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 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 struct rpc_cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 struct rpc_cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 struct rpc_cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 struct rpc_cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 struct rpc_cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 struct rpc_cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 404 {
 405         int res = 0;
 406
 407         might_sleep();
 408
 409         freezable_schedule_timeout_killable_unsafe(
 410                 nfs4_update_delay(timeout));
 411         if (fatal_signal_pending(current))
 412                 res = -ERESTARTSYS;
 413         return res;
 414 }
 415
 416 /* This is the error handling routine for processes that are allowed
 417  * to sleep.
 418  */
 419 static int nfs4_do_handle_exception(struct nfs_server *server,
 420                 int errorcode, struct nfs4_exception *exception)
 421 {
 422         struct nfs_client *clp = server->nfs_client;
 423         struct nfs4_state *state = exception->state;
 424         const nfs4_stateid *stateid = exception->stateid;
 425         struct inode *inode = exception->inode;
 426         int ret = errorcode;
 427
 428         exception->delay = 0;
 429         exception->recovering = 0;
 430         exception->retry = 0;
 431
 432         if (stateid == NULL && state != NULL)
 433                 stateid = &state->stateid;
 434
 435         switch(errorcode) {
 436                 case 0:
 437                         return 0;
 438                 case -NFS4ERR_BADHANDLE:
 439                 case -ESTALE:
 440                         if (inode != NULL && S_ISREG(inode->i_mode))
 441                                 pnfs_destroy_layout(NFS_I(inode));
 442                         break;
 443                 case -NFS4ERR_DELEG_REVOKED:
 444                 case -NFS4ERR_ADMIN_REVOKED:
 445                 case -NFS4ERR_EXPIRED:
 446                 case -NFS4ERR_BAD_STATEID:
 447                         if (inode != NULL && stateid != NULL) {
 448                                 nfs_inode_find_state_and_recover(inode,
 449                                                 stateid);
 450                                 goto wait_on_recovery;
 451                         }
 452                         /* Fall through */
 453                 case -NFS4ERR_OPENMODE:
 454                         if (inode) {
 455                                 int err;
 456
 457                                 err = nfs_async_inode_return_delegation(inode,
 458                                                 stateid);
 459                                 if (err == 0)
 460                                         goto wait_on_recovery;
 461                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 462                                         exception->retry = 1;
 463                                         break;
 464                                 }
 465                         }
 466                         if (state == NULL)
 467                                 break;
 468                         ret = nfs4_schedule_stateid_recovery(server, state);
 469                         if (ret < 0)
 470                                 break;
 471                         goto wait_on_recovery;
 472                 case -NFS4ERR_STALE_STATEID:
 473                 case -NFS4ERR_STALE_CLIENTID:
 474                         nfs4_schedule_lease_recovery(clp);
 475                         goto wait_on_recovery;
 476                 case -NFS4ERR_MOVED:
 477                         ret = nfs4_schedule_migration_recovery(server);
 478                         if (ret < 0)
 479                                 break;
 480                         goto wait_on_recovery;
 481                 case -NFS4ERR_LEASE_MOVED:
 482                         nfs4_schedule_lease_moved_recovery(clp);
 483                         goto wait_on_recovery;
 484 #if defined(CONFIG_NFS_V4_1)
 485                 case -NFS4ERR_BADSESSION:
 486                 case -NFS4ERR_BADSLOT:
 487                 case -NFS4ERR_BAD_HIGH_SLOT:
 488                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 489                 case -NFS4ERR_DEADSESSION:
 490                 case -NFS4ERR_SEQ_FALSE_RETRY:
 491                 case -NFS4ERR_SEQ_MISORDERED:
 492                         dprintk("%s ERROR: %d Reset session\n", __func__,
 493                                 errorcode);
 494                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 495                         goto wait_on_recovery;
 496 #endif /* defined(CONFIG_NFS_V4_1) */
 497                 case -NFS4ERR_FILE_OPEN:
 498                         if (exception->timeout > HZ) {
 499                                 /* We have retried a decent amount, time to
 500                                  * fail
 501                                  */
 502                                 ret = -EBUSY;
 503                                 break;
 504                         }
 505                         /* Fall through */
 506                 case -NFS4ERR_DELAY:
 507                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 508                         /* Fall through */
 509                 case -NFS4ERR_GRACE:
 510                 case -NFS4ERR_LAYOUTTRYLATER:
 511                 case -NFS4ERR_RECALLCONFLICT:
 512                         exception->delay = 1;
 513                         return 0;
 514
 515                 case -NFS4ERR_RETRY_UNCACHED_REP:
 516                 case -NFS4ERR_OLD_STATEID:
 517                         exception->retry = 1;
 518                         break;
 519                 case -NFS4ERR_BADOWNER:
 520                         /* The following works around a Linux server bug! */
 521                 case -NFS4ERR_BADNAME:
 522                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 523                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 524                                 exception->retry = 1;
 525                                 printk(KERN_WARNING "NFS: v4 server %s "
 526                                                 "does not accept raw "
 527                                                 "uid/gids. "
 528                                                 "Reenabling the idmapper.\n",
 529                                                 server->nfs_client->cl_hostname);
 530                         }
 531         }
 532         /* We failed to handle the error */
 533         return nfs4_map_errors(ret);
 534 wait_on_recovery:
 535         exception->recovering = 1;
 536         return 0;
 537 }
 538
 539 /* This is the error handling routine for processes that are allowed
 540  * to sleep.
 541  */
 542 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 543 {
 544         struct nfs_client *clp = server->nfs_client;
 545         int ret;
 546
 547         ret = nfs4_do_handle_exception(server, errorcode, exception);
 548         if (exception->delay) {
 549                 ret = nfs4_delay(server->client, &exception->timeout);
 550                 goto out_retry;
 551         }
 552         if (exception->recovering) {
 553                 ret = nfs4_wait_clnt_recover(clp);
 554                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 555                         return -EIO;
 556                 goto out_retry;
 557         }
 558         return ret;
 559 out_retry:
 560         if (ret == 0)
 561                 exception->retry = 1;
 562         return ret;
 563 }
 564
 565 static int
 566 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 567                 int errorcode, struct nfs4_exception *exception)
 568 {
 569         struct nfs_client *clp = server->nfs_client;
 570         int ret;
 571
 572         ret = nfs4_do_handle_exception(server, errorcode, exception);
 573         if (exception->delay) {
 574                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 575                 goto out_retry;
 576         }
 577         if (exception->recovering) {
 578                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 579                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 580                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 581                 goto out_retry;
 582         }
 583         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 584                 ret = -EIO;
 585         return ret;
 586 out_retry:
 587         if (ret == 0) {
 588                 exception->retry = 1;
 589                 /*
 590                  * For NFS4ERR_MOVED, the client transport will need to
 591                  * be recomputed after migration recovery has completed.
 592                  */
 593                 if (errorcode == -NFS4ERR_MOVED)
 594                         rpc_task_release_transport(task);
 595         }
 596         return ret;
 597 }
 598
 599 int
 600 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 601                         struct nfs4_state *state, long *timeout)
 602 {
 603         struct nfs4_exception exception = {
 604                 .state = state,
 605         };
 606
 607         if (task->tk_status >= 0)
 608                 return 0;
 609         if (timeout)
 610                 exception.timeout = *timeout;
 611         task->tk_status = nfs4_async_handle_exception(task, server,
 612                         task->tk_status,
 613                         &exception);
 614         if (exception.delay && timeout)
 615                 *timeout = exception.timeout;
 616         if (exception.retry)
 617                 return -EAGAIN;
 618         return 0;
 619 }
 620
 621 /*
 622  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 623  * or 'false' otherwise.
 624  */
 625 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 626 {
 627         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 628         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 629 }
 630
 631 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 632 {
 633         spin_lock(&clp->cl_lock);
 634         if (time_before(clp->cl_last_renewal,timestamp))
 635                 clp->cl_last_renewal = timestamp;
 636         spin_unlock(&clp->cl_lock);
 637 }
 638
 639 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 640 {
 641         struct nfs_client *clp = server->nfs_client;
 642
 643         if (!nfs4_has_session(clp))
 644                 do_renew_lease(clp, timestamp);
 645 }
 646
 647 struct nfs4_call_sync_data {
 648         const struct nfs_server *seq_server;
 649         struct nfs4_sequence_args *seq_args;
 650         struct nfs4_sequence_res *seq_res;
 651 };
 652
 653 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 654                         struct nfs4_sequence_res *res, int cache_reply,
 655                         int privileged)
 656 {
 657         args->sa_slot = NULL;
 658         args->sa_cache_this = cache_reply;
 659         args->sa_privileged = privileged;
 660
 661         res->sr_slot = NULL;
 662 }
 663
 664 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 665 {
 666         struct nfs4_slot *slot = res->sr_slot;
 667         struct nfs4_slot_table *tbl;
 668
 669         tbl = slot->table;
 670         spin_lock(&tbl->slot_tbl_lock);
 671         if (!nfs41_wake_and_assign_slot(tbl, slot))
 672                 nfs4_free_slot(tbl, slot);
 673         spin_unlock(&tbl->slot_tbl_lock);
 674
 675         res->sr_slot = NULL;
 676 }
 677
 678 static int nfs40_sequence_done(struct rpc_task *task,
 679                                struct nfs4_sequence_res *res)
 680 {
 681         if (res->sr_slot != NULL)
 682                 nfs40_sequence_free_slot(res);
 683         return 1;
 684 }
 685
 686 #if defined(CONFIG_NFS_V4_1)
 687
 688 static void nfs41_release_slot(struct nfs4_slot *slot)
 689 {
 690         struct nfs4_session *session;
 691         struct nfs4_slot_table *tbl;
 692         bool send_new_highest_used_slotid = false;
 693
 694         if (!slot)
 695                 return;
 696         tbl = slot->table;
 697         session = tbl->session;
 698
 699         /* Bump the slot sequence number */
 700         if (slot->seq_done)
 701                 slot->seq_nr++;
 702         slot->seq_done = 0;
 703
 704         spin_lock(&tbl->slot_tbl_lock);
 705         /* Be nice to the server: try to ensure that the last transmitted
 706          * value for highest_user_slotid <= target_highest_slotid
 707          */
 708         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 709                 send_new_highest_used_slotid = true;
 710
 711         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 712                 send_new_highest_used_slotid = false;
 713                 goto out_unlock;
 714         }
 715         nfs4_free_slot(tbl, slot);
 716
 717         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 718                 send_new_highest_used_slotid = false;
 719 out_unlock:
 720         spin_unlock(&tbl->slot_tbl_lock);
 721         if (send_new_highest_used_slotid)
 722                 nfs41_notify_server(session->clp);
 723         if (waitqueue_active(&tbl->slot_waitq))
 724                 wake_up_all(&tbl->slot_waitq);
 725 }
 726
 727 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 728 {
 729         nfs41_release_slot(res->sr_slot);
 730         res->sr_slot = NULL;
 731 }
 732
 733 static int nfs41_sequence_process(struct rpc_task *task,
 734                 struct nfs4_sequence_res *res)
 735 {
 736         struct nfs4_session *session;
 737         struct nfs4_slot *slot = res->sr_slot;
 738         struct nfs_client *clp;
 739         bool interrupted = false;
 740         int ret = 1;
 741
 742         if (slot == NULL)
 743                 goto out_noaction;
 744         /* don't increment the sequence number if the task wasn't sent */
 745         if (!RPC_WAS_SENT(task))
 746                 goto out;
 747
 748         session = slot->table->session;
 749
 750         if (slot->interrupted) {
 751                 if (res->sr_status != -NFS4ERR_DELAY)
 752                         slot->interrupted = 0;
 753                 interrupted = true;
 754         }
 755
 756         trace_nfs4_sequence_done(session, res);
 757         /* Check the SEQUENCE operation status */
 758         switch (res->sr_status) {
 759         case 0:
 760                 /* Update the slot's sequence and clientid lease timer */
 761                 slot->seq_done = 1;
 762                 clp = session->clp;
 763                 do_renew_lease(clp, res->sr_timestamp);
 764                 /* Check sequence flags */
 765                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 766                                 !!slot->privileged);
 767                 nfs41_update_target_slotid(slot->table, slot, res);
 768                 break;
 769         case 1:
 770                 /*
 771                  * sr_status remains 1 if an RPC level error occurred.
