
2123                 renew_lease(data->o_res.server, data->timestamp);
2124                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2125                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2126         }
2127         data->rpc_done = 1;
2128 }
2129
2130 static void nfs4_open_release(void *calldata)
2131 {
2132         struct nfs4_opendata *data = calldata;
2133         struct nfs4_state *state = NULL;
2134
2135         /* If this request hasn't been cancelled, do nothing */
2136         if (data->cancelled == 0)
2137                 goto out_free;
2138         /* In case of error, no cleanup! */
2139         if (data->rpc_status != 0 || !data->rpc_done)
2140                 goto out_free;
2141         /* In case we need an open_confirm, no cleanup! */
2142         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2143                 goto out_free;
2144         state = nfs4_opendata_to_nfs4_state(data);
2145         if (!IS_ERR(state))
2146                 nfs4_close_state(state, data->o_arg.fmode);
2147 out_free:
2148         nfs4_opendata_put(data);
2149 }
2150
2151 static const struct rpc_call_ops nfs4_open_ops = {
2152         .rpc_call_prepare = nfs4_open_prepare,
2153         .rpc_call_done = nfs4_open_done,
2154         .rpc_release = nfs4_open_release,
2155 };
2156
2157 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2158 {
2159         struct inode *dir = d_inode(data->dir);
2160         struct nfs_server *server = NFS_SERVER(dir);
2161         struct nfs_openargs *o_arg = &data->o_arg;
2162         struct nfs_openres *o_res = &data->o_res;
2163         struct rpc_task *task;
2164         struct rpc_message msg = {
2165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2166                 .rpc_argp = o_arg,
2167                 .rpc_resp = o_res,
2168                 .rpc_cred = data->owner->so_cred,
2169         };
2170         struct rpc_task_setup task_setup_data = {
2171                 .rpc_client = server->client,
2172                 .rpc_message = &msg,
2173                 .callback_ops = &nfs4_open_ops,
2174                 .callback_data = data,
2175                 .workqueue = nfsiod_workqueue,
2176                 .flags = RPC_TASK_ASYNC,
2177         };
2178         int status;
2179
2180         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2181         kref_get(&data->kref);
2182         data->rpc_done = 0;
2183         data->rpc_status = 0;
2184         data->cancelled = 0;
2185         data->is_recover = 0;
2186         if (isrecover) {
2187                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2188                 data->is_recover = 1;
2189         }
2190         task = rpc_run_task(&task_setup_data);
2191         if (IS_ERR(task))
2192                 return PTR_ERR(task);
2193         status = rpc_wait_for_completion_task(task);
2194         if (status != 0) {
2195                 data->cancelled = 1;
2196                 smp_wmb();
2197         } else
2198                 status = data->rpc_status;
2199         rpc_put_task(task);
2200
2201         return status;
2202 }
2203
2204 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2205 {
2206         struct inode *dir = d_inode(data->dir);
2207         struct nfs_openres *o_res = &data->o_res;
2208         int status;
2209
2210         status = nfs4_run_open_task(data, 1);
2211         if (status != 0 || !data->rpc_done)
2212                 return status;
2213
2214         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2215
2216         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2217                 status = _nfs4_proc_open_confirm(data);
2218
2219         return status;
2220 }
2221
2222 /*
2223  * Additional permission checks in order to distinguish between an
2224  * open for read, and an open for execute. This works around the
2225  * fact that NFSv4 OPEN treats read and execute permissions as being
2226  * the same.
2227  * Note that in the non-execute case, we want to turn off permission
2228  * checking if we just created a new file (POSIX open() semantics).
2229  */
2230 static int nfs4_opendata_access(struct rpc_cred *cred,
2231                                 struct nfs4_opendata *opendata,
2232                                 struct nfs4_state *state, fmode_t fmode,
2233                                 int openflags)
2234 {
2235         struct nfs_access_entry cache;
2236         u32 mask;
2237
2238         /* access call failed or for some reason the server doesn't
2239          * support any access modes -- defer access call until later */
2240         if (opendata->o_res.access_supported == 0)
2241                 return 0;
2242
2243         mask = 0;
2244         /*
2245          * Use openflags to check for exec, because fmode won't
2246          * always have FMODE_EXEC set when file open for exec.
2247          */
2248         if (openflags & __FMODE_EXEC) {
2249                 /* ONLY check for exec rights */
2250                 mask = MAY_EXEC;
2251         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2252                 mask = MAY_READ;
2253
2254         cache.cred = cred;
2255         cache.jiffies = jiffies;
2256         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2257         nfs_access_add_cache(state->inode, &cache);
2258
2259         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2260                 return 0;
2261
2262         return -EACCES;
2263 }
2264
2265 /*
2266  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2267  */
2268 static int _nfs4_proc_open(struct nfs4_opendata *data)
2269 {
2270         struct inode *dir = d_inode(data->dir);
2271         struct nfs_server *server = NFS_SERVER(dir);
2272         struct nfs_openargs *o_arg = &data->o_arg;
2273         struct nfs_openres *o_res = &data->o_res;
2274         int status;
2275
2276         status = nfs4_run_open_task(data, 0);
2277         if (!data->rpc_done)
2278                 return status;
2279         if (status != 0) {
2280                 if (status == -NFS4ERR_BADNAME &&
2281                                 !(o_arg->open_flags & O_CREAT))
2282                         return -ENOENT;
2283                 return status;
2284         }
2285
2286         nfs_fattr_map_and_free_names(server, &data->f_attr);
2287
2288         if (o_arg->open_flags & O_CREAT) {
2289                 if (o_arg->open_flags & O_EXCL)
2290                         data->file_created = 1;
2291                 else if (o_res->cinfo.before != o_res->cinfo.after)
2292                         data->file_created = 1;
2293                 if (data->file_created || dir->i_version != o_res->cinfo.after)
2294                         update_changeattr(dir, &o_res->cinfo,
2295                                         o_res->f_attr->time_start);
2296         }
2297         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2298                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2299         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2300                 status = _nfs4_proc_open_confirm(data);
2301                 if (status != 0)
2302                         return status;
2303         }
2304         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2305                 nfs4_sequence_free_slot(&o_res->seq_res);
2306                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2307         }
2308         return 0;
2309 }
2310
2311 /*
2312  * OPEN_EXPIRED:
2313  *      reclaim state on the server after a network partition.
2314  *      Assumes caller holds the appropriate lock
2315  */
2316 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2317 {
2318         struct nfs4_opendata *opendata;
2319         int ret;
2320
2321         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2322                         NFS4_OPEN_CLAIM_FH);
2323         if (IS_ERR(opendata))
2324                 return PTR_ERR(opendata);
2325         ret = nfs4_open_recover(opendata, state);
2326         if (ret == -ESTALE)
2327                 d_drop(ctx->dentry);
2328         nfs4_opendata_put(opendata);
2329         return ret;
2330 }
2331
2332 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2333 {
2334         struct nfs_server *server = NFS_SERVER(state->inode);
2335         struct nfs4_exception exception = { };
2336         int err;
2337
2338         do {
2339                 err = _nfs4_open_expired(ctx, state);
2340                 trace_nfs4_open_expired(ctx, 0, err);
2341                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2342                         continue;
2343                 switch (err) {
2344                 default:
2345                         goto out;
2346                 case -NFS4ERR_GRACE:
2347                 case -NFS4ERR_DELAY:
2348                         nfs4_handle_exception(server, err, &exception);
2349                         err = 0;
2350                 }
2351         } while (exception.retry);
2352 out:
2353         return err;
2354 }
2355
2356 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2357 {
2358         struct nfs_open_context *ctx;
2359         int ret;
2360
2361         ctx = nfs4_state_find_open_context(state);
2362         if (IS_ERR(ctx))
2363                 return -EAGAIN;
2364         ret = nfs4_do_open_expired(ctx, state);
2365         put_nfs_open_context(ctx);
2366         return ret;
2367 }
2368
2369 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2370                 const nfs4_stateid *stateid)
2371 {
2372         nfs_remove_bad_delegation(state->inode, stateid);
2373         write_seqlock(&state->seqlock);
2374         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2375         write_sequnlock(&state->seqlock);
2376         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2377 }
2378
2379 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2380 {
2381         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2382                 nfs_finish_clear_delegation_stateid(state, NULL);
2383 }
2384
2385 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2386 {
2387         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2388         nfs40_clear_delegation_stateid(state);
2389         return nfs4_open_expired(sp, state);
2390 }
2391
2392 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2393                 nfs4_stateid *stateid,
2394                 struct rpc_cred *cred)
2395 {
2396         return -NFS4ERR_BAD_STATEID;
2397 }
2398
2399 #if defined(CONFIG_NFS_V4_1)
2400 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2401                 nfs4_stateid *stateid,
2402                 struct rpc_cred *cred)
2403 {
2404         int status;
2405
2406         switch (stateid->type) {
2407         default:
2408                 break;
2409         case NFS4_INVALID_STATEID_TYPE:
2410         case NFS4_SPECIAL_STATEID_TYPE:
2411                 return -NFS4ERR_BAD_STATEID;
2412         case NFS4_REVOKED_STATEID_TYPE:
2413                 goto out_free;
2414         }
2415
2416         status = nfs41_test_stateid(server, stateid, cred);
2417         switch (status) {
2418         case -NFS4ERR_EXPIRED:
2419         case -NFS4ERR_ADMIN_REVOKED:
2420         case -NFS4ERR_DELEG_REVOKED:
2421                 break;
2422         default:
2423                 return status;
2424         }
2425 out_free:
2426         /* Ack the revoked state to the server */
2427         nfs41_free_stateid(server, stateid, cred, true);
2428         return -NFS4ERR_EXPIRED;
2429 }
2430
2431 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2432 {
2433         struct nfs_server *server = NFS_SERVER(state->inode);
2434         nfs4_stateid stateid;
2435         struct nfs_delegation *delegation;
2436         struct rpc_cred *cred;
2437         int status;
2438
2439         /* Get the delegation credential for use by test/free_stateid */
2440         rcu_read_lock();
2441         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2442         if (delegation == NULL) {
2443                 rcu_read_unlock();
2444                 return;
2445         }
2446
2447         nfs4_stateid_copy(&stateid, &delegation->stateid);
2448         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2449                 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2450                         &delegation->flags)) {
2451                 rcu_read_unlock();
2452                 nfs_finish_clear_delegation_stateid(state, &stateid);
2453                 return;
2454         }
2455
2456         cred = get_rpccred(delegation->cred);
2457         rcu_read_unlock();
2458         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2459         trace_nfs4_test_delegation_stateid(state, NULL, status);
2460         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2461                 nfs_finish_clear_delegation_stateid(state, &stateid);
2462
2463         put_rpccred(cred);
2464 }
2465
2466 /**
2467  * nfs41_check_expired_locks - possibly free a lock stateid
2468  *
2469  * @state: NFSv4 state for an inode
2470  *
2471  * Returns NFS_OK if recovery for this stateid is now finished.
2472  * Otherwise a negative NFS4ERR value is returned.
2473  */
2474 static int nfs41_check_expired_locks(struct nfs4_state *state)
2475 {
2476         int status, ret = NFS_OK;
2477         struct nfs4_lock_state *lsp, *prev = NULL;
2478         struct nfs_server *server = NFS_SERVER(state->inode);
2479
2480         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2481                 goto out;
2482
2483         spin_lock(&state->state_lock);
2484         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2485                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2486                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2487
2488                         atomic_inc(&lsp->ls_count);
2489                         spin_unlock(&state->state_lock);
2490
2491                         nfs4_put_lock_state(prev);
2492                         prev = lsp;
2493
2494                         status = nfs41_test_and_free_expired_stateid(server,
2495                                         &lsp->ls_stateid,
2496                                         cred);
2497                         trace_nfs4_test_lock_stateid(state, lsp, status);
2498                         if (status == -NFS4ERR_EXPIRED ||
2499                             status == -NFS4ERR_BAD_STATEID) {
2500                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2501                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2502                                 if (!recover_lost_locks)
2503                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2504                         } else if (status != NFS_OK) {
2505                                 ret = status;
2506                                 nfs4_put_lock_state(prev);
2507                                 goto out;
2508                         }
2509                         spin_lock(&state->state_lock);
2510                 }
2511         }
2512         spin_unlock(&state->state_lock);
2513         nfs4_put_lock_state(prev);
2514 out:
2515         return ret;
2516 }
2517
2518 /**
2519  * nfs41_check_open_stateid - possibly free an open stateid
2520  *
2521  * @state: NFSv4 state for an inode
2522  *
2523  * Returns NFS_OK if recovery for this stateid is now finished.
2524  * Otherwise a negative NFS4ERR value is returned.
2525  */
2526 static int nfs41_check_open_stateid(struct nfs4_state *state)
2527 {
2528         struct nfs_server *server = NFS_SERVER(state->inode);
2529         nfs4_stateid *stateid = &state->open_stateid;
2530         struct rpc_cred *cred = state->owner->so_cred;
2531         int status;
2532
2533         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2534                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2535                         if (nfs4_have_delegation(state->inode, state->state))
2536                                 return NFS_OK;
2537                         return -NFS4ERR_OPENMODE;
2538                 }
2539                 return -NFS4ERR_BAD_STATEID;
2540         }
2541         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2542         trace_nfs4_test_open_stateid(state, NULL, status);
2543         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2544                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2545                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2546                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2547                 clear_bit(NFS_OPEN_STATE, &state->flags);
2548                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2549         }
2550         if (status != NFS_OK)
2551                 return status;
2552         if (nfs_open_stateid_recover_openmode(state))
2553                 return -NFS4ERR_OPENMODE;
2554         return NFS_OK;
2555 }
2556
2557 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2558 {
2559         int status;
2560
2561         nfs41_check_delegation_stateid(state);
2562         status = nfs41_check_expired_locks(state);
2563         if (status != NFS_OK)
2564                 return status;
2565         status = nfs41_check_open_stateid(state);
2566         if (status != NFS_OK)
2567                 status = nfs4_open_expired(sp, state);
2568         return status;
2569 }
2570 #endif
2571
2572 /*
2573  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2574  * fields corresponding to attributes that were used to store the verifier.
2575  * Make sure we clobber those fields in the later setattr call
2576  */
2577 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2578                                 struct iattr *sattr, struct nfs4_label **label)
2579 {
2580         const u32 *attrset = opendata->o_res.attrset;
2581
2582         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2583             !(sattr->ia_valid & ATTR_ATIME_SET))
2584                 sattr->ia_valid |= ATTR_ATIME;
2585
2586         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2587             !(sattr->ia_valid & ATTR_MTIME_SET))
2588                 sattr->ia_valid |= ATTR_MTIME;
2589
2590         /* Except MODE, it seems harmless of setting twice. */
2591         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2592                 (attrset[1] & FATTR4_WORD1_MODE ||
2593                  attrset[2] & FATTR4_WORD2_MODE_UMASK))
2594                 sattr->ia_valid &= ~ATTR_MODE;
2595
2596         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2597                 *label = NULL;
2598 }
2599
2600 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2601                 fmode_t fmode,
2602                 int flags,
2603                 struct nfs_open_context *ctx)
2604 {
2605         struct nfs4_state_owner *sp = opendata->owner;
2606         struct nfs_server *server = sp->so_server;
2607         struct dentry *dentry;
2608         struct nfs4_state *state;
2609         unsigned int seq;
2610         int ret;
2611
2612         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2613
2614         ret = _nfs4_proc_open(opendata);
2615         if (ret != 0)
2616                 goto out;
2617
2618         state = nfs4_opendata_to_nfs4_state(opendata);
2619         ret = PTR_ERR(state);
2620         if (IS_ERR(state))
2621                 goto out;
2622         ctx->state = state;
2623         if (server->caps & NFS_CAP_POSIX_LOCK)
2624                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2625         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2626                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2627
2628         dentry = opendata->dentry;
2629         if (d_really_is_negative(dentry)) {
2630                 struct dentry *alias;
2631                 d_drop(dentry);
2632                 alias = d_exact_alias(dentry, state->inode);
2633                 if (!alias)
2634                         alias = d_splice_alias(igrab(state->inode), dentry);
2635                 /* d_splice_alias() can't fail here - it's a non-directory */
2636                 if (alias) {
2637                         dput(ctx->dentry);
2638                         ctx->dentry = dentry = alias;
2639                 }
2640                 nfs_set_verifier(dentry,
2641                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2642         }
2643
2644         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2645         if (ret != 0)
2646                 goto out;
2647
2648         if (d_inode(dentry) == state->inode) {
2649                 nfs_inode_attach_open_context(ctx);
2650                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2651                         nfs4_schedule_stateid_recovery(server, state);
2652         }
2653 out:
2654         return ret;
2655 }
2656
2657 /*
2658  * Returns a referenced nfs4_state
2659  */
2660 static int _nfs4_do_open(struct inode *dir,
2661                         struct nfs_open_context *ctx,
2662                         int flags,
2663                         struct iattr *sattr,
2664                         struct nfs4_label *label,
2665                         int *opened)
2666 {
2667         struct nfs4_state_owner  *sp;
2668         struct nfs4_state     *state = NULL;
2669         struct nfs_server       *server = NFS_SERVER(dir);
2670         struct nfs4_opendata *opendata;
2671         struct dentry *dentry = ctx->dentry;
2672         struct rpc_cred *cred = ctx->cred;
2673         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2674         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2675         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2676         struct nfs4_label *olabel = NULL;
2677         int status;
2678
2679         /* Protect against reboot recovery conflicts */
2680         status = -ENOMEM;
2681         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2682         if (sp == NULL) {
2683                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2684                 goto out_err;
2685         }
2686         status = nfs4_client_recover_expired_lease(server->nfs_client);
2687         if (status != 0)
2688                 goto err_put_state_owner;
2689         if (d_really_is_positive(dentry))
2690                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2691         status = -ENOMEM;
2692         if (d_really_is_positive(dentry))
2693                 claim = NFS4_OPEN_CLAIM_FH;
2694         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2695                         label, claim, GFP_KERNEL);
2696         if (opendata == NULL)
2697                 goto err_put_state_owner;
2698
2699         if (label) {
2700                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2701                 if (IS_ERR(olabel)) {
2702                         status = PTR_ERR(olabel);
2703                         goto err_opendata_put;
2704                 }
2705         }
2706
2707         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2708                 if (!opendata->f_attr.mdsthreshold) {
2709                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2710                         if (!opendata->f_attr.mdsthreshold)
2711                                 goto err_free_label;
2712                 }
2713                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2714         }
2715         if (d_really_is_positive(dentry))
2716                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2717
2718         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2719         if (status != 0)
2720                 goto err_free_label;
2721         state = ctx->state;
2722
2723         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2724             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2725                 nfs4_exclusive_attrset(opendata, sattr, &label);
2726                 /*
2727                  * send create attributes which was not set by open
2728                  * with an extra setattr.
2729                  */
2730                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2731                         nfs_fattr_init(opendata->o_res.f_attr);
2732                         status = nfs4_do_setattr(state->inode, cred,
2733                                         opendata->o_res.f_attr, sattr,
2734                                         ctx, label, olabel);
2735                         if (status == 0) {
2736                                 nfs_setattr_update_inode(state->inode, sattr,
2737                                                 opendata->o_res.f_attr);
2738                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2739                         }
2740                 }
2741         }
2742         if (opened && opendata->file_created)
2743                 *opened |= FILE_CREATED;
2744
2745         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2746                 *ctx_th = opendata->f_attr.mdsthreshold;
2747                 opendata->f_attr.mdsthreshold = NULL;
2748         }
2749
2750         nfs4_label_free(olabel);
2751
2752         nfs4_opendata_put(opendata);
2753         nfs4_put_state_owner(sp);
2754         return 0;
2755 err_free_label:
2756         nfs4_label_free(olabel);
2757 err_opendata_put:
2758         nfs4_opendata_put(opendata);
2759 err_put_state_owner:
2760         nfs4_put_state_owner(sp);
2761 out_err:
2762         return status;
2763 }
2764
2765
2766 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2767                                         struct nfs_open_context *ctx,
2768                                         int flags,
2769                                         struct iattr *sattr,
2770                                         struct nfs4_label *label,
2771                                         int *opened)
2772 {
2773         struct nfs_server *server = NFS_SERVER(dir);
2774         struct nfs4_exception exception = { };
2775         struct nfs4_state *res;
2776         int status;
2777
2778         do {
2779                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2780                 res = ctx->state;
2781                 trace_nfs4_open_file(ctx, flags, status);
2782                 if (status == 0)
2783                         break;
2784                 /* NOTE: BAD_SEQID means the server and client disagree about the
2785                  * book-keeping w.r.t. state-changing operations
2786                  * (OPEN/CLOSE/LOCK/LOCKU...)
2787                  * It is actually a sign of a bug on the client or on the server.
2788                  *
2789                  * If we receive a BAD_SEQID error in the particular case of
2790                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2791                  * have unhashed the old state_owner for us, and that we can
2792                  * therefore safely retry using a new one. We should still warn
2793                  * the user though...
2794                  */
2795                 if (status == -NFS4ERR_BAD_SEQID) {
2796                         pr_warn_ratelimited("NFS: v4 server %s "
2797                                         " returned a bad sequence-id error!\n",
2798                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2799                         exception.retry = 1;
2800                         continue;
2801                 }
2802                 /*
2803                  * BAD_STATEID on OPEN means that the server cancelled our
2804                  * state before it received the OPEN_CONFIRM.
2805                  * Recover by retrying the request as per the discussion
2806                  * on Page 181 of RFC3530.
2807                  */
2808                 if (status == -NFS4ERR_BAD_STATEID) {
2809                         exception.retry = 1;
2810                         continue;
2811                 }
2812                 if (status == -EAGAIN) {
2813                         /* We must have found a delegation */
2814                         exception.retry = 1;
2815                         continue;
2816                 }
2817                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2818                         continue;
2819                 res = ERR_PTR(nfs4_handle_exception(server,
2820                                         status, &exception));
2821         } while (exception.retry);
2822         return res;
2823 }
2824
2825 static int _nfs4_do_setattr(struct inode *inode,
2826                             struct nfs_setattrargs *arg,
2827                             struct nfs_setattrres *res,
2828                             struct rpc_cred *cred,
2829                             struct nfs_open_context *ctx)
2830 {
2831         struct nfs_server *server = NFS_SERVER(inode);
2832         struct rpc_message msg = {
2833                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2834                 .rpc_argp       = arg,
2835                 .rpc_resp       = res,
2836                 .rpc_cred       = cred,
2837         };
2838         struct rpc_cred *delegation_cred = NULL;
2839         unsigned long timestamp = jiffies;
2840         fmode_t fmode;
2841         bool truncate;
2842         int status;
2843
2844         nfs_fattr_init(res->fattr);
2845
2846         /* Servers should only apply open mode checks for file size changes */
2847         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2848         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2849
2850         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2851                 /* Use that stateid */
2852         } else if (truncate && ctx != NULL) {
2853                 struct nfs_lock_context *l_ctx;
2854                 if (!nfs4_valid_open_stateid(ctx->state))
2855                         return -EBADF;
2856                 l_ctx = nfs_get_lock_context(ctx);
2857                 if (IS_ERR(l_ctx))
2858                         return PTR_ERR(l_ctx);
2859                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2860                                                 &arg->stateid, &delegation_cred);
2861                 nfs_put_lock_context(l_ctx);
2862                 if (status == -EIO)
2863                         return -EBADF;
2864         } else
2865                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2866         if (delegation_cred)
2867                 msg.rpc_cred = delegation_cred;
2868
2869         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2870
2871         put_rpccred(delegation_cred);
2872         if (status == 0 && ctx != NULL)
2873                 renew_lease(server, timestamp);
2874         trace_nfs4_setattr(inode, &arg->stateid, status);
2875         return status;
2876 }
2877
2878 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2879                            struct nfs_fattr *fattr, struct iattr *sattr,
2880                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2881                            struct nfs4_label *olabel)
2882 {
2883         struct nfs_server *server = NFS_SERVER(inode);
2884         struct nfs4_state *state = ctx ? ctx->state : NULL;
2885         struct nfs_setattrargs  arg = {
2886                 .fh             = NFS_FH(inode),
2887                 .iap            = sattr,
2888                 .server         = server,
2889                 .bitmask = server->attr_bitmask,
2890                 .label          = ilabel,
2891         };
2892         struct nfs_setattrres  res = {
2893                 .fattr          = fattr,
2894                 .label          = olabel,
2895                 .server         = server,
2896         };
2897         struct nfs4_exception exception = {
2898                 .state = state,
2899                 .inode = inode,
2900                 .stateid = &arg.stateid,
2901         };
2902         int err;
2903
2904         arg.bitmask = nfs4_bitmask(server, ilabel);
2905         if (ilabel)
2906                 arg.bitmask = nfs4_bitmask(server, olabel);
2907
2908         do {
2909                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
2910                 switch (err) {
2911                 case -NFS4ERR_OPENMODE:
2912                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2913                                 pr_warn_once("NFSv4: server %s is incorrectly "
2914                                                 "applying open mode checks to "
2915                                                 "a SETATTR that is not "
2916                                                 "changing file size.\n",
2917                                                 server->nfs_client->cl_hostname);
2918                         }
2919                         if (state && !(state->state & FMODE_WRITE)) {
2920                                 err = -EBADF;
2921                                 if (sattr->ia_valid & ATTR_OPEN)
2922                                         err = -EACCES;
2923                                 goto out;
2924                         }
2925                 }
2926                 err = nfs4_handle_exception(server, err, &exception);
2927         } while (exception.retry);
2928 out:
2929         return err;
2930 }
2931
2932 static bool
2933 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2934 {
2935         if (inode == NULL || !nfs_have_layout(inode))
2936                 return false;
2937
2938         return pnfs_wait_on_layoutreturn(inode, task);
2939 }
2940
2941 struct nfs4_closedata {
2942         struct inode *inode;
2943         struct nfs4_state *state;
2944         struct nfs_closeargs arg;
2945         struct nfs_closeres res;
2946         struct {
2947                 struct nfs4_layoutreturn_args arg;
2948                 struct nfs4_layoutreturn_res res;
2949                 struct nfs4_xdr_opaque_data ld_private;
2950                 u32 roc_barrier;
2951                 bool roc;
2952         } lr;
2953         struct nfs_fattr fattr;
2954         unsigned long timestamp;
2955 };
2956
2957 static void nfs4_free_closedata(void *data)
2958 {
2959         struct nfs4_closedata *calldata = data;
2960         struct nfs4_state_owner *sp = calldata->state->owner;
2961         struct super_block *sb = calldata->state->inode->i_sb;
2962
2963         if (calldata->lr.roc)
2964                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
2965                                 calldata->res.lr_ret);
2966         nfs4_put_open_state(calldata->state);
2967         nfs_free_seqid(calldata->arg.seqid);
2968         nfs4_put_state_owner(sp);
2969         nfs_sb_deactive(sb);
2970         kfree(calldata);
2971 }
2972
2973 static void nfs4_close_done(struct rpc_task *task, void *data)
2974 {
2975         struct nfs4_closedata *calldata = data;
2976         struct nfs4_state *state = calldata->state;
2977         struct nfs_server *server = NFS_SERVER(calldata->inode);
2978         nfs4_stateid *res_stateid = NULL;
2979
2980         dprintk("%s: begin!\n", __func__);
2981         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2982                 return;
2983         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2984
2985         /* Handle Layoutreturn errors */
2986         if (calldata->arg.lr_args && task->tk_status != 0) {
2987                 switch (calldata->res.lr_ret) {
2988                 default:
2989                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
2990                         break;
2991                 case 0:
2992                         calldata->arg.lr_args = NULL;
2993                         calldata->res.lr_res = NULL;
2994                         break;
2995                 case -NFS4ERR_ADMIN_REVOKED:
2996                 case -NFS4ERR_DELEG_REVOKED:
2997                 case -NFS4ERR_EXPIRED:
2998                 case -NFS4ERR_BAD_STATEID:
2999                 case -NFS4ERR_OLD_STATEID:
3000                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3001                 case -NFS4ERR_WRONG_CRED:
3002                         calldata->arg.lr_args = NULL;
3003                         calldata->res.lr_res = NULL;
3004                         calldata->res.lr_ret = 0;
3005                         rpc_restart_call_prepare(task);
3006                         return;
3007                 }
3008         }
3009
3010         /* hmm. we are done with the inode, and in the process of freeing
3011          * the state_owner. we keep this around to process errors
3012          */
3013         switch (task->tk_status) {
3014                 case 0:
3015                         res_stateid = &calldata->res.stateid;
3016                         renew_lease(server, calldata->timestamp);
3017                         break;
3018                 case -NFS4ERR_ACCESS:
3019                         if (calldata->arg.bitmask != NULL) {
3020                                 calldata->arg.bitmask = NULL;
3021                                 calldata->res.fattr = NULL;
3022                                 task->tk_status = 0;
3023                                 rpc_restart_call_prepare(task);
3024                                 goto out_release;
3025
3026                         }
3027                         break;
3028                 case -NFS4ERR_ADMIN_REVOKED:
3029                 case -NFS4ERR_STALE_STATEID:
3030                 case -NFS4ERR_EXPIRED:
3031                         nfs4_free_revoked_stateid(server,
3032                                         &calldata->arg.stateid,
3033                                         task->tk_msg.rpc_cred);
3034                 case -NFS4ERR_OLD_STATEID:
3035                 case -NFS4ERR_BAD_STATEID:
3036                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3037                                                 &state->open_stateid)) {
3038                                 rpc_restart_call_prepare(task);
3039                                 goto out_release;
3040                         }
3041                         if (calldata->arg.fmode == 0)
3042                                 break;
3043                 default:
3044                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3045                                 rpc_restart_call_prepare(task);
3046                                 goto out_release;
3047                         }
3048         }
3049         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3050                         res_stateid, calldata->arg.fmode);
3051 out_release:
3052         nfs_release_seqid(calldata->arg.seqid);
3053         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3054         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3055 }
3056
3057 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3058 {
3059         struct nfs4_closedata *calldata = data;
3060         struct nfs4_state *state = calldata->state;
3061         struct inode *inode = calldata->inode;
3062         bool is_rdonly, is_wronly, is_rdwr;
3063         int call_close = 0;
3064
3065         dprintk("%s: begin!\n", __func__);
3066         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3067                 goto out_wait;
3068
3069         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3070         spin_lock(&state->owner->so_lock);
3071         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3072         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3073         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3074         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3075         /* Calculate the change in open mode */
3076         calldata->arg.fmode = 0;
3077         if (state->n_rdwr == 0) {
3078                 if (state->n_rdonly == 0)
3079                         call_close |= is_rdonly;
3080                 else if (is_rdonly)
3081                         calldata->arg.fmode |= FMODE_READ;
3082                 if (state->n_wronly == 0)
3083                         call_close |= is_wronly;
3084                 else if (is_wronly)
3085                         calldata->arg.fmode |= FMODE_WRITE;
3086                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3087                         call_close |= is_rdwr;
3088         } else if (is_rdwr)
3089                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3090
3091         if (!nfs4_valid_open_stateid(state) ||
3092             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3093                 call_close = 0;
3094         spin_unlock(&state->owner->so_lock);
3095
3096         if (!call_close) {
3097                 /* Note: exit _without_ calling nfs4_close_done */
3098                 goto out_no_action;
3099         }
3100
3101         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3102                 nfs_release_seqid(calldata->arg.seqid);
3103                 goto out_wait;
3104         }
3105
3106         if (calldata->arg.fmode == 0)
3107                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3108
3109         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3110                 /* Close-to-open cache consistency revalidation */
3111                 if (!nfs4_have_delegation(inode, FMODE_READ))
3112                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3113                 else
3114                         calldata->arg.bitmask = NULL;
3115         }
3116
3117         calldata->arg.share_access =
3118                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3119                                 calldata->arg.fmode, 0);
3120
3121         if (calldata->res.fattr == NULL)
3122                 calldata->arg.bitmask = NULL;
3123         else if (calldata->arg.bitmask == NULL)
3124                 calldata->res.fattr = NULL;
3125         calldata->timestamp = jiffies;
3126         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3127                                 &calldata->arg.seq_args,
3128                                 &calldata->res.seq_res,
3129                                 task) != 0)
3130                 nfs_release_seqid(calldata->arg.seqid);
3131         dprintk("%s: done!\n", __func__);
3132         return;
3133 out_no_action:
3134         task->tk_action = NULL;
3135 out_wait:
3136         nfs4_sequence_done(task, &calldata->res.seq_res);
3137 }
3138
3139 static const struct rpc_call_ops nfs4_close_ops = {
3140         .rpc_call_prepare = nfs4_close_prepare,
3141         .rpc_call_done = nfs4_close_done,
3142         .rpc_release = nfs4_free_closedata,
3143 };
3144
3145 /*
3146  * It is possible for data to be read/written from a mem-mapped file
3147  * after the sys_close call (which hits the vfs layer as a flush).
3148  * This means that we can't safely call nfsv4 close on a file until
3149  * the inode is cleared. This in turn means that we are not good
3150  * NFSv4 citizens - we do not indicate to the server to update the file's
3151  * share state even when we are done with one of the three share
3152  * stateid's in the inode.
3153  *
3154  * NOTE: Caller must be holding the sp->so_owner semaphore!
3155  */
3156 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3157 {
3158         struct nfs_server *server = NFS_SERVER(state->inode);
3159         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3160         struct nfs4_closedata *calldata;
3161         struct nfs4_state_owner *sp = state->owner;
3162         struct rpc_task *task;
3163         struct rpc_message msg = {
3164                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3165                 .rpc_cred = state->owner->so_cred,
3166         };
3167         struct rpc_task_setup task_setup_data = {
3168                 .rpc_client = server->client,
3169                 .rpc_message = &msg,
3170                 .callback_ops = &nfs4_close_ops,
3171                 .workqueue = nfsiod_workqueue,
3172                 .flags = RPC_TASK_ASYNC,
3173         };
3174         int status = -ENOMEM;
3175
3176         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3177                 &task_setup_data.rpc_client, &msg);
3178
3179         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3180         if (calldata == NULL)
3181                 goto out;
3182         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3183         calldata->inode = state->inode;
3184         calldata->state = state;
3185         calldata->arg.fh = NFS_FH(state->inode);
3186         /* Serialization for the sequence id */
3187         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3188         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3189         if (IS_ERR(calldata->arg.seqid))
3190                 goto out_free_calldata;
3191         nfs_fattr_init(&calldata->fattr);
3192         calldata->arg.fmode = 0;
3193         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3194         calldata->res.fattr = &calldata->fattr;
3195         calldata->res.seqid = calldata->arg.seqid;
3196         calldata->res.server = server;
3197         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3198         calldata->lr.roc = pnfs_roc(state->inode,
3199                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3200         if (calldata->lr.roc) {
3201                 calldata->arg.lr_args = &calldata->lr.arg;
3202                 calldata->res.lr_res = &calldata->lr.res;
3203         }
3204         nfs_sb_active(calldata->inode->i_sb);
3205
3206         msg.rpc_argp = &calldata->arg;
3207         msg.rpc_resp = &calldata->res;
3208         task_setup_data.callback_data = calldata;
3209         task = rpc_run_task(&task_setup_data);
3210         if (IS_ERR(task))
3211                 return PTR_ERR(task);
3212         status = 0;
3213         if (wait)
3214                 status = rpc_wait_for_completion_task(task);
3215         rpc_put_task(task);
3216         return status;
3217 out_free_calldata:
3218         kfree(calldata);
3219 out:
3220         nfs4_put_open_state(state);
3221         nfs4_put_state_owner(sp);
3222         return status;
3223 }
3224
3225 static struct inode *
3226 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3227                 int open_flags, struct iattr *attr, int *opened)
3228 {
3229         struct nfs4_state *state;
3230         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3231
3232         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3233
3234         /* Protect against concurrent sillydeletes */
3235         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3236
3237         nfs4_label_release_security(label);
3238
3239         if (IS_ERR(state))
3240                 return ERR_CAST(state);
3241         return state->inode;
3242 }
3243
3244 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3245 {
3246         if (ctx->state == NULL)
3247                 return;
3248         if (is_sync)
3249                 nfs4_close_sync(ctx->state, ctx->mode);
3250         else
3251                 nfs4_close_state(ctx->state, ctx->mode);
3252 }
3253
3254 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3255 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3256 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3257
3258 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3259 {
3260         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3261         struct nfs4_server_caps_arg args = {
3262                 .fhandle = fhandle,
3263                 .bitmask = bitmask,
3264         };
3265         struct nfs4_server_caps_res res = {};
3266         struct rpc_message msg = {
3267                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3268                 .rpc_argp = &args,
3269                 .rpc_resp = &res,
3270         };
3271         int status;
3272         int i;
3273
3274         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3275                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3276                      FATTR4_WORD0_LINK_SUPPORT |
3277                      FATTR4_WORD0_SYMLINK_SUPPORT |
3278                      FATTR4_WORD0_ACLSUPPORT;
3279         if (minorversion)
3280                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3281
3282         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3283         if (status == 0) {
3284                 /* Sanity check the server answers */
3285                 switch (minorversion) {
3286                 case 0:
3287                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3288                         res.attr_bitmask[2] = 0;
3289                         break;
3290                 case 1:
3291                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3292                         break;
3293                 case 2:
3294                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3295                 }
3296                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3297                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3298                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3299                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3300                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3301                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3302                                 NFS_CAP_SECURITY_LABEL);
3303                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3304                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3305                         server->caps |= NFS_CAP_ACLS;
3306                 if (res.has_links != 0)
3307                         server->caps |= NFS_CAP_HARDLINKS;
3308                 if (res.has_symlinks != 0)
3309                         server->caps |= NFS_CAP_SYMLINKS;
3310                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3311                         server->caps |= NFS_CAP_FILEID;
3312                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3313                         server->caps |= NFS_CAP_MODE;
3314                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3315                         server->caps |= NFS_CAP_NLINK;
3316                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3317                         server->caps |= NFS_CAP_OWNER;
3318                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3319                         server->caps |= NFS_CAP_OWNER_GROUP;
3320                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3321                         server->caps |= NFS_CAP_ATIME;
3322                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3323                         server->caps |= NFS_CAP_CTIME;
3324                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3325                         server->caps |= NFS_CAP_MTIME;
3326 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3327                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3328                         server->caps |= NFS_CAP_SECURITY_LABEL;
3329 #endif
3330                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3331                                 sizeof(server->attr_bitmask));
3332                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3333
3334                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3335                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3336                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3337                 server->cache_consistency_bitmask[2] = 0;
3338
3339                 /* Avoid a regression due to buggy server */
3340                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3341                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3342                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3343                         sizeof(server->exclcreat_bitmask));
3344
3345                 server->acl_bitmask = res.acl_bitmask;
3346                 server->fh_expire_type = res.fh_expire_type;
3347         }
3348
3349         return status;
3350 }
3351
3352 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3353 {
3354         struct nfs4_exception exception = { };
3355         int err;
3356         do {
3357                 err = nfs4_handle_exception(server,
3358                                 _nfs4_server_capabilities(server, fhandle),
3359                                 &exception);
3360         } while (exception.retry);
3361         return err;
3362 }
3363
3364 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3365                 struct nfs_fsinfo *info)
3366 {
3367         u32 bitmask[3];
3368         struct nfs4_lookup_root_arg args = {
3369                 .bitmask = bitmask,
3370         };
3371         struct nfs4_lookup_res res = {
3372                 .server = server,
3373                 .fattr = info->fattr,
3374                 .fh = fhandle,
3375         };
3376         struct rpc_message msg = {
3377                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3378                 .rpc_argp = &args,
3379                 .rpc_resp = &res,
3380         };
3381
3382         bitmask[0] = nfs4_fattr_bitmap[0];
3383         bitmask[1] = nfs4_fattr_bitmap[1];
3384         /*
3385          * Process the label in the upcoming getfattr
3386          */
3387         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3388
3389         nfs_fattr_init(info->fattr);
3390         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3391 }
3392
3393 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3394                 struct nfs_fsinfo *info)
3395 {
3396         struct nfs4_exception exception = { };
3397         int err;
3398         do {
3399                 err = _nfs4_lookup_root(server, fhandle, info);
3400                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3401                 switch (err) {
3402                 case 0:
3403                 case -NFS4ERR_WRONGSEC:
3404                         goto out;
3405                 default:
3406                         err = nfs4_handle_exception(server, err, &exception);
3407                 }
3408         } while (exception.retry);
3409 out:
3410         return err;
3411 }
3412
3413 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3414                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3415 {
3416         struct rpc_auth_create_args auth_args = {
3417                 .pseudoflavor = flavor,
3418         };
3419         struct rpc_auth *auth;
3420
3421         auth = rpcauth_create(&auth_args, server->client);
3422         if (IS_ERR(auth))
3423                 return -EACCES;
3424         return nfs4_lookup_root(server, fhandle, info);
3425 }
3426
3427 /*
3428  * Retry pseudoroot lookup with various security flavors.  We do this when:
3429  *
3430  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3431  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3432  *
3433  * Returns zero on success, or a negative NFS4ERR value, or a
3434  * negative errno value.
3435  */
3436 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3437                               struct nfs_fsinfo *info)
3438 {
3439         /* Per 3530bis 15.33.5 */
3440         static const rpc_authflavor_t flav_array[] = {
3441                 RPC_AUTH_GSS_KRB5P,
3442                 RPC_AUTH_GSS_KRB5I,
3443                 RPC_AUTH_GSS_KRB5,
3444                 RPC_AUTH_UNIX,                  /* courtesy */
3445                 RPC_AUTH_NULL,
3446         };
3447         int status = -EPERM;
3448         size_t i;
3449
3450         if (server->auth_info.flavor_len > 0) {
3451                 /* try each flavor specified by user */
3452                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3453                         status = nfs4_lookup_root_sec(server, fhandle, info,
3454                                                 server->auth_info.flavors[i]);
3455                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3456                                 continue;
3457                         break;
3458                 }
3459         } else {
3460                 /* no flavors specified by user, try default list */
3461                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3462                         status = nfs4_lookup_root_sec(server, fhandle, info,
3463                                                       flav_array[i]);
3464                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3465                                 continue;
3466                         break;
3467                 }
3468         }
3469
3470         /*
3471          * -EACCESS could mean that the user doesn't have correct permissions
3472          * to access the mount.  It could also mean that we tried to mount
3473          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3474          * existing mount programs don't handle -EACCES very well so it should
3475          * be mapped to -EPERM instead.
3476          */
3477         if (status == -EACCES)
3478                 status = -EPERM;
3479         return status;
3480 }
3481
3482 /**
3483  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3484  * @server: initialized nfs_server handle
3485  * @fhandle: we fill in the pseudo-fs root file handle
3486  * @info: we fill in an FSINFO struct
3487  * @auth_probe: probe the auth flavours
3488  *
3489  * Returns zero on success, or a negative errno.
3490  */
3491 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3492                          struct nfs_fsinfo *info,
3493                          bool auth_probe)
3494 {
3495         int status = 0;
3496
3497         if (!auth_probe)
3498                 status = nfs4_lookup_root(server, fhandle, info);
3499
3500         if (auth_probe || status == NFS4ERR_WRONGSEC)
3501                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3502                                 fhandle, info);
3503
3504         if (status == 0)
3505                 status = nfs4_server_capabilities(server, fhandle);
3506         if (status == 0)
3507                 status = nfs4_do_fsinfo(server, fhandle, info);
3508
3509         return nfs4_map_errors(status);
3510 }
3511
3512 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3513                               struct nfs_fsinfo *info)
3514 {
3515         int error;
3516         struct nfs_fattr *fattr = info->fattr;
3517         struct nfs4_label *label = NULL;
3518
3519         error = nfs4_server_capabilities(server, mntfh);
3520         if (error < 0) {
3521                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3522                 return error;
3523         }
3524
3525         label = nfs4_label_alloc(server, GFP_KERNEL);
3526         if (IS_ERR(label))
3527                 return PTR_ERR(label);
3528
3529         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3530         if (error < 0) {
3531                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3532                 goto err_free_label;
3533         }
3534
3535         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3536             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3537                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3538
3539 err_free_label:
3540         nfs4_label_free(label);
3541
3542         return error;
3543 }
3544
3545 /*
3546  * Get locations and (maybe) other attributes of a referral.
3547  * Note that we'll actually follow the referral later when
3548  * we detect fsid mismatch in inode revalidation
3549  */
3550 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3551                              const struct qstr *name, struct nfs_fattr *fattr,
3552                              struct nfs_fh *fhandle)
3553 {
3554         int status = -ENOMEM;
3555         struct page *page = NULL;
3556         struct nfs4_fs_locations *locations = NULL;
3557
3558         page = alloc_page(GFP_KERNEL);
3559         if (page == NULL)
3560                 goto out;
3561         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3562         if (locations == NULL)
3563                 goto out;
3564
3565         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3566         if (status != 0)
3567                 goto out;
3568
3569         /*
3570          * If the fsid didn't change, this is a migration event, not a
3571          * referral.  Cause us to drop into the exception handler, which
3572          * will kick off migration recovery.
3573          */
3574         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3575                 dprintk("%s: server did not return a different fsid for"
3576                         " a referral at %s\n", __func__, name->name);
3577                 status = -NFS4ERR_MOVED;
3578                 goto out;
3579         }
3580         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3581         nfs_fixup_referral_attributes(&locations->fattr);
3582
3583         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3584         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3585         memset(fhandle, 0, sizeof(struct nfs_fh));
3586 out:
3587         if (page)
3588                 __free_page(page);
3589         kfree(locations);
3590         return status;
3591 }
3592
3593 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3594                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3595 {
3596         struct nfs4_getattr_arg args = {
3597                 .fh = fhandle,
3598                 .bitmask = server->attr_bitmask,
3599         };
3600         struct nfs4_getattr_res res = {
3601                 .fattr = fattr,
3602                 .label = label,
3603                 .server = server,
3604         };
3605         struct rpc_message msg = {
3606                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3607                 .rpc_argp = &args,
3608                 .rpc_resp = &res,
3609         };
3610
3611         args.bitmask = nfs4_bitmask(server, label);
3612
3613         nfs_fattr_init(fattr);
3614         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3615 }
3616
3617 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3618                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3619 {
3620         struct nfs4_exception exception = { };
3621         int err;
3622         do {
3623                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3624                 trace_nfs4_getattr(server, fhandle, fattr, err);
3625                 err = nfs4_handle_exception(server, err,
3626                                 &exception);
3627         } while (exception.retry);
3628         return err;
3629 }
3630
3631 /*
3632  * The file is not closed if it is opened due to the a request to change
3633  * the size of the file. The open call will not be needed once the
3634  * VFS layer lookup-intents are implemented.
3635  *
3636  * Close is called when the inode is destroyed.
3637  * If we haven't opened the file for O_WRONLY, we
3638  * need to in the size_change case to obtain a stateid.
3639  *
3640  * Got race?
3641  * Because OPEN is always done by name in nfsv4, it is
3642  * possible that we opened a different file by the same
3643  * name.  We can recognize this race condition, but we
3644  * can't do anything about it besides returning an error.
3645  *
3646  * This will be fixed with VFS changes (lookup-intent).
3647  */
3648 static int
3649 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3650                   struct iattr *sattr)
3651 {
3652         struct inode *inode = d_inode(dentry);
3653         struct rpc_cred *cred = NULL;
3654         struct nfs_open_context *ctx = NULL;
3655         struct nfs4_label *label = NULL;
3656         int status;
3657
3658         if (pnfs_ld_layoutret_on_setattr(inode) &&
3659             sattr->ia_valid & ATTR_SIZE &&
3660             sattr->ia_size < i_size_read(inode))
3661                 pnfs_commit_and_return_layout(inode);
3662
3663         nfs_fattr_init(fattr);
3664
3665         /* Deal with open(O_TRUNC) */
3666         if (sattr->ia_valid & ATTR_OPEN)
3667                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3668
3669         /* Optimization: if the end result is no change, don't RPC */
3670         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3671                 return 0;
3672
3673         /* Search for an existing open(O_WRITE) file */
3674         if (sattr->ia_valid & ATTR_FILE) {
3675
3676                 ctx = nfs_file_open_context(sattr->ia_file);
3677                 if (ctx)
3678                         cred = ctx->cred;
3679         }
3680
3681         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3682         if (IS_ERR(label))
3683                 return PTR_ERR(label);
3684
3685         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3686         if (status == 0) {
3687                 nfs_setattr_update_inode(inode, sattr, fattr);
3688                 nfs_setsecurity(inode, fattr, label);
3689         }
3690         nfs4_label_free(label);
3691         return status;
3692 }
3693
3694 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3695                 const struct qstr *name, struct nfs_fh *fhandle,
3696                 struct nfs_fattr *fattr, struct nfs4_label *label)
3697 {
3698         struct nfs_server *server = NFS_SERVER(dir);
3699         int                    status;
3700         struct nfs4_lookup_arg args = {
3701                 .bitmask = server->attr_bitmask,
3702                 .dir_fh = NFS_FH(dir),
3703                 .name = name,
3704         };
3705         struct nfs4_lookup_res res = {
3706                 .server = server,
3707                 .fattr = fattr,
3708                 .label = label,
3709                 .fh = fhandle,
3710         };
3711         struct rpc_message msg = {
3712                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3713                 .rpc_argp = &args,
3714                 .rpc_resp = &res,
3715         };
3716
3717         args.bitmask = nfs4_bitmask(server, label);
3718
3719         nfs_fattr_init(fattr);
3720
3721         dprintk("NFS call  lookup %s\n", name->name);
3722         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3723         dprintk("NFS reply lookup: %d\n", status);
3724         return status;
3725 }
3726
3727 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3728 {
3729         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3730                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3731         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3732         fattr->nlink = 2;
3733 }
3734
3735 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3736                                    const struct qstr *name, struct nfs_fh *fhandle,
3737                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3738 {
3739         struct nfs4_exception exception = { };
3740         struct rpc_clnt *client = *clnt;
3741         int err;
3742         do {
3743                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3744                 trace_nfs4_lookup(dir, name, err);
3745                 switch (err) {
3746                 case -NFS4ERR_BADNAME:
3747                         err = -ENOENT;
3748                         goto out;
3749                 case -NFS4ERR_MOVED:
3750                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3751                         if (err == -NFS4ERR_MOVED)
3752                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3753                         goto out;
3754                 case -NFS4ERR_WRONGSEC:
3755                         err = -EPERM;
3756                         if (client != *clnt)
3757                                 goto out;
3758                         client = nfs4_negotiate_security(client, dir, name);
3759                         if (IS_ERR(client))
3760                                 return PTR_ERR(client);
3761
3762                         exception.retry = 1;
3763                         break;
3764                 default:
3765                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3766                 }
3767         } while (exception.retry);
3768
3769 out:
3770         if (err == 0)
3771                 *clnt = client;
3772         else if (client != *clnt)
3773                 rpc_shutdown_client(client);
3774
3775         return err;
3776 }
3777
3778 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3779                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3780                             struct nfs4_label *label)
3781 {
3782         int status;
3783         struct rpc_clnt *client = NFS_CLIENT(dir);
3784
3785         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3786         if (client != NFS_CLIENT(dir)) {
3787                 rpc_shutdown_client(client);
3788                 nfs_fixup_secinfo_attributes(fattr);
3789         }
3790         return status;
3791 }
3792
3793 struct rpc_clnt *
3794 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3795                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3796 {
3797         struct rpc_clnt *client = NFS_CLIENT(dir);
3798         int status;
3799
3800         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3801         if (status < 0)
3802                 return ERR_PTR(status);
3803         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3804 }
3805
3806 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3807 {
3808         struct nfs_server *server = NFS_SERVER(inode);
3809         struct nfs4_accessargs args = {
3810                 .fh = NFS_FH(inode),
3811                 .bitmask = server->cache_consistency_bitmask,
3812         };
3813         struct nfs4_accessres res = {
3814                 .server = server,
3815         };
3816         struct rpc_message msg = {
3817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3818                 .rpc_argp = &args,
3819                 .rpc_resp = &res,
3820                 .rpc_cred = entry->cred,
3821         };
3822         int mode = entry->mask;
3823         int status = 0;
3824
3825         /*
3826          * Determine which access bits we want to ask for...
3827          */
3828         if (mode & MAY_READ)
3829                 args.access |= NFS4_ACCESS_READ;
3830         if (S_ISDIR(inode->i_mode)) {
3831                 if (mode & MAY_WRITE)
3832                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3833                 if (mode & MAY_EXEC)
3834                         args.access |= NFS4_ACCESS_LOOKUP;
3835         } else {
3836                 if (mode & MAY_WRITE)
3837                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3838                 if (mode & MAY_EXEC)
3839                         args.access |= NFS4_ACCESS_EXECUTE;
3840         }
3841
3842         res.fattr = nfs_alloc_fattr();
3843         if (res.fattr == NULL)
3844                 return -ENOMEM;
3845
3846         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3847         if (!status) {
3848                 nfs_access_set_mask(entry, res.access);
3849                 nfs_refresh_inode(inode, res.fattr);
3850         }
3851         nfs_free_fattr(res.fattr);
3852         return status;
3853 }
3854
3855 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3856 {
3857         struct nfs4_exception exception = { };
3858         int err;
3859         do {
3860                 err = _nfs4_proc_access(inode, entry);
3861                 trace_nfs4_access(inode, err);
3862                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3863                                 &exception);
3864         } while (exception.retry);
3865         return err;
3866 }
3867
3868 /*
3869  * TODO: For the time being, we don't try to get any attributes
3870  * along with any of the zero-copy operations READ, READDIR,
3871  * READLINK, WRITE.
3872  *
3873  * In the case of the first three, we want to put the GETATTR
3874  * after the read-type operation -- this is because it is hard
3875  * to predict the length of a GETATTR response in v4, and thus
3876  * align the READ data correctly.  This means that the GETATTR
3877  * may end up partially falling into the page cache, and we should
3878  * shift it into the 'tail' of the xdr_buf before processing.
3879  * To do this efficiently, we need to know the total length
3880  * of data received, which doesn't seem to be available outside
3881  * of the RPC layer.
3882  *
3883  * In the case of WRITE, we also want to put the GETATTR after
3884  * the operation -- in this case because we want to make sure
3885  * we get the post-operation mtime and size.