 772                  * The server may or may not have processed the sequence
 773                  * operation..
 774                  * Mark the slot as having hosted an interrupted RPC call.
 775                  */
 776                 slot->interrupted = 1;
 777                 goto out;
 778         case -NFS4ERR_DELAY:
 779                 /* The server detected a resend of the RPC call and
 780                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 781                  * of RFC5661.
 782                  */
 783                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 784                         __func__,
 785                         slot->slot_nr,
 786                         slot->seq_nr);
 787                 goto out_retry;
 788         case -NFS4ERR_RETRY_UNCACHED_REP:
 789         case -NFS4ERR_SEQ_FALSE_RETRY:
 790                 /*
 791                  * The server thinks we tried to replay a request.
 792                  * Retry the call after bumping the sequence ID.
 793                  */
 794                 goto retry_new_seq;
 795         case -NFS4ERR_BADSLOT:
 796                 /*
 797                  * The slot id we used was probably retired. Try again
 798                  * using a different slot id.
 799                  */
 800                 if (slot->slot_nr < slot->table->target_highest_slotid)
 801                         goto session_recover;
 802                 goto retry_nowait;
 803         case -NFS4ERR_SEQ_MISORDERED:
 804                 /*
 805                  * Was the last operation on this sequence interrupted?
 806                  * If so, retry after bumping the sequence number.
 807                  */
 808                 if (interrupted)
 809                         goto retry_new_seq;
 810                 /*
 811                  * Could this slot have been previously retired?
 812                  * If so, then the server may be expecting seq_nr = 1!
 813                  */
 814                 if (slot->seq_nr != 1) {
 815                         slot->seq_nr = 1;
 816                         goto retry_nowait;
 817                 }
 818                 goto session_recover;
 819         default:
 820                 /* Just update the slot sequence no. */
 821                 slot->seq_done = 1;
 822         }
 823 out:
 824         /* The session may be reset by one of the error handlers. */
 825         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 826 out_noaction:
 827         return ret;
 828 session_recover:
 829         nfs4_schedule_session_recovery(session, res->sr_status);
 830         goto retry_nowait;
 831 retry_new_seq:
 832         ++slot->seq_nr;
 833 retry_nowait:
 834         if (rpc_restart_call_prepare(task)) {
 835                 nfs41_sequence_free_slot(res);
 836                 task->tk_status = 0;
 837                 ret = 0;
 838         }
 839         goto out;
 840 out_retry:
 841         if (!rpc_restart_call(task))
 842                 goto out;
 843         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 844         return 0;
 845 }
 846
 847 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 848 {
 849         if (!nfs41_sequence_process(task, res))
 850                 return 0;
 851         if (res->sr_slot != NULL)
 852                 nfs41_sequence_free_slot(res);
 853         return 1;
 854
 855 }
 856 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 857
 858 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 859 {
 860         if (res->sr_slot == NULL)
 861                 return 1;
 862         if (res->sr_slot->table->session != NULL)
 863                 return nfs41_sequence_process(task, res);
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL) {
 870                 if (res->sr_slot->table->session != NULL)
 871                         nfs41_sequence_free_slot(res);
 872                 else
 873                         nfs40_sequence_free_slot(res);
 874         }
 875 }
 876
 877 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 878 {
 879         if (res->sr_slot == NULL)
 880                 return 1;
 881         if (!res->sr_slot->table->session)
 882                 return nfs40_sequence_done(task, res);
 883         return nfs41_sequence_done(task, res);
 884 }
 885 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 886
 887 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 888 {
 889         struct nfs4_call_sync_data *data = calldata;
 890
 891         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 892
 893         nfs4_setup_sequence(data->seq_server->nfs_client,
 894                             data->seq_args, data->seq_res, task);
 895 }
 896
 897 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 898 {
 899         struct nfs4_call_sync_data *data = calldata;
 900
 901         nfs41_sequence_done(task, data->seq_res);
 902 }
 903
 904 static const struct rpc_call_ops nfs41_call_sync_ops = {
 905         .rpc_call_prepare = nfs41_call_sync_prepare,
 906         .rpc_call_done = nfs41_call_sync_done,
 907 };
 908
 909 static void
 910 nfs4_sequence_process_interrupted(struct nfs_client *client,
 911                 struct nfs4_slot *slot, struct rpc_cred *cred)
 912 {
 913         struct rpc_task *task;
 914
 915         task = _nfs41_proc_sequence(client, cred, slot, true);
 916         if (!IS_ERR(task))
 917                 rpc_put_task_async(task);
 918 }
 919
 920 #else   /* !CONFIG_NFS_V4_1 */
 921
 922 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 923 {
 924         return nfs40_sequence_done(task, res);
 925 }
 926
 927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot != NULL)
 930                 nfs40_sequence_free_slot(res);
 931 }
 932
 933 int nfs4_sequence_done(struct rpc_task *task,
 934                        struct nfs4_sequence_res *res)
 935 {
 936         return nfs40_sequence_done(task, res);
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 939
 940 static void
 941 nfs4_sequence_process_interrupted(struct nfs_client *client,
 942                 struct nfs4_slot *slot, struct rpc_cred *cred)
 943 {
 944         WARN_ON_ONCE(1);
 945         slot->interrupted = 0;
 946 }
 947
 948 #endif  /* !CONFIG_NFS_V4_1 */
 949
 950 static
 951 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 952                 struct nfs4_sequence_res *res,
 953                 struct nfs4_slot *slot)
 954 {
 955         if (!slot)
 956                 return;
 957         slot->privileged = args->sa_privileged ? 1 : 0;
 958         args->sa_slot = slot;
 959
 960         res->sr_slot = slot;
 961         res->sr_timestamp = jiffies;
 962         res->sr_status_flags = 0;
 963         res->sr_status = 1;
 964
 965 }
 966
 967 int nfs4_setup_sequence(struct nfs_client *client,
 968                         struct nfs4_sequence_args *args,
 969                         struct nfs4_sequence_res *res,
 970                         struct rpc_task *task)
 971 {
 972         struct nfs4_session *session = nfs4_get_session(client);
 973         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 974         struct nfs4_slot *slot;
 975
 976         /* slot already allocated? */
 977         if (res->sr_slot != NULL)
 978                 goto out_start;
 979
 980         if (session) {
 981                 tbl = &session->fc_slot_table;
 982                 task->tk_timeout = 0;
 983         }
 984
 985         for (;;) {
 986                 spin_lock(&tbl->slot_tbl_lock);
 987                 /* The state manager will wait until the slot table is empty */
 988                 if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 989                         goto out_sleep;
 990
 991                 slot = nfs4_alloc_slot(tbl);
 992                 if (IS_ERR(slot)) {
 993                         /* Try again in 1/4 second */
 994                         if (slot == ERR_PTR(-ENOMEM))
 995                                 task->tk_timeout = HZ >> 2;
 996                         goto out_sleep;
 997                 }
 998                 spin_unlock(&tbl->slot_tbl_lock);
 999
1000                 if (likely(!slot->interrupted))
1001                         break;
1002                 nfs4_sequence_process_interrupted(client,
1003                                 slot, task->tk_msg.rpc_cred);
1004         }
1005
1006         nfs4_sequence_attach_slot(args, res, slot);
1007
1008         trace_nfs4_setup_sequence(session, args);
1009 out_start:
1010         rpc_call_start(task);
1011         return 0;
1012
1013 out_sleep:
1014         if (args->sa_privileged)
1015                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1016                                 NULL, RPC_PRIORITY_PRIVILEGED);
1017         else
1018                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1019         spin_unlock(&tbl->slot_tbl_lock);
1020         return -EAGAIN;
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1023
1024 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1025 {
1026         struct nfs4_call_sync_data *data = calldata;
1027         nfs4_setup_sequence(data->seq_server->nfs_client,
1028                                 data->seq_args, data->seq_res, task);
1029 }
1030
1031 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1032 {
1033         struct nfs4_call_sync_data *data = calldata;
1034         nfs4_sequence_done(task, data->seq_res);
1035 }
1036
1037 static const struct rpc_call_ops nfs40_call_sync_ops = {
1038         .rpc_call_prepare = nfs40_call_sync_prepare,
1039         .rpc_call_done = nfs40_call_sync_done,
1040 };
1041
1042 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1043                                    struct nfs_server *server,
1044                                    struct rpc_message *msg,
1045                                    struct nfs4_sequence_args *args,
1046                                    struct nfs4_sequence_res *res)
1047 {
1048         int ret;
1049         struct rpc_task *task;
1050         struct nfs_client *clp = server->nfs_client;
1051         struct nfs4_call_sync_data data = {
1052                 .seq_server = server,
1053                 .seq_args = args,
1054                 .seq_res = res,
1055         };
1056         struct rpc_task_setup task_setup = {
1057                 .rpc_client = clnt,
1058                 .rpc_message = msg,
1059                 .callback_ops = clp->cl_mvops->call_sync_ops,
1060                 .callback_data = &data
1061         };
1062
1063         task = rpc_run_task(&task_setup);
1064         if (IS_ERR(task))
1065                 ret = PTR_ERR(task);
1066         else {
1067                 ret = task->tk_status;
1068                 rpc_put_task(task);
1069         }
1070         return ret;
1071 }
1072
1073 int nfs4_call_sync(struct rpc_clnt *clnt,
1074                    struct nfs_server *server,
1075                    struct rpc_message *msg,
1076                    struct nfs4_sequence_args *args,
1077                    struct nfs4_sequence_res *res,
1078                    int cache_reply)
1079 {
1080         nfs4_init_sequence(args, res, cache_reply, 0);
1081         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1082 }
1083
1084 static void
1085 nfs4_inc_nlink_locked(struct inode *inode)
1086 {
1087         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1088         inc_nlink(inode);
1089 }
1090
1091 static void
1092 nfs4_dec_nlink_locked(struct inode *inode)
1093 {
1094         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1095         drop_nlink(inode);
1096 }
1097
1098 static void
1099 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1100                 unsigned long timestamp, unsigned long cache_validity)
1101 {
1102         struct nfs_inode *nfsi = NFS_I(dir);
1103
1104         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1105                 | NFS_INO_INVALID_MTIME
1106                 | NFS_INO_INVALID_DATA
1107                 | cache_validity;
1108         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1109                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1110                 nfsi->attrtimeo_timestamp = jiffies;
1111         } else {
1112                 nfs_force_lookup_revalidate(dir);
1113                 if (cinfo->before != inode_peek_iversion_raw(dir))
1114                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1115                                 NFS_INO_INVALID_ACL;
1116         }
1117         inode_set_iversion_raw(dir, cinfo->after);
1118         nfsi->read_cache_jiffies = timestamp;
1119         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1120         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1121         nfs_fscache_invalidate(dir);
1122 }
1123
1124 static void
1125 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1126                 unsigned long timestamp, unsigned long cache_validity)
1127 {
1128         spin_lock(&dir->i_lock);
1129         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1130         spin_unlock(&dir->i_lock);