3886  *
3887  * Both of these changes to the XDR layer would in fact be quite
3888  * minor, but I decided to leave them for a subsequent patch.
3889  */
3890 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3891                 unsigned int pgbase, unsigned int pglen)
3892 {
3893         struct nfs4_readlink args = {
3894                 .fh       = NFS_FH(inode),
3895                 .pgbase   = pgbase,
3896                 .pglen    = pglen,
3897                 .pages    = &page,
3898         };
3899         struct nfs4_readlink_res res;
3900         struct rpc_message msg = {
3901                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3902                 .rpc_argp = &args,
3903                 .rpc_resp = &res,
3904         };
3905
3906         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3907 }
3908
3909 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3910                 unsigned int pgbase, unsigned int pglen)
3911 {
3912         struct nfs4_exception exception = { };
3913         int err;
3914         do {
3915                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3916                 trace_nfs4_readlink(inode, err);
3917                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3918                                 &exception);
3919         } while (exception.retry);
3920         return err;
3921 }
3922
3923 /*
3924  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3925  */
3926 static int
3927 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3928                  int flags)
3929 {
3930         struct nfs_server *server = NFS_SERVER(dir);
3931         struct nfs4_label l, *ilabel = NULL;
3932         struct nfs_open_context *ctx;
3933         struct nfs4_state *state;
3934         int status = 0;
3935
3936         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
3937         if (IS_ERR(ctx))
3938                 return PTR_ERR(ctx);
3939
3940         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3941
3942         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
3943                 sattr->ia_mode &= ~current_umask();
3944         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3945         if (IS_ERR(state)) {
3946                 status = PTR_ERR(state);
3947                 goto out;
3948         }
3949 out:
3950         nfs4_label_release_security(ilabel);
3951         put_nfs_open_context(ctx);
3952         return status;
3953 }
3954
3955 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3956 {
3957         struct nfs_server *server = NFS_SERVER(dir);
3958         struct nfs_removeargs args = {
3959                 .fh = NFS_FH(dir),
3960                 .name = *name,
3961         };
3962         struct nfs_removeres res = {
3963                 .server = server,
3964         };
3965         struct rpc_message msg = {
3966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3967                 .rpc_argp = &args,
3968                 .rpc_resp = &res,
3969         };
3970         unsigned long timestamp = jiffies;
3971         int status;
3972
3973         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3974         if (status == 0)
3975                 update_changeattr(dir, &res.cinfo, timestamp);
3976         return status;
3977 }
3978
3979 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3980 {
3981         struct nfs4_exception exception = { };
3982         int err;
3983         do {
3984                 err = _nfs4_proc_remove(dir, name);
3985                 trace_nfs4_remove(dir, name, err);
3986                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3987                                 &exception);
3988         } while (exception.retry);
3989         return err;
3990 }
3991
3992 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3993 {
3994         struct nfs_server *server = NFS_SERVER(dir);
3995         struct nfs_removeargs *args = msg->rpc_argp;
3996         struct nfs_removeres *res = msg->rpc_resp;
3997
3998         res->server = server;
3999         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4000         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4001
4002         nfs_fattr_init(res->dir_attr);
4003 }
4004
4005 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4006 {
4007         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4008                         &data->args.seq_args,
4009                         &data->res.seq_res,
4010                         task);
4011 }
4012
4013 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4014 {
4015         struct nfs_unlinkdata *data = task->tk_calldata;
4016         struct nfs_removeres *res = &data->res;
4017
4018         if (!nfs4_sequence_done(task, &res->seq_res))
4019                 return 0;
4020         if (nfs4_async_handle_error(task, res->server, NULL,
4021                                     &data->timeout) == -EAGAIN)
4022                 return 0;
4023         if (task->tk_status == 0)
4024                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4025         return 1;
4026 }
4027
4028 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4029 {
4030         struct nfs_server *server = NFS_SERVER(dir);
4031         struct nfs_renameargs *arg = msg->rpc_argp;
4032         struct nfs_renameres *res = msg->rpc_resp;
4033
4034         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4035         res->server = server;
4036         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4037 }
4038
4039 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4040 {
4041         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4042                         &data->args.seq_args,
4043                         &data->res.seq_res,
4044                         task);
4045 }
4046
4047 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4048                                  struct inode *new_dir)
4049 {
4050         struct nfs_renamedata *data = task->tk_calldata;
4051         struct nfs_renameres *res = &data->res;
4052
4053         if (!nfs4_sequence_done(task, &res->seq_res))
4054                 return 0;
4055         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4056                 return 0;
4057
4058         if (task->tk_status == 0) {
4059                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4060                 if (new_dir != old_dir)
4061                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4062         }
4063         return 1;
4064 }
4065
4066 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4067 {
4068         struct nfs_server *server = NFS_SERVER(inode);
4069         struct nfs4_link_arg arg = {
4070                 .fh     = NFS_FH(inode),
4071                 .dir_fh = NFS_FH(dir),
4072                 .name   = name,
4073                 .bitmask = server->attr_bitmask,
4074         };
4075         struct nfs4_link_res res = {
4076                 .server = server,
4077                 .label = NULL,
4078         };
4079         struct rpc_message msg = {
4080                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4081                 .rpc_argp = &arg,
4082                 .rpc_resp = &res,
4083         };
4084         int status = -ENOMEM;
4085
4086         res.fattr = nfs_alloc_fattr();
4087         if (res.fattr == NULL)
4088                 goto out;
4089
4090         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4091         if (IS_ERR(res.label)) {
4092                 status = PTR_ERR(res.label);
4093                 goto out;
4094         }
4095         arg.bitmask = nfs4_bitmask(server, res.label);
4096
4097         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4098         if (!status) {
4099                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4100                 status = nfs_post_op_update_inode(inode, res.fattr);
4101                 if (!status)
4102                         nfs_setsecurity(inode, res.fattr, res.label);
4103         }
4104
4105
4106         nfs4_label_free(res.label);
4107
4108 out:
4109         nfs_free_fattr(res.fattr);
4110         return status;
4111 }
4112
4113 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4114 {
4115         struct nfs4_exception exception = { };
4116         int err;
4117         do {
4118                 err = nfs4_handle_exception(NFS_SERVER(inode),
4119                                 _nfs4_proc_link(inode, dir, name),
4120                                 &exception);
4121         } while (exception.retry);
4122         return err;
4123 }
4124
4125 struct nfs4_createdata {
4126         struct rpc_message msg;
4127         struct nfs4_create_arg arg;
4128         struct nfs4_create_res res;
4129         struct nfs_fh fh;
4130         struct nfs_fattr fattr;
4131         struct nfs4_label *label;
4132 };
4133
4134 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4135                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4136 {
4137         struct nfs4_createdata *data;
4138
4139         data = kzalloc(sizeof(*data), GFP_KERNEL);
4140         if (data != NULL) {
4141                 struct nfs_server *server = NFS_SERVER(dir);
4142
4143                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4144                 if (IS_ERR(data->label))
4145                         goto out_free;
4146
4147                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4148                 data->msg.rpc_argp = &data->arg;
4149                 data->msg.rpc_resp = &data->res;
4150                 data->arg.dir_fh = NFS_FH(dir);
4151                 data->arg.server = server;
4152                 data->arg.name = name;
4153                 data->arg.attrs = sattr;
4154                 data->arg.ftype = ftype;
4155                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4156                 data->arg.umask = current_umask();
4157                 data->res.server = server;
4158                 data->res.fh = &data->fh;
4159                 data->res.fattr = &data->fattr;
4160                 data->res.label = data->label;
4161                 nfs_fattr_init(data->res.fattr);
4162         }
4163         return data;
4164 out_free:
4165         kfree(data);
4166         return NULL;
4167 }
4168
4169 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4170 {
4171         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4172                                     &data->arg.seq_args, &data->res.seq_res, 1);
4173         if (status == 0) {
4174                 update_changeattr(dir, &data->res.dir_cinfo,
4175                                 data->res.fattr->time_start);
4176                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4177         }
4178         return status;
4179 }
4180
4181 static void nfs4_free_createdata(struct nfs4_createdata *data)
4182 {
4183         nfs4_label_free(data->label);
4184         kfree(data);
4185 }
4186
4187 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4188                 struct page *page, unsigned int len, struct iattr *sattr,
4189                 struct nfs4_label *label)
4190 {
4191         struct nfs4_createdata *data;
4192         int status = -ENAMETOOLONG;
4193
4194         if (len > NFS4_MAXPATHLEN)
4195                 goto out;
4196
4197         status = -ENOMEM;
4198         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4199         if (data == NULL)
4200                 goto out;
4201
4202         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4203         data->arg.u.symlink.pages = &page;
4204         data->arg.u.symlink.len = len;
4205         data->arg.label = label;
4206
4207         status = nfs4_do_create(dir, dentry, data);
4208
4209         nfs4_free_createdata(data);
4210 out:
4211         return status;
4212 }
4213
4214 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4215                 struct page *page, unsigned int len, struct iattr *sattr)
4216 {
4217         struct nfs4_exception exception = { };
4218         struct nfs4_label l, *label = NULL;
4219         int err;
4220
4221         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4222
4223         do {
4224                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4225                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4226                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4227                                 &exception);
4228         } while (exception.retry);
4229
4230         nfs4_label_release_security(label);
4231         return err;
4232 }
4233
4234 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4235                 struct iattr *sattr, struct nfs4_label *label)
4236 {
4237         struct nfs4_createdata *data;
4238         int status = -ENOMEM;
4239
4240         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4241         if (data == NULL)
4242                 goto out;
4243
4244         data->arg.label = label;
4245         status = nfs4_do_create(dir, dentry, data);
4246
4247         nfs4_free_createdata(data);
4248 out:
4249         return status;
4250 }
4251
4252 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4253                 struct iattr *sattr)
4254 {
4255         struct nfs_server *server = NFS_SERVER(dir);
4256         struct nfs4_exception exception = { };
4257         struct nfs4_label l, *label = NULL;
4258         int err;
4259
4260         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4261
4262         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4263                 sattr->ia_mode &= ~current_umask();
4264         do {
4265                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4266                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4267                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4268                                 &exception);
4269         } while (exception.retry);
4270         nfs4_label_release_security(label);
4271
4272         return err;
4273 }
4274
4275 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4276                 u64 cookie, struct page **pages, unsigned int count, int plus)
4277 {
4278         struct inode            *dir = d_inode(dentry);
4279         struct nfs4_readdir_arg args = {
4280                 .fh = NFS_FH(dir),
4281                 .pages = pages,
4282                 .pgbase = 0,
4283                 .count = count,
4284                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4285                 .plus = plus,
4286         };
4287         struct nfs4_readdir_res res;
4288         struct rpc_message msg = {
4289                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4290                 .rpc_argp = &args,
4291                 .rpc_resp = &res,
4292                 .rpc_cred = cred,
4293         };
4294         int                     status;
4295
4296         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4297                         dentry,
4298                         (unsigned long long)cookie);
4299         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4300         res.pgbase = args.pgbase;
4301         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4302         if (status >= 0) {
4303                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4304                 status += args.pgbase;
4305         }
4306
4307         nfs_invalidate_atime(dir);
4308
4309         dprintk("%s: returns %d\n", __func__, status);
4310         return status;
4311 }
4312
4313 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4314                 u64 cookie, struct page **pages, unsigned int count, int plus)
4315 {
4316         struct nfs4_exception exception = { };
4317         int err;
4318         do {
4319                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4320                                 pages, count, plus);
4321                 trace_nfs4_readdir(d_inode(dentry), err);
4322                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4323                                 &exception);
4324         } while (exception.retry);
4325         return err;
4326 }
4327
4328 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4329                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4330 {
4331         struct nfs4_createdata *data;
4332         int mode = sattr->ia_mode;
4333         int status = -ENOMEM;
4334
4335         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4336         if (data == NULL)
4337                 goto out;
4338
4339         if (S_ISFIFO(mode))
4340                 data->arg.ftype = NF4FIFO;
4341         else if (S_ISBLK(mode)) {
4342                 data->arg.ftype = NF4BLK;
4343                 data->arg.u.device.specdata1 = MAJOR(rdev);
4344                 data->arg.u.device.specdata2 = MINOR(rdev);
4345         }
4346         else if (S_ISCHR(mode)) {
4347                 data->arg.ftype = NF4CHR;
4348                 data->arg.u.device.specdata1 = MAJOR(rdev);
4349                 data->arg.u.device.specdata2 = MINOR(rdev);
4350         } else if (!S_ISSOCK(mode)) {
4351                 status = -EINVAL;
4352                 goto out_free;
4353         }
4354
4355         data->arg.label = label;
4356         status = nfs4_do_create(dir, dentry, data);
4357 out_free:
4358         nfs4_free_createdata(data);
4359 out:
4360         return status;
4361 }
4362
4363 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4364                 struct iattr *sattr, dev_t rdev)
4365 {
4366         struct nfs_server *server = NFS_SERVER(dir);
4367         struct nfs4_exception exception = { };
4368         struct nfs4_label l, *label = NULL;
4369         int err;
4370
4371         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4372
4373         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4374                 sattr->ia_mode &= ~current_umask();
4375         do {
4376                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4377                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4378                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4379                                 &exception);
4380         } while (exception.retry);
4381
4382         nfs4_label_release_security(label);
4383
4384         return err;
4385 }
4386
4387 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4388                  struct nfs_fsstat *fsstat)
4389 {
4390         struct nfs4_statfs_arg args = {
4391                 .fh = fhandle,
4392                 .bitmask = server->attr_bitmask,
4393         };
4394         struct nfs4_statfs_res res = {
4395                 .fsstat = fsstat,
4396         };
4397         struct rpc_message msg = {
4398                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4399                 .rpc_argp = &args,
4400                 .rpc_resp = &res,
4401         };
4402
4403         nfs_fattr_init(fsstat->fattr);
4404         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4405 }
4406
4407 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4408 {
4409         struct nfs4_exception exception = { };
4410         int err;
4411         do {
4412                 err = nfs4_handle_exception(server,
4413                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4414                                 &exception);
4415         } while (exception.retry);
4416         return err;
4417 }
4418
4419 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4420                 struct nfs_fsinfo *fsinfo)
4421 {
4422         struct nfs4_fsinfo_arg args = {
4423                 .fh = fhandle,
4424                 .bitmask = server->attr_bitmask,
4425         };
4426         struct nfs4_fsinfo_res res = {
4427                 .fsinfo = fsinfo,
4428         };
4429         struct rpc_message msg = {
4430                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4431                 .rpc_argp = &args,
4432                 .rpc_resp = &res,
4433         };
4434
4435         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4436 }
4437
4438 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4439 {
4440         struct nfs4_exception exception = { };
4441         unsigned long now = jiffies;
4442         int err;
4443
4444         do {
4445                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4446                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4447                 if (err == 0) {
4448                         nfs4_set_lease_period(server->nfs_client,
4449                                         fsinfo->lease_time * HZ,
4450                                         now);
4451                         break;
4452                 }
4453                 err = nfs4_handle_exception(server, err, &exception);
4454         } while (exception.retry);
4455         return err;
4456 }
4457
4458 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4459 {
4460         int error;
4461
4462         nfs_fattr_init(fsinfo->fattr);
4463         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4464         if (error == 0) {
4465                 /* block layout checks this! */
4466                 server->pnfs_blksize = fsinfo->blksize;
4467                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4468         }
4469
4470         return error;
4471 }
4472
4473 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4474                 struct nfs_pathconf *pathconf)
4475 {
4476         struct nfs4_pathconf_arg args = {
4477                 .fh = fhandle,
4478                 .bitmask = server->attr_bitmask,
4479         };
4480         struct nfs4_pathconf_res res = {
4481                 .pathconf = pathconf,
4482         };
4483         struct rpc_message msg = {
4484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4485                 .rpc_argp = &args,
4486                 .rpc_resp = &res,
4487         };
4488
4489         /* None of the pathconf attributes are mandatory to implement */
4490         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4491                 memset(pathconf, 0, sizeof(*pathconf));
4492                 return 0;
4493         }
4494
4495         nfs_fattr_init(pathconf->fattr);
4496         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4497 }
4498
4499 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4500                 struct nfs_pathconf *pathconf)
4501 {
4502         struct nfs4_exception exception = { };
4503         int err;
4504
4505         do {
4506                 err = nfs4_handle_exception(server,
4507                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4508                                 &exception);
4509         } while (exception.retry);
4510         return err;
4511 }
4512
4513 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4514                 const struct nfs_open_context *ctx,
4515                 const struct nfs_lock_context *l_ctx,
4516                 fmode_t fmode)
4517 {
4518         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4519 }
4520 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4521
4522 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4523                 const struct nfs_open_context *ctx,
4524                 const struct nfs_lock_context *l_ctx,
4525                 fmode_t fmode)
4526 {
4527         nfs4_stateid current_stateid;
4528
4529         /* If the current stateid represents a lost lock, then exit */
4530         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4531                 return true;
4532         return nfs4_stateid_match(stateid, &current_stateid);
4533 }
4534
4535 static bool nfs4_error_stateid_expired(int err)
4536 {
4537         switch (err) {
4538         case -NFS4ERR_DELEG_REVOKED:
4539         case -NFS4ERR_ADMIN_REVOKED:
4540         case -NFS4ERR_BAD_STATEID:
4541         case -NFS4ERR_STALE_STATEID:
4542         case -NFS4ERR_OLD_STATEID:
4543         case -NFS4ERR_OPENMODE:
4544         case -NFS4ERR_EXPIRED:
4545                 return true;
4546         }
4547         return false;
4548 }
4549
4550 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4551 {
4552         struct nfs_server *server = NFS_SERVER(hdr->inode);
4553
4554         trace_nfs4_read(hdr, task->tk_status);
4555         if (task->tk_status < 0) {
4556                 struct nfs4_exception exception = {
4557                         .inode = hdr->inode,
4558                         .state = hdr->args.context->state,
4559                         .stateid = &hdr->args.stateid,
4560                 };
4561                 task->tk_status = nfs4_async_handle_exception(task,
4562                                 server, task->tk_status, &exception);
4563                 if (exception.retry) {
4564                         rpc_restart_call_prepare(task);
4565                         return -EAGAIN;
4566                 }
4567         }
4568
4569         if (task->tk_status > 0)
4570                 renew_lease(server, hdr->timestamp);
4571         return 0;
4572 }
4573
4574 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4575                 struct nfs_pgio_args *args)
4576 {
4577
4578         if (!nfs4_error_stateid_expired(task->tk_status) ||
4579                 nfs4_stateid_is_current(&args->stateid,
4580                                 args->context,
4581                                 args->lock_context,
4582                                 FMODE_READ))
4583                 return false;
4584         rpc_restart_call_prepare(task);
4585         return true;
4586 }
4587
4588 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4589 {
4590
4591         dprintk("--> %s\n", __func__);
4592
4593         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4594                 return -EAGAIN;
4595         if (nfs4_read_stateid_changed(task, &hdr->args))
4596                 return -EAGAIN;
4597         if (task->tk_status > 0)
4598                 nfs_invalidate_atime(hdr->inode);
4599         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4600                                     nfs4_read_done_cb(task, hdr);
4601 }
4602
4603 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4604                                  struct rpc_message *msg)
4605 {
4606         hdr->timestamp   = jiffies;
4607         if (!hdr->pgio_done_cb)
4608                 hdr->pgio_done_cb = nfs4_read_done_cb;
4609         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4610         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4611 }
4612
4613 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4614                                       struct nfs_pgio_header *hdr)
4615 {
4616         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4617                         &hdr->args.seq_args,
4618                         &hdr->res.seq_res,
4619                         task))
4620                 return 0;
4621         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4622                                 hdr->args.lock_context,
4623                                 hdr->rw_mode) == -EIO)
4624                 return -EIO;
4625         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4626                 return -EIO;
4627         return 0;
4628 }
4629
4630 static int nfs4_write_done_cb(struct rpc_task *task,
4631                               struct nfs_pgio_header *hdr)
4632 {
4633         struct inode *inode = hdr->inode;
4634
4635         trace_nfs4_write(hdr, task->tk_status);
4636         if (task->tk_status < 0) {
4637                 struct nfs4_exception exception = {
4638                         .inode = hdr->inode,
4639                         .state = hdr->args.context->state,
4640                         .stateid = &hdr->args.stateid,
4641                 };
4642                 task->tk_status = nfs4_async_handle_exception(task,
4643                                 NFS_SERVER(inode), task->tk_status,
4644                                 &exception);
4645                 if (exception.retry) {
4646                         rpc_restart_call_prepare(task);
4647                         return -EAGAIN;
4648                 }
4649         }
4650         if (task->tk_status >= 0) {
4651                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4652                 nfs_writeback_update_inode(hdr);
4653         }
4654         return 0;
4655 }
4656
4657 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4658                 struct nfs_pgio_args *args)
4659 {
4660
4661         if (!nfs4_error_stateid_expired(task->tk_status) ||
4662                 nfs4_stateid_is_current(&args->stateid,
4663                                 args->context,
4664                                 args->lock_context,
4665                                 FMODE_WRITE))
4666                 return false;
4667         rpc_restart_call_prepare(task);
4668         return true;
4669 }
4670
4671 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4672 {
4673         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4674                 return -EAGAIN;
4675         if (nfs4_write_stateid_changed(task, &hdr->args))
4676                 return -EAGAIN;
4677         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4678                 nfs4_write_done_cb(task, hdr);
4679 }
4680
4681 static
4682 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4683 {
4684         /* Don't request attributes for pNFS or O_DIRECT writes */
4685         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4686                 return false;
4687         /* Otherwise, request attributes if and only if we don't hold
4688          * a delegation
4689          */
4690         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4691 }
4692
4693 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4694                                   struct rpc_message *msg)
4695 {
4696         struct nfs_server *server = NFS_SERVER(hdr->inode);
4697
4698         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4699                 hdr->args.bitmask = NULL;
4700                 hdr->res.fattr = NULL;
4701         } else
4702                 hdr->args.bitmask = server->cache_consistency_bitmask;
4703
4704         if (!hdr->pgio_done_cb)
4705                 hdr->pgio_done_cb = nfs4_write_done_cb;
4706         hdr->res.server = server;
4707         hdr->timestamp   = jiffies;
4708
4709         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4710         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4711 }
4712
4713 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4714 {
4715         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
4716                         &data->args.seq_args,
4717                         &data->res.seq_res,
4718                         task);
4719 }
4720
4721 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4722 {
4723         struct inode *inode = data->inode;
4724
4725         trace_nfs4_commit(data, task->tk_status);
4726         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4727                                     NULL, NULL) == -EAGAIN) {
4728                 rpc_restart_call_prepare(task);
4729                 return -EAGAIN;
4730         }
4731         return 0;
4732 }
4733
4734 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4735 {
4736         if (!nfs4_sequence_done(task, &data->res.seq_res))
4737                 return -EAGAIN;
4738         return data->commit_done_cb(task, data);
4739 }
4740
4741 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4742 {
4743         struct nfs_server *server = NFS_SERVER(data->inode);
4744
4745         if (data->commit_done_cb == NULL)
4746                 data->commit_done_cb = nfs4_commit_done_cb;
4747         data->res.server = server;
4748         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4749         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4750 }
4751
4752 struct nfs4_renewdata {
4753         struct nfs_client       *client;
4754         unsigned long           timestamp;
4755 };
4756
4757 /*
4758  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4759  * standalone procedure for queueing an asynchronous RENEW.
4760  */
4761 static void nfs4_renew_release(void *calldata)
4762 {
4763         struct nfs4_renewdata *data = calldata;
4764         struct nfs_client *clp = data->client;
4765
4766         if (atomic_read(&clp->cl_count) > 1)
4767                 nfs4_schedule_state_renewal(clp);
4768         nfs_put_client(clp);
4769         kfree(data);
4770 }
4771
4772 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4773 {
4774         struct nfs4_renewdata *data = calldata;
4775         struct nfs_client *clp = data->client;
4776         unsigned long timestamp = data->timestamp;
4777
4778         trace_nfs4_renew_async(clp, task->tk_status);
4779         switch (task->tk_status) {
4780         case 0:
4781                 break;
4782         case -NFS4ERR_LEASE_MOVED:
4783                 nfs4_schedule_lease_moved_recovery(clp);
4784                 break;
4785         default:
4786                 /* Unless we're shutting down, schedule state recovery! */
4787                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4788                         return;
4789                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4790                         nfs4_schedule_lease_recovery(clp);
4791                         return;
4792                 }
4793                 nfs4_schedule_path_down_recovery(clp);
4794         }
4795         do_renew_lease(clp, timestamp);
4796 }
4797
4798 static const struct rpc_call_ops nfs4_renew_ops = {
4799         .rpc_call_done = nfs4_renew_done,
4800         .rpc_release = nfs4_renew_release,
4801 };
4802
4803 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4804 {
4805         struct rpc_message msg = {
4806                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4807                 .rpc_argp       = clp,
4808                 .rpc_cred       = cred,
4809         };
4810         struct nfs4_renewdata *data;
4811
4812         if (renew_flags == 0)
4813                 return 0;
4814         if (!atomic_inc_not_zero(&clp->cl_count))
4815                 return -EIO;
4816         data = kmalloc(sizeof(*data), GFP_NOFS);
4817         if (data == NULL) {
4818                 nfs_put_client(clp);
4819                 return -ENOMEM;
4820         }
4821         data->client = clp;
4822         data->timestamp = jiffies;
4823         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4824                         &nfs4_renew_ops, data);
4825 }
4826
4827 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4828 {
4829         struct rpc_message msg = {
4830                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4831                 .rpc_argp       = clp,
4832                 .rpc_cred       = cred,
4833         };
4834         unsigned long now = jiffies;
4835         int status;
4836
4837         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4838         if (status < 0)
4839                 return status;
4840         do_renew_lease(clp, now);
4841         return 0;
4842 }
4843
4844 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4845 {
4846         return server->caps & NFS_CAP_ACLS;
4847 }
4848
4849 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4850  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4851  * the stack.
4852  */
4853 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4854
4855 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4856                 struct page **pages)
4857 {
4858         struct page *newpage, **spages;
4859         int rc = 0;
4860         size_t len;
4861         spages = pages;
4862
4863         do {
4864                 len = min_t(size_t, PAGE_SIZE, buflen);
4865                 newpage = alloc_page(GFP_KERNEL);
4866
4867                 if (newpage == NULL)
4868                         goto unwind;
4869                 memcpy(page_address(newpage), buf, len);
4870                 buf += len;
4871                 buflen -= len;
4872                 *pages++ = newpage;
4873                 rc++;
4874         } while (buflen != 0);
4875
4876         return rc;
4877
4878 unwind:
4879         for(; rc > 0; rc--)
4880                 __free_page(spages[rc-1]);
4881         return -ENOMEM;
4882 }
4883
4884 struct nfs4_cached_acl {
4885         int cached;
4886         size_t len;
4887         char data[0];
4888 };
4889
4890 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4891 {
4892         struct nfs_inode *nfsi = NFS_I(inode);
4893
4894         spin_lock(&inode->i_lock);
4895         kfree(nfsi->nfs4_acl);
4896         nfsi->nfs4_acl = acl;
4897         spin_unlock(&inode->i_lock);
4898 }
4899
4900 static void nfs4_zap_acl_attr(struct inode *inode)
4901 {
4902         nfs4_set_cached_acl(inode, NULL);
4903 }
4904
4905 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4906 {
4907         struct nfs_inode *nfsi = NFS_I(inode);
4908         struct nfs4_cached_acl *acl;
4909         int ret = -ENOENT;
4910
4911         spin_lock(&inode->i_lock);
4912         acl = nfsi->nfs4_acl;
4913         if (acl == NULL)
4914                 goto out;
4915         if (buf == NULL) /* user is just asking for length */
4916                 goto out_len;
4917         if (acl->cached == 0)
4918                 goto out;
4919         ret = -ERANGE; /* see getxattr(2) man page */
4920         if (acl->len > buflen)
4921                 goto out;
4922         memcpy(buf, acl->data, acl->len);
4923 out_len:
4924         ret = acl->len;
4925 out:
4926         spin_unlock(&inode->i_lock);
4927         return ret;
4928 }
4929
4930 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4931 {
4932         struct nfs4_cached_acl *acl;
4933         size_t buflen = sizeof(*acl) + acl_len;
4934
4935         if (buflen <= PAGE_SIZE) {
4936                 acl = kmalloc(buflen, GFP_KERNEL);
4937                 if (acl == NULL)
4938                         goto out;
4939                 acl->cached = 1;
4940                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4941         } else {
4942                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4943                 if (acl == NULL)
4944                         goto out;
4945                 acl->cached = 0;
4946         }
4947         acl->len = acl_len;
4948 out:
4949         nfs4_set_cached_acl(inode, acl);
4950 }
4951
4952 /*
4953  * The getxattr API returns the required buffer length when called with a
4954  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4955  * the required buf.  On a NULL buf, we send a page of data to the server
4956  * guessing that the ACL request can be serviced by a page. If so, we cache
4957  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4958  * the cache. If not so, we throw away the page, and cache the required
4959  * length. The next getxattr call will then produce another round trip to
4960  * the server, this time with the input buf of the required size.
4961  */
4962 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4963 {
4964         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
4965         struct nfs_getaclargs args = {
4966                 .fh = NFS_FH(inode),
4967                 .acl_pages = pages,
4968                 .acl_len = buflen,
4969         };
4970         struct nfs_getaclres res = {
4971                 .acl_len = buflen,
4972         };
4973         struct rpc_message msg = {
4974                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4975                 .rpc_argp = &args,
4976                 .rpc_resp = &res,
4977         };
4978         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
4979         int ret = -ENOMEM, i;
4980
4981         if (npages > ARRAY_SIZE(pages))
4982                 return -ERANGE;
4983
4984         for (i = 0; i < npages; i++) {
4985                 pages[i] = alloc_page(GFP_KERNEL);
4986                 if (!pages[i])
4987                         goto out_free;
4988         }
4989
4990         /* for decoding across pages */
4991         res.acl_scratch = alloc_page(GFP_KERNEL);
4992         if (!res.acl_scratch)
4993                 goto out_free;
4994
4995         args.acl_len = npages * PAGE_SIZE;
4996
4997         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4998                 __func__, buf, buflen, npages, args.acl_len);
4999         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5000                              &msg, &args.seq_args, &res.seq_res, 0);
5001         if (ret)
5002                 goto out_free;
5003
5004         /* Handle the case where the passed-in buffer is too short */
5005         if (res.acl_flags & NFS4_ACL_TRUNC) {
5006                 /* Did the user only issue a request for the acl length? */
5007                 if (buf == NULL)
5008                         goto out_ok;
5009                 ret = -ERANGE;
5010                 goto out_free;
5011         }
5012         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5013         if (buf) {
5014                 if (res.acl_len > buflen) {
5015                         ret = -ERANGE;
5016                         goto out_free;
5017                 }
5018                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5019         }
5020 out_ok:
5021         ret = res.acl_len;
5022 out_free:
5023         for (i = 0; i < npages; i++)
5024                 if (pages[i])
5025                         __free_page(pages[i]);
5026         if (res.acl_scratch)
5027                 __free_page(res.acl_scratch);
5028         return ret;
5029 }
5030
5031 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5032 {
5033         struct nfs4_exception exception = { };
5034         ssize_t ret;
5035         do {
5036                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5037                 trace_nfs4_get_acl(inode, ret);
5038                 if (ret >= 0)
5039                         break;
5040                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5041         } while (exception.retry);
5042         return ret;
5043 }
5044
5045 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5046 {
5047         struct nfs_server *server = NFS_SERVER(inode);
5048         int ret;
5049
5050         if (!nfs4_server_supports_acls(server))
5051                 return -EOPNOTSUPP;
5052         ret = nfs_revalidate_inode(server, inode);
5053         if (ret < 0)
5054                 return ret;
5055         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5056                 nfs_zap_acl_cache(inode);
5057         ret = nfs4_read_cached_acl(inode, buf, buflen);
5058         if (ret != -ENOENT)
5059                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5060                  * but no cached acl data, just the acl length */
5061                 return ret;
5062         return nfs4_get_acl_uncached(inode, buf, buflen);
5063 }
5064
5065 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5066 {
5067         struct nfs_server *server = NFS_SERVER(inode);
5068         struct page *pages[NFS4ACL_MAXPAGES];
5069         struct nfs_setaclargs arg = {
5070                 .fh             = NFS_FH(inode),
5071                 .acl_pages      = pages,
5072                 .acl_len        = buflen,
5073         };
5074         struct nfs_setaclres res;
5075         struct rpc_message msg = {
5076                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5077                 .rpc_argp       = &arg,
5078                 .rpc_resp       = &res,
5079         };
5080         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5081         int ret, i;
5082
5083         if (!nfs4_server_supports_acls(server))
5084                 return -EOPNOTSUPP;
5085         if (npages > ARRAY_SIZE(pages))
5086                 return -ERANGE;
5087         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5088         if (i < 0)
5089                 return i;
5090         nfs4_inode_return_delegation(inode);
5091         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5092
5093         /*
5094          * Free each page after tx, so the only ref left is
5095          * held by the network stack
5096          */
5097         for (; i > 0; i--)
5098                 put_page(pages[i-1]);
5099
5100         /*
5101          * Acl update can result in inode attribute update.
5102          * so mark the attribute cache invalid.
5103          */
5104         spin_lock(&inode->i_lock);
5105         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5106         spin_unlock(&inode->i_lock);
5107         nfs_access_zap_cache(inode);
5108         nfs_zap_acl_cache(inode);
5109         return ret;
5110 }
5111
5112 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5113 {
5114         struct nfs4_exception exception = { };
5115         int err;
5116         do {
5117                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5118                 trace_nfs4_set_acl(inode, err);
5119                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5120                                 &exception);
5121         } while (exception.retry);
5122         return err;
5123 }
5124
5125 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5126 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5127                                         size_t buflen)
5128 {
5129         struct nfs_server *server = NFS_SERVER(inode);
5130         struct nfs_fattr fattr;
5131         struct nfs4_label label = {0, 0, buflen, buf};
5132
5133         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5134         struct nfs4_getattr_arg arg = {
5135                 .fh             = NFS_FH(inode),
5136                 .bitmask        = bitmask,
5137         };
5138         struct nfs4_getattr_res res = {
5139                 .fattr          = &fattr,
5140                 .label          = &label,
5141                 .server         = server,
5142         };
5143         struct rpc_message msg = {
5144                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5145                 .rpc_argp       = &arg,
5146                 .rpc_resp       = &res,
5147         };
5148         int ret;
5149
5150         nfs_fattr_init(&fattr);
5151
5152         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5153         if (ret)
5154                 return ret;
5155         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5156                 return -ENOENT;
5157         if (buflen < label.len)
5158                 return -ERANGE;
5159         return 0;
5160 }
5161
5162 static int nfs4_get_security_label(struct inode *inode, void *buf,
5163                                         size_t buflen)
5164 {
5165         struct nfs4_exception exception = { };
5166         int err;
5167
5168         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5169                 return -EOPNOTSUPP;
5170
5171         do {
5172                 err = _nfs4_get_security_label(inode, buf, buflen);
5173                 trace_nfs4_get_security_label(inode, err);
5174                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5175                                 &exception);
5176         } while (exception.retry);
5177         return err;
5178 }
5179
5180 static int _nfs4_do_set_security_label(struct inode *inode,
5181                 struct nfs4_label *ilabel,
5182                 struct nfs_fattr *fattr,
5183                 struct nfs4_label *olabel)
5184 {
5185
5186         struct iattr sattr = {0};
5187         struct nfs_server *server = NFS_SERVER(inode);
5188         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5189         struct nfs_setattrargs arg = {
5190                 .fh             = NFS_FH(inode),
5191                 .iap            = &sattr,
5192                 .server         = server,
5193                 .bitmask        = bitmask,
5194                 .label          = ilabel,
5195         };
5196         struct nfs_setattrres res = {
5197                 .fattr          = fattr,
5198                 .label          = olabel,
5199                 .server         = server,
5200         };
5201         struct rpc_message msg = {
5202                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5203                 .rpc_argp       = &arg,
5204                 .rpc_resp       = &res,
5205         };
5206         int status;
5207
5208         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5209
5210         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5211         if (status)
5212                 dprintk("%s failed: %d\n", __func__, status);
5213
5214         return status;
5215 }
5216
5217 static int nfs4_do_set_security_label(struct inode *inode,
5218                 struct nfs4_label *ilabel,
5219                 struct nfs_fattr *fattr,
5220                 struct nfs4_label *olabel)
5221 {
5222         struct nfs4_exception exception = { };
5223         int err;
5224
5225         do {
5226                 err = _nfs4_do_set_security_label(inode, ilabel,
5227                                 fattr, olabel);
5228                 trace_nfs4_set_security_label(inode, err);
5229                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5230                                 &exception);
5231         } while (exception.retry);
5232         return err;
5233 }
5234
5235 static int
5236 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5237 {
5238         struct nfs4_label ilabel, *olabel = NULL;
5239         struct nfs_fattr fattr;
5240         struct rpc_cred *cred;
5241         int status;
5242
5243         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5244                 return -EOPNOTSUPP;
5245
5246         nfs_fattr_init(&fattr);
5247
5248         ilabel.pi = 0;
5249         ilabel.lfs = 0;
5250         ilabel.label = (char *)buf;
5251         ilabel.len = buflen;
5252
5253         cred = rpc_lookup_cred();
5254         if (IS_ERR(cred))
5255                 return PTR_ERR(cred);
5256
5257         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5258         if (IS_ERR(olabel)) {
5259                 status = -PTR_ERR(olabel);
5260                 goto out;
5261         }
5262
5263         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5264         if (status == 0)
5265                 nfs_setsecurity(inode, &fattr, olabel);
5266
5267         nfs4_label_free(olabel);
5268 out:
5269         put_rpccred(cred);
5270         return status;
5271 }
5272 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5273
5274
5275 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5276                                     nfs4_verifier *bootverf)
5277 {
5278         __be32 verf[2];
5279
5280         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5281                 /* An impossible timestamp guarantees this value
5282                  * will never match a generated boot time. */
5283                 verf[0] = cpu_to_be32(U32_MAX);
5284                 verf[1] = cpu_to_be32(U32_MAX);
5285         } else {
5286                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5287                 u64 ns = ktime_to_ns(nn->boot_time);
5288
5289                 verf[0] = cpu_to_be32(ns >> 32);
5290                 verf[1] = cpu_to_be32(ns);
5291         }
5292         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5293 }
5294
5295 static int
5296 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5297 {
5298         size_t len;
5299         char *str;
5300
5301         if (clp->cl_owner_id != NULL)
5302                 return 0;
5303
5304         rcu_read_lock();
5305         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5306                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5307                 1 +
5308                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5309                 1;
5310         rcu_read_unlock();
5311
5312         if (len > NFS4_OPAQUE_LIMIT + 1)
5313                 return -EINVAL;
5314
5315         /*
5316          * Since this string is allocated at mount time, and held until the
5317          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5318          * about a memory-reclaim deadlock.
5319          */
5320         str = kmalloc(len, GFP_KERNEL);
5321         if (!str)
5322                 return -ENOMEM;
5323
5324         rcu_read_lock();
5325         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5326                         clp->cl_ipaddr,
5327                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5328                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5329         rcu_read_unlock();
5330
5331         clp->cl_owner_id = str;
5332         return 0;
5333 }
5334
5335 static int
5336 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5337 {
5338         size_t len;
5339         char *str;
5340
5341         len = 10 + 10 + 1 + 10 + 1 +
5342                 strlen(nfs4_client_id_uniquifier) + 1 +
5343                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5344
5345         if (len > NFS4_OPAQUE_LIMIT + 1)
5346                 return -EINVAL;
5347
5348         /*
5349          * Since this string is allocated at mount time, and held until the
5350          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5351          * about a memory-reclaim deadlock.
5352          */
5353         str = kmalloc(len, GFP_KERNEL);
5354         if (!str)
5355                 return -ENOMEM;
5356
5357         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5358                         clp->rpc_ops->version, clp->cl_minorversion,
5359                         nfs4_client_id_uniquifier,
5360                         clp->cl_rpcclient->cl_nodename);
5361         clp->cl_owner_id = str;
5362         return 0;
5363 }
5364
5365 static int
5366 nfs4_init_uniform_client_string(struct nfs_client *clp)
5367 {
5368         size_t len;
5369         char *str;
5370
5371         if (clp->cl_owner_id != NULL)
5372                 return 0;
5373
5374         if (nfs4_client_id_uniquifier[0] != '\0')
5375                 return nfs4_init_uniquifier_client_string(clp);
5376
5377         len = 10 + 10 + 1 + 10 + 1 +
5378                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5379
5380         if (len > NFS4_OPAQUE_LIMIT + 1)
5381                 return -EINVAL;
5382
5383         /*
5384          * Since this string is allocated at mount time, and held until the
5385          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5386          * about a memory-reclaim deadlock.
5387          */
5388         str = kmalloc(len, GFP_KERNEL);
5389         if (!str)
5390                 return -ENOMEM;
5391
5392         scnprintf(str, len, "Linux NFSv%u.%u %s",
5393                         clp->rpc_ops->version, clp->cl_minorversion,
5394                         clp->cl_rpcclient->cl_nodename);
5395         clp->cl_owner_id = str;
5396         return 0;
5397 }
5398
5399 /*
5400  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5401  * services.  Advertise one based on the address family of the
5402  * clientaddr.
5403  */
5404 static unsigned int
5405 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5406 {
5407         if (strchr(clp->cl_ipaddr, ':') != NULL)
5408                 return scnprintf(buf, len, "tcp6");
5409         else
5410                 return scnprintf(buf, len, "tcp");
5411 }
5412
5413 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5414 {
5415         struct nfs4_setclientid *sc = calldata;
5416
5417         if (task->tk_status == 0)
5418                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5419 }
5420
5421 static const struct rpc_call_ops nfs4_setclientid_ops = {
5422         .rpc_call_done = nfs4_setclientid_done,
5423 };
5424
5425 /**
5426  * nfs4_proc_setclientid - Negotiate client ID
5427  * @clp: state data structure
5428  * @program: RPC program for NFSv4 callback service
5429  * @port: IP port number for NFS4 callback service
5430  * @cred: RPC credential to use for this call
5431  * @res: where to place the result
5432  *
5433  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5434  */
5435 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5436                 unsigned short port, struct rpc_cred *cred,
5437                 struct nfs4_setclientid_res *res)
5438 {
5439         nfs4_verifier sc_verifier;
5440         struct nfs4_setclientid setclientid = {
5441                 .sc_verifier = &sc_verifier,
5442                 .sc_prog = program,
5443                 .sc_clnt = clp,
5444         };
5445         struct rpc_message msg = {
5446                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5447                 .rpc_argp = &setclientid,
5448                 .rpc_resp = res,
5449                 .rpc_cred = cred,
5450         };
5451         struct rpc_task *task;
5452         struct rpc_task_setup task_setup_data = {
5453                 .rpc_client = clp->cl_rpcclient,
5454                 .rpc_message = &msg,
5455                 .callback_ops = &nfs4_setclientid_ops,
5456                 .callback_data = &setclientid,
5457                 .flags = RPC_TASK_TIMEOUT,
5458         };
5459         int status;
5460
5461         /* nfs_client_id4 */
5462         nfs4_init_boot_verifier(clp, &sc_verifier);
5463
5464         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5465                 status = nfs4_init_uniform_client_string(clp);
5466         else
5467                 status = nfs4_init_nonuniform_client_string(clp);
5468
5469         if (status)
5470                 goto out;
5471
5472         /* cb_client4 */
5473         setclientid.sc_netid_len =
5474                                 nfs4_init_callback_netid(clp,
5475                                                 setclientid.sc_netid,
5476                                                 sizeof(setclientid.sc_netid));
5477         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5478                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5479                                 clp->cl_ipaddr, port >> 8, port & 255);
5480
5481         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5482                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5483                 clp->cl_owner_id);
5484         task = rpc_run_task(&task_setup_data);
5485         if (IS_ERR(task)) {
5486                 status = PTR_ERR(task);
5487                 goto out;
5488         }
5489         status = task->tk_status;
5490         if (setclientid.sc_cred) {
5491                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5492                 put_rpccred(setclientid.sc_cred);
5493         }
5494         rpc_put_task(task);
5495 out:
5496         trace_nfs4_setclientid(clp, status);
5497         dprintk("NFS reply setclientid: %d\n", status);
5498         return status;
5499 }
5500
5501 /**
5502  * nfs4_proc_setclientid_confirm - Confirm client ID
5503  * @clp: state data structure
5504  * @res: result of a previous SETCLIENTID
5505  * @cred: RPC credential to use for this call
5506  *
5507  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5508  */
5509 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5510                 struct nfs4_setclientid_res *arg,
5511                 struct rpc_cred *cred)
5512 {
5513         struct rpc_message msg = {
5514                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5515                 .rpc_argp = arg,
5516                 .rpc_cred = cred,
5517         };
5518         int status;
5519
5520         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5521                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5522                 clp->cl_clientid);
5523         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5524         trace_nfs4_setclientid_confirm(clp, status);
5525         dprintk("NFS reply setclientid_confirm: %d\n", status);
5526         return status;
5527 }
5528
5529 struct nfs4_delegreturndata {
5530         struct nfs4_delegreturnargs args;
5531         struct nfs4_delegreturnres res;
5532         struct nfs_fh fh;
5533         nfs4_stateid stateid;
5534         unsigned long timestamp;
5535         struct {
5536                 struct nfs4_layoutreturn_args arg;
5537                 struct nfs4_layoutreturn_res res;
5538                 struct nfs4_xdr_opaque_data ld_private;
5539                 u32 roc_barrier;
5540                 bool roc;
5541         } lr;
5542         struct nfs_fattr fattr;
5543         int rpc_status;
5544         struct inode *inode;
5545 };
5546
5547 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5548 {
5549         struct nfs4_delegreturndata *data = calldata;
5550
5551         if (!nfs4_sequence_done(task, &data->res.seq_res))
5552                 return;
5553
5554         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5555
5556         /* Handle Layoutreturn errors */
5557         if (data->args.lr_args && task->tk_status != 0) {
5558                 switch(data->res.lr_ret) {
5559                 default:
5560                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5561                         break;
5562                 case 0:
5563                         data->args.lr_args = NULL;
5564                         data->res.lr_res = NULL;
5565                         break;
5566                 case -NFS4ERR_ADMIN_REVOKED:
5567                 case -NFS4ERR_DELEG_REVOKED:
5568                 case -NFS4ERR_EXPIRED:
5569                 case -NFS4ERR_BAD_STATEID:
5570                 case -NFS4ERR_OLD_STATEID:
5571                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5572                 case -NFS4ERR_WRONG_CRED:
5573                         data->args.lr_args = NULL;
5574                         data->res.lr_res = NULL;
5575                         data->res.lr_ret = 0;
5576                         rpc_restart_call_prepare(task);
5577                         return;
5578                 }
5579         }
5580
5581         switch (task->tk_status) {
5582         case 0:
5583                 renew_lease(data->res.server, data->timestamp);
5584                 break;
5585         case -NFS4ERR_ADMIN_REVOKED:
5586         case -NFS4ERR_DELEG_REVOKED:
5587         case -NFS4ERR_EXPIRED:
5588                 nfs4_free_revoked_stateid(data->res.server,
5589                                 data->args.stateid,
5590                                 task->tk_msg.rpc_cred);
5591         case -NFS4ERR_BAD_STATEID:
5592         case -NFS4ERR_OLD_STATEID:
5593         case -NFS4ERR_STALE_STATEID:
5594                 task->tk_status = 0;
5595                 break;
5596         case -NFS4ERR_ACCESS:
5597                 if (data->args.bitmask) {
5598                         data->args.bitmask = NULL;
5599                         data->res.fattr = NULL;
5600                         task->tk_status = 0;
5601                         rpc_restart_call_prepare(task);
5602                         return;
5603                 }
5604         default:
5605                 if (nfs4_async_handle_error(task, data->res.server,
5606                                             NULL, NULL) == -EAGAIN) {
5607                         rpc_restart_call_prepare(task);
5608                         return;
5609                 }
5610         }
5611         data->rpc_status = task->tk_status;
5612 }
5613
5614 static void nfs4_delegreturn_release(void *calldata)
5615 {
5616         struct nfs4_delegreturndata *data = calldata;
5617         struct inode *inode = data->inode;
5618
5619         if (inode) {
5620                 if (data->lr.roc)
5621                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5622                                         data->res.lr_ret);
5623                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5624                 nfs_iput_and_deactive(inode);
5625         }
5626         kfree(calldata);
5627 }
5628
5629 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5630 {
5631         struct nfs4_delegreturndata *d_data;
5632
5633         d_data = (struct nfs4_delegreturndata *)data;
5634
5635         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5636                 return;
5637
5638         nfs4_setup_sequence(d_data->res.server->nfs_client,
5639                         &d_data->args.seq_args,
5640                         &d_data->res.seq_res,
5641                         task);
5642 }
5643
5644 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5645         .rpc_call_prepare = nfs4_delegreturn_prepare,
5646         .rpc_call_done = nfs4_delegreturn_done,
5647         .rpc_release = nfs4_delegreturn_release,
5648 };
5649
5650 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5651 {
5652         struct nfs4_delegreturndata *data;
5653         struct nfs_server *server = NFS_SERVER(inode);
5654         struct rpc_task *task;
5655         struct rpc_message msg = {
5656                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5657                 .rpc_cred = cred,
5658         };
5659         struct rpc_task_setup task_setup_data = {
5660                 .rpc_client = server->client,
5661                 .rpc_message = &msg,
5662                 .callback_ops = &nfs4_delegreturn_ops,
5663                 .flags = RPC_TASK_ASYNC,
5664         };
5665         int status = 0;
5666
5667         data = kzalloc(sizeof(*data), GFP_NOFS);
5668         if (data == NULL)
5669                 return -ENOMEM;
5670         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5671
5672         nfs4_state_protect(server->nfs_client,
5673                         NFS_SP4_MACH_CRED_CLEANUP,
5674                         &task_setup_data.rpc_client, &msg);
5675
5676         data->args.fhandle = &data->fh;
5677         data->args.stateid = &data->stateid;
5678         data->args.bitmask = server->cache_consistency_bitmask;
5679         nfs_copy_fh(&data->fh, NFS_FH(inode));
5680         nfs4_stateid_copy(&data->stateid, stateid);
5681         data->res.fattr = &data->fattr;
5682         data->res.server = server;
5683         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5684         data->lr.arg.ld_private = &data->lr.ld_private;
5685         nfs_fattr_init(data->res.fattr);
5686         data->timestamp = jiffies;
5687         data->rpc_status = 0;
5688         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5689         data->inode = nfs_igrab_and_active(inode);
5690         if (data->inode) {
5691                 if (data->lr.roc) {
5692                         data->args.lr_args = &data->lr.arg;
5693                         data->res.lr_res = &data->lr.res;
5694                 }
5695         } else if (data->lr.roc) {
5696                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5697                 data->lr.roc = false;
5698         }
5699
5700         task_setup_data.callback_data = data;
5701         msg.rpc_argp = &data->args;
5702         msg.rpc_resp = &data->res;
5703         task = rpc_run_task(&task_setup_data);
5704         if (IS_ERR(task))
5705                 return PTR_ERR(task);
5706         if (!issync)
5707                 goto out;
5708         status = rpc_wait_for_completion_task(task);
5709         if (status != 0)
5710                 goto out;
5711         status = data->rpc_status;
5712 out:
5713         rpc_put_task(task);
5714         return status;
5715 }
5716
5717 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5718 {
5719         struct nfs_server *server = NFS_SERVER(inode);
5720         struct nfs4_exception exception = { };
5721         int err;
5722         do {
5723                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5724                 trace_nfs4_delegreturn(inode, stateid, err);
5725                 switch (err) {
5726                         case -NFS4ERR_STALE_STATEID:
5727                         case -NFS4ERR_EXPIRED:
5728                         case 0:
5729                                 return 0;
5730                 }
5731                 err = nfs4_handle_exception(server, err, &exception);
5732         } while (exception.retry);
5733         return err;
5734 }
5735
5736 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5737 {
5738         struct inode *inode = state->inode;
5739         struct nfs_server *server = NFS_SERVER(inode);
5740         struct nfs_client *clp = server->nfs_client;
5741         struct nfs_lockt_args arg = {
5742                 .fh = NFS_FH(inode),
5743                 .fl = request,
5744         };
5745         struct nfs_lockt_res res = {
5746                 .denied = request,
5747         };
5748         struct rpc_message msg = {
5749                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5750                 .rpc_argp       = &arg,
5751                 .rpc_resp       = &res,
5752                 .rpc_cred       = state->owner->so_cred,
5753         };
5754         struct nfs4_lock_state *lsp;
5755         int status;
5756
5757         arg.lock_owner.clientid = clp->cl_clientid;
5758         status = nfs4_set_lock_state(state, request);
5759         if (status != 0)
5760                 goto out;
5761         lsp = request->fl_u.nfs4_fl.owner;
5762         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5763         arg.lock_owner.s_dev = server->s_dev;
5764         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5765         switch (status) {
5766                 case 0:
5767                         request->fl_type = F_UNLCK;
5768                         break;
5769                 case -NFS4ERR_DENIED:
5770                         status = 0;
5771         }
5772         request->fl_ops->fl_release_private(request);
5773         request->fl_ops = NULL;
5774 out:
5775         return status;
5776 }
5777
5778 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5779 {
5780         struct nfs4_exception exception = { };
5781         int err;
5782
5783         do {
5784                 err = _nfs4_proc_getlk(state, cmd, request);
5785                 trace_nfs4_get_lock(request, state, cmd, err);
5786                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5787                                 &exception);
5788         } while (exception.retry);
5789         return err;
5790 }
5791
5792 struct nfs4_unlockdata {
5793         struct nfs_locku_args arg;
5794         struct nfs_locku_res res;
5795         struct nfs4_lock_state *lsp;
5796         struct nfs_open_context *ctx;
5797         struct nfs_lock_context *l_ctx;
5798         struct file_lock fl;
5799         struct nfs_server *server;
5800         unsigned long timestamp;
5801 };
5802
5803 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5804                 struct nfs_open_context *ctx,
5805                 struct nfs4_lock_state *lsp,
5806                 struct nfs_seqid *seqid)
5807 {
5808         struct nfs4_unlockdata *p;
5809         struct inode *inode = lsp->ls_state->inode;
5810
5811         p = kzalloc(sizeof(*p), GFP_NOFS);
5812         if (p == NULL)
5813                 return NULL;
5814         p->arg.fh = NFS_FH(inode);
5815         p->arg.fl = &p->fl;
5816         p->arg.seqid = seqid;
5817         p->res.seqid = seqid;
5818         p->lsp = lsp;
5819         atomic_inc(&lsp->ls_count);
5820         /* Ensure we don't close file until we're done freeing locks! */
5821         p->ctx = get_nfs_open_context(ctx);
5822         p->l_ctx = nfs_get_lock_context(ctx);
5823         memcpy(&p->fl, fl, sizeof(p->fl));
5824         p->server = NFS_SERVER(inode);
5825         return p;
5826 }
5827
5828 static void nfs4_locku_release_calldata(void *data)
5829 {
5830         struct nfs4_unlockdata *calldata = data;
5831         nfs_free_seqid(calldata->arg.seqid);
5832         nfs4_put_lock_state(calldata->lsp);
5833         nfs_put_lock_context(calldata->l_ctx);
5834         put_nfs_open_context(calldata->ctx);
5835         kfree(calldata);
5836 }
5837
5838 static void nfs4_locku_done(struct rpc_task *task, void *data)
5839 {
5840         struct nfs4_unlockdata *calldata = data;
5841
5842         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5843                 return;
5844         switch (task->tk_status) {
5845                 case 0:
5846                         renew_lease(calldata->server, calldata->timestamp);
5847                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5848                         if (nfs4_update_lock_stateid(calldata->lsp,
5849                                         &calldata->res.stateid))
5850                                 break;
5851                 case -NFS4ERR_ADMIN_REVOKED:
5852                 case -NFS4ERR_EXPIRED:
5853                         nfs4_free_revoked_stateid(calldata->server,
5854                                         &calldata->arg.stateid,
5855                                         task->tk_msg.rpc_cred);
5856                 case -NFS4ERR_BAD_STATEID:
5857                 case -NFS4ERR_OLD_STATEID:
5858                 case -NFS4ERR_STALE_STATEID:
5859                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5860                                                 &calldata->lsp->ls_stateid))
5861                                 rpc_restart_call_prepare(task);
5862                         break;
5863                 default:
5864                         if (nfs4_async_handle_error(task, calldata->server,
5865                                                     NULL, NULL) == -EAGAIN)
5866                                 rpc_restart_call_prepare(task);
5867         }
5868         nfs_release_seqid(calldata->arg.seqid);
5869 }
5870
5871 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5872 {
5873         struct nfs4_unlockdata *calldata = data;
5874
5875         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
5876                 nfs_async_iocounter_wait(task, calldata->l_ctx))
5877                 return;
5878
5879         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5880                 goto out_wait;
5881         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5882         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5883                 /* Note: exit _without_ running nfs4_locku_done */
5884                 goto out_no_action;
5885         }
5886         calldata->timestamp = jiffies;
5887         if (nfs4_setup_sequence(calldata->server->nfs_client,
5888                                 &calldata->arg.seq_args,
5889                                 &calldata->res.seq_res,
5890                                 task) != 0)
5891                 nfs_release_seqid(calldata->arg.seqid);
5892         return;
5893 out_no_action:
5894         task->tk_action = NULL;
5895 out_wait:
5896         nfs4_sequence_done(task, &calldata->res.seq_res);
5897 }
5898
5899 static const struct rpc_call_ops nfs4_locku_ops = {
5900         .rpc_call_prepare = nfs4_locku_prepare,
5901         .rpc_call_done = nfs4_locku_done,
5902         .rpc_release = nfs4_locku_release_calldata,
5903 };
5904
5905 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5906                 struct nfs_open_context *ctx,
5907                 struct nfs4_lock_state *lsp,
5908                 struct nfs_seqid *seqid)
5909 {
5910         struct nfs4_unlockdata *data;
5911         struct rpc_message msg = {
5912                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5913                 .rpc_cred = ctx->cred,
5914         };
5915         struct rpc_task_setup task_setup_data = {
5916                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5917                 .rpc_message = &msg,
5918                 .callback_ops = &nfs4_locku_ops,
5919                 .workqueue = nfsiod_workqueue,
5920                 .flags = RPC_TASK_ASYNC,
5921         };
5922
5923         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5924                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5925
5926         /* Ensure this is an unlock - when canceling a lock, the
5927          * canceled lock is passed in, and it won't be an unlock.