1131 }
1132
1133 struct nfs4_open_createattrs {
1134         struct nfs4_label *label;
1135         struct iattr *sattr;
1136         const __u32 verf[2];
1137 };
1138
1139 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1140                 int err, struct nfs4_exception *exception)
1141 {
1142         if (err != -EINVAL)
1143                 return false;
1144         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1145                 return false;
1146         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1147         exception->retry = 1;
1148         return true;
1149 }
1150
1151 static u32
1152 nfs4_map_atomic_open_share(struct nfs_server *server,
1153                 fmode_t fmode, int openflags)
1154 {
1155         u32 res = 0;
1156
1157         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1158         case FMODE_READ:
1159                 res = NFS4_SHARE_ACCESS_READ;
1160                 break;
1161         case FMODE_WRITE:
1162                 res = NFS4_SHARE_ACCESS_WRITE;
1163                 break;
1164         case FMODE_READ|FMODE_WRITE:
1165                 res = NFS4_SHARE_ACCESS_BOTH;
1166         }
1167         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1168                 goto out;
1169         /* Want no delegation if we're using O_DIRECT */
1170         if (openflags & O_DIRECT)
1171                 res |= NFS4_SHARE_WANT_NO_DELEG;
1172 out:
1173         return res;
1174 }
1175
1176 static enum open_claim_type4
1177 nfs4_map_atomic_open_claim(struct nfs_server *server,
1178                 enum open_claim_type4 claim)
1179 {
1180         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1181                 return claim;
1182         switch (claim) {
1183         default:
1184                 return claim;
1185         case NFS4_OPEN_CLAIM_FH:
1186                 return NFS4_OPEN_CLAIM_NULL;
1187         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1188                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1189         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1190                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1191         }
1192 }
1193
1194 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1195 {
1196         p->o_res.f_attr = &p->f_attr;
1197         p->o_res.f_label = p->f_label;
1198         p->o_res.seqid = p->o_arg.seqid;
1199         p->c_res.seqid = p->c_arg.seqid;
1200         p->o_res.server = p->o_arg.server;
1201         p->o_res.access_request = p->o_arg.access;
1202         nfs_fattr_init(&p->f_attr);
1203         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1204 }
1205
1206 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1207                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1208                 const struct nfs4_open_createattrs *c,
1209                 enum open_claim_type4 claim,
1210                 gfp_t gfp_mask)
1211 {
1212         struct dentry *parent = dget_parent(dentry);
1213         struct inode *dir = d_inode(parent);
1214         struct nfs_server *server = NFS_SERVER(dir);
1215         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1216         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1217         struct nfs4_opendata *p;
1218
1219         p = kzalloc(sizeof(*p), gfp_mask);
1220         if (p == NULL)
1221                 goto err;
1222
1223         p->f_label = nfs4_label_alloc(server, gfp_mask);
1224         if (IS_ERR(p->f_label))
1225                 goto err_free_p;
1226
1227         p->a_label = nfs4_label_alloc(server, gfp_mask);
1228         if (IS_ERR(p->a_label))
1229                 goto err_free_f;
1230
1231         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1232         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1233         if (IS_ERR(p->o_arg.seqid))
1234                 goto err_free_label;
1235         nfs_sb_active(dentry->d_sb);
1236         p->dentry = dget(dentry);
1237         p->dir = parent;
1238         p->owner = sp;
1239         atomic_inc(&sp->so_count);
1240         p->o_arg.open_flags = flags;
1241         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1242         p->o_arg.umask = current_umask();
1243         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1244         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1245                         fmode, flags);
1246         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1247          * will return permission denied for all bits until close */
1248         if (!(flags & O_EXCL)) {
1249                 /* ask server to check for all possible rights as results
1250                  * are cached */
1251                 switch (p->o_arg.claim) {
1252                 default:
1253                         break;
1254                 case NFS4_OPEN_CLAIM_NULL:
1255                 case NFS4_OPEN_CLAIM_FH:
1256                         p->o_arg.access = NFS4_ACCESS_READ |
1257                                 NFS4_ACCESS_MODIFY |
1258                                 NFS4_ACCESS_EXTEND |
1259                                 NFS4_ACCESS_EXECUTE;
1260                 }
1261         }
1262         p->o_arg.clientid = server->nfs_client->cl_clientid;
1263         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1264         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1265         p->o_arg.name = &dentry->d_name;
1266         p->o_arg.server = server;
1267         p->o_arg.bitmask = nfs4_bitmask(server, label);
1268         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1269         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1270         switch (p->o_arg.claim) {
1271         case NFS4_OPEN_CLAIM_NULL:
1272         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1273         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1274                 p->o_arg.fh = NFS_FH(dir);
1275                 break;
1276         case NFS4_OPEN_CLAIM_PREVIOUS:
1277         case NFS4_OPEN_CLAIM_FH:
1278         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1279         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1280                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1281         }
1282         if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1283                 p->o_arg.u.attrs = &p->attrs;
1284                 memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1285
1286                 memcpy(p->o_arg.u.verifier.data, c->verf,
1287                                 sizeof(p->o_arg.u.verifier.data));
1288         }
1289         p->c_arg.fh = &p->o_res.fh;
1290         p->c_arg.stateid = &p->o_res.stateid;
1291         p->c_arg.seqid = p->o_arg.seqid;
1292         nfs4_init_opendata_res(p);
1293         kref_init(&p->kref);
1294         return p;
1295
1296 err_free_label:
1297         nfs4_label_free(p->a_label);
1298 err_free_f:
1299         nfs4_label_free(p->f_label);
1300 err_free_p:
1301         kfree(p);
1302 err:
1303         dput(parent);
1304         return NULL;
1305 }
1306
1307 static void nfs4_opendata_free(struct kref *kref)
1308 {
1309         struct nfs4_opendata *p = container_of(kref,
1310                         struct nfs4_opendata, kref);
1311         struct super_block *sb = p->dentry->d_sb;
1312
1313         nfs4_lgopen_release(p->lgp);
1314         nfs_free_seqid(p->o_arg.seqid);
1315         nfs4_sequence_free_slot(&p->o_res.seq_res);
1316         if (p->state != NULL)
1317                 nfs4_put_open_state(p->state);
1318         nfs4_put_state_owner(p->owner);
1319
1320         nfs4_label_free(p->a_label);
1321         nfs4_label_free(p->f_label);
1322
1323         dput(p->dir);
1324         dput(p->dentry);
1325         nfs_sb_deactive(sb);
1326         nfs_fattr_free_names(&p->f_attr);
1327         kfree(p->f_attr.mdsthreshold);
1328         kfree(p);
1329 }
1330
1331 static void nfs4_opendata_put(struct nfs4_opendata *p)
1332 {
1333         if (p != NULL)
1334                 kref_put(&p->kref, nfs4_opendata_free);
1335 }
1336
1337 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1338                 fmode_t fmode)
1339 {
1340         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1341         case FMODE_READ|FMODE_WRITE:
1342                 return state->n_rdwr != 0;
1343         case FMODE_WRITE:
1344                 return state->n_wronly != 0;
1345         case FMODE_READ:
1346                 return state->n_rdonly != 0;
1347         }
1348         WARN_ON_ONCE(1);
1349         return false;
1350 }
1351
1352 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1353 {
1354         int ret = 0;
1355
1356         if (open_mode & (O_EXCL|O_TRUNC))
1357                 goto out;
1358         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1359                 case FMODE_READ:
1360                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1361                                 && state->n_rdonly != 0;
1362                         break;
1363                 case FMODE_WRITE:
1364                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1365                                 && state->n_wronly != 0;
1366                         break;
1367                 case FMODE_READ|FMODE_WRITE:
1368                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1369                                 && state->n_rdwr != 0;
1370         }
1371 out:
1372         return ret;
1373 }
1374
1375 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1376                 enum open_claim_type4 claim)
1377 {
1378         if (delegation == NULL)
1379                 return 0;
1380         if ((delegation->type & fmode) != fmode)
1381                 return 0;
1382         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1383                 return 0;
1384         switch (claim) {
1385         case NFS4_OPEN_CLAIM_NULL:
1386         case NFS4_OPEN_CLAIM_FH:
1387                 break;
1388         case NFS4_OPEN_CLAIM_PREVIOUS:
1389                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1390                         break;
1391                 /* Fall through */
1392         default:
1393                 return 0;
1394         }
1395         nfs_mark_delegation_referenced(delegation);
1396         return 1;
1397 }
1398
1399 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1400 {
1401         switch (fmode) {
1402                 case FMODE_WRITE:
1403                         state->n_wronly++;
1404                         break;
1405                 case FMODE_READ:
1406                         state->n_rdonly++;
1407                         break;
1408                 case FMODE_READ|FMODE_WRITE:
1409                         state->n_rdwr++;
1410         }
1411         nfs4_state_set_mode_locked(state, state->state | fmode);
1412 }
1413
1414 #ifdef CONFIG_NFS_V4_1
1415 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1416 {
1417         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1418                 return true;
1419         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1420                 return true;
1421         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1422                 return true;
1423         return false;
1424 }
1425 #endif /* CONFIG_NFS_V4_1 */
1426
1427 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1428 {
1429         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1430                 wake_up_all(&state->waitq);
1431 }
1432
1433 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1434                 const nfs4_stateid *stateid)
1435 {
1436         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1437         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1438
1439         if (stateid_seqid == state_seqid + 1U ||
1440             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1441                 nfs_state_log_update_open_stateid(state);
1442         else
1443                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1444 }
1445
1446 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1447 {
1448         struct nfs_client *clp = state->owner->so_server->nfs_client;
1449         bool need_recover = false;
1450
1451         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1452                 need_recover = true;
1453         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1454                 need_recover = true;
1455         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1456                 need_recover = true;
1457         if (need_recover)
1458                 nfs4_state_mark_reclaim_nograce(clp, state);
1459 }
1460
1461 /*
1462  * Check for whether or not the caller may update the open stateid
1463  * to the value passed in by stateid.
1464  *
1465  * Note: This function relies heavily on the server implementing
1466  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1467  * correctly.
1468  * i.e. The stateid seqids have to be initialised to 1, and
1469  * are then incremented on every state transition.
1470  */
1471 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1472                 const nfs4_stateid *stateid)
1473 {
1474         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1475             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1476                 if (stateid->seqid == cpu_to_be32(1))
1477                         nfs_state_log_update_open_stateid(state);
1478                 else
1479                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1480                 return true;
1481         }
1482
1483         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1484                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1485                 return true;
1486         }
1487         return false;
1488 }
1489
1490 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1491 {
1492         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1493                 return;
1494         if (state->n_wronly)
1495                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1496         if (state->n_rdonly)
1497                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1498         if (state->n_rdwr)
1499                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1500         set_bit(NFS_OPEN_STATE, &state->flags);
1501 }
1502
1503 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1504                 nfs4_stateid *stateid, fmode_t fmode)
1505 {
1506         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1507         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1508         case FMODE_WRITE:
1509                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1510                 break;
1511         case FMODE_READ:
1512                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1513                 break;
1514         case 0:
1515                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1516                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1517                 clear_bit(NFS_OPEN_STATE, &state->flags);
1518         }
1519         if (stateid == NULL)
1520                 return;
1521         /* Handle OPEN+OPEN_DOWNGRADE races */
1522         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1523             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1524                 nfs_resync_open_stateid_locked(state);
1525                 goto out;
1526         }
1527         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1528                 nfs4_stateid_copy(&state->stateid, stateid);
1529         nfs4_stateid_copy(&state->open_stateid, stateid);
1530         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1531 out:
1532         nfs_state_log_update_open_stateid(state);
1533 }
1534
1535 static void nfs_clear_open_stateid(struct nfs4_state *state,
1536         nfs4_stateid *arg_stateid,
1537         nfs4_stateid *stateid, fmode_t fmode)
1538 {
1539         write_seqlock(&state->seqlock);
1540         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1541         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1542                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1543         write_sequnlock(&state->seqlock);
1544         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1545                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1546 }
1547
1548 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1549                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1550 {
1551         DEFINE_WAIT(wait);
1552         int status = 0;
1553         for (;;) {
1554
1555                 if (!nfs_need_update_open_stateid(state, stateid))
1556                         return;
1557                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1558                         break;
1559                 if (status)
1560                         break;
1561                 /* Rely on seqids for serialisation with NFSv4.0 */
1562                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1563                         break;
1564
1565                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1566                 /*
1567                  * Ensure we process the state changes in the same order
1568                  * in which the server processed them by delaying the
1569                  * update of the stateid until we are in sequence.
1570                  */
1571                 write_sequnlock(&state->seqlock);
1572                 spin_unlock(&state->owner->so_lock);
1573                 rcu_read_unlock();
1574                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1575                 if (!signal_pending(current)) {
1576                         if (schedule_timeout(5*HZ) == 0)
1577                                 status = -EAGAIN;
1578                         else
1579                                 status = 0;
1580                 } else
1581                         status = -EINTR;
1582                 finish_wait(&state->waitq, &wait);
1583                 rcu_read_lock();
1584                 spin_lock(&state->owner->so_lock);
1585                 write_seqlock(&state->seqlock);
1586         }
1587
1588         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1589             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1590                 nfs4_stateid_copy(freeme, &state->open_stateid);
1591                 nfs_test_and_clear_all_open_stateid(state);
1592         }
1593
1594         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1595                 nfs4_stateid_copy(&state->stateid, stateid);
1596         nfs4_stateid_copy(&state->open_stateid, stateid);
1597         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1598         nfs_state_log_update_open_stateid(state);
1599 }
1600
1601 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1602                 const nfs4_stateid *open_stateid,
1603                 fmode_t fmode,
1604                 nfs4_stateid *freeme)
1605 {
1606         /*
1607          * Protect the call to nfs4_state_set_mode_locked and
1608          * serialise the stateid update
1609          */
1610         write_seqlock(&state->seqlock);
1611         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1612         switch (fmode) {
1613         case FMODE_READ:
1614                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1615                 break;
1616         case FMODE_WRITE:
1617                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1618                 break;
1619         case FMODE_READ|FMODE_WRITE:
1620                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1621         }
1622         set_bit(NFS_OPEN_STATE, &state->flags);
1623         write_sequnlock(&state->seqlock);
1624 }
1625
1626 static void nfs_state_set_delegation(struct nfs4_state *state,
1627                 const nfs4_stateid *deleg_stateid,
1628                 fmode_t fmode)
1629 {
1630         /*
1631          * Protect the call to nfs4_state_set_mode_locked and
1632          * serialise the stateid update
1633          */
1634         write_seqlock(&state->seqlock);
1635         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1636         set_bit(NFS_DELEGATED_STATE, &state->flags);
1637         write_sequnlock(&state->seqlock);
1638 }
1639
1640 static void nfs_state_clear_delegation(struct nfs4_state *state)
1641 {
1642         write_seqlock(&state->seqlock);
1643         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1644         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1645         write_sequnlock(&state->seqlock);
1646 }
1647
1648 static int update_open_stateid(struct nfs4_state *state,
1649                 const nfs4_stateid *open_stateid,
1650                 const nfs4_stateid *delegation,
1651                 fmode_t fmode)
1652 {
1653         struct nfs_server *server = NFS_SERVER(state->inode);
1654         struct nfs_client *clp = server->nfs_client;
1655         struct nfs_inode *nfsi = NFS_I(state->inode);
1656         struct nfs_delegation *deleg_cur;
1657         nfs4_stateid freeme = { };
1658         int ret = 0;
1659
1660         fmode &= (FMODE_READ|FMODE_WRITE);
1661
1662         rcu_read_lock();
1663         spin_lock(&state->owner->so_lock);
1664         if (open_stateid != NULL) {
1665                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1666                 ret = 1;
1667         }
1668
1669         deleg_cur = rcu_dereference(nfsi->delegation);
1670         if (deleg_cur == NULL)
1671                 goto no_delegation;
1672
1673         spin_lock(&deleg_cur->lock);
1674         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1675            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1676             (deleg_cur->type & fmode) != fmode)
1677                 goto no_delegation_unlock;
1678
1679         if (delegation == NULL)
1680                 delegation = &deleg_cur->stateid;
1681         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1682                 goto no_delegation_unlock;
1683
1684         nfs_mark_delegation_referenced(deleg_cur);
1685         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1686         ret = 1;
1687 no_delegation_unlock:
1688         spin_unlock(&deleg_cur->lock);
1689 no_delegation:
1690         if (ret)
1691                 update_open_stateflags(state, fmode);
1692         spin_unlock(&state->owner->so_lock);
1693         rcu_read_unlock();
1694
1695         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1696                 nfs4_schedule_state_manager(clp);
1697         if (freeme.type != 0)
1698                 nfs4_test_and_free_stateid(server, &freeme,
1699                                 state->owner->so_cred);
1700
1701         return ret;
1702 }
1703
1704 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1705                 const nfs4_stateid *stateid)
1706 {
1707         struct nfs4_state *state = lsp->ls_state;
1708         bool ret = false;
1709
1710         spin_lock(&state->state_lock);
1711         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1712                 goto out_noupdate;
1713         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1714                 goto out_noupdate;
1715         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1716         ret = true;
1717 out_noupdate:
1718         spin_unlock(&state->state_lock);
1719         return ret;
1720 }
1721
1722 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1723 {
1724         struct nfs_delegation *delegation;
1725
1726         fmode &= FMODE_READ|FMODE_WRITE;
1727         rcu_read_lock();
1728         delegation = rcu_dereference(NFS_I(inode)->delegation);
1729         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1730                 rcu_read_unlock();
1731                 return;
1732         }
1733         rcu_read_unlock();
1734         nfs4_inode_return_delegation(inode);
1735 }
1736
1737 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1738 {
1739         struct nfs4_state *state = opendata->state;
1740         struct nfs_inode *nfsi = NFS_I(state->inode);
1741         struct nfs_delegation *delegation;
1742         int open_mode = opendata->o_arg.open_flags;
1743         fmode_t fmode = opendata->o_arg.fmode;
1744         enum open_claim_type4 claim = opendata->o_arg.claim;
1745         nfs4_stateid stateid;
1746         int ret = -EAGAIN;
1747
1748         for (;;) {
1749                 spin_lock(&state->owner->so_lock);
1750                 if (can_open_cached(state, fmode, open_mode)) {
1751                         update_open_stateflags(state, fmode);
1752                         spin_unlock(&state->owner->so_lock);
1753                         goto out_return_state;
1754                 }
1755                 spin_unlock(&state->owner->so_lock);
1756                 rcu_read_lock();
1757                 delegation = rcu_dereference(nfsi->delegation);
1758                 if (!can_open_delegated(delegation, fmode, claim)) {
1759                         rcu_read_unlock();
1760                         break;
1761                 }
1762                 /* Save the delegation */
1763                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1764                 rcu_read_unlock();
1765                 nfs_release_seqid(opendata->o_arg.seqid);
1766                 if (!opendata->is_recover) {
1767                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1768                         if (ret != 0)
1769                                 goto out;
1770                 }
1771                 ret = -EAGAIN;
1772
1773                 /* Try to update the stateid using the delegation */
1774                 if (update_open_stateid(state, NULL, &stateid, fmode))
1775                         goto out_return_state;
1776         }
1777 out:
1778         return ERR_PTR(ret);
1779 out_return_state:
1780         atomic_inc(&state->count);
1781         return state;
1782 }
1783
1784 static void
1785 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1786 {
1787         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1788         struct nfs_delegation *delegation;
1789         int delegation_flags = 0;
1790
1791         rcu_read_lock();
1792         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1793         if (delegation)
1794                 delegation_flags = delegation->flags;
1795         rcu_read_unlock();
1796         switch (data->o_arg.claim) {
1797         default:
1798                 break;
1799         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1800         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1801                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1802                                    "returning a delegation for "
1803                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1804                                    clp->cl_hostname);
1805                 return;
1806         }
1807         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1808                 nfs_inode_set_delegation(state->inode,
1809                                 data->owner->so_cred,
1810                                 data->o_res.delegation_type,
1811                                 &data->o_res.delegation,
1812                                 data->o_res.pagemod_limit);
1813         else
1814                 nfs_inode_reclaim_delegation(state->inode,
1815                                 data->owner->so_cred,
1816                                 data->o_res.delegation_type,
1817                                 &data->o_res.delegation,
1818                                 data->o_res.pagemod_limit);
1819
1820         if (data->o_res.do_recall)
1821                 nfs_async_inode_return_delegation(state->inode,
1822                                                   &data->o_res.delegation);
1823 }
1824
1825 /*
1826  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1827  * and update the nfs4_state.