5928          */
5929         fl->fl_type = F_UNLCK;
5930         if (fl->fl_flags & FL_CLOSE)
5931                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
5932
5933         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5934         if (data == NULL) {
5935                 nfs_free_seqid(seqid);
5936                 return ERR_PTR(-ENOMEM);
5937         }
5938
5939         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5940         msg.rpc_argp = &data->arg;
5941         msg.rpc_resp = &data->res;
5942         task_setup_data.callback_data = data;
5943         return rpc_run_task(&task_setup_data);
5944 }
5945
5946 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5947 {
5948         struct inode *inode = state->inode;
5949         struct nfs4_state_owner *sp = state->owner;
5950         struct nfs_inode *nfsi = NFS_I(inode);
5951         struct nfs_seqid *seqid;
5952         struct nfs4_lock_state *lsp;
5953         struct rpc_task *task;
5954         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5955         int status = 0;
5956         unsigned char fl_flags = request->fl_flags;
5957
5958         status = nfs4_set_lock_state(state, request);
5959         /* Unlock _before_ we do the RPC call */
5960         request->fl_flags |= FL_EXISTS;
5961         /* Exclude nfs_delegation_claim_locks() */
5962         mutex_lock(&sp->so_delegreturn_mutex);
5963         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5964         down_read(&nfsi->rwsem);
5965         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
5966                 up_read(&nfsi->rwsem);
5967                 mutex_unlock(&sp->so_delegreturn_mutex);
5968                 goto out;
5969         }
5970         up_read(&nfsi->rwsem);
5971         mutex_unlock(&sp->so_delegreturn_mutex);
5972         if (status != 0)
5973                 goto out;
5974         /* Is this a delegated lock? */
5975         lsp = request->fl_u.nfs4_fl.owner;
5976         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5977                 goto out;
5978         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5979         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5980         status = -ENOMEM;
5981         if (IS_ERR(seqid))
5982                 goto out;
5983         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5984         status = PTR_ERR(task);
5985         if (IS_ERR(task))
5986                 goto out;
5987         status = rpc_wait_for_completion_task(task);
5988         rpc_put_task(task);
5989 out:
5990         request->fl_flags = fl_flags;
5991         trace_nfs4_unlock(request, state, F_SETLK, status);
5992         return status;
5993 }
5994
5995 struct nfs4_lockdata {
5996         struct nfs_lock_args arg;
5997         struct nfs_lock_res res;
5998         struct nfs4_lock_state *lsp;
5999         struct nfs_open_context *ctx;
6000         struct file_lock fl;
6001         unsigned long timestamp;
6002         int rpc_status;
6003         int cancelled;
6004         struct nfs_server *server;
6005 };
6006
6007 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6008                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6009                 gfp_t gfp_mask)
6010 {
6011         struct nfs4_lockdata *p;
6012         struct inode *inode = lsp->ls_state->inode;
6013         struct nfs_server *server = NFS_SERVER(inode);
6014         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6015
6016         p = kzalloc(sizeof(*p), gfp_mask);
6017         if (p == NULL)
6018                 return NULL;
6019
6020         p->arg.fh = NFS_FH(inode);
6021         p->arg.fl = &p->fl;
6022         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6023         if (IS_ERR(p->arg.open_seqid))
6024                 goto out_free;
6025         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6026         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6027         if (IS_ERR(p->arg.lock_seqid))
6028                 goto out_free_seqid;
6029         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6030         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6031         p->arg.lock_owner.s_dev = server->s_dev;
6032         p->res.lock_seqid = p->arg.lock_seqid;
6033         p->lsp = lsp;
6034         p->server = server;
6035         atomic_inc(&lsp->ls_count);
6036         p->ctx = get_nfs_open_context(ctx);
6037         memcpy(&p->fl, fl, sizeof(p->fl));
6038         return p;
6039 out_free_seqid:
6040         nfs_free_seqid(p->arg.open_seqid);
6041 out_free:
6042         kfree(p);
6043         return NULL;
6044 }
6045
6046 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6047 {
6048         struct nfs4_lockdata *data = calldata;
6049         struct nfs4_state *state = data->lsp->ls_state;
6050
6051         dprintk("%s: begin!\n", __func__);
6052         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6053                 goto out_wait;
6054         /* Do we need to do an open_to_lock_owner? */
6055         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6056                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6057                         goto out_release_lock_seqid;
6058                 }
6059                 nfs4_stateid_copy(&data->arg.open_stateid,
6060                                 &state->open_stateid);
6061                 data->arg.new_lock_owner = 1;
6062                 data->res.open_seqid = data->arg.open_seqid;
6063         } else {
6064                 data->arg.new_lock_owner = 0;
6065                 nfs4_stateid_copy(&data->arg.lock_stateid,
6066                                 &data->lsp->ls_stateid);
6067         }
6068         if (!nfs4_valid_open_stateid(state)) {
6069                 data->rpc_status = -EBADF;
6070                 task->tk_action = NULL;
6071                 goto out_release_open_seqid;
6072         }
6073         data->timestamp = jiffies;
6074         if (nfs4_setup_sequence(data->server->nfs_client,
6075                                 &data->arg.seq_args,
6076                                 &data->res.seq_res,
6077                                 task) == 0)
6078                 return;
6079 out_release_open_seqid:
6080         nfs_release_seqid(data->arg.open_seqid);
6081 out_release_lock_seqid:
6082         nfs_release_seqid(data->arg.lock_seqid);
6083 out_wait:
6084         nfs4_sequence_done(task, &data->res.seq_res);
6085         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6086 }
6087
6088 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6089 {
6090         struct nfs4_lockdata *data = calldata;
6091         struct nfs4_lock_state *lsp = data->lsp;
6092
6093         dprintk("%s: begin!\n", __func__);
6094
6095         if (!nfs4_sequence_done(task, &data->res.seq_res))
6096                 return;
6097
6098         data->rpc_status = task->tk_status;
6099         switch (task->tk_status) {
6100         case 0:
6101                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6102                                 data->timestamp);
6103                 if (data->arg.new_lock) {
6104                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6105                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6106                                 rpc_restart_call_prepare(task);
6107                                 break;
6108                         }
6109                 }
6110                 if (data->arg.new_lock_owner != 0) {
6111                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6112                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6113                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6114                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6115                         rpc_restart_call_prepare(task);
6116                 break;
6117         case -NFS4ERR_BAD_STATEID:
6118         case -NFS4ERR_OLD_STATEID:
6119         case -NFS4ERR_STALE_STATEID:
6120         case -NFS4ERR_EXPIRED:
6121                 if (data->arg.new_lock_owner != 0) {
6122                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6123                                                 &lsp->ls_state->open_stateid))
6124                                 rpc_restart_call_prepare(task);
6125                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6126                                                 &lsp->ls_stateid))
6127                                 rpc_restart_call_prepare(task);
6128         }
6129         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6130 }
6131
6132 static void nfs4_lock_release(void *calldata)
6133 {
6134         struct nfs4_lockdata *data = calldata;
6135
6136         dprintk("%s: begin!\n", __func__);
6137         nfs_free_seqid(data->arg.open_seqid);
6138         if (data->cancelled != 0) {
6139                 struct rpc_task *task;
6140                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6141                                 data->arg.lock_seqid);
6142                 if (!IS_ERR(task))
6143                         rpc_put_task_async(task);
6144                 dprintk("%s: cancelling lock!\n", __func__);
6145         } else
6146                 nfs_free_seqid(data->arg.lock_seqid);
6147         nfs4_put_lock_state(data->lsp);
6148         put_nfs_open_context(data->ctx);
6149         kfree(data);
6150         dprintk("%s: done!\n", __func__);
6151 }
6152
6153 static const struct rpc_call_ops nfs4_lock_ops = {
6154         .rpc_call_prepare = nfs4_lock_prepare,
6155         .rpc_call_done = nfs4_lock_done,
6156         .rpc_release = nfs4_lock_release,
6157 };
6158
6159 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6160 {
6161         switch (error) {
6162         case -NFS4ERR_ADMIN_REVOKED:
6163         case -NFS4ERR_EXPIRED:
6164         case -NFS4ERR_BAD_STATEID:
6165                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6166                 if (new_lock_owner != 0 ||
6167                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6168                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6169                 break;
6170         case -NFS4ERR_STALE_STATEID:
6171                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6172                 nfs4_schedule_lease_recovery(server->nfs_client);
6173         };
6174 }
6175
6176 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6177 {
6178         struct nfs4_lockdata *data;
6179         struct rpc_task *task;
6180         struct rpc_message msg = {
6181                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6182                 .rpc_cred = state->owner->so_cred,
6183         };
6184         struct rpc_task_setup task_setup_data = {
6185                 .rpc_client = NFS_CLIENT(state->inode),
6186                 .rpc_message = &msg,
6187                 .callback_ops = &nfs4_lock_ops,
6188                 .workqueue = nfsiod_workqueue,
6189                 .flags = RPC_TASK_ASYNC,
6190         };
6191         int ret;
6192
6193         dprintk("%s: begin!\n", __func__);
6194         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6195                         fl->fl_u.nfs4_fl.owner,
6196                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6197         if (data == NULL)
6198                 return -ENOMEM;
6199         if (IS_SETLKW(cmd))
6200                 data->arg.block = 1;
6201         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6202         msg.rpc_argp = &data->arg;
6203         msg.rpc_resp = &data->res;
6204         task_setup_data.callback_data = data;
6205         if (recovery_type > NFS_LOCK_NEW) {
6206                 if (recovery_type == NFS_LOCK_RECLAIM)
6207                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6208                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6209         } else
6210                 data->arg.new_lock = 1;
6211         task = rpc_run_task(&task_setup_data);
6212         if (IS_ERR(task))
6213                 return PTR_ERR(task);
6214         ret = rpc_wait_for_completion_task(task);
6215         if (ret == 0) {
6216                 ret = data->rpc_status;
6217                 if (ret)
6218                         nfs4_handle_setlk_error(data->server, data->lsp,
6219                                         data->arg.new_lock_owner, ret);
6220         } else
6221                 data->cancelled = 1;
6222         rpc_put_task(task);
6223         dprintk("%s: done, ret = %d!\n", __func__, ret);
6224         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6225         return ret;
6226 }
6227
6228 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6229 {
6230         struct nfs_server *server = NFS_SERVER(state->inode);
6231         struct nfs4_exception exception = {
6232                 .inode = state->inode,
6233         };
6234         int err;
6235
6236         do {
6237                 /* Cache the lock if possible... */
6238                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6239                         return 0;
6240                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6241                 if (err != -NFS4ERR_DELAY)
6242                         break;
6243                 nfs4_handle_exception(server, err, &exception);
6244         } while (exception.retry);
6245         return err;
6246 }
6247
6248 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6249 {
6250         struct nfs_server *server = NFS_SERVER(state->inode);
6251         struct nfs4_exception exception = {
6252                 .inode = state->inode,
6253         };
6254         int err;
6255
6256         err = nfs4_set_lock_state(state, request);
6257         if (err != 0)
6258                 return err;
6259         if (!recover_lost_locks) {
6260                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6261                 return 0;
6262         }
6263         do {
6264                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6265                         return 0;
6266                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6267                 switch (err) {
6268                 default:
6269                         goto out;
6270                 case -NFS4ERR_GRACE:
6271                 case -NFS4ERR_DELAY:
6272                         nfs4_handle_exception(server, err, &exception);
6273                         err = 0;
6274                 }
6275         } while (exception.retry);
6276 out:
6277         return err;
6278 }
6279
6280 #if defined(CONFIG_NFS_V4_1)
6281 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6282 {
6283         struct nfs4_lock_state *lsp;
6284         int status;
6285
6286         status = nfs4_set_lock_state(state, request);
6287         if (status != 0)
6288                 return status;
6289         lsp = request->fl_u.nfs4_fl.owner;
6290         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6291             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6292                 return 0;
6293         return nfs4_lock_expired(state, request);
6294 }
6295 #endif
6296
6297 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6298 {
6299         struct nfs_inode *nfsi = NFS_I(state->inode);
6300         struct nfs4_state_owner *sp = state->owner;
6301         unsigned char fl_flags = request->fl_flags;
6302         int status;
6303
6304         request->fl_flags |= FL_ACCESS;
6305         status = locks_lock_inode_wait(state->inode, request);
6306         if (status < 0)
6307                 goto out;
6308         mutex_lock(&sp->so_delegreturn_mutex);
6309         down_read(&nfsi->rwsem);
6310         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6311                 /* Yes: cache locks! */
6312                 /* ...but avoid races with delegation recall... */
6313                 request->fl_flags = fl_flags & ~FL_SLEEP;
6314                 status = locks_lock_inode_wait(state->inode, request);
6315                 up_read(&nfsi->rwsem);
6316                 mutex_unlock(&sp->so_delegreturn_mutex);
6317                 goto out;
6318         }
6319         up_read(&nfsi->rwsem);
6320         mutex_unlock(&sp->so_delegreturn_mutex);
6321         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6322 out:
6323         request->fl_flags = fl_flags;
6324         return status;
6325 }
6326
6327 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6328 {
6329         struct nfs4_exception exception = {
6330                 .state = state,
6331                 .inode = state->inode,
6332         };
6333         int err;
6334
6335         do {
6336                 err = _nfs4_proc_setlk(state, cmd, request);
6337                 if (err == -NFS4ERR_DENIED)
6338                         err = -EAGAIN;
6339                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6340                                 err, &exception);
6341         } while (exception.retry);
6342         return err;
6343 }
6344
6345 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6346 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6347
6348 static int
6349 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6350                         struct file_lock *request)
6351 {
6352         int             status = -ERESTARTSYS;
6353         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6354
6355         while(!signalled()) {
6356                 status = nfs4_proc_setlk(state, cmd, request);
6357                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6358                         break;
6359                 freezable_schedule_timeout_interruptible(timeout);
6360                 timeout *= 2;
6361                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6362                 status = -ERESTARTSYS;
6363         }
6364         return status;
6365 }
6366
6367 #ifdef CONFIG_NFS_V4_1
6368 struct nfs4_lock_waiter {
6369         struct task_struct      *task;
6370         struct inode            *inode;
6371         struct nfs_lowner       *owner;
6372         bool                    notified;
6373 };
6374
6375 static int
6376 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6377 {
6378         int ret;
6379         struct cb_notify_lock_args *cbnl = key;
6380         struct nfs4_lock_waiter *waiter = wait->private;
6381         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6382                                 *wowner = waiter->owner;
6383
6384         /* Only wake if the callback was for the same owner */
6385         if (lowner->clientid != wowner->clientid ||
6386             lowner->id != wowner->id             ||
6387             lowner->s_dev != wowner->s_dev)
6388                 return 0;
6389
6390         /* Make sure it's for the right inode */
6391         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6392                 return 0;
6393
6394         waiter->notified = true;
6395
6396         /* override "private" so we can use default_wake_function */
6397         wait->private = waiter->task;
6398         ret = autoremove_wake_function(wait, mode, flags, key);
6399         wait->private = waiter;
6400         return ret;
6401 }
6402
6403 static int
6404 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6405 {
6406         int status = -ERESTARTSYS;
6407         unsigned long flags;
6408         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6409         struct nfs_server *server = NFS_SERVER(state->inode);
6410         struct nfs_client *clp = server->nfs_client;
6411         wait_queue_head_t *q = &clp->cl_lock_waitq;
6412         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6413                                     .id = lsp->ls_seqid.owner_id,
6414                                     .s_dev = server->s_dev };
6415         struct nfs4_lock_waiter waiter = { .task  = current,
6416                                            .inode = state->inode,
6417                                            .owner = &owner,
6418                                            .notified = false };
6419         wait_queue_entry_t wait;
6420
6421         /* Don't bother with waitqueue if we don't expect a callback */
6422         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6423                 return nfs4_retry_setlk_simple(state, cmd, request);
6424
6425         init_wait(&wait);
6426         wait.private = &waiter;
6427         wait.func = nfs4_wake_lock_waiter;
6428         add_wait_queue(q, &wait);
6429
6430         while(!signalled()) {
6431                 status = nfs4_proc_setlk(state, cmd, request);
6432                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6433                         break;
6434
6435                 status = -ERESTARTSYS;
6436                 spin_lock_irqsave(&q->lock, flags);
6437                 if (waiter.notified) {
6438                         spin_unlock_irqrestore(&q->lock, flags);
6439                         continue;
6440                 }
6441                 set_current_state(TASK_INTERRUPTIBLE);
6442                 spin_unlock_irqrestore(&q->lock, flags);
6443
6444                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6445         }
6446
6447         finish_wait(q, &wait);
6448         return status;
6449 }
6450 #else /* !CONFIG_NFS_V4_1 */
6451 static inline int
6452 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6453 {
6454         return nfs4_retry_setlk_simple(state, cmd, request);
6455 }
6456 #endif
6457
6458 static int
6459 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6460 {
6461         struct nfs_open_context *ctx;
6462         struct nfs4_state *state;
6463         int status;
6464
6465         /* verify open state */
6466         ctx = nfs_file_open_context(filp);
6467         state = ctx->state;
6468
6469         if (IS_GETLK(cmd)) {
6470                 if (state != NULL)
6471                         return nfs4_proc_getlk(state, F_GETLK, request);
6472                 return 0;
6473         }
6474
6475         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6476                 return -EINVAL;
6477
6478         if (request->fl_type == F_UNLCK) {
6479                 if (state != NULL)
6480                         return nfs4_proc_unlck(state, cmd, request);
6481                 return 0;
6482         }
6483
6484         if (state == NULL)
6485                 return -ENOLCK;
6486
6487         if ((request->fl_flags & FL_POSIX) &&
6488             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6489                 return -ENOLCK;
6490
6491         status = nfs4_set_lock_state(state, request);
6492         if (status != 0)
6493                 return status;
6494
6495         return nfs4_retry_setlk(state, cmd, request);
6496 }
6497
6498 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6499 {
6500         struct nfs_server *server = NFS_SERVER(state->inode);
6501         int err;
6502
6503         err = nfs4_set_lock_state(state, fl);
6504         if (err != 0)
6505                 return err;
6506         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6507         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6508 }
6509
6510 struct nfs_release_lockowner_data {
6511         struct nfs4_lock_state *lsp;
6512         struct nfs_server *server;
6513         struct nfs_release_lockowner_args args;
6514         struct nfs_release_lockowner_res res;
6515         unsigned long timestamp;
6516 };
6517
6518 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6519 {
6520         struct nfs_release_lockowner_data *data = calldata;
6521         struct nfs_server *server = data->server;
6522         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
6523                            &data->res.seq_res, task);
6524         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6525         data->timestamp = jiffies;
6526 }
6527
6528 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6529 {
6530         struct nfs_release_lockowner_data *data = calldata;
6531         struct nfs_server *server = data->server;
6532
6533         nfs40_sequence_done(task, &data->res.seq_res);
6534
6535         switch (task->tk_status) {
6536         case 0:
6537                 renew_lease(server, data->timestamp);
6538                 break;
6539         case -NFS4ERR_STALE_CLIENTID:
6540         case -NFS4ERR_EXPIRED:
6541                 nfs4_schedule_lease_recovery(server->nfs_client);
6542                 break;
6543         case -NFS4ERR_LEASE_MOVED:
6544         case -NFS4ERR_DELAY:
6545                 if (nfs4_async_handle_error(task, server,
6546                                             NULL, NULL) == -EAGAIN)
6547                         rpc_restart_call_prepare(task);
6548         }
6549 }
6550
6551 static void nfs4_release_lockowner_release(void *calldata)
6552 {
6553         struct nfs_release_lockowner_data *data = calldata;
6554         nfs4_free_lock_state(data->server, data->lsp);
6555         kfree(calldata);
6556 }
6557
6558 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6559         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6560         .rpc_call_done = nfs4_release_lockowner_done,
6561         .rpc_release = nfs4_release_lockowner_release,
6562 };
6563
6564 static void
6565 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6566 {
6567         struct nfs_release_lockowner_data *data;
6568         struct rpc_message msg = {
6569                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6570         };
6571
6572         if (server->nfs_client->cl_mvops->minor_version != 0)
6573                 return;
6574
6575         data = kmalloc(sizeof(*data), GFP_NOFS);
6576         if (!data)
6577                 return;
6578         data->lsp = lsp;
6579         data->server = server;
6580         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6581         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6582         data->args.lock_owner.s_dev = server->s_dev;
6583
6584         msg.rpc_argp = &data->args;
6585         msg.rpc_resp = &data->res;
6586         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6587         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6588 }
6589
6590 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6591
6592 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6593                                    struct dentry *unused, struct inode *inode,
6594                                    const char *key, const void *buf,
6595                                    size_t buflen, int flags)
6596 {
6597         return nfs4_proc_set_acl(inode, buf, buflen);
6598 }
6599
6600 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6601                                    struct dentry *unused, struct inode *inode,
6602                                    const char *key, void *buf, size_t buflen)
6603 {
6604         return nfs4_proc_get_acl(inode, buf, buflen);
6605 }
6606
6607 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6608 {
6609         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6610 }
6611
6612 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6613
6614 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6615                                      struct dentry *unused, struct inode *inode,
6616                                      const char *key, const void *buf,
6617                                      size_t buflen, int flags)
6618 {
6619         if (security_ismaclabel(key))
6620                 return nfs4_set_security_label(inode, buf, buflen);
6621
6622         return -EOPNOTSUPP;
6623 }
6624
6625 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6626                                      struct dentry *unused, struct inode *inode,
6627                                      const char *key, void *buf, size_t buflen)
6628 {
6629         if (security_ismaclabel(key))
6630                 return nfs4_get_security_label(inode, buf, buflen);
6631         return -EOPNOTSUPP;
6632 }
6633
6634 static ssize_t
6635 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6636 {
6637         int len = 0;
6638
6639         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6640                 len = security_inode_listsecurity(inode, list, list_len);
6641                 if (list_len && len > list_len)
6642                         return -ERANGE;
6643         }
6644         return len;
6645 }
6646
6647 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6648         .prefix = XATTR_SECURITY_PREFIX,
6649         .get    = nfs4_xattr_get_nfs4_label,
6650         .set    = nfs4_xattr_set_nfs4_label,
6651 };
6652
6653 #else
6654
6655 static ssize_t
6656 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6657 {
6658         return 0;
6659 }
6660
6661 #endif
6662
6663 /*
6664  * nfs_fhget will use either the mounted_on_fileid or the fileid
6665  */
6666 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6667 {
6668         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6669                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6670               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6671               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6672                 return;
6673
6674         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6675                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6676         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6677         fattr->nlink = 2;
6678 }
6679
6680 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6681                                    const struct qstr *name,
6682                                    struct nfs4_fs_locations *fs_locations,
6683                                    struct page *page)
6684 {
6685         struct nfs_server *server = NFS_SERVER(dir);
6686         u32 bitmask[3] = {
6687                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6688         };
6689         struct nfs4_fs_locations_arg args = {
6690                 .dir_fh = NFS_FH(dir),
6691                 .name = name,
6692                 .page = page,
6693                 .bitmask = bitmask,
6694         };
6695         struct nfs4_fs_locations_res res = {
6696                 .fs_locations = fs_locations,
6697         };
6698         struct rpc_message msg = {
6699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6700                 .rpc_argp = &args,
6701                 .rpc_resp = &res,
6702         };
6703         int status;
6704
6705         dprintk("%s: start\n", __func__);
6706
6707         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6708          * is not supported */
6709         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6710                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6711         else
6712                 bitmask[0] |= FATTR4_WORD0_FILEID;
6713
6714         nfs_fattr_init(&fs_locations->fattr);
6715         fs_locations->server = server;
6716         fs_locations->nlocations = 0;
6717         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6718         dprintk("%s: returned status = %d\n", __func__, status);
6719         return status;
6720 }
6721
6722 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6723                            const struct qstr *name,
6724                            struct nfs4_fs_locations *fs_locations,
6725                            struct page *page)
6726 {
6727         struct nfs4_exception exception = { };
6728         int err;
6729         do {
6730                 err = _nfs4_proc_fs_locations(client, dir, name,
6731                                 fs_locations, page);
6732                 trace_nfs4_get_fs_locations(dir, name, err);
6733                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6734                                 &exception);
6735         } while (exception.retry);
6736         return err;
6737 }
6738
6739 /*
6740  * This operation also signals the server that this client is
6741  * performing migration recovery.  The server can stop returning
6742  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6743  * appended to this compound to identify the client ID which is
6744  * performing recovery.