1828  */
1829 static struct nfs4_state *
1830 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1831 {
1832         struct inode *inode = data->state->inode;
1833         struct nfs4_state *state = data->state;
1834         int ret;
1835
1836         if (!data->rpc_done) {
1837                 if (data->rpc_status)
1838                         return ERR_PTR(data->rpc_status);
1839                 /* cached opens have already been processed */
1840                 goto update;
1841         }
1842
1843         ret = nfs_refresh_inode(inode, &data->f_attr);
1844         if (ret)
1845                 return ERR_PTR(ret);
1846
1847         if (data->o_res.delegation_type != 0)
1848                 nfs4_opendata_check_deleg(data, state);
1849 update:
1850         update_open_stateid(state, &data->o_res.stateid, NULL,
1851                             data->o_arg.fmode);
1852         atomic_inc(&state->count);
1853
1854         return state;
1855 }
1856
1857 static struct inode *
1858 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1859 {
1860         struct inode *inode;
1861
1862         switch (data->o_arg.claim) {
1863         case NFS4_OPEN_CLAIM_NULL:
1864         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1865         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1866                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1867                         return ERR_PTR(-EAGAIN);
1868                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1869                                 &data->f_attr, data->f_label);
1870                 break;
1871         default:
1872                 inode = d_inode(data->dentry);
1873                 ihold(inode);
1874                 nfs_refresh_inode(inode, &data->f_attr);
1875         }
1876         return inode;
1877 }
1878
1879 static struct nfs4_state *
1880 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1881 {
1882         struct nfs4_state *state;
1883         struct inode *inode;
1884
1885         inode = nfs4_opendata_get_inode(data);
1886         if (IS_ERR(inode))
1887                 return ERR_CAST(inode);
1888         if (data->state != NULL && data->state->inode == inode) {
1889                 state = data->state;
1890                 atomic_inc(&state->count);
1891         } else
1892                 state = nfs4_get_open_state(inode, data->owner);
1893         iput(inode);
1894         if (state == NULL)
1895                 state = ERR_PTR(-ENOMEM);
1896         return state;
1897 }
1898
1899 static struct nfs4_state *
1900 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1901 {
1902         struct nfs4_state *state;
1903
1904         if (!data->rpc_done) {
1905                 state = nfs4_try_open_cached(data);
1906                 trace_nfs4_cached_open(data->state);
1907                 goto out;
1908         }
1909
1910         state = nfs4_opendata_find_nfs4_state(data);
1911         if (IS_ERR(state))
1912                 goto out;
1913
1914         if (data->o_res.delegation_type != 0)
1915                 nfs4_opendata_check_deleg(data, state);
1916         update_open_stateid(state, &data->o_res.stateid, NULL,
1917                         data->o_arg.fmode);
1918 out:
1919         nfs_release_seqid(data->o_arg.seqid);
1920         return state;
1921 }
1922
1923 static struct nfs4_state *
1924 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1925 {
1926         struct nfs4_state *ret;
1927
1928         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1929                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1930         else
1931                 ret = _nfs4_opendata_to_nfs4_state(data);
1932         nfs4_sequence_free_slot(&data->o_res.seq_res);
1933         return ret;
1934 }
1935
1936 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1937 {
1938         struct nfs_inode *nfsi = NFS_I(state->inode);
1939         struct nfs_open_context *ctx;
1940
1941         spin_lock(&state->inode->i_lock);
1942         list_for_each_entry(ctx, &nfsi->open_files, list) {
1943                 if (ctx->state != state)
1944                         continue;
1945                 get_nfs_open_context(ctx);
1946                 spin_unlock(&state->inode->i_lock);
1947                 return ctx;
1948         }
1949         spin_unlock(&state->inode->i_lock);
1950         return ERR_PTR(-ENOENT);
1951 }
1952
1953 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1954                 struct nfs4_state *state, enum open_claim_type4 claim)
1955 {
1956         struct nfs4_opendata *opendata;
1957
1958         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1959                         NULL, claim, GFP_NOFS);
1960         if (opendata == NULL)
1961                 return ERR_PTR(-ENOMEM);
1962         opendata->state = state;
1963         atomic_inc(&state->count);
1964         return opendata;
1965 }
1966
1967 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1968                 fmode_t fmode)
1969 {
1970         struct nfs4_state *newstate;
1971         int ret;
1972
1973         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1974                 return 0;
1975         opendata->o_arg.open_flags = 0;
1976         opendata->o_arg.fmode = fmode;
1977         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1978                         NFS_SB(opendata->dentry->d_sb),
1979                         fmode, 0);
1980         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1981         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1982         nfs4_init_opendata_res(opendata);
1983         ret = _nfs4_recover_proc_open(opendata);
1984         if (ret != 0)
1985                 return ret;
1986         newstate = nfs4_opendata_to_nfs4_state(opendata);
1987         if (IS_ERR(newstate))
1988                 return PTR_ERR(newstate);
1989         if (newstate != opendata->state)
1990                 ret = -ESTALE;
1991         nfs4_close_state(newstate, fmode);
1992         return ret;
1993 }
1994
1995 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1996 {
1997         int ret;
1998
1999         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
2000         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2001         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2002         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2003         /* memory barrier prior to reading state->n_* */
2004         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2005         clear_bit(NFS_OPEN_STATE, &state->flags);
2006         smp_rmb();
2007         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2008         if (ret != 0)
2009                 return ret;
2010         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2011         if (ret != 0)
2012                 return ret;
2013         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2014         if (ret != 0)
2015                 return ret;
2016         /*
2017          * We may have performed cached opens for all three recoveries.
2018          * Check if we need to update the current stateid.
2019          */
2020         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2021             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2022                 write_seqlock(&state->seqlock);
2023                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2024                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2025                 write_sequnlock(&state->seqlock);
2026         }
2027         return 0;
2028 }
2029
2030 /*
2031  * OPEN_RECLAIM:
2032  *      reclaim state on the server after a reboot.
2033  */
2034 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2035 {
2036         struct nfs_delegation *delegation;
2037         struct nfs4_opendata *opendata;
2038         fmode_t delegation_type = 0;
2039         int status;
2040
2041         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2042                         NFS4_OPEN_CLAIM_PREVIOUS);
2043         if (IS_ERR(opendata))
2044                 return PTR_ERR(opendata);
2045         rcu_read_lock();
2046         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2047         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2048                 delegation_type = delegation->type;
2049         rcu_read_unlock();
2050         opendata->o_arg.u.delegation_type = delegation_type;
2051         status = nfs4_open_recover(opendata, state);
2052         nfs4_opendata_put(opendata);
2053         return status;
2054 }
2055
2056 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2057 {
2058         struct nfs_server *server = NFS_SERVER(state->inode);
2059         struct nfs4_exception exception = { };
2060         int err;
2061         do {
2062                 err = _nfs4_do_open_reclaim(ctx, state);
2063                 trace_nfs4_open_reclaim(ctx, 0, err);
2064                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2065                         continue;
2066                 if (err != -NFS4ERR_DELAY)
2067                         break;
2068                 nfs4_handle_exception(server, err, &exception);
2069         } while (exception.retry);
2070         return err;
2071 }
2072
2073 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2074 {
2075         struct nfs_open_context *ctx;
2076         int ret;
2077
2078         ctx = nfs4_state_find_open_context(state);
2079         if (IS_ERR(ctx))
2080                 return -EAGAIN;
2081         ret = nfs4_do_open_reclaim(ctx, state);
2082         put_nfs_open_context(ctx);
2083         return ret;
2084 }
2085
2086 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2087 {
2088         switch (err) {
2089                 default:
2090                         printk(KERN_ERR "NFS: %s: unhandled error "
2091                                         "%d.\n", __func__, err);
2092                 case 0:
2093                 case -ENOENT:
2094                 case -EAGAIN:
2095                 case -ESTALE:
2096                         break;
2097                 case -NFS4ERR_BADSESSION:
2098                 case -NFS4ERR_BADSLOT:
2099                 case -NFS4ERR_BAD_HIGH_SLOT:
2100                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2101                 case -NFS4ERR_DEADSESSION:
2102                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2103                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2104                         return -EAGAIN;
2105                 case -NFS4ERR_STALE_CLIENTID:
2106                 case -NFS4ERR_STALE_STATEID:
2107                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2108                         /* Don't recall a delegation if it was lost */
2109                         nfs4_schedule_lease_recovery(server->nfs_client);
2110                         return -EAGAIN;
2111                 case -NFS4ERR_MOVED:
2112                         nfs4_schedule_migration_recovery(server);
2113                         return -EAGAIN;
2114                 case -NFS4ERR_LEASE_MOVED:
2115                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2116                         return -EAGAIN;
2117                 case -NFS4ERR_DELEG_REVOKED:
2118                 case -NFS4ERR_ADMIN_REVOKED:
2119                 case -NFS4ERR_EXPIRED:
2120                 case -NFS4ERR_BAD_STATEID:
2121                 case -NFS4ERR_OPENMODE:
2122                         nfs_inode_find_state_and_recover(state->inode,
2123                                         stateid);
2124                         nfs4_schedule_stateid_recovery(server, state);
2125                         return -EAGAIN;
2126                 case -NFS4ERR_DELAY:
2127                 case -NFS4ERR_GRACE:
2128                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2129                         ssleep(1);
2130                         return -EAGAIN;
2131                 case -ENOMEM:
2132                 case -NFS4ERR_DENIED:
2133                         if (fl) {
2134                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2135                                 if (lsp)
2136                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2137                         }
2138                         return 0;
2139         }
2140         return err;
2141 }
2142
2143 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2144                 struct nfs4_state *state, const nfs4_stateid *stateid,
2145                 fmode_t type)
2146 {
2147         struct nfs_server *server = NFS_SERVER(state->inode);
2148         struct nfs4_opendata *opendata;
2149         int err = 0;
2150
2151         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2152                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2153         if (IS_ERR(opendata))
2154                 return PTR_ERR(opendata);
2155         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2156         nfs_state_clear_delegation(state);
2157         switch (type & (FMODE_READ|FMODE_WRITE)) {
2158         case FMODE_READ|FMODE_WRITE:
2159         case FMODE_WRITE:
2160                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2161                 if (err)
2162                         break;
2163                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2164                 if (err)
2165                         break;
2166                 /* Fall through */
2167         case FMODE_READ:
2168                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2169         }
2170         nfs4_opendata_put(opendata);
2171         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2172 }
2173
2174 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2175 {
2176         struct nfs4_opendata *data = calldata;
2177
2178         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2179                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2180 }
2181
2182 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2183 {
2184         struct nfs4_opendata *data = calldata;
2185
2186         nfs40_sequence_done(task, &data->c_res.seq_res);
2187
2188         data->rpc_status = task->tk_status;
2189         if (data->rpc_status == 0) {
2190                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2191                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2192                 renew_lease(data->o_res.server, data->timestamp);
2193                 data->rpc_done = true;
2194         }
2195 }
2196
2197 static void nfs4_open_confirm_release(void *calldata)
2198 {
2199         struct nfs4_opendata *data = calldata;
2200         struct nfs4_state *state = NULL;
2201
2202         /* If this request hasn't been cancelled, do nothing */
2203         if (!data->cancelled)
2204                 goto out_free;
2205         /* In case of error, no cleanup! */
2206         if (!data->rpc_done)
2207                 goto out_free;
2208         state = nfs4_opendata_to_nfs4_state(data);
2209         if (!IS_ERR(state))
2210                 nfs4_close_state(state, data->o_arg.fmode);
2211 out_free:
2212         nfs4_opendata_put(data);
2213 }
2214
2215 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2216         .rpc_call_prepare = nfs4_open_confirm_prepare,
2217         .rpc_call_done = nfs4_open_confirm_done,
2218         .rpc_release = nfs4_open_confirm_release,
2219 };
2220
2221 /*
2222  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2223  */
2224 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2225 {
2226         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2227         struct rpc_task *task;
2228         struct  rpc_message msg = {
2229                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2230                 .rpc_argp = &data->c_arg,
2231                 .rpc_resp = &data->c_res,
2232                 .rpc_cred = data->owner->so_cred,
2233         };
2234         struct rpc_task_setup task_setup_data = {
2235                 .rpc_client = server->client,
2236                 .rpc_message = &msg,
2237                 .callback_ops = &nfs4_open_confirm_ops,
2238                 .callback_data = data,
2239                 .workqueue = nfsiod_workqueue,
2240                 .flags = RPC_TASK_ASYNC,
2241         };
2242         int status;
2243
2244         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2245                                 data->is_recover);
2246         kref_get(&data->kref);
2247         data->rpc_done = false;
2248         data->rpc_status = 0;
2249         data->timestamp = jiffies;
2250         task = rpc_run_task(&task_setup_data);
2251         if (IS_ERR(task))
2252                 return PTR_ERR(task);
2253         status = rpc_wait_for_completion_task(task);
2254         if (status != 0) {
2255                 data->cancelled = true;
2256                 smp_wmb();
2257         } else
2258                 status = data->rpc_status;
2259         rpc_put_task(task);
2260         return status;
2261 }
2262
2263 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2264 {
2265         struct nfs4_opendata *data = calldata;
2266         struct nfs4_state_owner *sp = data->owner;
2267         struct nfs_client *clp = sp->so_server->nfs_client;
2268         enum open_claim_type4 claim = data->o_arg.claim;
2269
2270         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2271                 goto out_wait;
2272         /*
2273          * Check if we still need to send an OPEN call, or if we can use
2274          * a delegation instead.
2275          */
2276         if (data->state != NULL) {
2277                 struct nfs_delegation *delegation;
2278
2279                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2280                         goto out_no_action;
2281                 rcu_read_lock();
2282                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2283                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2284                         goto unlock_no_action;
2285                 rcu_read_unlock();
2286         }
2287         /* Update client id. */
2288         data->o_arg.clientid = clp->cl_clientid;
2289         switch (claim) {
2290         default:
2291                 break;
2292         case NFS4_OPEN_CLAIM_PREVIOUS:
2293         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2294         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2295                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2296                 /* Fall through */
2297         case NFS4_OPEN_CLAIM_FH:
2298                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2299         }
2300         data->timestamp = jiffies;
2301         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2302                                 &data->o_arg.seq_args,
2303                                 &data->o_res.seq_res,
2304                                 task) != 0)
2305                 nfs_release_seqid(data->o_arg.seqid);
2306
2307         /* Set the create mode (note dependency on the session type) */
2308         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2309         if (data->o_arg.open_flags & O_EXCL) {
2310                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2311                 if (nfs4_has_persistent_session(clp))
2312                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2313                 else if (clp->cl_mvops->minor_version > 0)
2314                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2315         }
2316         return;
2317 unlock_no_action:
2318         trace_nfs4_cached_open(data->state);
2319         rcu_read_unlock();
2320 out_no_action:
2321         task->tk_action = NULL;
2322 out_wait:
2323         nfs4_sequence_done(task, &data->o_res.seq_res);
2324 }
2325
2326 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2327 {
2328         struct nfs4_opendata *data = calldata;
2329
2330         data->rpc_status = task->tk_status;
2331
2332         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2333                 return;
2334
2335         if (task->tk_status == 0) {
2336                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2337                         switch (data->o_res.f_attr->mode & S_IFMT) {
2338                         case S_IFREG:
2339                                 break;
2340                         case S_IFLNK:
2341                                 data->rpc_status = -ELOOP;
2342                                 break;
2343                         case S_IFDIR:
2344                                 data->rpc_status = -EISDIR;
2345                                 break;
2346                         default:
2347                                 data->rpc_status = -ENOTDIR;
2348                         }
2349                 }
2350                 renew_lease(data->o_res.server, data->timestamp);
2351                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2352                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2353         }
2354         data->rpc_done = true;
2355 }
2356
2357 static void nfs4_open_release(void *calldata)
2358 {
2359         struct nfs4_opendata *data = calldata;
2360         struct nfs4_state *state = NULL;
2361
2362         /* If this request hasn't been cancelled, do nothing */
2363         if (!data->cancelled)
2364                 goto out_free;
2365         /* In case of error, no cleanup! */
2366         if (data->rpc_status != 0 || !data->rpc_done)
2367                 goto out_free;
2368         /* In case we need an open_confirm, no cleanup! */
2369         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2370                 goto out_free;
2371         state = nfs4_opendata_to_nfs4_state(data);
2372         if (!IS_ERR(state))
2373                 nfs4_close_state(state, data->o_arg.fmode);
2374 out_free:
2375         nfs4_opendata_put(data);
2376 }
2377
2378 static const struct rpc_call_ops nfs4_open_ops = {
2379         .rpc_call_prepare = nfs4_open_prepare,
2380         .rpc_call_done = nfs4_open_done,
2381         .rpc_release = nfs4_open_release,
2382 };
2383
2384 static int nfs4_run_open_task(struct nfs4_opendata *data,
2385                               struct nfs_open_context *ctx)
2386 {
2387         struct inode *dir = d_inode(data->dir);
2388         struct nfs_server *server = NFS_SERVER(dir);
2389         struct nfs_openargs *o_arg = &data->o_arg;
2390         struct nfs_openres *o_res = &data->o_res;
2391         struct rpc_task *task;
2392         struct rpc_message msg = {
2393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2394                 .rpc_argp = o_arg,
2395                 .rpc_resp = o_res,
2396                 .rpc_cred = data->owner->so_cred,
2397         };
2398         struct rpc_task_setup task_setup_data = {
2399                 .rpc_client = server->client,
2400                 .rpc_message = &msg,
2401                 .callback_ops = &nfs4_open_ops,
2402                 .callback_data = data,
2403                 .workqueue = nfsiod_workqueue,
2404                 .flags = RPC_TASK_ASYNC,
2405         };
2406         int status;
2407
2408         kref_get(&data->kref);
2409         data->rpc_done = false;
2410         data->rpc_status = 0;
2411         data->cancelled = false;
2412         data->is_recover = false;
2413         if (!ctx) {
2414                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2415                 data->is_recover = true;
2416         } else {
2417                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2418                 pnfs_lgopen_prepare(data, ctx);
2419         }
2420         task = rpc_run_task(&task_setup_data);
2421         if (IS_ERR(task))
2422                 return PTR_ERR(task);
2423         status = rpc_wait_for_completion_task(task);
2424         if (status != 0) {
2425                 data->cancelled = true;
2426                 smp_wmb();
2427         } else
2428                 status = data->rpc_status;
2429         rpc_put_task(task);
2430
2431         return status;
2432 }
2433
2434 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2435 {
2436         struct inode *dir = d_inode(data->dir);
2437         struct nfs_openres *o_res = &data->o_res;
2438         int status;
2439
2440         status = nfs4_run_open_task(data, NULL);
2441         if (status != 0 || !data->rpc_done)
2442                 return status;
2443
2444         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2445
2446         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2447                 status = _nfs4_proc_open_confirm(data);
2448
2449         return status;
2450 }
2451
2452 /*
2453  * Additional permission checks in order to distinguish between an
2454  * open for read, and an open for execute. This works around the
2455  * fact that NFSv4 OPEN treats read and execute permissions as being
2456  * the same.
2457  * Note that in the non-execute case, we want to turn off permission
2458  * checking if we just created a new file (POSIX open() semantics).
2459  */
2460 static int nfs4_opendata_access(struct rpc_cred *cred,
2461                                 struct nfs4_opendata *opendata,
2462                                 struct nfs4_state *state, fmode_t fmode,
2463                                 int openflags)
2464 {
2465         struct nfs_access_entry cache;
2466         u32 mask, flags;
2467
2468         /* access call failed or for some reason the server doesn't
2469          * support any access modes -- defer access call until later */
2470         if (opendata->o_res.access_supported == 0)
2471                 return 0;
2472
2473         mask = 0;
2474         /*
2475          * Use openflags to check for exec, because fmode won't
2476          * always have FMODE_EXEC set when file open for exec.
2477          */
2478         if (openflags & __FMODE_EXEC) {
2479                 /* ONLY check for exec rights */
2480                 if (S_ISDIR(state->inode->i_mode))
2481                         mask = NFS4_ACCESS_LOOKUP;
2482                 else
2483                         mask = NFS4_ACCESS_EXECUTE;
2484         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2485                 mask = NFS4_ACCESS_READ;
2486
2487         cache.cred = cred;
2488         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2489         nfs_access_add_cache(state->inode, &cache);
2490
2491         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2492         if ((mask & ~cache.mask & flags) == 0)
2493                 return 0;
2494
2495         return -EACCES;
2496 }
2497
2498 /*
2499  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2500  */
2501 static int _nfs4_proc_open(struct nfs4_opendata *data,
2502                            struct nfs_open_context *ctx)
2503 {
2504         struct inode *dir = d_inode(data->dir);
2505         struct nfs_server *server = NFS_SERVER(dir);
2506         struct nfs_openargs *o_arg = &data->o_arg;
2507         struct nfs_openres *o_res = &data->o_res;
2508         int status;
2509
2510         status = nfs4_run_open_task(data, ctx);
2511         if (!data->rpc_done)
2512                 return status;
2513         if (status != 0) {
2514                 if (status == -NFS4ERR_BADNAME &&
2515                                 !(o_arg->open_flags & O_CREAT))
2516                         return -ENOENT;
2517                 return status;
2518         }
2519
2520         nfs_fattr_map_and_free_names(server, &data->f_attr);
2521
2522         if (o_arg->open_flags & O_CREAT) {
2523                 if (o_arg->open_flags & O_EXCL)
2524                         data->file_created = true;
2525                 else if (o_res->cinfo.before != o_res->cinfo.after)
2526                         data->file_created = true;
2527                 if (data->file_created ||
2528                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2529                         update_changeattr(dir, &o_res->cinfo,
2530                                         o_res->f_attr->time_start, 0);
2531         }
2532         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2533                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2534         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2535                 status = _nfs4_proc_open_confirm(data);
2536                 if (status != 0)
2537                         return status;
2538         }
2539         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2540                 nfs4_sequence_free_slot(&o_res->seq_res);
2541                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2542                                 o_res->f_label, NULL);
2543         }
2544         return 0;
2545 }
2546
2547 /*
2548  * OPEN_EXPIRED:
2549  *      reclaim state on the server after a network partition.