6745  */
6746 static int _nfs40_proc_get_locations(struct inode *inode,
6747                                      struct nfs4_fs_locations *locations,
6748                                      struct page *page, struct rpc_cred *cred)
6749 {
6750         struct nfs_server *server = NFS_SERVER(inode);
6751         struct rpc_clnt *clnt = server->client;
6752         u32 bitmask[2] = {
6753                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6754         };
6755         struct nfs4_fs_locations_arg args = {
6756                 .clientid       = server->nfs_client->cl_clientid,
6757                 .fh             = NFS_FH(inode),
6758                 .page           = page,
6759                 .bitmask        = bitmask,
6760                 .migration      = 1,            /* skip LOOKUP */
6761                 .renew          = 1,            /* append RENEW */
6762         };
6763         struct nfs4_fs_locations_res res = {
6764                 .fs_locations   = locations,
6765                 .migration      = 1,
6766                 .renew          = 1,
6767         };
6768         struct rpc_message msg = {
6769                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6770                 .rpc_argp       = &args,
6771                 .rpc_resp       = &res,
6772                 .rpc_cred       = cred,
6773         };
6774         unsigned long now = jiffies;
6775         int status;
6776
6777         nfs_fattr_init(&locations->fattr);
6778         locations->server = server;
6779         locations->nlocations = 0;
6780
6781         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6782         nfs4_set_sequence_privileged(&args.seq_args);
6783         status = nfs4_call_sync_sequence(clnt, server, &msg,
6784                                         &args.seq_args, &res.seq_res);
6785         if (status)
6786                 return status;
6787
6788         renew_lease(server, now);
6789         return 0;
6790 }
6791
6792 #ifdef CONFIG_NFS_V4_1
6793
6794 /*
6795  * This operation also signals the server that this client is
6796  * performing migration recovery.  The server can stop asserting
6797  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6798  * performing this operation is identified in the SEQUENCE
6799  * operation in this compound.
6800  *
6801  * When the client supports GETATTR(fs_locations_info), it can
6802  * be plumbed in here.
6803  */
6804 static int _nfs41_proc_get_locations(struct inode *inode,
6805                                      struct nfs4_fs_locations *locations,
6806                                      struct page *page, struct rpc_cred *cred)
6807 {
6808         struct nfs_server *server = NFS_SERVER(inode);
6809         struct rpc_clnt *clnt = server->client;
6810         u32 bitmask[2] = {
6811                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6812         };
6813         struct nfs4_fs_locations_arg args = {
6814                 .fh             = NFS_FH(inode),
6815                 .page           = page,
6816                 .bitmask        = bitmask,
6817                 .migration      = 1,            /* skip LOOKUP */
6818         };
6819         struct nfs4_fs_locations_res res = {
6820                 .fs_locations   = locations,
6821                 .migration      = 1,
6822         };
6823         struct rpc_message msg = {
6824                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6825                 .rpc_argp       = &args,
6826                 .rpc_resp       = &res,
6827                 .rpc_cred       = cred,
6828         };
6829         int status;
6830
6831         nfs_fattr_init(&locations->fattr);
6832         locations->server = server;
6833         locations->nlocations = 0;
6834
6835         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6836         nfs4_set_sequence_privileged(&args.seq_args);
6837         status = nfs4_call_sync_sequence(clnt, server, &msg,
6838                                         &args.seq_args, &res.seq_res);
6839         if (status == NFS4_OK &&
6840             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6841                 status = -NFS4ERR_LEASE_MOVED;
6842         return status;
6843 }
6844
6845 #endif  /* CONFIG_NFS_V4_1 */
6846
6847 /**
6848  * nfs4_proc_get_locations - discover locations for a migrated FSID
6849  * @inode: inode on FSID that is migrating
6850  * @locations: result of query
6851  * @page: buffer
6852  * @cred: credential to use for this operation
6853  *
6854  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6855  * operation failed, or a negative errno if a local error occurred.
6856  *
6857  * On success, "locations" is filled in, but if the server has
6858  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6859  * asserted.
6860  *
6861  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6862  * from this client that require migration recovery.
6863  */
6864 int nfs4_proc_get_locations(struct inode *inode,
6865                             struct nfs4_fs_locations *locations,
6866                             struct page *page, struct rpc_cred *cred)
6867 {
6868         struct nfs_server *server = NFS_SERVER(inode);
6869         struct nfs_client *clp = server->nfs_client;
6870         const struct nfs4_mig_recovery_ops *ops =
6871                                         clp->cl_mvops->mig_recovery_ops;
6872         struct nfs4_exception exception = { };
6873         int status;
6874
6875         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6876                 (unsigned long long)server->fsid.major,
6877                 (unsigned long long)server->fsid.minor,
6878                 clp->cl_hostname);
6879         nfs_display_fhandle(NFS_FH(inode), __func__);
6880
6881         do {
6882                 status = ops->get_locations(inode, locations, page, cred);
6883                 if (status != -NFS4ERR_DELAY)
6884                         break;
6885                 nfs4_handle_exception(server, status, &exception);
6886         } while (exception.retry);
6887         return status;
6888 }
6889
6890 /*
6891  * This operation also signals the server that this client is
6892  * performing "lease moved" recovery.  The server can stop
6893  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6894  * is appended to this compound to identify the client ID which is
6895  * performing recovery.
6896  */
6897 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6898 {
6899         struct nfs_server *server = NFS_SERVER(inode);
6900         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6901         struct rpc_clnt *clnt = server->client;
6902         struct nfs4_fsid_present_arg args = {
6903                 .fh             = NFS_FH(inode),
6904                 .clientid       = clp->cl_clientid,
6905                 .renew          = 1,            /* append RENEW */
6906         };
6907         struct nfs4_fsid_present_res res = {
6908                 .renew          = 1,
6909         };
6910         struct rpc_message msg = {
6911                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6912                 .rpc_argp       = &args,
6913                 .rpc_resp       = &res,
6914                 .rpc_cred       = cred,
6915         };
6916         unsigned long now = jiffies;
6917         int status;
6918
6919         res.fh = nfs_alloc_fhandle();
6920         if (res.fh == NULL)
6921                 return -ENOMEM;
6922
6923         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6924         nfs4_set_sequence_privileged(&args.seq_args);
6925         status = nfs4_call_sync_sequence(clnt, server, &msg,
6926                                                 &args.seq_args, &res.seq_res);
6927         nfs_free_fhandle(res.fh);
6928         if (status)
6929                 return status;
6930
6931         do_renew_lease(clp, now);
6932         return 0;
6933 }
6934
6935 #ifdef CONFIG_NFS_V4_1
6936
6937 /*
6938  * This operation also signals the server that this client is
6939  * performing "lease moved" recovery.  The server can stop asserting
6940  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6941  * this operation is identified in the SEQUENCE operation in this
6942  * compound.
6943  */
6944 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6945 {
6946         struct nfs_server *server = NFS_SERVER(inode);
6947         struct rpc_clnt *clnt = server->client;
6948         struct nfs4_fsid_present_arg args = {
6949                 .fh             = NFS_FH(inode),
6950         };
6951         struct nfs4_fsid_present_res res = {
6952         };
6953         struct rpc_message msg = {
6954                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6955                 .rpc_argp       = &args,
6956                 .rpc_resp       = &res,
6957                 .rpc_cred       = cred,
6958         };
6959         int status;
6960
6961         res.fh = nfs_alloc_fhandle();
6962         if (res.fh == NULL)
6963                 return -ENOMEM;
6964
6965         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6966         nfs4_set_sequence_privileged(&args.seq_args);
6967         status = nfs4_call_sync_sequence(clnt, server, &msg,
6968                                                 &args.seq_args, &res.seq_res);
6969         nfs_free_fhandle(res.fh);
6970         if (status == NFS4_OK &&
6971             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6972                 status = -NFS4ERR_LEASE_MOVED;
6973         return status;
6974 }
6975
6976 #endif  /* CONFIG_NFS_V4_1 */
6977
6978 /**
6979  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6980  * @inode: inode on FSID to check
6981  * @cred: credential to use for this operation
6982  *
6983  * Server indicates whether the FSID is present, moved, or not
6984  * recognized.  This operation is necessary to clear a LEASE_MOVED
6985  * condition for this client ID.
6986  *
6987  * Returns NFS4_OK if the FSID is present on this server,
6988  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6989  *  NFS4ERR code if some error occurred on the server, or a
6990  *  negative errno if a local failure occurred.
6991  */
6992 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6993 {
6994         struct nfs_server *server = NFS_SERVER(inode);
6995         struct nfs_client *clp = server->nfs_client;
6996         const struct nfs4_mig_recovery_ops *ops =
6997                                         clp->cl_mvops->mig_recovery_ops;
6998         struct nfs4_exception exception = { };
6999         int status;
7000
7001         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7002                 (unsigned long long)server->fsid.major,
7003                 (unsigned long long)server->fsid.minor,
7004                 clp->cl_hostname);
7005         nfs_display_fhandle(NFS_FH(inode), __func__);
7006
7007         do {
7008                 status = ops->fsid_present(inode, cred);
7009                 if (status != -NFS4ERR_DELAY)
7010                         break;
7011                 nfs4_handle_exception(server, status, &exception);
7012         } while (exception.retry);
7013         return status;
7014 }
7015
7016 /**
7017  * If 'use_integrity' is true and the state managment nfs_client
7018  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7019  * and the machine credential as per RFC3530bis and RFC5661 Security
7020  * Considerations sections. Otherwise, just use the user cred with the
7021  * filesystem's rpc_client.
7022  */
7023 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7024 {
7025         int status;
7026         struct nfs4_secinfo_arg args = {
7027                 .dir_fh = NFS_FH(dir),
7028                 .name   = name,
7029         };
7030         struct nfs4_secinfo_res res = {
7031                 .flavors     = flavors,
7032         };
7033         struct rpc_message msg = {
7034                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7035                 .rpc_argp = &args,
7036                 .rpc_resp = &res,
7037         };
7038         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7039         struct rpc_cred *cred = NULL;
7040
7041         if (use_integrity) {
7042                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7043                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7044                 msg.rpc_cred = cred;
7045         }
7046
7047         dprintk("NFS call  secinfo %s\n", name->name);
7048
7049         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7050                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7051
7052         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7053                                 &res.seq_res, 0);
7054         dprintk("NFS reply  secinfo: %d\n", status);
7055
7056         if (cred)
7057                 put_rpccred(cred);
7058
7059         return status;
7060 }
7061
7062 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7063                       struct nfs4_secinfo_flavors *flavors)
7064 {
7065         struct nfs4_exception exception = { };
7066         int err;
7067         do {
7068                 err = -NFS4ERR_WRONGSEC;
7069
7070                 /* try to use integrity protection with machine cred */
7071                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7072                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7073
7074                 /*
7075                  * if unable to use integrity protection, or SECINFO with
7076                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7077                  * disallowed by spec, but exists in deployed servers) use
7078                  * the current filesystem's rpc_client and the user cred.
7079                  */
7080                 if (err == -NFS4ERR_WRONGSEC)
7081                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7082
7083                 trace_nfs4_secinfo(dir, name, err);
7084                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7085                                 &exception);
7086         } while (exception.retry);
7087         return err;
7088 }
7089
7090 #ifdef CONFIG_NFS_V4_1
7091 /*
7092  * Check the exchange flags returned by the server for invalid flags, having
7093  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7094  * DS flags set.
7095  */
7096 static int nfs4_check_cl_exchange_flags(u32 flags)
7097 {
7098         if (flags & ~EXCHGID4_FLAG_MASK_R)
7099                 goto out_inval;
7100         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7101             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7102                 goto out_inval;
7103         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7104                 goto out_inval;
7105         return NFS_OK;
7106 out_inval:
7107         return -NFS4ERR_INVAL;
7108 }
7109
7110 static bool
7111 nfs41_same_server_scope(struct nfs41_server_scope *a,
7112                         struct nfs41_server_scope *b)
7113 {
7114         if (a->server_scope_sz != b->server_scope_sz)
7115                 return false;
7116         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7117 }
7118
7119 static void
7120 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7121 {
7122 }
7123
7124 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7125         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7126 };
7127
7128 /*
7129  * nfs4_proc_bind_one_conn_to_session()
7130  *
7131  * The 4.1 client currently uses the same TCP connection for the
7132  * fore and backchannel.
7133  */
7134 static
7135 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7136                 struct rpc_xprt *xprt,
7137                 struct nfs_client *clp,
7138                 struct rpc_cred *cred)
7139 {
7140         int status;
7141         struct nfs41_bind_conn_to_session_args args = {
7142                 .client = clp,
7143                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7144         };
7145         struct nfs41_bind_conn_to_session_res res;
7146         struct rpc_message msg = {
7147                 .rpc_proc =
7148                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7149                 .rpc_argp = &args,
7150                 .rpc_resp = &res,
7151                 .rpc_cred = cred,
7152         };
7153         struct rpc_task_setup task_setup_data = {
7154                 .rpc_client = clnt,
7155                 .rpc_xprt = xprt,
7156                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7157                 .rpc_message = &msg,
7158                 .flags = RPC_TASK_TIMEOUT,
7159         };
7160         struct rpc_task *task;
7161
7162         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7163         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7164                 args.dir = NFS4_CDFC4_FORE;
7165
7166         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7167         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7168                 args.dir = NFS4_CDFC4_FORE;
7169
7170         task = rpc_run_task(&task_setup_data);
7171         if (!IS_ERR(task)) {
7172                 status = task->tk_status;
7173                 rpc_put_task(task);
7174         } else
7175                 status = PTR_ERR(task);
7176         trace_nfs4_bind_conn_to_session(clp, status);
7177         if (status == 0) {
7178                 if (memcmp(res.sessionid.data,
7179                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7180                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7181                         return -EIO;
7182                 }
7183                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7184                         dprintk("NFS: %s: Unexpected direction from server\n",
7185                                 __func__);
7186                         return -EIO;
7187                 }
7188                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7189                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7190                                 __func__);
7191                         return -EIO;
7192                 }
7193         }
7194
7195         return status;
7196 }
7197
7198 struct rpc_bind_conn_calldata {
7199         struct nfs_client *clp;
7200         struct rpc_cred *cred;
7201 };
7202
7203 static int
7204 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7205                 struct rpc_xprt *xprt,
7206                 void *calldata)
7207 {
7208         struct rpc_bind_conn_calldata *p = calldata;
7209
7210         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7211 }
7212
7213 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7214 {
7215         struct rpc_bind_conn_calldata data = {
7216                 .clp = clp,
7217                 .cred = cred,
7218         };
7219         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7220                         nfs4_proc_bind_conn_to_session_callback, &data);
7221 }
7222
7223 /*
7224  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7225  * and operations we'd like to see to enable certain features in the allow map
7226  */
7227 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7228         .how = SP4_MACH_CRED,
7229         .enforce.u.words = {
7230                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7231                       1 << (OP_EXCHANGE_ID - 32) |
7232                       1 << (OP_CREATE_SESSION - 32) |
7233                       1 << (OP_DESTROY_SESSION - 32) |
7234                       1 << (OP_DESTROY_CLIENTID - 32)
7235         },
7236         .allow.u.words = {
7237                 [0] = 1 << (OP_CLOSE) |
7238                       1 << (OP_OPEN_DOWNGRADE) |
7239                       1 << (OP_LOCKU) |
7240                       1 << (OP_DELEGRETURN) |
7241                       1 << (OP_COMMIT),
7242                 [1] = 1 << (OP_SECINFO - 32) |
7243                       1 << (OP_SECINFO_NO_NAME - 32) |
7244                       1 << (OP_LAYOUTRETURN - 32) |
7245                       1 << (OP_TEST_STATEID - 32) |
7246                       1 << (OP_FREE_STATEID - 32) |
7247                       1 << (OP_WRITE - 32)
7248         }
7249 };
7250
7251 /*
7252  * Select the state protection mode for client `clp' given the server results
7253  * from exchange_id in `sp'.
7254  *
7255  * Returns 0 on success, negative errno otherwise.
7256  */
7257 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7258                                  struct nfs41_state_protection *sp)
7259 {
7260         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7261                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7262                       1 << (OP_EXCHANGE_ID - 32) |
7263                       1 << (OP_CREATE_SESSION - 32) |
7264                       1 << (OP_DESTROY_SESSION - 32) |
7265                       1 << (OP_DESTROY_CLIENTID - 32)
7266         };
7267         unsigned int i;
7268
7269         if (sp->how == SP4_MACH_CRED) {
7270                 /* Print state protect result */
7271                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7272                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7273                         if (test_bit(i, sp->enforce.u.longs))
7274                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7275                         if (test_bit(i, sp->allow.u.longs))
7276                                 dfprintk(MOUNT, "  allow op %d\n", i);
7277                 }
7278
7279                 /* make sure nothing is on enforce list that isn't supported */
7280                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7281                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7282                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7283                                 return -EINVAL;
7284                         }
7285                 }
7286
7287                 /*
7288                  * Minimal mode - state operations are allowed to use machine
7289                  * credential.  Note this already happens by default, so the
7290                  * client doesn't have to do anything more than the negotiation.
7291                  *
7292                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7293                  *       we're already using the machine cred for exchange_id
7294                  *       and will never use a different cred.
7295                  */
7296                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7297                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7298                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7299                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7300                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7301                         dfprintk(MOUNT, "  minimal mode enabled\n");
7302                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7303                 } else {
7304                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7305                         return -EINVAL;
7306                 }
7307
7308                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7309                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7310                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7311                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7312                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7313                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7314                 }
7315
7316                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7317                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7318                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7319                                 &clp->cl_sp4_flags);
7320                 }
7321
7322                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7323                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7324                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7325                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7326                 }
7327
7328                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7329                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7330                         dfprintk(MOUNT, "  stateid mode enabled\n");
7331                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7332                 }
7333
7334                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7335                         dfprintk(MOUNT, "  write mode enabled\n");
7336                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7337                 }
7338
7339                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7340                         dfprintk(MOUNT, "  commit mode enabled\n");
7341                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7342                 }
7343         }
7344
7345         return 0;
7346 }
7347
7348 struct nfs41_exchange_id_data {
7349         struct nfs41_exchange_id_res res;
7350         struct nfs41_exchange_id_args args;
7351         struct rpc_xprt *xprt;
7352         int rpc_status;
7353 };
7354
7355 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7356 {
7357         struct nfs41_exchange_id_data *cdata =
7358                                         (struct nfs41_exchange_id_data *)data;
7359         struct nfs_client *clp = cdata->args.client;
7360         int status = task->tk_status;
7361
7362         trace_nfs4_exchange_id(clp, status);
7363
7364         if (status == 0)
7365                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7366
7367         if (cdata->xprt && status == 0) {
7368                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7369                                                       cdata->xprt);
7370                 goto out;
7371         }
7372
7373         if (status  == 0)
7374                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7375
7376         if (status == 0) {
7377                 clp->cl_clientid = cdata->res.clientid;
7378                 clp->cl_exchange_flags = cdata->res.flags;
7379                 /* Client ID is not confirmed */
7380                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7381                         clear_bit(NFS4_SESSION_ESTABLISHED,
7382                         &clp->cl_session->session_state);
7383                         clp->cl_seqid = cdata->res.seqid;
7384                 }
7385
7386                 kfree(clp->cl_serverowner);
7387                 clp->cl_serverowner = cdata->res.server_owner;
7388                 cdata->res.server_owner = NULL;
7389
7390                 /* use the most recent implementation id */
7391                 kfree(clp->cl_implid);
7392                 clp->cl_implid = cdata->res.impl_id;
7393                 cdata->res.impl_id = NULL;
7394
7395                 if (clp->cl_serverscope != NULL &&
7396                     !nfs41_same_server_scope(clp->cl_serverscope,
7397                                         cdata->res.server_scope)) {
7398                         dprintk("%s: server_scope mismatch detected\n",
7399                                 __func__);
7400                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7401                         kfree(clp->cl_serverscope);
7402                         clp->cl_serverscope = NULL;
7403                 }
7404
7405                 if (clp->cl_serverscope == NULL) {
7406                         clp->cl_serverscope = cdata->res.server_scope;
7407                         cdata->res.server_scope = NULL;
7408                 }
7409                 /* Save the EXCHANGE_ID verifier session trunk tests */
7410                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7411                        sizeof(clp->cl_confirm.data));
7412         }
7413 out:
7414         cdata->rpc_status = status;
7415         return;
7416 }
7417
7418 static void nfs4_exchange_id_release(void *data)
7419 {
7420         struct nfs41_exchange_id_data *cdata =
7421                                         (struct nfs41_exchange_id_data *)data;
7422
7423         if (cdata->xprt) {
7424                 xprt_put(cdata->xprt);
7425                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7426         }
7427         nfs_put_client(cdata->args.client);
7428         kfree(cdata->res.impl_id);
7429         kfree(cdata->res.server_scope);
7430         kfree(cdata->res.server_owner);
7431         kfree(cdata);
7432 }
7433
7434 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7435         .rpc_call_done = nfs4_exchange_id_done,
7436         .rpc_release = nfs4_exchange_id_release,
7437 };
7438
7439 /*
7440  * _nfs4_proc_exchange_id()
7441  *
7442  * Wrapper for EXCHANGE_ID operation.
7443  */
7444 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7445                         u32 sp4_how, struct rpc_xprt *xprt)
7446 {
7447         nfs4_verifier verifier;
7448         struct rpc_message msg = {
7449                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7450                 .rpc_cred = cred,
7451         };
7452         struct rpc_task_setup task_setup_data = {
7453                 .rpc_client = clp->cl_rpcclient,
7454                 .callback_ops = &nfs4_exchange_id_call_ops,
7455                 .rpc_message = &msg,
7456                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7457         };
7458         struct nfs41_exchange_id_data *calldata;
7459         struct rpc_task *task;
7460         int status;
7461
7462         if (!atomic_inc_not_zero(&clp->cl_count))
7463                 return -EIO;
7464
7465         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7466         if (!calldata) {
7467                 nfs_put_client(clp);
7468                 return -ENOMEM;
7469         }
7470
7471         if (!xprt)
7472                 nfs4_init_boot_verifier(clp, &verifier);
7473
7474         status = nfs4_init_uniform_client_string(clp);
7475         if (status)
7476                 goto out_calldata;
7477
7478         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7479                                                 GFP_NOFS);
7480         status = -ENOMEM;
7481         if (unlikely(calldata->res.server_owner == NULL))
7482                 goto out_calldata;
7483
7484         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7485                                         GFP_NOFS);
7486         if (unlikely(calldata->res.server_scope == NULL))
7487                 goto out_server_owner;
7488
7489         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7490         if (unlikely(calldata->res.impl_id == NULL))
7491                 goto out_server_scope;
7492
7493         switch (sp4_how) {
7494         case SP4_NONE:
7495                 calldata->args.state_protect.how = SP4_NONE;
7496                 break;
7497
7498         case SP4_MACH_CRED:
7499                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7500                 break;
7501
7502         default:
7503                 /* unsupported! */
7504                 WARN_ON_ONCE(1);
7505                 status = -EINVAL;
7506                 goto out_impl_id;
7507         }
7508         if (xprt) {
7509                 calldata->xprt = xprt;
7510                 task_setup_data.rpc_xprt = xprt;
7511                 task_setup_data.flags =
7512                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7513                 calldata->args.verifier = &clp->cl_confirm;
7514         } else {
7515                 calldata->args.verifier = &verifier;
7516         }
7517         calldata->args.client = clp;
7518 #ifdef CONFIG_NFS_V4_1_MIGRATION
7519         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7520         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7521         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7522 #else
7523         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7524         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7525 #endif
7526         msg.rpc_argp = &calldata->args;
7527         msg.rpc_resp = &calldata->res;
7528         task_setup_data.callback_data = calldata;
7529
7530         task = rpc_run_task(&task_setup_data);
7531         if (IS_ERR(task))
7532                 return PTR_ERR(task);
7533
7534         if (!xprt) {
7535                 status = rpc_wait_for_completion_task(task);
7536                 if (!status)
7537                         status = calldata->rpc_status;
7538         } else  /* session trunking test */
7539                 status = calldata->rpc_status;
7540
7541         rpc_put_task(task);
7542 out:
7543         return status;
7544
7545 out_impl_id:
7546         kfree(calldata->res.impl_id);
7547 out_server_scope:
7548         kfree(calldata->res.server_scope);
7549 out_server_owner:
7550         kfree(calldata->res.server_owner);
7551 out_calldata:
7552         kfree(calldata);
7553         nfs_put_client(clp);
7554         goto out;
7555 }
7556
7557 /*
7558  * nfs4_proc_exchange_id()
7559  *
7560  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7561  *
7562  * Since the clientid has expired, all compounds using sessions
7563  * associated with the stale clientid will be returning
7564  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7565  * be in some phase of session reset.
7566  *
7567  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7568  */
7569 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7570 {
7571         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7572         int status;
7573
7574         /* try SP4_MACH_CRED if krb5i/p */
7575         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7576             authflavor == RPC_AUTH_GSS_KRB5P) {
7577                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7578                 if (!status)
7579                         return 0;
7580         }
7581
7582         /* try SP4_NONE */
7583         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7584 }
7585
7586 /**
7587  * nfs4_test_session_trunk
7588  *
7589  * This is an add_xprt_test() test function called from
7590  * rpc_clnt_setup_test_and_add_xprt.
7591  *
7592  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7593  * and is dereferrenced in nfs4_exchange_id_release
7594  *
7595  * Upon success, add the new transport to the rpc_clnt
7596  *
7597  * @clnt: struct rpc_clnt to get new transport
7598  * @xprt: the rpc_xprt to test
7599  * @data: call data for _nfs4_proc_exchange_id.