2550  *      Assumes caller holds the appropriate lock
2551  */
2552 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2553 {
2554         struct nfs4_opendata *opendata;
2555         int ret;
2556
2557         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2558                         NFS4_OPEN_CLAIM_FH);
2559         if (IS_ERR(opendata))
2560                 return PTR_ERR(opendata);
2561         ret = nfs4_open_recover(opendata, state);
2562         if (ret == -ESTALE)
2563                 d_drop(ctx->dentry);
2564         nfs4_opendata_put(opendata);
2565         return ret;
2566 }
2567
2568 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2569 {
2570         struct nfs_server *server = NFS_SERVER(state->inode);
2571         struct nfs4_exception exception = { };
2572         int err;
2573
2574         do {
2575                 err = _nfs4_open_expired(ctx, state);
2576                 trace_nfs4_open_expired(ctx, 0, err);
2577                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2578                         continue;
2579                 switch (err) {
2580                 default:
2581                         goto out;
2582                 case -NFS4ERR_GRACE:
2583                 case -NFS4ERR_DELAY:
2584                         nfs4_handle_exception(server, err, &exception);
2585                         err = 0;
2586                 }
2587         } while (exception.retry);
2588 out:
2589         return err;
2590 }
2591
2592 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2593 {
2594         struct nfs_open_context *ctx;
2595         int ret;
2596
2597         ctx = nfs4_state_find_open_context(state);
2598         if (IS_ERR(ctx))
2599                 return -EAGAIN;
2600         ret = nfs4_do_open_expired(ctx, state);
2601         put_nfs_open_context(ctx);
2602         return ret;
2603 }
2604
2605 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2606                 const nfs4_stateid *stateid)
2607 {
2608         nfs_remove_bad_delegation(state->inode, stateid);
2609         nfs_state_clear_delegation(state);
2610 }
2611
2612 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2613 {
2614         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2615                 nfs_finish_clear_delegation_stateid(state, NULL);
2616 }
2617
2618 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2619 {
2620         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2621         nfs40_clear_delegation_stateid(state);
2622         return nfs4_open_expired(sp, state);
2623 }
2624
2625 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2626                 nfs4_stateid *stateid,
2627                 struct rpc_cred *cred)
2628 {
2629         return -NFS4ERR_BAD_STATEID;
2630 }
2631
2632 #if defined(CONFIG_NFS_V4_1)
2633 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2634                 nfs4_stateid *stateid,
2635                 struct rpc_cred *cred)
2636 {
2637         int status;
2638
2639         switch (stateid->type) {
2640         default:
2641                 break;
2642         case NFS4_INVALID_STATEID_TYPE:
2643         case NFS4_SPECIAL_STATEID_TYPE:
2644                 return -NFS4ERR_BAD_STATEID;
2645         case NFS4_REVOKED_STATEID_TYPE:
2646                 goto out_free;
2647         }
2648
2649         status = nfs41_test_stateid(server, stateid, cred);
2650         switch (status) {
2651         case -NFS4ERR_EXPIRED:
2652         case -NFS4ERR_ADMIN_REVOKED:
2653         case -NFS4ERR_DELEG_REVOKED:
2654                 break;
2655         default:
2656                 return status;
2657         }
2658 out_free:
2659         /* Ack the revoked state to the server */
2660         nfs41_free_stateid(server, stateid, cred, true);
2661         return -NFS4ERR_EXPIRED;
2662 }
2663
2664 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2665 {
2666         struct nfs_server *server = NFS_SERVER(state->inode);
2667         nfs4_stateid stateid;
2668         struct nfs_delegation *delegation;
2669         struct rpc_cred *cred;
2670         int status;
2671
2672         /* Get the delegation credential for use by test/free_stateid */
2673         rcu_read_lock();
2674         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2675         if (delegation == NULL) {
2676                 rcu_read_unlock();
2677                 nfs_state_clear_delegation(state);
2678                 return;
2679         }
2680
2681         nfs4_stateid_copy(&stateid, &delegation->stateid);
2682         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2683                 rcu_read_unlock();
2684                 nfs_state_clear_delegation(state);
2685                 return;
2686         }
2687
2688         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2689                                 &delegation->flags)) {
2690                 rcu_read_unlock();
2691                 return;
2692         }
2693
2694         cred = get_rpccred(delegation->cred);
2695         rcu_read_unlock();
2696         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2697         trace_nfs4_test_delegation_stateid(state, NULL, status);
2698         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2699                 nfs_finish_clear_delegation_stateid(state, &stateid);
2700
2701         put_rpccred(cred);
2702 }
2703
2704 /**
2705  * nfs41_check_expired_locks - possibly free a lock stateid
2706  *
2707  * @state: NFSv4 state for an inode
2708  *
2709  * Returns NFS_OK if recovery for this stateid is now finished.
2710  * Otherwise a negative NFS4ERR value is returned.
2711  */
2712 static int nfs41_check_expired_locks(struct nfs4_state *state)
2713 {
2714         int status, ret = NFS_OK;
2715         struct nfs4_lock_state *lsp, *prev = NULL;
2716         struct nfs_server *server = NFS_SERVER(state->inode);
2717
2718         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2719                 goto out;
2720
2721         spin_lock(&state->state_lock);
2722         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2723                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2724                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2725
2726                         refcount_inc(&lsp->ls_count);
2727                         spin_unlock(&state->state_lock);
2728
2729                         nfs4_put_lock_state(prev);
2730                         prev = lsp;
2731
2732                         status = nfs41_test_and_free_expired_stateid(server,
2733                                         &lsp->ls_stateid,
2734                                         cred);
2735                         trace_nfs4_test_lock_stateid(state, lsp, status);
2736                         if (status == -NFS4ERR_EXPIRED ||
2737                             status == -NFS4ERR_BAD_STATEID) {
2738                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2739                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2740                                 if (!recover_lost_locks)
2741                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2742                         } else if (status != NFS_OK) {
2743                                 ret = status;
2744                                 nfs4_put_lock_state(prev);
2745                                 goto out;
2746                         }
2747                         spin_lock(&state->state_lock);
2748                 }
2749         }
2750         spin_unlock(&state->state_lock);
2751         nfs4_put_lock_state(prev);
2752 out:
2753         return ret;
2754 }
2755
2756 /**
2757  * nfs41_check_open_stateid - possibly free an open stateid
2758  *
2759  * @state: NFSv4 state for an inode
2760  *
2761  * Returns NFS_OK if recovery for this stateid is now finished.
2762  * Otherwise a negative NFS4ERR value is returned.
2763  */
2764 static int nfs41_check_open_stateid(struct nfs4_state *state)
2765 {
2766         struct nfs_server *server = NFS_SERVER(state->inode);
2767         nfs4_stateid *stateid = &state->open_stateid;
2768         struct rpc_cred *cred = state->owner->so_cred;
2769         int status;
2770
2771         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2772                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2773                         if (nfs4_have_delegation(state->inode, state->state))
2774                                 return NFS_OK;
2775                         return -NFS4ERR_OPENMODE;
2776                 }
2777                 return -NFS4ERR_BAD_STATEID;
2778         }
2779         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2780         trace_nfs4_test_open_stateid(state, NULL, status);
2781         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2782                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2783                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2784                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2785                 clear_bit(NFS_OPEN_STATE, &state->flags);
2786                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2787                 return status;
2788         }
2789         if (nfs_open_stateid_recover_openmode(state))
2790                 return -NFS4ERR_OPENMODE;
2791         return NFS_OK;
2792 }
2793
2794 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2795 {
2796         int status;
2797
2798         nfs41_check_delegation_stateid(state);
2799         status = nfs41_check_expired_locks(state);
2800         if (status != NFS_OK)
2801                 return status;
2802         status = nfs41_check_open_stateid(state);
2803         if (status != NFS_OK)
2804                 status = nfs4_open_expired(sp, state);
2805         return status;
2806 }
2807 #endif
2808
2809 /*
2810  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2811  * fields corresponding to attributes that were used to store the verifier.
2812  * Make sure we clobber those fields in the later setattr call
2813  */
2814 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2815                                 struct iattr *sattr, struct nfs4_label **label)
2816 {
2817         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2818         __u32 attrset[3];
2819         unsigned ret;
2820         unsigned i;
2821
2822         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2823                 attrset[i] = opendata->o_res.attrset[i];
2824                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2825                         attrset[i] &= ~bitmask[i];
2826         }
2827
2828         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2829                 sattr->ia_valid : 0;
2830
2831         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2832                 if (sattr->ia_valid & ATTR_ATIME_SET)
2833                         ret |= ATTR_ATIME_SET;
2834                 else
2835                         ret |= ATTR_ATIME;
2836         }
2837
2838         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2839                 if (sattr->ia_valid & ATTR_MTIME_SET)
2840                         ret |= ATTR_MTIME_SET;
2841                 else
2842                         ret |= ATTR_MTIME;
2843         }
2844
2845         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2846                 *label = NULL;
2847         return ret;
2848 }
2849
2850 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2851                 fmode_t fmode,
2852                 int flags,
2853                 struct nfs_open_context *ctx)
2854 {
2855         struct nfs4_state_owner *sp = opendata->owner;
2856         struct nfs_server *server = sp->so_server;
2857         struct dentry *dentry;
2858         struct nfs4_state *state;
2859         unsigned int seq;
2860         int ret;
2861
2862         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2863
2864         ret = _nfs4_proc_open(opendata, ctx);
2865         if (ret != 0)
2866                 goto out;
2867
2868         state = _nfs4_opendata_to_nfs4_state(opendata);
2869         ret = PTR_ERR(state);
2870         if (IS_ERR(state))
2871                 goto out;
2872         ctx->state = state;
2873         if (server->caps & NFS_CAP_POSIX_LOCK)
2874                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2875         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2876                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2877
2878         dentry = opendata->dentry;
2879         if (d_really_is_negative(dentry)) {
2880                 struct dentry *alias;
2881                 d_drop(dentry);
2882                 alias = d_exact_alias(dentry, state->inode);
2883                 if (!alias)
2884                         alias = d_splice_alias(igrab(state->inode), dentry);
2885                 /* d_splice_alias() can't fail here - it's a non-directory */
2886                 if (alias) {
2887                         dput(ctx->dentry);
2888                         ctx->dentry = dentry = alias;
2889                 }
2890                 nfs_set_verifier(dentry,
2891                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2892         }
2893
2894         /* Parse layoutget results before we check for access */
2895         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2896
2897         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2898         if (ret != 0)
2899                 goto out;
2900
2901         if (d_inode(dentry) == state->inode) {
2902                 nfs_inode_attach_open_context(ctx);
2903                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2904                         nfs4_schedule_stateid_recovery(server, state);
2905         }
2906
2907 out:
2908         nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2909         return ret;
2910 }
2911
2912 /*
2913  * Returns a referenced nfs4_state
2914  */
2915 static int _nfs4_do_open(struct inode *dir,
2916                         struct nfs_open_context *ctx,
2917                         int flags,
2918                         const struct nfs4_open_createattrs *c,
2919                         int *opened)
2920 {
2921         struct nfs4_state_owner  *sp;
2922         struct nfs4_state     *state = NULL;
2923         struct nfs_server       *server = NFS_SERVER(dir);
2924         struct nfs4_opendata *opendata;
2925         struct dentry *dentry = ctx->dentry;
2926         struct rpc_cred *cred = ctx->cred;
2927         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2928         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2929         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2930         struct iattr *sattr = c->sattr;
2931         struct nfs4_label *label = c->label;
2932         struct nfs4_label *olabel = NULL;
2933         int status;
2934
2935         /* Protect against reboot recovery conflicts */
2936         status = -ENOMEM;
2937         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2938         if (sp == NULL) {
2939                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2940                 goto out_err;
2941         }
2942         status = nfs4_client_recover_expired_lease(server->nfs_client);
2943         if (status != 0)
2944                 goto err_put_state_owner;
2945         if (d_really_is_positive(dentry))
2946                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2947         status = -ENOMEM;
2948         if (d_really_is_positive(dentry))
2949                 claim = NFS4_OPEN_CLAIM_FH;
2950         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2951                         c, claim, GFP_KERNEL);
2952         if (opendata == NULL)
2953                 goto err_put_state_owner;
2954
2955         if (label) {
2956                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2957                 if (IS_ERR(olabel)) {
2958                         status = PTR_ERR(olabel);
2959                         goto err_opendata_put;
2960                 }
2961         }
2962
2963         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2964                 if (!opendata->f_attr.mdsthreshold) {
2965                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2966                         if (!opendata->f_attr.mdsthreshold)
2967                                 goto err_free_label;
2968                 }
2969                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2970         }
2971         if (d_really_is_positive(dentry))
2972                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2973
2974         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2975         if (status != 0)
2976                 goto err_free_label;
2977         state = ctx->state;
2978
2979         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2980             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2981                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
2982                 /*
2983                  * send create attributes which was not set by open
2984                  * with an extra setattr.