7600  */
7601 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7602                             void *data)
7603 {
7604         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7605         u32 sp4_how;
7606
7607         dprintk("--> %s try %s\n", __func__,
7608                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7609
7610         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7611
7612         /* Test connection for session trunking. Async exchange_id call */
7613         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7614 }
7615 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7616
7617 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7618                 struct rpc_cred *cred)
7619 {
7620         struct rpc_message msg = {
7621                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7622                 .rpc_argp = clp,
7623                 .rpc_cred = cred,
7624         };
7625         int status;
7626
7627         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7628         trace_nfs4_destroy_clientid(clp, status);
7629         if (status)
7630                 dprintk("NFS: Got error %d from the server %s on "
7631                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7632         return status;
7633 }
7634
7635 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7636                 struct rpc_cred *cred)
7637 {
7638         unsigned int loop;
7639         int ret;
7640
7641         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7642                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7643                 switch (ret) {
7644                 case -NFS4ERR_DELAY:
7645                 case -NFS4ERR_CLIENTID_BUSY:
7646                         ssleep(1);
7647                         break;
7648                 default:
7649                         return ret;
7650                 }
7651         }
7652         return 0;
7653 }
7654
7655 int nfs4_destroy_clientid(struct nfs_client *clp)
7656 {
7657         struct rpc_cred *cred;
7658         int ret = 0;
7659
7660         if (clp->cl_mvops->minor_version < 1)
7661                 goto out;
7662         if (clp->cl_exchange_flags == 0)
7663                 goto out;
7664         if (clp->cl_preserve_clid)
7665                 goto out;
7666         cred = nfs4_get_clid_cred(clp);
7667         ret = nfs4_proc_destroy_clientid(clp, cred);
7668         if (cred)
7669                 put_rpccred(cred);
7670         switch (ret) {
7671         case 0:
7672         case -NFS4ERR_STALE_CLIENTID:
7673                 clp->cl_exchange_flags = 0;
7674         }
7675 out:
7676         return ret;
7677 }
7678
7679 struct nfs4_get_lease_time_data {
7680         struct nfs4_get_lease_time_args *args;
7681         struct nfs4_get_lease_time_res *res;
7682         struct nfs_client *clp;
7683 };
7684
7685 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7686                                         void *calldata)
7687 {
7688         struct nfs4_get_lease_time_data *data =
7689                         (struct nfs4_get_lease_time_data *)calldata;
7690
7691         dprintk("--> %s\n", __func__);
7692         /* just setup sequence, do not trigger session recovery
7693            since we're invoked within one */
7694         nfs4_setup_sequence(data->clp,
7695                         &data->args->la_seq_args,
7696                         &data->res->lr_seq_res,
7697                         task);
7698         dprintk("<-- %s\n", __func__);
7699 }
7700
7701 /*
7702  * Called from nfs4_state_manager thread for session setup, so don't recover
7703  * from sequence operation or clientid errors.
7704  */
7705 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7706 {
7707         struct nfs4_get_lease_time_data *data =
7708                         (struct nfs4_get_lease_time_data *)calldata;
7709
7710         dprintk("--> %s\n", __func__);
7711         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7712                 return;
7713         switch (task->tk_status) {
7714         case -NFS4ERR_DELAY:
7715         case -NFS4ERR_GRACE:
7716                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7717                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7718                 task->tk_status = 0;
7719                 /* fall through */
7720         case -NFS4ERR_RETRY_UNCACHED_REP:
7721                 rpc_restart_call_prepare(task);
7722                 return;
7723         }
7724         dprintk("<-- %s\n", __func__);
7725 }
7726
7727 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7728         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7729         .rpc_call_done = nfs4_get_lease_time_done,
7730 };
7731
7732 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7733 {
7734         struct rpc_task *task;
7735         struct nfs4_get_lease_time_args args;
7736         struct nfs4_get_lease_time_res res = {
7737                 .lr_fsinfo = fsinfo,
7738         };
7739         struct nfs4_get_lease_time_data data = {
7740                 .args = &args,
7741                 .res = &res,
7742                 .clp = clp,
7743         };
7744         struct rpc_message msg = {
7745                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7746                 .rpc_argp = &args,
7747                 .rpc_resp = &res,
7748         };
7749         struct rpc_task_setup task_setup = {
7750                 .rpc_client = clp->cl_rpcclient,
7751                 .rpc_message = &msg,
7752                 .callback_ops = &nfs4_get_lease_time_ops,
7753                 .callback_data = &data,
7754                 .flags = RPC_TASK_TIMEOUT,
7755         };
7756         int status;
7757
7758         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7759         nfs4_set_sequence_privileged(&args.la_seq_args);
7760         task = rpc_run_task(&task_setup);
7761
7762         if (IS_ERR(task))
7763                 return PTR_ERR(task);
7764
7765         status = task->tk_status;
7766         rpc_put_task(task);
7767         return status;
7768 }
7769
7770 /*
7771  * Initialize the values to be used by the client in CREATE_SESSION
7772  * If nfs4_init_session set the fore channel request and response sizes,
7773  * use them.
7774  *
7775  * Set the back channel max_resp_sz_cached to zero to force the client to
7776  * always set csa_cachethis to FALSE because the current implementation
7777  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7778  */
7779 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7780                                     struct rpc_clnt *clnt)
7781 {
7782         unsigned int max_rqst_sz, max_resp_sz;
7783         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7784
7785         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7786         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7787
7788         /* Fore channel attributes */
7789         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7790         args->fc_attrs.max_resp_sz = max_resp_sz;
7791         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7792         args->fc_attrs.max_reqs = max_session_slots;
7793
7794         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7795                 "max_ops=%u max_reqs=%u\n",
7796                 __func__,
7797                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7798                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7799
7800         /* Back channel attributes */
7801         args->bc_attrs.max_rqst_sz = max_bc_payload;
7802         args->bc_attrs.max_resp_sz = max_bc_payload;
7803         args->bc_attrs.max_resp_sz_cached = 0;
7804         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7805         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7806
7807         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7808                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7809                 __func__,
7810                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7811                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7812                 args->bc_attrs.max_reqs);
7813 }
7814
7815 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7816                 struct nfs41_create_session_res *res)
7817 {
7818         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7819         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7820
7821         if (rcvd->max_resp_sz > sent->max_resp_sz)
7822                 return -EINVAL;
7823         /*
7824          * Our requested max_ops is the minimum we need; we're not
7825          * prepared to break up compounds into smaller pieces than that.
7826          * So, no point even trying to continue if the server won't
7827          * cooperate:
7828          */
7829         if (rcvd->max_ops < sent->max_ops)
7830                 return -EINVAL;
7831         if (rcvd->max_reqs == 0)
7832                 return -EINVAL;
7833         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7834                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7835         return 0;
7836 }
7837
7838 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7839                 struct nfs41_create_session_res *res)
7840 {
7841         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7842         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7843
7844         if (!(res->flags & SESSION4_BACK_CHAN))
7845                 goto out;
7846         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7847                 return -EINVAL;
7848         if (rcvd->max_resp_sz < sent->max_resp_sz)
7849                 return -EINVAL;
7850         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7851                 return -EINVAL;
7852         if (rcvd->max_ops > sent->max_ops)
7853                 return -EINVAL;
7854         if (rcvd->max_reqs > sent->max_reqs)
7855                 return -EINVAL;
7856 out:
7857         return 0;
7858 }
7859
7860 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7861                                      struct nfs41_create_session_res *res)
7862 {
7863         int ret;
7864
7865         ret = nfs4_verify_fore_channel_attrs(args, res);
7866         if (ret)
7867                 return ret;
7868         return nfs4_verify_back_channel_attrs(args, res);
7869 }
7870
7871 static void nfs4_update_session(struct nfs4_session *session,
7872                 struct nfs41_create_session_res *res)
7873 {
7874         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7875         /* Mark client id and session as being confirmed */
7876         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7877         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7878         session->flags = res->flags;
7879         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7880         if (res->flags & SESSION4_BACK_CHAN)
7881                 memcpy(&session->bc_attrs, &res->bc_attrs,
7882                                 sizeof(session->bc_attrs));
7883 }
7884
7885 static int _nfs4_proc_create_session(struct nfs_client *clp,
7886                 struct rpc_cred *cred)
7887 {
7888         struct nfs4_session *session = clp->cl_session;
7889         struct nfs41_create_session_args args = {
7890                 .client = clp,
7891                 .clientid = clp->cl_clientid,
7892                 .seqid = clp->cl_seqid,
7893                 .cb_program = NFS4_CALLBACK,
7894         };
7895         struct nfs41_create_session_res res;
7896
7897         struct rpc_message msg = {
7898                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7899                 .rpc_argp = &args,
7900                 .rpc_resp = &res,
7901                 .rpc_cred = cred,
7902         };
7903         int status;
7904
7905         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7906         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7907
7908         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7909         trace_nfs4_create_session(clp, status);
7910
7911         switch (status) {
7912         case -NFS4ERR_STALE_CLIENTID:
7913         case -NFS4ERR_DELAY:
7914         case -ETIMEDOUT:
7915         case -EACCES:
7916         case -EAGAIN:
7917                 goto out;
7918         };
7919
7920         clp->cl_seqid++;
7921         if (!status) {
7922                 /* Verify the session's negotiated channel_attrs values */
7923                 status = nfs4_verify_channel_attrs(&args, &res);
7924                 /* Increment the clientid slot sequence id */
7925                 if (status)
7926                         goto out;
7927                 nfs4_update_session(session, &res);
7928         }
7929 out:
7930         return status;
7931 }
7932
7933 /*
7934  * Issues a CREATE_SESSION operation to the server.
7935  * It is the responsibility of the caller to verify the session is
7936  * expired before calling this routine.
7937  */
7938 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7939 {
7940         int status;
7941         unsigned *ptr;
7942         struct nfs4_session *session = clp->cl_session;
7943
7944         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7945
7946         status = _nfs4_proc_create_session(clp, cred);
7947         if (status)
7948                 goto out;
7949
7950         /* Init or reset the session slot tables */
7951         status = nfs4_setup_session_slot_tables(session);
7952         dprintk("slot table setup returned %d\n", status);
7953         if (status)
7954                 goto out;
7955
7956         ptr = (unsigned *)&session->sess_id.data[0];
7957         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7958                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7959 out:
7960         dprintk("<-- %s\n", __func__);
7961         return status;
7962 }
7963
7964 /*
7965  * Issue the over-the-wire RPC DESTROY_SESSION.
7966  * The caller must serialize access to this routine.
7967  */
7968 int nfs4_proc_destroy_session(struct nfs4_session *session,
7969                 struct rpc_cred *cred)
7970 {
7971         struct rpc_message msg = {
7972                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7973                 .rpc_argp = session,
7974                 .rpc_cred = cred,
7975         };
7976         int status = 0;
7977
7978         dprintk("--> nfs4_proc_destroy_session\n");
7979
7980         /* session is still being setup */
7981         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7982                 return 0;
7983
7984         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7985         trace_nfs4_destroy_session(session->clp, status);
7986
7987         if (status)
7988                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7989                         "Session has been destroyed regardless...\n", status);
7990
7991         dprintk("<-- nfs4_proc_destroy_session\n");
7992         return status;
7993 }
7994
7995 /*
7996  * Renew the cl_session lease.
7997  */
7998 struct nfs4_sequence_data {
7999         struct nfs_client *clp;
8000         struct nfs4_sequence_args args;
8001         struct nfs4_sequence_res res;
8002 };
8003
8004 static void nfs41_sequence_release(void *data)
8005 {
8006         struct nfs4_sequence_data *calldata = data;
8007         struct nfs_client *clp = calldata->clp;
8008
8009         if (atomic_read(&clp->cl_count) > 1)
8010                 nfs4_schedule_state_renewal(clp);
8011         nfs_put_client(clp);
8012         kfree(calldata);
8013 }
8014
8015 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8016 {
8017         switch(task->tk_status) {
8018         case -NFS4ERR_DELAY:
8019                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8020                 return -EAGAIN;
8021         default:
8022                 nfs4_schedule_lease_recovery(clp);
8023         }
8024         return 0;
8025 }
8026
8027 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8028 {
8029         struct nfs4_sequence_data *calldata = data;
8030         struct nfs_client *clp = calldata->clp;
8031
8032         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8033                 return;
8034
8035         trace_nfs4_sequence(clp, task->tk_status);
8036         if (task->tk_status < 0) {
8037                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8038                 if (atomic_read(&clp->cl_count) == 1)
8039                         goto out;
8040
8041                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8042                         rpc_restart_call_prepare(task);
8043                         return;
8044                 }
8045         }
8046         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8047 out:
8048         dprintk("<-- %s\n", __func__);
8049 }
8050
8051 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8052 {
8053         struct nfs4_sequence_data *calldata = data;
8054         struct nfs_client *clp = calldata->clp;
8055         struct nfs4_sequence_args *args;
8056         struct nfs4_sequence_res *res;
8057
8058         args = task->tk_msg.rpc_argp;
8059         res = task->tk_msg.rpc_resp;
8060
8061         nfs4_setup_sequence(clp, args, res, task);
8062 }
8063
8064 static const struct rpc_call_ops nfs41_sequence_ops = {
8065         .rpc_call_done = nfs41_sequence_call_done,
8066         .rpc_call_prepare = nfs41_sequence_prepare,
8067         .rpc_release = nfs41_sequence_release,
8068 };
8069
8070 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8071                 struct rpc_cred *cred,
8072                 bool is_privileged)
8073 {
8074         struct nfs4_sequence_data *calldata;
8075         struct rpc_message msg = {
8076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8077                 .rpc_cred = cred,
8078         };
8079         struct rpc_task_setup task_setup_data = {
8080                 .rpc_client = clp->cl_rpcclient,
8081                 .rpc_message = &msg,
8082                 .callback_ops = &nfs41_sequence_ops,
8083                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8084         };
8085
8086         if (!atomic_inc_not_zero(&clp->cl_count))
8087                 return ERR_PTR(-EIO);
8088         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8089         if (calldata == NULL) {
8090                 nfs_put_client(clp);
8091                 return ERR_PTR(-ENOMEM);
8092         }
8093         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8094         if (is_privileged)
8095                 nfs4_set_sequence_privileged(&calldata->args);
8096         msg.rpc_argp = &calldata->args;
8097         msg.rpc_resp = &calldata->res;
8098         calldata->clp = clp;
8099         task_setup_data.callback_data = calldata;
8100
8101         return rpc_run_task(&task_setup_data);
8102 }
8103
8104 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8105 {
8106         struct rpc_task *task;
8107         int ret = 0;
8108
8109         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8110                 return -EAGAIN;
8111         task = _nfs41_proc_sequence(clp, cred, false);
8112         if (IS_ERR(task))
8113                 ret = PTR_ERR(task);
8114         else
8115                 rpc_put_task_async(task);
8116         dprintk("<-- %s status=%d\n", __func__, ret);
8117         return ret;
8118 }
8119
8120 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8121 {
8122         struct rpc_task *task;
8123         int ret;
8124
8125         task = _nfs41_proc_sequence(clp, cred, true);
8126         if (IS_ERR(task)) {
8127                 ret = PTR_ERR(task);
8128                 goto out;
8129         }
8130         ret = rpc_wait_for_completion_task(task);
8131         if (!ret)
8132                 ret = task->tk_status;
8133         rpc_put_task(task);
8134 out:
8135         dprintk("<-- %s status=%d\n", __func__, ret);
8136         return ret;
8137 }
8138
8139 struct nfs4_reclaim_complete_data {
8140         struct nfs_client *clp;
8141         struct nfs41_reclaim_complete_args arg;
8142         struct nfs41_reclaim_complete_res res;
8143 };
8144
8145 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8146 {
8147         struct nfs4_reclaim_complete_data *calldata = data;
8148
8149         nfs4_setup_sequence(calldata->clp,
8150                         &calldata->arg.seq_args,
8151                         &calldata->res.seq_res,
8152                         task);
8153 }
8154
8155 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8156 {
8157         switch(task->tk_status) {
8158         case 0:
8159         case -NFS4ERR_COMPLETE_ALREADY:
8160         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8161                 break;
8162         case -NFS4ERR_DELAY:
8163                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8164                 /* fall through */
8165         case -NFS4ERR_RETRY_UNCACHED_REP:
8166                 return -EAGAIN;
8167         case -NFS4ERR_BADSESSION:
8168         case -NFS4ERR_DEADSESSION:
8169         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8170                 nfs4_schedule_session_recovery(clp->cl_session,
8171                                 task->tk_status);
8172                 break;
8173         default:
8174                 nfs4_schedule_lease_recovery(clp);
8175         }
8176         return 0;
8177 }
8178
8179 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8180 {
8181         struct nfs4_reclaim_complete_data *calldata = data;
8182         struct nfs_client *clp = calldata->clp;
8183         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8184
8185         dprintk("--> %s\n", __func__);
8186         if (!nfs41_sequence_done(task, res))
8187                 return;
8188
8189         trace_nfs4_reclaim_complete(clp, task->tk_status);
8190         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8191                 rpc_restart_call_prepare(task);
8192                 return;
8193         }
8194         dprintk("<-- %s\n", __func__);
8195 }
8196
8197 static void nfs4_free_reclaim_complete_data(void *data)
8198 {
8199         struct nfs4_reclaim_complete_data *calldata = data;
8200
8201         kfree(calldata);
8202 }
8203
8204 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8205         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8206         .rpc_call_done = nfs4_reclaim_complete_done,
8207         .rpc_release = nfs4_free_reclaim_complete_data,
8208 };
8209
8210 /*
8211  * Issue a global reclaim complete.
8212  */
8213 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8214                 struct rpc_cred *cred)
8215 {
8216         struct nfs4_reclaim_complete_data *calldata;
8217         struct rpc_task *task;
8218         struct rpc_message msg = {
8219                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8220                 .rpc_cred = cred,
8221         };
8222         struct rpc_task_setup task_setup_data = {
8223                 .rpc_client = clp->cl_rpcclient,
8224                 .rpc_message = &msg,
8225                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8226                 .flags = RPC_TASK_ASYNC,
8227         };
8228         int status = -ENOMEM;
8229
8230         dprintk("--> %s\n", __func__);
8231         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8232         if (calldata == NULL)
8233                 goto out;
8234         calldata->clp = clp;
8235         calldata->arg.one_fs = 0;
8236
8237         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8238         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8239         msg.rpc_argp = &calldata->arg;
8240         msg.rpc_resp = &calldata->res;
8241         task_setup_data.callback_data = calldata;
8242         task = rpc_run_task(&task_setup_data);
8243         if (IS_ERR(task)) {
8244                 status = PTR_ERR(task);
8245                 goto out;
8246         }
8247         status = rpc_wait_for_completion_task(task);
8248         if (status == 0)
8249                 status = task->tk_status;
8250         rpc_put_task(task);
8251 out:
8252         dprintk("<-- %s status=%d\n", __func__, status);
8253         return status;
8254 }
8255
8256 static void
8257 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8258 {
8259         struct nfs4_layoutget *lgp = calldata;
8260         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8261
8262         dprintk("--> %s\n", __func__);
8263         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8264                                 &lgp->res.seq_res, task);
8265         dprintk("<-- %s\n", __func__);
8266 }
8267
8268 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8269 {
8270         struct nfs4_layoutget *lgp = calldata;
8271
8272         dprintk("--> %s\n", __func__);
8273         nfs41_sequence_process(task, &lgp->res.seq_res);
8274         dprintk("<-- %s\n", __func__);
8275 }
8276
8277 static int
8278 nfs4_layoutget_handle_exception(struct rpc_task *task,
8279                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8280 {
8281         struct inode *inode = lgp->args.inode;
8282         struct nfs_server *server = NFS_SERVER(inode);
8283         struct pnfs_layout_hdr *lo;
8284         int nfs4err = task->tk_status;
8285         int err, status = 0;
8286         LIST_HEAD(head);
8287
8288         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8289
8290         switch (nfs4err) {
8291         case 0:
8292                 goto out;
8293
8294         /*
8295          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8296          * on the file. set tk_status to -ENODATA to tell upper layer to
8297          * retry go inband.
8298          */
8299         case -NFS4ERR_LAYOUTUNAVAILABLE:
8300                 status = -ENODATA;
8301                 goto out;
8302         /*
8303          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8304          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8305          */
8306         case -NFS4ERR_BADLAYOUT:
8307                 status = -EOVERFLOW;
8308                 goto out;
8309         /*
8310          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8311          * (or clients) writing to the same RAID stripe except when
8312          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8313          *
8314          * Treat it like we would RECALLCONFLICT -- we retry for a little
8315          * while, and then eventually give up.
8316          */
8317         case -NFS4ERR_LAYOUTTRYLATER:
8318                 if (lgp->args.minlength == 0) {
8319                         status = -EOVERFLOW;
8320                         goto out;
8321                 }
8322                 status = -EBUSY;
8323                 break;
8324         case -NFS4ERR_RECALLCONFLICT:
8325                 status = -ERECALLCONFLICT;
8326                 break;
8327         case -NFS4ERR_DELEG_REVOKED:
8328         case -NFS4ERR_ADMIN_REVOKED:
8329         case -NFS4ERR_EXPIRED:
8330         case -NFS4ERR_BAD_STATEID:
8331                 exception->timeout = 0;
8332                 spin_lock(&inode->i_lock);
8333                 lo = NFS_I(inode)->layout;
8334                 /* If the open stateid was bad, then recover it. */
8335                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8336                     nfs4_stateid_match_other(&lgp->args.stateid,
8337                                         &lgp->args.ctx->state->stateid)) {
8338                         spin_unlock(&inode->i_lock);
8339                         exception->state = lgp->args.ctx->state;
8340                         exception->stateid = &lgp->args.stateid;
8341                         break;
8342                 }
8343
8344                 /*
8345                  * Mark the bad layout state as invalid, then retry
8346                  */
8347                 pnfs_mark_layout_stateid_invalid(lo, &head);
8348                 spin_unlock(&inode->i_lock);
8349                 nfs_commit_inode(inode, 0);
8350                 pnfs_free_lseg_list(&head);
8351                 status = -EAGAIN;
8352                 goto out;
8353         }
8354
8355         nfs4_sequence_free_slot(&lgp->res.seq_res);
8356         err = nfs4_handle_exception(server, nfs4err, exception);
8357         if (!status) {
8358                 if (exception->retry)
8359                         status = -EAGAIN;
8360                 else
8361                         status = err;
8362         }
8363 out:
8364         dprintk("<-- %s\n", __func__);
8365         return status;
8366 }
8367
8368 static size_t max_response_pages(struct nfs_server *server)
8369 {
8370         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8371         return nfs_page_array_len(0, max_resp_sz);
8372 }
8373
8374 static void nfs4_free_pages(struct page **pages, size_t size)
8375 {
8376         int i;
8377
8378         if (!pages)
8379                 return;
8380
8381         for (i = 0; i < size; i++) {
8382                 if (!pages[i])
8383                         break;
8384                 __free_page(pages[i]);
8385         }
8386         kfree(pages);
8387 }
8388
8389 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8390 {
8391         struct page **pages;
8392         int i;
8393
8394         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8395         if (!pages) {
8396                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8397                 return NULL;
8398         }
8399
8400         for (i = 0; i < size; i++) {
8401                 pages[i] = alloc_page(gfp_flags);
8402                 if (!pages[i]) {
8403                         dprintk("%s: failed to allocate page\n", __func__);
8404                         nfs4_free_pages(pages, size);
8405                         return NULL;
8406                 }
8407         }
8408
8409         return pages;
8410 }
8411
8412 static void nfs4_layoutget_release(void *calldata)
8413 {
8414         struct nfs4_layoutget *lgp = calldata;
8415         struct inode *inode = lgp->args.inode;
8416         struct nfs_server *server = NFS_SERVER(inode);
8417         size_t max_pages = max_response_pages(server);
8418
8419         dprintk("--> %s\n", __func__);
8420         nfs4_sequence_free_slot(&lgp->res.seq_res);
8421         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8422         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8423         put_nfs_open_context(lgp->args.ctx);
8424         kfree(calldata);
8425         dprintk("<-- %s\n", __func__);
8426 }
8427
8428 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8429         .rpc_call_prepare = nfs4_layoutget_prepare,
8430         .rpc_call_done = nfs4_layoutget_done,
8431         .rpc_release = nfs4_layoutget_release,
8432 };
8433
8434 struct pnfs_layout_segment *
8435 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8436 {
8437         struct inode *inode = lgp->args.inode;
8438         struct nfs_server *server = NFS_SERVER(inode);
8439         size_t max_pages = max_response_pages(server);
8440         struct rpc_task *task;
8441         struct rpc_message msg = {
8442                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8443                 .rpc_argp = &lgp->args,
8444                 .rpc_resp = &lgp->res,
8445                 .rpc_cred = lgp->cred,
8446         };
8447         struct rpc_task_setup task_setup_data = {
8448                 .rpc_client = server->client,
8449                 .rpc_message = &msg,
8450                 .callback_ops = &nfs4_layoutget_call_ops,
8451                 .callback_data = lgp,
8452                 .flags = RPC_TASK_ASYNC,
8453         };
8454         struct pnfs_layout_segment *lseg = NULL;
8455         struct nfs4_exception exception = {
8456                 .inode = inode,
8457                 .timeout = *timeout,
8458         };
8459         int status = 0;
8460
8461         dprintk("--> %s\n", __func__);
8462
8463         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8464         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8465
8466         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8467         if (!lgp->args.layout.pages) {
8468                 nfs4_layoutget_release(lgp);
8469                 return ERR_PTR(-ENOMEM);
8470         }
8471         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8472
8473         lgp->res.layoutp = &lgp->args.layout;
8474         lgp->res.seq_res.sr_slot = NULL;
8475         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8476
8477         task = rpc_run_task(&task_setup_data);
8478         if (IS_ERR(task))
8479                 return ERR_CAST(task);
8480         status = rpc_wait_for_completion_task(task);
8481         if (status == 0) {
8482                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8483                 *timeout = exception.timeout;
8484         }
8485
8486         trace_nfs4_layoutget(lgp->args.ctx,
8487                         &lgp->args.range,
8488                         &lgp->res.range,
8489                         &lgp->res.stateid,
8490                         status);
8491
8492         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8493         if (status == 0 && lgp->res.layoutp->len)
8494                 lseg = pnfs_layout_process(lgp);
8495         rpc_put_task(task);
8496         dprintk("<-- %s status=%d\n", __func__, status);
8497         if (status)
8498                 return ERR_PTR(status);
8499         return lseg;
8500 }
8501
8502 static void
8503 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8504 {
8505         struct nfs4_layoutreturn *lrp = calldata;
8506
8507         dprintk("--> %s\n", __func__);
8508         nfs4_setup_sequence(lrp->clp,
8509                         &lrp->args.seq_args,
8510                         &lrp->res.seq_res,
8511                         task);
8512 }
8513
8514 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8515 {
8516         struct nfs4_layoutreturn *lrp = calldata;
8517         struct nfs_server *server;
8518
8519         dprintk("--> %s\n", __func__);
8520
8521         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8522                 return;
8523
8524         server = NFS_SERVER(lrp->args.inode);
8525         switch (task->tk_status) {
8526         default:
8527                 task->tk_status = 0;
8528         case 0:
8529                 break;
8530         case -NFS4ERR_DELAY:
8531                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8532                         break;
8533                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8534                 rpc_restart_call_prepare(task);
8535                 return;
8536         }
8537         dprintk("<-- %s\n", __func__);
8538 }
8539
8540 static void nfs4_layoutreturn_release(void *calldata)
8541 {
8542         struct nfs4_layoutreturn *lrp = calldata;
8543         struct pnfs_layout_hdr *lo = lrp->args.layout;
8544
8545         dprintk("--> %s\n", __func__);
8546         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8547                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8548         nfs4_sequence_free_slot(&lrp->res.seq_res);
8549         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8550                 lrp->ld_private.ops->free(&lrp->ld_private);
8551         pnfs_put_layout_hdr(lrp->args.layout);
8552         nfs_iput_and_deactive(lrp->inode);
8553         kfree(calldata);
8554         dprintk("<-- %s\n", __func__);
8555 }
8556
8557 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8558         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8559         .rpc_call_done = nfs4_layoutreturn_done,
8560         .rpc_release = nfs4_layoutreturn_release,
8561 };
8562
8563 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8564 {
8565         struct rpc_task *task;
8566         struct rpc_message msg = {
8567                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8568                 .rpc_argp = &lrp->args,
8569                 .rpc_resp = &lrp->res,
8570                 .rpc_cred = lrp->cred,
8571         };
8572         struct rpc_task_setup task_setup_data = {
8573                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8574                 .rpc_message = &msg,
8575                 .callback_ops = &nfs4_layoutreturn_call_ops,
8576                 .callback_data = lrp,
8577         };
8578         int status = 0;
8579
8580         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8581                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8582                         &task_setup_data.rpc_client, &msg);
8583
8584         dprintk("--> %s\n", __func__);
8585         if (!sync) {
8586                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8587                 if (!lrp->inode) {
8588                         nfs4_layoutreturn_release(lrp);
8589                         return -EAGAIN;
8590                 }
8591                 task_setup_data.flags |= RPC_TASK_ASYNC;
8592         }
8593         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8594         task = rpc_run_task(&task_setup_data);
8595         if (IS_ERR(task))
8596                 return PTR_ERR(task);
8597         if (sync)
8598                 status = task->tk_status;
8599         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8600         dprintk("<-- %s status=%d\n", __func__, status);
8601         rpc_put_task(task);
8602         return status;
8603 }
8604
8605 static int
8606 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8607                 struct pnfs_device *pdev,
8608                 struct rpc_cred *cred)
8609 {
8610         struct nfs4_getdeviceinfo_args args = {
8611                 .pdev = pdev,
8612                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8613                         NOTIFY_DEVICEID4_DELETE,
8614         };
8615         struct nfs4_getdeviceinfo_res res = {
8616                 .pdev = pdev,
8617         };
8618         struct rpc_message msg = {
8619                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8620                 .rpc_argp = &args,
8621                 .rpc_resp = &res,
8622                 .rpc_cred = cred,
8623         };
8624         int status;
8625
8626         dprintk("--> %s\n", __func__);
8627         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8628         if (res.notification & ~args.notify_types)
8629                 dprintk("%s: unsupported notification\n", __func__);
8630         if (res.notification != args.notify_types)
8631                 pdev->nocache = 1;
8632
8633         dprintk("<-- %s status=%d\n", __func__, status);
8634
8635         return status;
8636 }
8637
8638 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8639                 struct pnfs_device *pdev,
8640                 struct rpc_cred *cred)
8641 {
8642         struct nfs4_exception exception = { };
8643         int err;
8644
8645         do {
8646                 err = nfs4_handle_exception(server,
8647                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8648                                         &exception);
8649         } while (exception.retry);
8650         return err;
8651 }
8652 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8653
8654 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8655 {
8656         struct nfs4_layoutcommit_data *data = calldata;
8657         struct nfs_server *server = NFS_SERVER(data->args.inode);
8658
8659         nfs4_setup_sequence(server->nfs_client,
8660                         &data->args.seq_args,
8661                         &data->res.seq_res,
8662                         task);
8663 }
8664
8665 static void
8666 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8667 {
8668         struct nfs4_layoutcommit_data *data = calldata;
8669         struct nfs_server *server = NFS_SERVER(data->args.inode);
8670
8671         if (!nfs41_sequence_done(task, &data->res.seq_res))
8672                 return;
8673
8674         switch (task->tk_status) { /* Just ignore these failures */
8675         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8676         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8677         case -NFS4ERR_BADLAYOUT:     /* no layout */
8678         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8679                 task->tk_status = 0;
8680         case 0:
8681                 break;
8682         default:
8683                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8684                         rpc_restart_call_prepare(task);
8685                         return;
8686                 }
8687         }
8688 }
8689
8690 static void nfs4_layoutcommit_release(void *calldata)
8691 {
8692         struct nfs4_layoutcommit_data *data = calldata;
8693
8694         pnfs_cleanup_layoutcommit(data);
8695         nfs_post_op_update_inode_force_wcc(data->args.inode,
8696                                            data->res.fattr);
8697         put_rpccred(data->cred);
8698         nfs_iput_and_deactive(data->inode);
8699         kfree(data);
8700 }
8701
8702 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8703         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8704         .rpc_call_done = nfs4_layoutcommit_done,
8705         .rpc_release = nfs4_layoutcommit_release,
8706 };
8707
8708 int
8709 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8710 {
8711         struct rpc_message msg = {
8712                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8713                 .rpc_argp = &data->args,
8714                 .rpc_resp = &data->res,
8715                 .rpc_cred = data->cred,
8716         };
8717         struct rpc_task_setup task_setup_data = {
8718                 .task = &data->task,
8719                 .rpc_client = NFS_CLIENT(data->args.inode),
8720                 .rpc_message = &msg,
8721                 .callback_ops = &nfs4_layoutcommit_ops,
8722                 .callback_data = data,
8723         };
8724         struct rpc_task *task;
8725         int status = 0;
8726
8727         dprintk("NFS: initiating layoutcommit call. sync %d "
8728                 "lbw: %llu inode %lu\n", sync,
8729                 data->args.lastbytewritten,
8730                 data->args.inode->i_ino);
8731
8732         if (!sync) {
8733                 data->inode = nfs_igrab_and_active(data->args.inode);
8734                 if (data->inode == NULL) {
8735                         nfs4_layoutcommit_release(data);
8736                         return -EAGAIN;
8737                 }
8738                 task_setup_data.flags = RPC_TASK_ASYNC;
8739         }
8740         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8741         task = rpc_run_task(&task_setup_data);
8742         if (IS_ERR(task))
8743                 return PTR_ERR(task);
8744         if (sync)
8745                 status = task->tk_status;
8746         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8747         dprintk("%s: status %d\n", __func__, status);
8748         rpc_put_task(task);
8749         return status;
8750 }
8751
8752 /**
8753  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8754  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8755  */
8756 static int
8757 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8758                     struct nfs_fsinfo *info,
8759                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8760 {
8761         struct nfs41_secinfo_no_name_args args = {
8762                 .style = SECINFO_STYLE_CURRENT_FH,
8763         };
8764         struct nfs4_secinfo_res res = {
8765                 .flavors = flavors,
8766         };
8767         struct rpc_message msg = {
8768                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8769                 .rpc_argp = &args,
8770                 .rpc_resp = &res,
8771         };
8772         struct rpc_clnt *clnt = server->client;
8773         struct rpc_cred *cred = NULL;
8774         int status;
8775
8776         if (use_integrity) {
8777                 clnt = server->nfs_client->cl_rpcclient;
8778                 cred = nfs4_get_clid_cred(server->nfs_client);
8779                 msg.rpc_cred = cred;
8780         }
8781
8782         dprintk("--> %s\n", __func__);
8783         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8784                                 &res.seq_res, 0);
8785         dprintk("<-- %s status=%d\n", __func__, status);
8786
8787         if (cred)
8788                 put_rpccred(cred);
8789
8790         return status;
8791 }
8792
8793 static int
8794 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8795                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8796 {
8797         struct nfs4_exception exception = { };
8798         int err;
8799         do {
8800                 /* first try using integrity protection */
8801                 err = -NFS4ERR_WRONGSEC;
8802
8803                 /* try to use integrity protection with machine cred */
8804                 if (_nfs4_is_integrity_protected(server->nfs_client))
8805                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8806                                                           flavors, true);
8807
8808                 /*
8809                  * if unable to use integrity protection, or SECINFO with
8810                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8811                  * disallowed by spec, but exists in deployed servers) use
8812                  * the current filesystem's rpc_client and the user cred.
8813                  */
8814                 if (err == -NFS4ERR_WRONGSEC)
8815                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8816                                                           flavors, false);
8817
8818                 switch (err) {
8819                 case 0:
8820                 case -NFS4ERR_WRONGSEC:
8821                 case -ENOTSUPP:
8822                         goto out;
8823                 default:
8824                         err = nfs4_handle_exception(server, err, &exception);
8825                 }
8826         } while (exception.retry);
8827 out:
8828         return err;
8829 }
8830
8831 static int
8832 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8833                     struct nfs_fsinfo *info)
8834 {
8835         int err;
8836         struct page *page;
8837         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8838         struct nfs4_secinfo_flavors *flavors;
8839         struct nfs4_secinfo4 *secinfo;
8840         int i;
8841
8842         page = alloc_page(GFP_KERNEL);
8843         if (!page) {
8844                 err = -ENOMEM;
8845                 goto out;
8846         }
8847
8848         flavors = page_address(page);
8849         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8850
8851         /*
8852          * Fall back on "guess and check" method if
8853          * the server doesn't support SECINFO_NO_NAME
8854          */
8855         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8856                 err = nfs4_find_root_sec(server, fhandle, info);
8857                 goto out_freepage;
8858         }
8859         if (err)
8860                 goto out_freepage;
8861
8862         for (i = 0; i < flavors->num_flavors; i++) {
8863                 secinfo = &flavors->flavors[i];
8864
8865                 switch (secinfo->flavor) {
8866                 case RPC_AUTH_NULL:
8867                 case RPC_AUTH_UNIX:
8868                 case RPC_AUTH_GSS:
8869                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8870                                         &secinfo->flavor_info);
8871                         break;
8872                 default:
8873                         flavor = RPC_AUTH_MAXFLAVOR;
8874                         break;
8875                 }
8876
8877                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8878                         flavor = RPC_AUTH_MAXFLAVOR;
8879
8880                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8881                         err = nfs4_lookup_root_sec(server, fhandle,
8882                                                    info, flavor);
8883                         if (!err)
8884                                 break;
8885                 }
8886         }
8887
8888         if (flavor == RPC_AUTH_MAXFLAVOR)
8889                 err = -EPERM;
8890
8891 out_freepage:
8892         put_page(page);
8893         if (err == -EACCES)
8894                 return -EPERM;
8895 out:
8896         return err;
8897 }
8898
8899 static int _nfs41_test_stateid(struct nfs_server *server,
8900                 nfs4_stateid *stateid,
8901                 struct rpc_cred *cred)
8902 {
8903         int status;
8904         struct nfs41_test_stateid_args args = {
8905                 .stateid = stateid,
8906         };
8907         struct nfs41_test_stateid_res res;
8908         struct rpc_message msg = {
8909                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8910                 .rpc_argp = &args,
8911                 .rpc_resp = &res,
8912                 .rpc_cred = cred,
8913         };
8914         struct rpc_clnt *rpc_client = server->client;
8915
8916         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8917                 &rpc_client, &msg);
8918
8919         dprintk("NFS call  test_stateid %p\n", stateid);
8920         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8921         nfs4_set_sequence_privileged(&args.seq_args);
8922         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8923                         &args.seq_args, &res.seq_res);
8924         if (status != NFS_OK) {
8925                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8926                 return status;
8927         }
8928         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8929         return -res.status;
8930 }
8931
8932 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8933                 int err, struct nfs4_exception *exception)
8934 {
8935         exception->retry = 0;
8936         switch(err) {
8937         case -NFS4ERR_DELAY:
8938         case -NFS4ERR_RETRY_UNCACHED_REP:
8939                 nfs4_handle_exception(server, err, exception);
8940                 break;
8941         case -NFS4ERR_BADSESSION:
8942         case -NFS4ERR_BADSLOT:
8943         case -NFS4ERR_BAD_HIGH_SLOT:
8944         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8945         case -NFS4ERR_DEADSESSION:
8946                 nfs4_do_handle_exception(server, err, exception);
8947         }
8948 }
8949
8950 /**
8951  * nfs41_test_stateid - perform a TEST_STATEID operation
8952  *
8953  * @server: server / transport on which to perform the operation
8954  * @stateid: state ID to test
8955  * @cred: credential
8956  *
8957  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8958  * Otherwise a negative NFS4ERR value is returned if the operation
8959  * failed or the state ID is not currently valid.
8960  */
8961 static int nfs41_test_stateid(struct nfs_server *server,
8962                 nfs4_stateid *stateid,
8963                 struct rpc_cred *cred)
8964 {
8965         struct nfs4_exception exception = { };
8966         int err;
8967         do {
8968                 err = _nfs41_test_stateid(server, stateid, cred);
8969                 nfs4_handle_delay_or_session_error(server, err, &exception);
8970         } while (exception.retry);
8971         return err;
8972 }
8973
8974 struct nfs_free_stateid_data {
8975         struct nfs_server *server;
8976         struct nfs41_free_stateid_args args;
8977         struct nfs41_free_stateid_res res;
8978 };
8979
8980 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8981 {
8982         struct nfs_free_stateid_data *data = calldata;
8983         nfs4_setup_sequence(data->server->nfs_client,
8984                         &data->args.seq_args,
8985                         &data->res.seq_res,
8986                         task);
8987 }
8988
8989 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8990 {
8991         struct nfs_free_stateid_data *data = calldata;
8992
8993         nfs41_sequence_done(task, &data->res.seq_res);
8994
8995         switch (task->tk_status) {
8996         case -NFS4ERR_DELAY:
8997                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8998                         rpc_restart_call_prepare(task);
8999         }
9000 }
9001
9002 static void nfs41_free_stateid_release(void *calldata)
9003 {
9004         kfree(calldata);
9005 }
9006
9007 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9008         .rpc_call_prepare = nfs41_free_stateid_prepare,
9009         .rpc_call_done = nfs41_free_stateid_done,
9010         .rpc_release = nfs41_free_stateid_release,
9011 };
9012
9013 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9014                 const nfs4_stateid *stateid,
9015                 struct rpc_cred *cred,
9016                 bool privileged)
9017 {
9018         struct rpc_message msg = {
9019                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9020                 .rpc_cred = cred,
9021         };
9022         struct rpc_task_setup task_setup = {
9023                 .rpc_client = server->client,
9024                 .rpc_message = &msg,
9025                 .callback_ops = &nfs41_free_stateid_ops,
9026                 .flags = RPC_TASK_ASYNC,
9027         };
9028         struct nfs_free_stateid_data *data;
9029
9030         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9031                 &task_setup.rpc_client, &msg);
9032
9033         dprintk("NFS call  free_stateid %p\n", stateid);
9034         data = kmalloc(sizeof(*data), GFP_NOFS);
9035         if (!data)
9036                 return ERR_PTR(-ENOMEM);
9037         data->server = server;
9038         nfs4_stateid_copy(&data->args.stateid, stateid);
9039
9040         task_setup.callback_data = data;
9041
9042         msg.rpc_argp = &data->args;
9043         msg.rpc_resp = &data->res;
9044         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9045         if (privileged)
9046                 nfs4_set_sequence_privileged(&data->args.seq_args);
9047
9048         return rpc_run_task(&task_setup);
9049 }
9050
9051 /**
9052  * nfs41_free_stateid - perform a FREE_STATEID operation
9053  *
9054  * @server: server / transport on which to perform the operation
9055  * @stateid: state ID to release
9056  * @cred: credential
9057  * @is_recovery: set to true if this call needs to be privileged
9058  *
9059  * Note: this function is always asynchronous.
9060  */
9061 static int nfs41_free_stateid(struct nfs_server *server,
9062                 const nfs4_stateid *stateid,
9063                 struct rpc_cred *cred,
9064                 bool is_recovery)
9065 {
9066         struct rpc_task *task;
9067
9068         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9069         if (IS_ERR(task))
9070                 return PTR_ERR(task);
9071         rpc_put_task(task);
9072         return 0;
9073 }
9074
9075 static void
9076 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9077 {
9078         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9079
9080         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9081         nfs4_free_lock_state(server, lsp);
9082 }
9083
9084 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9085                 const nfs4_stateid *s2)
9086 {
9087         if (s1->type != s2->type)
9088                 return false;
9089
9090         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9091                 return false;
9092
9093         if (s1->seqid == s2->seqid)
9094                 return true;
9095
9096         return s1->seqid == 0 || s2->seqid == 0;
9097 }
9098
9099 #endif /* CONFIG_NFS_V4_1 */
9100
9101 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9102                 const nfs4_stateid *s2)
9103 {
9104         return nfs4_stateid_match(s1, s2);
9105 }
9106
9107
9108 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9109         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9110         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9111         .recover_open   = nfs4_open_reclaim,
9112         .recover_lock   = nfs4_lock_reclaim,
9113         .establish_clid = nfs4_init_clientid,
9114         .detect_trunking = nfs40_discover_server_trunking,
9115 };
9116
9117 #if defined(CONFIG_NFS_V4_1)
9118 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9119         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9120         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9121         .recover_open   = nfs4_open_reclaim,
9122         .recover_lock   = nfs4_lock_reclaim,
9123         .establish_clid = nfs41_init_clientid,
9124         .reclaim_complete = nfs41_proc_reclaim_complete,
9125         .detect_trunking = nfs41_discover_server_trunking,
9126 };
9127 #endif /* CONFIG_NFS_V4_1 */
9128
9129 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9130         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9131         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9132         .recover_open   = nfs40_open_expired,
9133         .recover_lock   = nfs4_lock_expired,
9134         .establish_clid = nfs4_init_clientid,
9135 };
9136
9137 #if defined(CONFIG_NFS_V4_1)
9138 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9139         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9140         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9141         .recover_open   = nfs41_open_expired,
9142         .recover_lock   = nfs41_lock_expired,
9143         .establish_clid = nfs41_init_clientid,
9144 };
9145 #endif /* CONFIG_NFS_V4_1 */
9146
9147 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9148         .sched_state_renewal = nfs4_proc_async_renew,
9149         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9150         .renew_lease = nfs4_proc_renew,
9151 };
9152
9153 #if defined(CONFIG_NFS_V4_1)
9154 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9155         .sched_state_renewal = nfs41_proc_async_sequence,
9156         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9157         .renew_lease = nfs4_proc_sequence,
9158 };
9159 #endif
9160
9161 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9162         .get_locations = _nfs40_proc_get_locations,
9163         .fsid_present = _nfs40_proc_fsid_present,
9164 };
9165
9166 #if defined(CONFIG_NFS_V4_1)
9167 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9168         .get_locations = _nfs41_proc_get_locations,
9169         .fsid_present = _nfs41_proc_fsid_present,
9170 };
9171 #endif  /* CONFIG_NFS_V4_1 */
9172
9173 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9174         .minor_version = 0,
9175         .init_caps = NFS_CAP_READDIRPLUS
9176                 | NFS_CAP_ATOMIC_OPEN
9177                 | NFS_CAP_POSIX_LOCK,
9178         .init_client = nfs40_init_client,
9179         .shutdown_client = nfs40_shutdown_client,
9180         .match_stateid = nfs4_match_stateid,
9181         .find_root_sec = nfs4_find_root_sec,
9182         .free_lock_state = nfs4_release_lockowner,
9183         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9184         .alloc_seqid = nfs_alloc_seqid,
9185         .call_sync_ops = &nfs40_call_sync_ops,
9186         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9187         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9188         .state_renewal_ops = &nfs40_state_renewal_ops,
9189         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9190 };
9191
9192 #if defined(CONFIG_NFS_V4_1)
9193 static struct nfs_seqid *
9194 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9195 {
9196         return NULL;
9197 }
9198
9199 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9200         .minor_version = 1,
9201         .init_caps = NFS_CAP_READDIRPLUS
9202                 | NFS_CAP_ATOMIC_OPEN
9203                 | NFS_CAP_POSIX_LOCK
9204                 | NFS_CAP_STATEID_NFSV41
9205                 | NFS_CAP_ATOMIC_OPEN_V1,
9206         .init_client = nfs41_init_client,
9207         .shutdown_client = nfs41_shutdown_client,
9208         .match_stateid = nfs41_match_stateid,
9209         .find_root_sec = nfs41_find_root_sec,
9210         .free_lock_state = nfs41_free_lock_state,
9211         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9212         .alloc_seqid = nfs_alloc_no_seqid,
9213         .session_trunk = nfs4_test_session_trunk,
9214         .call_sync_ops = &nfs41_call_sync_ops,
9215         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9216         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9217         .state_renewal_ops = &nfs41_state_renewal_ops,
9218         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9219 };
9220 #endif
9221
9222 #if defined(CONFIG_NFS_V4_2)
9223 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9224         .minor_version = 2,
9225         .init_caps = NFS_CAP_READDIRPLUS
9226                 | NFS_CAP_ATOMIC_OPEN
9227                 | NFS_CAP_POSIX_LOCK
9228                 | NFS_CAP_STATEID_NFSV41
9229                 | NFS_CAP_ATOMIC_OPEN_V1
9230                 | NFS_CAP_ALLOCATE
9231                 | NFS_CAP_COPY
9232                 | NFS_CAP_DEALLOCATE
9233                 | NFS_CAP_SEEK
9234                 | NFS_CAP_LAYOUTSTATS
9235                 | NFS_CAP_CLONE,
9236         .init_client = nfs41_init_client,
9237         .shutdown_client = nfs41_shutdown_client,
9238         .match_stateid = nfs41_match_stateid,
9239         .find_root_sec = nfs41_find_root_sec,
9240         .free_lock_state = nfs41_free_lock_state,
9241         .call_sync_ops = &nfs41_call_sync_ops,
9242         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9243         .alloc_seqid = nfs_alloc_no_seqid,
9244         .session_trunk = nfs4_test_session_trunk,
9245         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9246         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9247         .state_renewal_ops = &nfs41_state_renewal_ops,
9248         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9249 };
9250 #endif
9251
9252 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9253         [0] = &nfs_v4_0_minor_ops,
9254 #if defined(CONFIG_NFS_V4_1)
9255         [1] = &nfs_v4_1_minor_ops,
9256 #endif
9257 #if defined(CONFIG_NFS_V4_2)
9258         [2] = &nfs_v4_2_minor_ops,
9259 #endif
9260 };
9261
9262 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9263 {
9264         ssize_t error, error2;
9265
9266         error = generic_listxattr(dentry, list, size);
9267         if (error < 0)
9268                 return error;
9269         if (list) {
9270                 list += error;
9271                 size -= error;
9272         }
9273
9274         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9275         if (error2 < 0)
9276                 return error2;
9277         return error + error2;
9278 }
9279
9280 static const struct inode_operations nfs4_dir_inode_operations = {
9281         .create         = nfs_create,
9282         .lookup         = nfs_lookup,
9283         .atomic_open    = nfs_atomic_open,
9284         .link           = nfs_link,
9285         .unlink         = nfs_unlink,
9286         .symlink        = nfs_symlink,
9287         .mkdir          = nfs_mkdir,
9288         .rmdir          = nfs_rmdir,
9289         .mknod          = nfs_mknod,
9290         .rename         = nfs_rename,
9291         .permission     = nfs_permission,
9292         .getattr        = nfs_getattr,
9293         .setattr        = nfs_setattr,
9294         .listxattr      = nfs4_listxattr,
9295 };
9296
9297 static const struct inode_operations nfs4_file_inode_operations = {
9298         .permission     = nfs_permission,
9299         .getattr        = nfs_getattr,
9300         .setattr        = nfs_setattr,
9301         .listxattr      = nfs4_listxattr,
9302 };
9303
9304 const struct nfs_rpc_ops nfs_v4_clientops = {
9305         .version        = 4,                    /* protocol version */
9306         .dentry_ops     = &nfs4_dentry_operations,
9307         .dir_inode_ops  = &nfs4_dir_inode_operations,
9308         .file_inode_ops = &nfs4_file_inode_operations,
9309         .file_ops       = &nfs4_file_operations,
9310         .getroot        = nfs4_proc_get_root,
9311         .submount       = nfs4_submount,
9312         .try_mount      = nfs4_try_mount,
9313         .getattr        = nfs4_proc_getattr,
9314         .setattr        = nfs4_proc_setattr,
9315         .lookup         = nfs4_proc_lookup,
9316         .access         = nfs4_proc_access,
9317         .readlink       = nfs4_proc_readlink,
9318         .create         = nfs4_proc_create,
9319         .remove         = nfs4_proc_remove,
9320         .unlink_setup   = nfs4_proc_unlink_setup,
9321         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9322         .unlink_done    = nfs4_proc_unlink_done,
9323         .rename_setup   = nfs4_proc_rename_setup,
9324         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9325         .rename_done    = nfs4_proc_rename_done,
9326         .link           = nfs4_proc_link,
9327         .symlink        = nfs4_proc_symlink,
9328         .mkdir          = nfs4_proc_mkdir,
9329         .rmdir          = nfs4_proc_remove,
9330         .readdir        = nfs4_proc_readdir,
9331         .mknod          = nfs4_proc_mknod,
9332         .statfs         = nfs4_proc_statfs,
9333         .fsinfo         = nfs4_proc_fsinfo,
9334         .pathconf       = nfs4_proc_pathconf,
9335         .set_capabilities = nfs4_server_capabilities,
9336         .decode_dirent  = nfs4_decode_dirent,
9337         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9338         .read_setup     = nfs4_proc_read_setup,
9339         .read_done      = nfs4_read_done,
9340         .write_setup    = nfs4_proc_write_setup,
9341         .write_done     = nfs4_write_done,
9342         .commit_setup   = nfs4_proc_commit_setup,
9343         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9344         .commit_done    = nfs4_commit_done,
9345         .lock           = nfs4_proc_lock,
9346         .clear_acl_cache = nfs4_zap_acl_attr,
9347         .close_context  = nfs4_close_context,
9348         .open_context   = nfs4_atomic_open,
9349         .have_delegation = nfs4_have_delegation,
9350         .return_delegation = nfs4_inode_return_delegation,
9351         .alloc_client   = nfs4_alloc_client,
9352         .init_client    = nfs4_init_client,
9353         .free_client    = nfs4_free_client,
9354         .create_server  = nfs4_create_server,
9355         .clone_server   = nfs_clone_server,
9356 };
9357
9358 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9359         .name   = XATTR_NAME_NFSV4_ACL,
9360         .list   = nfs4_xattr_list_nfs4_acl,
9361         .get    = nfs4_xattr_get_nfs4_acl,
9362         .set    = nfs4_xattr_set_nfs4_acl,
9363 };
9364
9365 const struct xattr_handler *nfs4_xattr_handlers[] = {
9366         &nfs4_xattr_nfs4_acl_handler,
9367 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9368         &nfs4_xattr_nfs4_label_handler,
9369 #endif
9370         NULL
9371 };
9372
9373 /*
9374  * Local variables:
9375  *  c-basic-offset: 8
9376  * End:
9377  */