2985                  */
2986                 if (attrs || label) {
2987                         unsigned ia_old = sattr->ia_valid;
2988
2989                         sattr->ia_valid = attrs;
2990                         nfs_fattr_init(opendata->o_res.f_attr);
2991                         status = nfs4_do_setattr(state->inode, cred,
2992                                         opendata->o_res.f_attr, sattr,
2993                                         ctx, label, olabel);
2994                         if (status == 0) {
2995                                 nfs_setattr_update_inode(state->inode, sattr,
2996                                                 opendata->o_res.f_attr);
2997                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2998                         }
2999                         sattr->ia_valid = ia_old;
3000                 }
3001         }
3002         if (opened && opendata->file_created)
3003                 *opened = 1;
3004
3005         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3006                 *ctx_th = opendata->f_attr.mdsthreshold;
3007                 opendata->f_attr.mdsthreshold = NULL;
3008         }
3009
3010         nfs4_label_free(olabel);
3011
3012         nfs4_opendata_put(opendata);
3013         nfs4_put_state_owner(sp);
3014         return 0;
3015 err_free_label:
3016         nfs4_label_free(olabel);
3017 err_opendata_put:
3018         nfs4_opendata_put(opendata);
3019 err_put_state_owner:
3020         nfs4_put_state_owner(sp);
3021 out_err:
3022         return status;
3023 }
3024
3025
3026 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3027                                         struct nfs_open_context *ctx,
3028                                         int flags,
3029                                         struct iattr *sattr,
3030                                         struct nfs4_label *label,
3031                                         int *opened)
3032 {
3033         struct nfs_server *server = NFS_SERVER(dir);
3034         struct nfs4_exception exception = { };
3035         struct nfs4_state *res;
3036         struct nfs4_open_createattrs c = {
3037                 .label = label,
3038                 .sattr = sattr,
3039                 .verf = {
3040                         [0] = (__u32)jiffies,
3041                         [1] = (__u32)current->pid,
3042                 },
3043         };
3044         int status;
3045
3046         do {
3047                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3048                 res = ctx->state;
3049                 trace_nfs4_open_file(ctx, flags, status);
3050                 if (status == 0)
3051                         break;
3052                 /* NOTE: BAD_SEQID means the server and client disagree about the
3053                  * book-keeping w.r.t. state-changing operations
3054                  * (OPEN/CLOSE/LOCK/LOCKU...)
3055                  * It is actually a sign of a bug on the client or on the server.
3056                  *
3057                  * If we receive a BAD_SEQID error in the particular case of
3058                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3059                  * have unhashed the old state_owner for us, and that we can
3060                  * therefore safely retry using a new one. We should still warn
3061                  * the user though...
3062                  */
3063                 if (status == -NFS4ERR_BAD_SEQID) {
3064                         pr_warn_ratelimited("NFS: v4 server %s "
3065                                         " returned a bad sequence-id error!\n",
3066                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3067                         exception.retry = 1;
3068                         continue;
3069                 }
3070                 /*
3071                  * BAD_STATEID on OPEN means that the server cancelled our
3072                  * state before it received the OPEN_CONFIRM.
3073                  * Recover by retrying the request as per the discussion
3074                  * on Page 181 of RFC3530.
3075                  */
3076                 if (status == -NFS4ERR_BAD_STATEID) {
3077                         exception.retry = 1;
3078                         continue;
3079                 }
3080                 if (status == -EAGAIN) {
3081                         /* We must have found a delegation */
3082                         exception.retry = 1;
3083                         continue;
3084                 }
3085                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3086                         continue;
3087                 res = ERR_PTR(nfs4_handle_exception(server,
3088                                         status, &exception));
3089         } while (exception.retry);
3090         return res;
3091 }
3092
3093 static int _nfs4_do_setattr(struct inode *inode,
3094                             struct nfs_setattrargs *arg,
3095                             struct nfs_setattrres *res,
3096                             struct rpc_cred *cred,
3097                             struct nfs_open_context *ctx)
3098 {
3099         struct nfs_server *server = NFS_SERVER(inode);
3100         struct rpc_message msg = {
3101                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3102                 .rpc_argp       = arg,
3103                 .rpc_resp       = res,
3104                 .rpc_cred       = cred,
3105         };
3106         struct rpc_cred *delegation_cred = NULL;
3107         unsigned long timestamp = jiffies;
3108         bool truncate;
3109         int status;
3110
3111         nfs_fattr_init(res->fattr);
3112
3113         /* Servers should only apply open mode checks for file size changes */
3114         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3115         if (!truncate)
3116                 goto zero_stateid;
3117
3118         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3119                 /* Use that stateid */
3120         } else if (ctx != NULL) {
3121                 struct nfs_lock_context *l_ctx;
3122                 if (!nfs4_valid_open_stateid(ctx->state))
3123                         return -EBADF;
3124                 l_ctx = nfs_get_lock_context(ctx);
3125                 if (IS_ERR(l_ctx))
3126                         return PTR_ERR(l_ctx);
3127                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3128                                                 &arg->stateid, &delegation_cred);
3129                 nfs_put_lock_context(l_ctx);
3130                 if (status == -EIO)
3131                         return -EBADF;
3132         } else {
3133 zero_stateid:
3134                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3135         }
3136         if (delegation_cred)
3137                 msg.rpc_cred = delegation_cred;
3138
3139         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3140
3141         put_rpccred(delegation_cred);
3142         if (status == 0 && ctx != NULL)
3143                 renew_lease(server, timestamp);
3144         trace_nfs4_setattr(inode, &arg->stateid, status);
3145         return status;
3146 }
3147
3148 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
3149                            struct nfs_fattr *fattr, struct iattr *sattr,
3150                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3151                            struct nfs4_label *olabel)
3152 {
3153         struct nfs_server *server = NFS_SERVER(inode);
3154         __u32 bitmask[NFS4_BITMASK_SZ];
3155         struct nfs4_state *state = ctx ? ctx->state : NULL;
3156         struct nfs_setattrargs  arg = {
3157                 .fh             = NFS_FH(inode),
3158                 .iap            = sattr,
3159                 .server         = server,
3160                 .bitmask = bitmask,
3161                 .label          = ilabel,
3162         };
3163         struct nfs_setattrres  res = {
3164                 .fattr          = fattr,
3165                 .label          = olabel,
3166                 .server         = server,
3167         };
3168         struct nfs4_exception exception = {
3169                 .state = state,
3170                 .inode = inode,
3171                 .stateid = &arg.stateid,
3172         };
3173         int err;
3174
3175         do {
3176                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3177                                 nfs4_bitmask(server, olabel),
3178                                 inode);
3179
3180                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3181                 switch (err) {
3182                 case -NFS4ERR_OPENMODE:
3183                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3184                                 pr_warn_once("NFSv4: server %s is incorrectly "
3185                                                 "applying open mode checks to "
3186                                                 "a SETATTR that is not "
3187                                                 "changing file size.\n",
3188                                                 server->nfs_client->cl_hostname);
3189                         }
3190                         if (state && !(state->state & FMODE_WRITE)) {
3191                                 err = -EBADF;
3192                                 if (sattr->ia_valid & ATTR_OPEN)
3193                                         err = -EACCES;
3194                                 goto out;
3195                         }
3196                 }
3197                 err = nfs4_handle_exception(server, err, &exception);
3198         } while (exception.retry);
3199 out:
3200         return err;
3201 }
3202
3203 static bool
3204 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3205 {
3206         if (inode == NULL || !nfs_have_layout(inode))
3207                 return false;
3208
3209         return pnfs_wait_on_layoutreturn(inode, task);
3210 }
3211
3212 struct nfs4_closedata {
3213         struct inode *inode;
3214         struct nfs4_state *state;
3215         struct nfs_closeargs arg;
3216         struct nfs_closeres res;
3217         struct {
3218                 struct nfs4_layoutreturn_args arg;
3219                 struct nfs4_layoutreturn_res res;
3220                 struct nfs4_xdr_opaque_data ld_private;
3221                 u32 roc_barrier;
3222                 bool roc;
3223         } lr;
3224         struct nfs_fattr fattr;
3225         unsigned long timestamp;
3226 };
3227
3228 static void nfs4_free_closedata(void *data)
3229 {
3230         struct nfs4_closedata *calldata = data;
3231         struct nfs4_state_owner *sp = calldata->state->owner;
3232         struct super_block *sb = calldata->state->inode->i_sb;
3233
3234         if (calldata->lr.roc)
3235                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3236                                 calldata->res.lr_ret);
3237         nfs4_put_open_state(calldata->state);
3238         nfs_free_seqid(calldata->arg.seqid);
3239         nfs4_put_state_owner(sp);
3240         nfs_sb_deactive(sb);
3241         kfree(calldata);
3242 }
3243
3244 static void nfs4_close_done(struct rpc_task *task, void *data)
3245 {
3246         struct nfs4_closedata *calldata = data;
3247         struct nfs4_state *state = calldata->state;
3248         struct nfs_server *server = NFS_SERVER(calldata->inode);
3249         nfs4_stateid *res_stateid = NULL;
3250         struct nfs4_exception exception = {
3251                 .state = state,
3252                 .inode = calldata->inode,
3253                 .stateid = &calldata->arg.stateid,
3254         };
3255
3256         dprintk("%s: begin!\n", __func__);
3257         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3258                 return;
3259         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3260
3261         /* Handle Layoutreturn errors */
3262         if (calldata->arg.lr_args && task->tk_status != 0) {
3263                 switch (calldata->res.lr_ret) {
3264                 default:
3265                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3266                         break;
3267                 case 0:
3268                         calldata->arg.lr_args = NULL;
3269                         calldata->res.lr_res = NULL;
3270                         break;
3271                 case -NFS4ERR_OLD_STATEID:
3272                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3273                                                 &calldata->arg.lr_args->range,
3274                                                 calldata->inode))
3275                                 goto lr_restart;
3276                         /* Fallthrough */
3277                 case -NFS4ERR_ADMIN_REVOKED:
3278                 case -NFS4ERR_DELEG_REVOKED:
3279                 case -NFS4ERR_EXPIRED:
3280                 case -NFS4ERR_BAD_STATEID: