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, const struct 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                 const struct cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 const struct cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 const struct 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                 const struct 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                 const struct 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                 const struct 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_killable(long *timeout)
 404 {
 405         might_sleep();
 406
 407         freezable_schedule_timeout_killable_unsafe(
 408                 nfs4_update_delay(timeout));
 409         if (!__fatal_signal_pending(current))
 410                 return 0;
 411         return -EINTR;
 412 }
 413
 414 static int nfs4_delay_interruptible(long *timeout)
 415 {
 416         might_sleep();
 417
 418         freezable_schedule_timeout_interruptible(nfs4_update_delay(timeout));
 419         if (!signal_pending(current))
 420                 return 0;
 421         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 422 }
 423
 424 static int nfs4_delay(long *timeout, bool interruptible)
 425 {
 426         if (interruptible)
 427                 return nfs4_delay_interruptible(timeout);
 428         return nfs4_delay_killable(timeout);
 429 }
 430
 431 static const nfs4_stateid *
 432 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 433 {
 434         if (!stateid)
 435                 return NULL;
 436         switch (stateid->type) {
 437         case NFS4_OPEN_STATEID_TYPE:
 438         case NFS4_LOCK_STATEID_TYPE:
 439         case NFS4_DELEGATION_STATEID_TYPE:
 440                 return stateid;
 441         default:
 442                 break;
 443         }
 444         return NULL;
 445 }
 446
 447 /* This is the error handling routine for processes that are allowed
 448  * to sleep.
 449  */
 450 static int nfs4_do_handle_exception(struct nfs_server *server,
 451                 int errorcode, struct nfs4_exception *exception)
 452 {
 453         struct nfs_client *clp = server->nfs_client;
 454         struct nfs4_state *state = exception->state;
 455         const nfs4_stateid *stateid;
 456         struct inode *inode = exception->inode;
 457         int ret = errorcode;
 458
 459         exception->delay = 0;
 460         exception->recovering = 0;
 461         exception->retry = 0;
 462
 463         stateid = nfs4_recoverable_stateid(exception->stateid);
 464         if (stateid == NULL && state != NULL)
 465                 stateid = nfs4_recoverable_stateid(&state->stateid);
 466
 467         switch(errorcode) {
 468                 case 0:
 469                         return 0;
 470                 case -NFS4ERR_BADHANDLE:
 471                 case -ESTALE:
 472                         if (inode != NULL && S_ISREG(inode->i_mode))
 473                                 pnfs_destroy_layout(NFS_I(inode));
 474                         break;
 475                 case -NFS4ERR_DELEG_REVOKED:
 476                 case -NFS4ERR_ADMIN_REVOKED:
 477                 case -NFS4ERR_EXPIRED:
 478                 case -NFS4ERR_BAD_STATEID:
 479                         if (inode != NULL && stateid != NULL) {
 480                                 nfs_inode_find_state_and_recover(inode,
 481                                                 stateid);
 482                                 goto wait_on_recovery;
 483                         }
 484                         /* Fall through */
 485                 case -NFS4ERR_OPENMODE:
 486                         if (inode) {
 487                                 int err;
 488
 489                                 err = nfs_async_inode_return_delegation(inode,
 490                                                 stateid);
 491                                 if (err == 0)
 492                                         goto wait_on_recovery;
 493                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 494                                         exception->retry = 1;
 495                                         break;
 496                                 }
 497                         }
 498                         if (state == NULL)
 499                                 break;
 500                         ret = nfs4_schedule_stateid_recovery(server, state);
 501                         if (ret < 0)
 502                                 break;
 503                         goto wait_on_recovery;
 504                 case -NFS4ERR_STALE_STATEID:
 505                 case -NFS4ERR_STALE_CLIENTID:
 506                         nfs4_schedule_lease_recovery(clp);
 507                         goto wait_on_recovery;
 508                 case -NFS4ERR_MOVED:
 509                         ret = nfs4_schedule_migration_recovery(server);
 510                         if (ret < 0)
 511                                 break;
 512                         goto wait_on_recovery;
 513                 case -NFS4ERR_LEASE_MOVED:
 514                         nfs4_schedule_lease_moved_recovery(clp);
 515                         goto wait_on_recovery;
 516 #if defined(CONFIG_NFS_V4_1)
 517                 case -NFS4ERR_BADSESSION:
 518                 case -NFS4ERR_BADSLOT:
 519                 case -NFS4ERR_BAD_HIGH_SLOT:
 520                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 521                 case -NFS4ERR_DEADSESSION:
 522                 case -NFS4ERR_SEQ_FALSE_RETRY:
 523                 case -NFS4ERR_SEQ_MISORDERED:
 524                         dprintk("%s ERROR: %d Reset session\n", __func__,
 525                                 errorcode);
 526                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 527                         goto wait_on_recovery;
 528 #endif /* defined(CONFIG_NFS_V4_1) */
 529                 case -NFS4ERR_FILE_OPEN:
 530                         if (exception->timeout > HZ) {
 531                                 /* We have retried a decent amount, time to
 532                                  * fail
 533                                  */
 534                                 ret = -EBUSY;
 535                                 break;
 536                         }
 537                         /* Fall through */
 538                 case -NFS4ERR_DELAY:
 539                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 540                         /* Fall through */
 541                 case -NFS4ERR_GRACE:
 542                 case -NFS4ERR_LAYOUTTRYLATER:
 543                 case -NFS4ERR_RECALLCONFLICT:
 544                         exception->delay = 1;
 545                         return 0;
 546
 547                 case -NFS4ERR_RETRY_UNCACHED_REP:
 548                 case -NFS4ERR_OLD_STATEID:
 549                         exception->retry = 1;
 550                         break;
 551                 case -NFS4ERR_BADOWNER:
 552                         /* The following works around a Linux server bug! */
 553                 case -NFS4ERR_BADNAME:
 554                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 555                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 556                                 exception->retry = 1;
 557                                 printk(KERN_WARNING "NFS: v4 server %s "
 558                                                 "does not accept raw "
 559                                                 "uid/gids. "
 560                                                 "Reenabling the idmapper.\n",
 561                                                 server->nfs_client->cl_hostname);
 562                         }
 563         }
 564         /* We failed to handle the error */
 565         return nfs4_map_errors(ret);
 566 wait_on_recovery:
 567         exception->recovering = 1;
 568         return 0;
 569 }
 570
 571 /* This is the error handling routine for processes that are allowed
 572  * to sleep.
 573  */
 574 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 575 {
 576         struct nfs_client *clp = server->nfs_client;
 577         int ret;
 578
 579         ret = nfs4_do_handle_exception(server, errorcode, exception);
 580         if (exception->delay) {
 581                 ret = nfs4_delay(&exception->timeout,
 582                                 exception->interruptible);
 583                 goto out_retry;
 584         }
 585         if (exception->recovering) {
 586                 ret = nfs4_wait_clnt_recover(clp);
 587                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 588                         return -EIO;
 589                 goto out_retry;
 590         }
 591         return ret;
 592 out_retry:
 593         if (ret == 0)
 594                 exception->retry = 1;
 595         return ret;
 596 }
 597
 598 static int
 599 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 600                 int errorcode, struct nfs4_exception *exception)
 601 {
 602         struct nfs_client *clp = server->nfs_client;
 603         int ret;
 604
 605         ret = nfs4_do_handle_exception(server, errorcode, exception);
 606         if (exception->delay) {
 607                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 608                 goto out_retry;
 609         }
 610         if (exception->recovering) {
 611                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 612                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 613                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 614                 goto out_retry;
 615         }
 616         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 617                 ret = -EIO;
 618         return ret;
 619 out_retry:
 620         if (ret == 0) {
 621                 exception->retry = 1;
 622                 /*
 623                  * For NFS4ERR_MOVED, the client transport will need to
 624                  * be recomputed after migration recovery has completed.
 625                  */
 626                 if (errorcode == -NFS4ERR_MOVED)
 627                         rpc_task_release_transport(task);
 628         }
 629         return ret;
 630 }
 631
 632 int
 633 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 634                         struct nfs4_state *state, long *timeout)
 635 {
 636         struct nfs4_exception exception = {
 637                 .state = state,
 638         };
 639
 640         if (task->tk_status >= 0)
 641                 return 0;
 642         if (timeout)
 643                 exception.timeout = *timeout;
 644         task->tk_status = nfs4_async_handle_exception(task, server,
 645                         task->tk_status,
 646                         &exception);
 647         if (exception.delay && timeout)
 648                 *timeout = exception.timeout;
 649         if (exception.retry)
 650                 return -EAGAIN;
 651         return 0;
 652 }
 653
 654 /*
 655  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 656  * or 'false' otherwise.
 657  */
 658 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 659 {
 660         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 661         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 662 }
 663
 664 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 665 {
 666         spin_lock(&clp->cl_lock);
 667         if (time_before(clp->cl_last_renewal,timestamp))
 668                 clp->cl_last_renewal = timestamp;
 669         spin_unlock(&clp->cl_lock);
 670 }
 671
 672 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 673 {
 674         struct nfs_client *clp = server->nfs_client;
 675
 676         if (!nfs4_has_session(clp))
 677                 do_renew_lease(clp, timestamp);
 678 }
 679
 680 struct nfs4_call_sync_data {
 681         const struct nfs_server *seq_server;
 682         struct nfs4_sequence_args *seq_args;
 683         struct nfs4_sequence_res *seq_res;
 684 };
 685
 686 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 687                         struct nfs4_sequence_res *res, int cache_reply,
 688                         int privileged)
 689 {
 690         args->sa_slot = NULL;
 691         args->sa_cache_this = cache_reply;
 692         args->sa_privileged = privileged;
 693
 694         res->sr_slot = NULL;
 695 }
 696
 697 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 698 {
 699         struct nfs4_slot *slot = res->sr_slot;
 700         struct nfs4_slot_table *tbl;
 701
 702         tbl = slot->table;
 703         spin_lock(&tbl->slot_tbl_lock);
 704         if (!nfs41_wake_and_assign_slot(tbl, slot))
 705                 nfs4_free_slot(tbl, slot);
 706         spin_unlock(&tbl->slot_tbl_lock);
 707
 708         res->sr_slot = NULL;
 709 }
 710
 711 static int nfs40_sequence_done(struct rpc_task *task,
 712                                struct nfs4_sequence_res *res)
 713 {
 714         if (res->sr_slot != NULL)
 715                 nfs40_sequence_free_slot(res);
 716         return 1;
 717 }
 718
 719 #if defined(CONFIG_NFS_V4_1)
 720
 721 static void nfs41_release_slot(struct nfs4_slot *slot)
 722 {
 723         struct nfs4_session *session;
 724         struct nfs4_slot_table *tbl;
 725         bool send_new_highest_used_slotid = false;
 726
 727         if (!slot)
 728                 return;
 729         tbl = slot->table;
 730         session = tbl->session;
 731
 732         /* Bump the slot sequence number */
 733         if (slot->seq_done)
 734                 slot->seq_nr++;
 735         slot->seq_done = 0;
 736
 737         spin_lock(&tbl->slot_tbl_lock);
 738         /* Be nice to the server: try to ensure that the last transmitted
 739          * value for highest_user_slotid <= target_highest_slotid
 740          */
 741         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 742                 send_new_highest_used_slotid = true;
 743
 744         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 745                 send_new_highest_used_slotid = false;
 746                 goto out_unlock;
 747         }
 748         nfs4_free_slot(tbl, slot);
 749
 750         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 751                 send_new_highest_used_slotid = false;
 752 out_unlock:
 753         spin_unlock(&tbl->slot_tbl_lock);
 754         if (send_new_highest_used_slotid)
 755                 nfs41_notify_server(session->clp);
 756         if (waitqueue_active(&tbl->slot_waitq))
 757                 wake_up_all(&tbl->slot_waitq);
 758 }
 759
 760 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 761 {
 762         nfs41_release_slot(res->sr_slot);
 763         res->sr_slot = NULL;
 764 }
 765
 766 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 767                 u32 seqnr)
 768 {
 769         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 770                 slot->seq_nr_highest_sent = seqnr;
 771 }
 772 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 773                 u32 seqnr)
 774 {
 775         slot->seq_nr_highest_sent = seqnr;
 776         slot->seq_nr_last_acked = seqnr;
 777 }
 778
 779 static int nfs41_sequence_process(struct rpc_task *task,
 780                 struct nfs4_sequence_res *res)
 781 {
 782         struct nfs4_session *session;
 783         struct nfs4_slot *slot = res->sr_slot;
 784         struct nfs_client *clp;
 785         int ret = 1;
 786
 787         if (slot == NULL)
 788                 goto out_noaction;
 789         /* don't increment the sequence number if the task wasn't sent */
 790         if (!RPC_WAS_SENT(task) || slot->seq_done)
 791                 goto out;
 792
 793         session = slot->table->session;
 794
 795         trace_nfs4_sequence_done(session, res);
 796         /* Check the SEQUENCE operation status */
 797         switch (res->sr_status) {
 798         case 0:
 799                 /* Mark this sequence number as having been acked */
 800                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 801                 /* Update the slot's sequence and clientid lease timer */
 802                 slot->seq_done = 1;
 803                 clp = session->clp;
 804                 do_renew_lease(clp, res->sr_timestamp);
 805                 /* Check sequence flags */
 806                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 807                                 !!slot->privileged);
 808                 nfs41_update_target_slotid(slot->table, slot, res);
 809                 break;
 810         case 1:
 811                 /*
 812                  * sr_status remains 1 if an RPC level error occurred.
 813                  * The server may or may not have processed the sequence
 814                  * operation..
 815                  */
 816                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 817                 slot->seq_done = 1;
 818                 goto out;
 819         case -NFS4ERR_DELAY:
 820                 /* The server detected a resend of the RPC call and
 821                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 822                  * of RFC5661.
 823                  */
 824                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 825                         __func__,
 826                         slot->slot_nr,
 827                         slot->seq_nr);
 828                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 829                 goto out_retry;
 830         case -NFS4ERR_RETRY_UNCACHED_REP:
 831         case -NFS4ERR_SEQ_FALSE_RETRY:
 832                 /*
 833                  * The server thinks we tried to replay a request.
 834                  * Retry the call after bumping the sequence ID.
 835                  */
 836                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 837                 goto retry_new_seq;
 838         case -NFS4ERR_BADSLOT:
 839                 /*
 840                  * The slot id we used was probably retired. Try again
 841                  * using a different slot id.
 842                  */
 843                 if (slot->slot_nr < slot->table->target_highest_slotid)
 844                         goto session_recover;
 845                 goto retry_nowait;
 846         case -NFS4ERR_SEQ_MISORDERED:
 847                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 848                 /*
 849                  * Were one or more calls using this slot interrupted?
 850                  * If the server never received the request, then our
 851                  * transmitted slot sequence number may be too high.
 852                  */
 853                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 854                         slot->seq_nr--;
 855                         goto retry_nowait;
 856                 }
 857                 /*
 858                  * RFC5661:
 859                  * A retry might be sent while the original request is
 860                  * still in progress on the replier. The replier SHOULD
 861                  * deal with the issue by returning NFS4ERR_DELAY as the
 862                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 863                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 864                  *
 865                  * Restart the search after a delay.
 866                  */
 867                 slot->seq_nr = slot->seq_nr_highest_sent;
 868                 goto out_retry;
 869         default:
 870                 /* Just update the slot sequence no. */
 871                 slot->seq_done = 1;
 872         }
 873 out:
 874         /* The session may be reset by one of the error handlers. */
 875         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 876 out_noaction:
 877         return ret;
 878 session_recover:
 879         nfs4_schedule_session_recovery(session, res->sr_status);
 880         goto retry_nowait;
 881 retry_new_seq:
 882         ++slot->seq_nr;
 883 retry_nowait:
 884         if (rpc_restart_call_prepare(task)) {
 885                 nfs41_sequence_free_slot(res);
 886                 task->tk_status = 0;
 887                 ret = 0;
 888         }
 889         goto out;
 890 out_retry:
 891         if (!rpc_restart_call(task))
 892                 goto out;
 893         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 894         return 0;
 895 }
 896
 897 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 898 {
 899         if (!nfs41_sequence_process(task, res))
 900                 return 0;
 901         if (res->sr_slot != NULL)
 902                 nfs41_sequence_free_slot(res);
 903         return 1;
 904
 905 }
 906 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 907
 908 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 909 {
 910         if (res->sr_slot == NULL)
 911                 return 1;
 912         if (res->sr_slot->table->session != NULL)
 913                 return nfs41_sequence_process(task, res);
 914         return nfs40_sequence_done(task, res);
 915 }
 916
 917 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 918 {
 919         if (res->sr_slot != NULL) {
 920                 if (res->sr_slot->table->session != NULL)
 921                         nfs41_sequence_free_slot(res);
 922                 else
 923                         nfs40_sequence_free_slot(res);
 924         }
 925 }
 926
 927 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot == NULL)
 930                 return 1;
 931         if (!res->sr_slot->table->session)
 932                 return nfs40_sequence_done(task, res);
 933         return nfs41_sequence_done(task, res);
 934 }
 935 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 936
 937 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 938 {
 939         struct nfs4_call_sync_data *data = calldata;
 940
 941         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 942
 943         nfs4_setup_sequence(data->seq_server->nfs_client,
 944                             data->seq_args, data->seq_res, task);
 945 }
 946
 947 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950
 951         nfs41_sequence_done(task, data->seq_res);
 952 }
 953
 954 static const struct rpc_call_ops nfs41_call_sync_ops = {
 955         .rpc_call_prepare = nfs41_call_sync_prepare,
 956         .rpc_call_done = nfs41_call_sync_done,
 957 };
 958
 959 #else   /* !CONFIG_NFS_V4_1 */
 960
 961 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 962 {
 963         return nfs40_sequence_done(task, res);
 964 }
 965
 966 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 967 {
 968         if (res->sr_slot != NULL)
 969                 nfs40_sequence_free_slot(res);
 970 }
 971
 972 int nfs4_sequence_done(struct rpc_task *task,
 973                        struct nfs4_sequence_res *res)
 974 {
 975         return nfs40_sequence_done(task, res);
 976 }
 977 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 978
 979 #endif  /* !CONFIG_NFS_V4_1 */
 980
 981 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
 982 {
 983         res->sr_timestamp = jiffies;
 984         res->sr_status_flags = 0;
 985         res->sr_status = 1;
 986 }
 987
 988 static
 989 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 990                 struct nfs4_sequence_res *res,
 991                 struct nfs4_slot *slot)
 992 {
 993         if (!slot)
 994                 return;
 995         slot->privileged = args->sa_privileged ? 1 : 0;
 996         args->sa_slot = slot;
 997
 998         res->sr_slot = slot;
 999 }
1000
1001 int nfs4_setup_sequence(struct nfs_client *client,
1002                         struct nfs4_sequence_args *args,
1003                         struct nfs4_sequence_res *res,
1004                         struct rpc_task *task)
1005 {
1006         struct nfs4_session *session = nfs4_get_session(client);
1007         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1008         struct nfs4_slot *slot;
1009
1010         /* slot already allocated? */
1011         if (res->sr_slot != NULL)
1012                 goto out_start;
1013
1014         if (session)
1015                 tbl = &session->fc_slot_table;
1016
1017         spin_lock(&tbl->slot_tbl_lock);
1018         /* The state manager will wait until the slot table is empty */
1019         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1020                 goto out_sleep;
1021
1022         slot = nfs4_alloc_slot(tbl);
1023         if (IS_ERR(slot)) {
1024                 if (slot == ERR_PTR(-ENOMEM))
1025                         goto out_sleep_timeout;
1026                 goto out_sleep;
1027         }
1028         spin_unlock(&tbl->slot_tbl_lock);
1029
1030         nfs4_sequence_attach_slot(args, res, slot);
1031
1032         trace_nfs4_setup_sequence(session, args);
1033 out_start:
1034         nfs41_sequence_res_init(res);
1035         rpc_call_start(task);
1036         return 0;
1037 out_sleep_timeout:
1038         /* Try again in 1/4 second */
1039         if (args->sa_privileged)
1040                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1041                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1042         else
1043                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1044                                 NULL, jiffies + (HZ >> 2));
1045         spin_unlock(&tbl->slot_tbl_lock);
1046         return -EAGAIN;
1047 out_sleep:
1048         if (args->sa_privileged)
1049                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1050                                 RPC_PRIORITY_PRIVILEGED);
1051         else
1052                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1053         spin_unlock(&tbl->slot_tbl_lock);
1054         return -EAGAIN;
1055 }
1056 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1057
1058 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1059 {
1060         struct nfs4_call_sync_data *data = calldata;
1061         nfs4_setup_sequence(data->seq_server->nfs_client,
1062                                 data->seq_args, data->seq_res, task);
1063 }
1064
1065 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1066 {
1067         struct nfs4_call_sync_data *data = calldata;
1068         nfs4_sequence_done(task, data->seq_res);
1069 }
1070
1071 static const struct rpc_call_ops nfs40_call_sync_ops = {
1072         .rpc_call_prepare = nfs40_call_sync_prepare,
1073         .rpc_call_done = nfs40_call_sync_done,
1074 };
1075
1076 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1077 {
1078         int ret;
1079         struct rpc_task *task;
1080
1081         task = rpc_run_task(task_setup);
1082         if (IS_ERR(task))
1083                 return PTR_ERR(task);
1084
1085         ret = task->tk_status;
1086         rpc_put_task(task);
1087         return ret;
1088 }
1089
1090 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1091                                    struct nfs_server *server,
1092                                    struct rpc_message *msg,
1093                                    struct nfs4_sequence_args *args,
1094                                    struct nfs4_sequence_res *res)
1095 {
1096         struct nfs_client *clp = server->nfs_client;
1097         struct nfs4_call_sync_data data = {
1098                 .seq_server = server,
1099                 .seq_args = args,
1100                 .seq_res = res,
1101         };
1102         struct rpc_task_setup task_setup = {
1103                 .rpc_client = clnt,
1104                 .rpc_message = msg,
1105                 .callback_ops = clp->cl_mvops->call_sync_ops,
1106                 .callback_data = &data
1107         };
1108
1109         return nfs4_call_sync_custom(&task_setup);
1110 }
1111
1112 int nfs4_call_sync(struct rpc_clnt *clnt,
1113                    struct nfs_server *server,
1114                    struct rpc_message *msg,
1115                    struct nfs4_sequence_args *args,
1116                    struct nfs4_sequence_res *res,
1117                    int cache_reply)
1118 {
1119         nfs4_init_sequence(args, res, cache_reply, 0);
1120         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1121 }
1122
1123 static void
1124 nfs4_inc_nlink_locked(struct inode *inode)
1125 {
1126         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1127         inc_nlink(inode);
1128 }
1129
1130 static void
1131 nfs4_dec_nlink_locked(struct inode *inode)
1132 {
1133         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1134         drop_nlink(inode);
1135 }
1136
1137 static void
1138 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1139                 unsigned long timestamp, unsigned long cache_validity)
1140 {
1141         struct nfs_inode *nfsi = NFS_I(dir);
1142
1143         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1144                 | NFS_INO_INVALID_MTIME
1145                 | NFS_INO_INVALID_DATA
1146                 | cache_validity;
1147         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1148                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1149                 nfsi->attrtimeo_timestamp = jiffies;
1150         } else {
1151                 nfs_force_lookup_revalidate(dir);
1152                 if (cinfo->before != inode_peek_iversion_raw(dir))
1153                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1154                                 NFS_INO_INVALID_ACL;
1155         }
1156         inode_set_iversion_raw(dir, cinfo->after);
1157         nfsi->read_cache_jiffies = timestamp;
1158         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1159         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1160         nfs_fscache_invalidate(dir);
1161 }
1162
1163 static void
1164 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1165                 unsigned long timestamp, unsigned long cache_validity)
1166 {
1167         spin_lock(&dir->i_lock);
1168         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1169         spin_unlock(&dir->i_lock);
1170 }
1171
1172 struct nfs4_open_createattrs {
1173         struct nfs4_label *label;
1174         struct iattr *sattr;
1175         const __u32 verf[2];
1176 };
1177
1178 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1179                 int err, struct nfs4_exception *exception)
1180 {
1181         if (err != -EINVAL)
1182                 return false;
1183         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1184                 return false;
1185         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1186         exception->retry = 1;
1187         return true;
1188 }
1189
1190 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1191 {
1192          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1193 }
1194
1195 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1196 {
1197         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1198
1199         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1200 }
1201
1202 static u32
1203 nfs4_map_atomic_open_share(struct nfs_server *server,
1204                 fmode_t fmode, int openflags)
1205 {
1206         u32 res = 0;
1207
1208         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1209         case FMODE_READ:
1210                 res = NFS4_SHARE_ACCESS_READ;
1211                 break;
1212         case FMODE_WRITE:
1213                 res = NFS4_SHARE_ACCESS_WRITE;
1214                 break;
1215         case FMODE_READ|FMODE_WRITE:
1216                 res = NFS4_SHARE_ACCESS_BOTH;
1217         }
1218         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1219                 goto out;
1220         /* Want no delegation if we're using O_DIRECT */
1221         if (openflags & O_DIRECT)
1222                 res |= NFS4_SHARE_WANT_NO_DELEG;
1223 out:
1224         return res;
1225 }
1226
1227 static enum open_claim_type4
1228 nfs4_map_atomic_open_claim(struct nfs_server *server,
1229                 enum open_claim_type4 claim)
1230 {
1231         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1232                 return claim;
1233         switch (claim) {
1234         default:
1235                 return claim;
1236         case NFS4_OPEN_CLAIM_FH:
1237                 return NFS4_OPEN_CLAIM_NULL;
1238         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1239                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1240         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1241                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1242         }
1243 }
1244
1245 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1246 {
1247         p->o_res.f_attr = &p->f_attr;
1248         p->o_res.f_label = p->f_label;
1249         p->o_res.seqid = p->o_arg.seqid;
1250         p->c_res.seqid = p->c_arg.seqid;
1251         p->o_res.server = p->o_arg.server;
1252         p->o_res.access_request = p->o_arg.access;
1253         nfs_fattr_init(&p->f_attr);
1254         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1255 }
1256
1257 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1258                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1259                 const struct nfs4_open_createattrs *c,
1260                 enum open_claim_type4 claim,
1261                 gfp_t gfp_mask)
1262 {
1263         struct dentry *parent = dget_parent(dentry);
1264         struct inode *dir = d_inode(parent);
1265         struct nfs_server *server = NFS_SERVER(dir);
1266         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1267         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1268         struct nfs4_opendata *p;
1269
1270         p = kzalloc(sizeof(*p), gfp_mask);
1271         if (p == NULL)
1272                 goto err;
1273
1274         p->f_label = nfs4_label_alloc(server, gfp_mask);
1275         if (IS_ERR(p->f_label))
1276                 goto err_free_p;
1277
1278         p->a_label = nfs4_label_alloc(server, gfp_mask);
1279         if (IS_ERR(p->a_label))
1280                 goto err_free_f;
1281
1282         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1283         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1284         if (IS_ERR(p->o_arg.seqid))
1285                 goto err_free_label;
1286         nfs_sb_active(dentry->d_sb);
1287         p->dentry = dget(dentry);
1288         p->dir = parent;
1289         p->owner = sp;
1290         atomic_inc(&sp->so_count);
1291         p->o_arg.open_flags = flags;
1292         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1293         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1294         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1295                         fmode, flags);
1296         if (flags & O_CREAT) {
1297                 p->o_arg.umask = current_umask();
1298                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1299                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1300                         p->o_arg.u.attrs = &p->attrs;
1301                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1302
1303                         memcpy(p->o_arg.u.verifier.data, c->verf,
1304                                         sizeof(p->o_arg.u.verifier.data));
1305                 }
1306         }
1307         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1308          * will return permission denied for all bits until close */
1309         if (!(flags & O_EXCL)) {
1310                 /* ask server to check for all possible rights as results
1311                  * are cached */
1312                 switch (p->o_arg.claim) {
1313                 default:
1314                         break;
1315                 case NFS4_OPEN_CLAIM_NULL:
1316                 case NFS4_OPEN_CLAIM_FH:
1317                         p->o_arg.access = NFS4_ACCESS_READ |
1318                                 NFS4_ACCESS_MODIFY |
1319                                 NFS4_ACCESS_EXTEND |
1320                                 NFS4_ACCESS_EXECUTE;
1321                 }
1322         }
1323         p->o_arg.clientid = server->nfs_client->cl_clientid;
1324         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1325         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1326         p->o_arg.name = &dentry->d_name;
1327         p->o_arg.server = server;
1328         p->o_arg.bitmask = nfs4_bitmask(server, label);
1329         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1330         switch (p->o_arg.claim) {
1331         case NFS4_OPEN_CLAIM_NULL:
1332         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1333         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1334                 p->o_arg.fh = NFS_FH(dir);
1335                 break;
1336         case NFS4_OPEN_CLAIM_PREVIOUS:
1337         case NFS4_OPEN_CLAIM_FH:
1338         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1339         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1340                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1341         }
1342         p->c_arg.fh = &p->o_res.fh;
1343         p->c_arg.stateid = &p->o_res.stateid;
1344         p->c_arg.seqid = p->o_arg.seqid;
1345         nfs4_init_opendata_res(p);
1346         kref_init(&p->kref);
1347         return p;
1348
1349 err_free_label:
1350         nfs4_label_free(p->a_label);
1351 err_free_f:
1352         nfs4_label_free(p->f_label);
1353 err_free_p:
1354         kfree(p);
1355 err:
1356         dput(parent);
1357         return NULL;
1358 }
1359
1360 static void nfs4_opendata_free(struct kref *kref)
1361 {
1362         struct nfs4_opendata *p = container_of(kref,
1363                         struct nfs4_opendata, kref);
1364         struct super_block *sb = p->dentry->d_sb;
1365
1366         nfs4_lgopen_release(p->lgp);
1367         nfs_free_seqid(p->o_arg.seqid);
1368         nfs4_sequence_free_slot(&p->o_res.seq_res);
1369         if (p->state != NULL)
1370                 nfs4_put_open_state(p->state);
1371         nfs4_put_state_owner(p->owner);
1372
1373         nfs4_label_free(p->a_label);
1374         nfs4_label_free(p->f_label);
1375
1376         dput(p->dir);
1377         dput(p->dentry);
1378         nfs_sb_deactive(sb);
1379         nfs_fattr_free_names(&p->f_attr);
1380         kfree(p->f_attr.mdsthreshold);
1381         kfree(p);
1382 }
1383
1384 static void nfs4_opendata_put(struct nfs4_opendata *p)
1385 {
1386         if (p != NULL)
1387                 kref_put(&p->kref, nfs4_opendata_free);
1388 }
1389
1390 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1391                 fmode_t fmode)
1392 {
1393         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1394         case FMODE_READ|FMODE_WRITE:
1395                 return state->n_rdwr != 0;
1396         case FMODE_WRITE:
1397                 return state->n_wronly != 0;
1398         case FMODE_READ:
1399                 return state->n_rdonly != 0;
1400         }
1401         WARN_ON_ONCE(1);
1402         return false;
1403 }
1404
1405 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1406                 int open_mode, enum open_claim_type4 claim)
1407 {
1408         int ret = 0;
1409
1410         if (open_mode & (O_EXCL|O_TRUNC))
1411                 goto out;
1412         switch (claim) {
1413         case NFS4_OPEN_CLAIM_NULL:
1414         case NFS4_OPEN_CLAIM_FH:
1415                 goto out;
1416         default:
1417                 break;
1418         }
1419         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1420                 case FMODE_READ:
1421                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1422                                 && state->n_rdonly != 0;
1423                         break;
1424                 case FMODE_WRITE:
1425                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1426                                 && state->n_wronly != 0;
1427                         break;
1428                 case FMODE_READ|FMODE_WRITE:
1429                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1430                                 && state->n_rdwr != 0;
1431         }
1432 out:
1433         return ret;
1434 }
1435
1436 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1437                 enum open_claim_type4 claim)
1438 {
1439         if (delegation == NULL)
1440                 return 0;
1441         if ((delegation->type & fmode) != fmode)
1442                 return 0;
1443         switch (claim) {
1444         case NFS4_OPEN_CLAIM_NULL:
1445         case NFS4_OPEN_CLAIM_FH:
1446                 break;
1447         case NFS4_OPEN_CLAIM_PREVIOUS:
1448                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1449                         break;
1450                 /* Fall through */
1451         default:
1452                 return 0;
1453         }
1454         nfs_mark_delegation_referenced(delegation);
1455         return 1;
1456 }
1457
1458 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1459 {
1460         switch (fmode) {
1461                 case FMODE_WRITE:
1462                         state->n_wronly++;
1463                         break;
1464                 case FMODE_READ:
1465                         state->n_rdonly++;
1466                         break;
1467                 case FMODE_READ|FMODE_WRITE:
1468                         state->n_rdwr++;
1469         }
1470         nfs4_state_set_mode_locked(state, state->state | fmode);
1471 }
1472
1473 #ifdef CONFIG_NFS_V4_1
1474 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1475 {
1476         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1477                 return true;
1478         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1479                 return true;
1480         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1481                 return true;
1482         return false;
1483 }
1484 #endif /* CONFIG_NFS_V4_1 */
1485
1486 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1487 {
1488         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1489                 wake_up_all(&state->waitq);
1490 }
1491
1492 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1493                 const nfs4_stateid *stateid)
1494 {
1495         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1496         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1497
1498         if (stateid_seqid == state_seqid + 1U ||
1499             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1500                 nfs_state_log_update_open_stateid(state);
1501         else
1502                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1503 }
1504
1505 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1506 {
1507         struct nfs_client *clp = state->owner->so_server->nfs_client;
1508         bool need_recover = false;
1509
1510         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1511                 need_recover = true;
1512         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1513                 need_recover = true;
1514         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1515                 need_recover = true;
1516         if (need_recover)
1517                 nfs4_state_mark_reclaim_nograce(clp, state);
1518 }
1519
1520 /*
1521  * Check for whether or not the caller may update the open stateid
1522  * to the value passed in by stateid.
1523  *
1524  * Note: This function relies heavily on the server implementing
1525  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1526  * correctly.
1527  * i.e. The stateid seqids have to be initialised to 1, and
1528  * are then incremented on every state transition.
1529  */
1530 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1531                 const nfs4_stateid *stateid)
1532 {
1533         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1534             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1535                 if (stateid->seqid == cpu_to_be32(1))
1536                         nfs_state_log_update_open_stateid(state);
1537                 else
1538                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1539                 return true;
1540         }
1541
1542         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1543                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1544                 return true;
1545         }
1546         return false;
1547 }
1548
1549 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1550 {
1551         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1552                 return;
1553         if (state->n_wronly)
1554                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1555         if (state->n_rdonly)
1556                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1557         if (state->n_rdwr)
1558                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1559         set_bit(NFS_OPEN_STATE, &state->flags);
1560 }
1561
1562 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1563                 nfs4_stateid *stateid, fmode_t fmode)
1564 {
1565         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1566         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1567         case FMODE_WRITE:
1568                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1569                 break;
1570         case FMODE_READ:
1571                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1572                 break;
1573         case 0:
1574                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1575                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1576                 clear_bit(NFS_OPEN_STATE, &state->flags);
1577         }
1578         if (stateid == NULL)
1579                 return;
1580         /* Handle OPEN+OPEN_DOWNGRADE races */
1581         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1582             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1583                 nfs_resync_open_stateid_locked(state);
1584                 goto out;
1585         }
1586         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1587                 nfs4_stateid_copy(&state->stateid, stateid);
1588         nfs4_stateid_copy(&state->open_stateid, stateid);
1589         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1590 out:
1591         nfs_state_log_update_open_stateid(state);
1592 }
1593
1594 static void nfs_clear_open_stateid(struct nfs4_state *state,
1595         nfs4_stateid *arg_stateid,
1596         nfs4_stateid *stateid, fmode_t fmode)
1597 {
1598         write_seqlock(&state->seqlock);
1599         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1600         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1601                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1602         write_sequnlock(&state->seqlock);
1603         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1604                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1605 }
1606
1607 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1608                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1609         __must_hold(&state->owner->so_lock)
1610         __must_hold(&state->seqlock)
1611         __must_hold(RCU)
1612
1613 {
1614         DEFINE_WAIT(wait);
1615         int status = 0;
1616         for (;;) {
1617
1618                 if (!nfs_need_update_open_stateid(state, stateid))
1619                         return;
1620                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1621                         break;
1622                 if (status)
1623                         break;
1624                 /* Rely on seqids for serialisation with NFSv4.0 */
1625                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1626                         break;
1627
1628                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1629                 /*
1630                  * Ensure we process the state changes in the same order
1631                  * in which the server processed them by delaying the
1632                  * update of the stateid until we are in sequence.
1633                  */
1634                 write_sequnlock(&state->seqlock);
1635                 spin_unlock(&state->owner->so_lock);
1636                 rcu_read_unlock();
1637                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1638                 if (!signal_pending(current)) {
1639                         if (schedule_timeout(5*HZ) == 0)
1640                                 status = -EAGAIN;
1641                         else
1642                                 status = 0;
1643                 } else
1644                         status = -EINTR;
1645                 finish_wait(&state->waitq, &wait);
1646                 rcu_read_lock();
1647                 spin_lock(&state->owner->so_lock);
1648                 write_seqlock(&state->seqlock);
1649         }
1650
1651         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1652             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1653                 nfs4_stateid_copy(freeme, &state->open_stateid);
1654                 nfs_test_and_clear_all_open_stateid(state);
1655         }
1656
1657         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1658                 nfs4_stateid_copy(&state->stateid, stateid);
1659         nfs4_stateid_copy(&state->open_stateid, stateid);
1660         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1661         nfs_state_log_update_open_stateid(state);
1662 }
1663
1664 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1665                 const nfs4_stateid *open_stateid,
1666                 fmode_t fmode,
1667                 nfs4_stateid *freeme)
1668 {
1669         /*
1670          * Protect the call to nfs4_state_set_mode_locked and
1671          * serialise the stateid update
1672          */
1673         write_seqlock(&state->seqlock);
1674         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1675         switch (fmode) {
1676         case FMODE_READ:
1677                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1678                 break;
1679         case FMODE_WRITE:
1680                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1681                 break;
1682         case FMODE_READ|FMODE_WRITE:
1683                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1684         }
1685         set_bit(NFS_OPEN_STATE, &state->flags);
1686         write_sequnlock(&state->seqlock);
1687 }
1688
1689 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1690 {
1691         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1692         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1693         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1694         clear_bit(NFS_OPEN_STATE, &state->flags);
1695 }
1696
1697 static void nfs_state_set_delegation(struct nfs4_state *state,
1698                 const nfs4_stateid *deleg_stateid,
1699                 fmode_t fmode)
1700 {
1701         /*
1702          * Protect the call to nfs4_state_set_mode_locked and
1703          * serialise the stateid update
1704          */
1705         write_seqlock(&state->seqlock);
1706         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1707         set_bit(NFS_DELEGATED_STATE, &state->flags);
1708         write_sequnlock(&state->seqlock);
1709 }
1710
1711 static void nfs_state_clear_delegation(struct nfs4_state *state)
1712 {
1713         write_seqlock(&state->seqlock);
1714         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1715         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1716         write_sequnlock(&state->seqlock);
1717 }
1718
1719 static int update_open_stateid(struct nfs4_state *state,
1720                 const nfs4_stateid *open_stateid,
1721                 const nfs4_stateid *delegation,
1722                 fmode_t fmode)
1723 {
1724         struct nfs_server *server = NFS_SERVER(state->inode);
1725         struct nfs_client *clp = server->nfs_client;
1726         struct nfs_inode *nfsi = NFS_I(state->inode);
1727         struct nfs_delegation *deleg_cur;
1728         nfs4_stateid freeme = { };
1729         int ret = 0;
1730
1731         fmode &= (FMODE_READ|FMODE_WRITE);
1732
1733         rcu_read_lock();
1734         spin_lock(&state->owner->so_lock);
1735         if (open_stateid != NULL) {
1736                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1737                 ret = 1;
1738         }
1739
1740         deleg_cur = rcu_dereference(nfsi->delegation);
1741         if (deleg_cur == NULL)
1742                 goto no_delegation;
1743
1744         spin_lock(&deleg_cur->lock);
1745         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1746            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1747             (deleg_cur->type & fmode) != fmode)
1748                 goto no_delegation_unlock;
1749
1750         if (delegation == NULL)
1751                 delegation = &deleg_cur->stateid;
1752         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1753                 goto no_delegation_unlock;
1754
1755         nfs_mark_delegation_referenced(deleg_cur);
1756         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1757         ret = 1;
1758 no_delegation_unlock:
1759         spin_unlock(&deleg_cur->lock);
1760 no_delegation:
1761         if (ret)
1762                 update_open_stateflags(state, fmode);
1763         spin_unlock(&state->owner->so_lock);
1764         rcu_read_unlock();
1765
1766         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1767                 nfs4_schedule_state_manager(clp);
1768         if (freeme.type != 0)
1769                 nfs4_test_and_free_stateid(server, &freeme,
1770                                 state->owner->so_cred);
1771
1772         return ret;
1773 }
1774
1775 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1776                 const nfs4_stateid *stateid)
1777 {
1778         struct nfs4_state *state = lsp->ls_state;
1779         bool ret = false;
1780
1781         spin_lock(&state->state_lock);
1782         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1783                 goto out_noupdate;
1784         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1785                 goto out_noupdate;
1786         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1787         ret = true;
1788 out_noupdate:
1789         spin_unlock(&state->state_lock);
1790         return ret;
1791 }
1792
1793 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1794 {
1795         struct nfs_delegation *delegation;
1796
1797         fmode &= FMODE_READ|FMODE_WRITE;
1798         rcu_read_lock();
1799         delegation = rcu_dereference(NFS_I(inode)->delegation);
1800         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1801                 rcu_read_unlock();
1802                 return;
1803         }
1804         rcu_read_unlock();
1805         nfs4_inode_return_delegation(inode);
1806 }
1807
1808 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1809 {
1810         struct nfs4_state *state = opendata->state;
1811         struct nfs_delegation *delegation;
1812         int open_mode = opendata->o_arg.open_flags;
1813         fmode_t fmode = opendata->o_arg.fmode;
1814         enum open_claim_type4 claim = opendata->o_arg.claim;
1815         nfs4_stateid stateid;
1816         int ret = -EAGAIN;
1817
1818         for (;;) {
1819                 spin_lock(&state->owner->so_lock);
1820                 if (can_open_cached(state, fmode, open_mode, claim)) {
1821                         update_open_stateflags(state, fmode);
1822                         spin_unlock(&state->owner->so_lock);
1823                         goto out_return_state;
1824                 }
1825                 spin_unlock(&state->owner->so_lock);
1826                 rcu_read_lock();
1827                 delegation = nfs4_get_valid_delegation(state->inode);
1828                 if (!can_open_delegated(delegation, fmode, claim)) {
1829                         rcu_read_unlock();
1830                         break;
1831                 }
1832                 /* Save the delegation */
1833                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1834                 rcu_read_unlock();
1835                 nfs_release_seqid(opendata->o_arg.seqid);
1836                 if (!opendata->is_recover) {
1837                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1838                         if (ret != 0)
1839                                 goto out;
1840                 }
1841                 ret = -EAGAIN;
1842
1843                 /* Try to update the stateid using the delegation */
1844                 if (update_open_stateid(state, NULL, &stateid, fmode))
1845                         goto out_return_state;
1846         }
1847 out:
1848         return ERR_PTR(ret);
1849 out_return_state:
1850         refcount_inc(&state->count);
1851         return state;
1852 }
1853
1854 static void
1855 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1856 {
1857         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1858         struct nfs_delegation *delegation;
1859         int delegation_flags = 0;
1860
1861         rcu_read_lock();
1862         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1863         if (delegation)
1864                 delegation_flags = delegation->flags;
1865         rcu_read_unlock();
1866         switch (data->o_arg.claim) {
1867         default:
1868                 break;
1869         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1870         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1871                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1872                                    "returning a delegation for "
1873                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1874                                    clp->cl_hostname);
1875                 return;
1876         }
1877         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1878                 nfs_inode_set_delegation(state->inode,
1879                                 data->owner->so_cred,
1880                                 data->o_res.delegation_type,
1881                                 &data->o_res.delegation,
1882                                 data->o_res.pagemod_limit);
1883         else
1884                 nfs_inode_reclaim_delegation(state->inode,
1885                                 data->owner->so_cred,
1886                                 data->o_res.delegation_type,
1887                                 &data->o_res.delegation,
1888                                 data->o_res.pagemod_limit);
1889
1890         if (data->o_res.do_recall)
1891                 nfs_async_inode_return_delegation(state->inode,
1892                                                   &data->o_res.delegation);
1893 }
1894
1895 /*
1896  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1897  * and update the nfs4_state.
1898  */
1899 static struct nfs4_state *
1900 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1901 {
1902         struct inode *inode = data->state->inode;
1903         struct nfs4_state *state = data->state;
1904         int ret;
1905
1906         if (!data->rpc_done) {
1907                 if (data->rpc_status)
1908                         return ERR_PTR(data->rpc_status);
1909                 /* cached opens have already been processed */
1910                 goto update;
1911         }
1912
1913         ret = nfs_refresh_inode(inode, &data->f_attr);
1914         if (ret)
1915                 return ERR_PTR(ret);
1916
1917         if (data->o_res.delegation_type != 0)
1918                 nfs4_opendata_check_deleg(data, state);
1919 update:
1920         if (!update_open_stateid(state, &data->o_res.stateid,
1921                                 NULL, data->o_arg.fmode))
1922                 return ERR_PTR(-EAGAIN);
1923         refcount_inc(&state->count);
1924
1925         return state;
1926 }
1927
1928 static struct inode *
1929 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1930 {
1931         struct inode *inode;
1932
1933         switch (data->o_arg.claim) {
1934         case NFS4_OPEN_CLAIM_NULL:
1935         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1936         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1937                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1938                         return ERR_PTR(-EAGAIN);
1939                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1940                                 &data->f_attr, data->f_label);
1941                 break;
1942         default:
1943                 inode = d_inode(data->dentry);
1944                 ihold(inode);
1945                 nfs_refresh_inode(inode, &data->f_attr);
1946         }
1947         return inode;
1948 }
1949
1950 static struct nfs4_state *
1951 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1952 {
1953         struct nfs4_state *state;
1954         struct inode *inode;
1955
1956         inode = nfs4_opendata_get_inode(data);
1957         if (IS_ERR(inode))
1958                 return ERR_CAST(inode);
1959         if (data->state != NULL && data->state->inode == inode) {
1960                 state = data->state;
1961                 refcount_inc(&state->count);
1962         } else
1963                 state = nfs4_get_open_state(inode, data->owner);
1964         iput(inode);
1965         if (state == NULL)
1966                 state = ERR_PTR(-ENOMEM);
1967         return state;
1968 }
1969
1970 static struct nfs4_state *
1971 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1972 {
1973         struct nfs4_state *state;
1974
1975         if (!data->rpc_done) {
1976                 state = nfs4_try_open_cached(data);
1977                 trace_nfs4_cached_open(data->state);
1978                 goto out;
1979         }
1980
1981         state = nfs4_opendata_find_nfs4_state(data);
1982         if (IS_ERR(state))
1983                 goto out;
1984
1985         if (data->o_res.delegation_type != 0)
1986                 nfs4_opendata_check_deleg(data, state);
1987         if (!update_open_stateid(state, &data->o_res.stateid,
1988                                 NULL, data->o_arg.fmode)) {
1989                 nfs4_put_open_state(state);
1990                 state = ERR_PTR(-EAGAIN);
1991         }
1992 out:
1993         nfs_release_seqid(data->o_arg.seqid);
1994         return state;
1995 }
1996
1997 static struct nfs4_state *
1998 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1999 {
2000         struct nfs4_state *ret;
2001
2002         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2003                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2004         else
2005                 ret = _nfs4_opendata_to_nfs4_state(data);
2006         nfs4_sequence_free_slot(&data->o_res.seq_res);
2007         return ret;
2008 }
2009
2010 static struct nfs_open_context *
2011 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2012 {
2013         struct nfs_inode *nfsi = NFS_I(state->inode);
2014         struct nfs_open_context *ctx;
2015
2016         rcu_read_lock();
2017         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2018                 if (ctx->state != state)
2019                         continue;
2020                 if ((ctx->mode & mode) != mode)
2021                         continue;
2022                 if (!get_nfs_open_context(ctx))
2023                         continue;
2024                 rcu_read_unlock();
2025                 return ctx;
2026         }
2027         rcu_read_unlock();
2028         return ERR_PTR(-ENOENT);
2029 }
2030
2031 static struct nfs_open_context *
2032 nfs4_state_find_open_context(struct nfs4_state *state)
2033 {
2034         struct nfs_open_context *ctx;
2035
2036         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2037         if (!IS_ERR(ctx))
2038                 return ctx;
2039         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2040         if (!IS_ERR(ctx))
2041                 return ctx;
2042         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2043 }
2044
2045 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2046                 struct nfs4_state *state, enum open_claim_type4 claim)
2047 {
2048         struct nfs4_opendata *opendata;
2049
2050         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2051                         NULL, claim, GFP_NOFS);
2052         if (opendata == NULL)
2053                 return ERR_PTR(-ENOMEM);
2054         opendata->state = state;
2055         refcount_inc(&state->count);
2056         return opendata;
2057 }
2058
2059 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2060                 fmode_t fmode)
2061 {
2062         struct nfs4_state *newstate;
2063         int ret;
2064
2065         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2066                 return 0;
2067         opendata->o_arg.open_flags = 0;
2068         opendata->o_arg.fmode = fmode;
2069         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2070                         NFS_SB(opendata->dentry->d_sb),
2071                         fmode, 0);
2072         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2073         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2074         nfs4_init_opendata_res(opendata);
2075         ret = _nfs4_recover_proc_open(opendata);
2076         if (ret != 0)
2077                 return ret;
2078         newstate = nfs4_opendata_to_nfs4_state(opendata);
2079         if (IS_ERR(newstate))
2080                 return PTR_ERR(newstate);
2081         if (newstate != opendata->state)
2082                 ret = -ESTALE;
2083         nfs4_close_state(newstate, fmode);
2084         return ret;
2085 }
2086
2087 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2088 {
2089         int ret;
2090
2091         /* memory barrier prior to reading state->n_* */
2092         smp_rmb();
2093         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2094         if (ret != 0)
2095                 return ret;
2096         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2097         if (ret != 0)
2098                 return ret;
2099         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2100         if (ret != 0)
2101                 return ret;
2102         /*
2103          * We may have performed cached opens for all three recoveries.
2104          * Check if we need to update the current stateid.
2105          */
2106         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2107             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2108                 write_seqlock(&state->seqlock);
2109                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2110                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2111                 write_sequnlock(&state->seqlock);
2112         }
2113         return 0;
2114 }
2115
2116 /*
2117  * OPEN_RECLAIM:
2118  *      reclaim state on the server after a reboot.
2119  */
2120 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2121 {
2122         struct nfs_delegation *delegation;
2123         struct nfs4_opendata *opendata;
2124         fmode_t delegation_type = 0;
2125         int status;
2126
2127         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2128                         NFS4_OPEN_CLAIM_PREVIOUS);
2129         if (IS_ERR(opendata))
2130                 return PTR_ERR(opendata);
2131         rcu_read_lock();
2132         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2133         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2134                 delegation_type = delegation->type;
2135         rcu_read_unlock();
2136         opendata->o_arg.u.delegation_type = delegation_type;
2137         status = nfs4_open_recover(opendata, state);
2138         nfs4_opendata_put(opendata);
2139         return status;
2140 }
2141
2142 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2143 {
2144         struct nfs_server *server = NFS_SERVER(state->inode);
2145         struct nfs4_exception exception = { };
2146         int err;
2147         do {
2148                 err = _nfs4_do_open_reclaim(ctx, state);
2149                 trace_nfs4_open_reclaim(ctx, 0, err);
2150                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2151                         continue;
2152                 if (err != -NFS4ERR_DELAY)
2153                         break;
2154                 nfs4_handle_exception(server, err, &exception);
2155         } while (exception.retry);
2156         return err;
2157 }
2158
2159 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2160 {
2161         struct nfs_open_context *ctx;
2162         int ret;
2163
2164         ctx = nfs4_state_find_open_context(state);
2165         if (IS_ERR(ctx))
2166                 return -EAGAIN;
2167         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2168         nfs_state_clear_open_state_flags(state);
2169         ret = nfs4_do_open_reclaim(ctx, state);
2170         put_nfs_open_context(ctx);
2171         return ret;
2172 }
2173
2174 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)
2175 {
2176         switch (err) {
2177                 default:
2178                         printk(KERN_ERR "NFS: %s: unhandled error "
2179                                         "%d.\n", __func__, err);
2180                 case 0:
2181                 case -ENOENT:
2182                 case -EAGAIN:
2183                 case -ESTALE:
2184                 case -ETIMEDOUT:
2185                         break;
2186                 case -NFS4ERR_BADSESSION:
2187                 case -NFS4ERR_BADSLOT:
2188                 case -NFS4ERR_BAD_HIGH_SLOT:
2189                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2190                 case -NFS4ERR_DEADSESSION:
2191                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2192                         return -EAGAIN;
2193                 case -NFS4ERR_STALE_CLIENTID:
2194                 case -NFS4ERR_STALE_STATEID:
2195                         /* Don't recall a delegation if it was lost */
2196                         nfs4_schedule_lease_recovery(server->nfs_client);
2197                         return -EAGAIN;
2198                 case -NFS4ERR_MOVED:
2199                         nfs4_schedule_migration_recovery(server);
2200                         return -EAGAIN;
2201                 case -NFS4ERR_LEASE_MOVED:
2202                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2203                         return -EAGAIN;
2204                 case -NFS4ERR_DELEG_REVOKED:
2205                 case -NFS4ERR_ADMIN_REVOKED:
2206                 case -NFS4ERR_EXPIRED:
2207                 case -NFS4ERR_BAD_STATEID:
2208                 case -NFS4ERR_OPENMODE:
2209                         nfs_inode_find_state_and_recover(state->inode,
2210                                         stateid);
2211                         nfs4_schedule_stateid_recovery(server, state);
2212                         return -EAGAIN;
2213                 case -NFS4ERR_DELAY:
2214                 case -NFS4ERR_GRACE:
2215                         ssleep(1);
2216                         return -EAGAIN;
2217                 case -ENOMEM:
2218                 case -NFS4ERR_DENIED:
2219                         if (fl) {
2220                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2221                                 if (lsp)
2222                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2223                         }
2224                         return 0;
2225         }
2226         return err;
2227 }
2228
2229 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2230                 struct nfs4_state *state, const nfs4_stateid *stateid)
2231 {
2232         struct nfs_server *server = NFS_SERVER(state->inode);
2233         struct nfs4_opendata *opendata;
2234         int err = 0;
2235
2236         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2237                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2238         if (IS_ERR(opendata))
2239                 return PTR_ERR(opendata);
2240         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2241         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2242                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2243                 if (err)
2244                         goto out;
2245         }
2246         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2247                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2248                 if (err)
2249                         goto out;
2250         }
2251         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2252                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2253                 if (err)
2254                         goto out;
2255         }
2256         nfs_state_clear_delegation(state);
2257 out:
2258         nfs4_opendata_put(opendata);
2259         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2260 }
2261
2262 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2263 {
2264         struct nfs4_opendata *data = calldata;
2265
2266         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2267                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2268 }
2269
2270 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2271 {
2272         struct nfs4_opendata *data = calldata;
2273
2274         nfs40_sequence_done(task, &data->c_res.seq_res);
2275
2276         data->rpc_status = task->tk_status;
2277         if (data->rpc_status == 0) {
2278                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2279                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2280                 renew_lease(data->o_res.server, data->timestamp);
2281                 data->rpc_done = true;
2282         }
2283 }
2284
2285 static void nfs4_open_confirm_release(void *calldata)
2286 {
2287         struct nfs4_opendata *data = calldata;
2288         struct nfs4_state *state = NULL;
2289
2290         /* If this request hasn't been cancelled, do nothing */
2291         if (!data->cancelled)
2292                 goto out_free;
2293         /* In case of error, no cleanup! */
2294         if (!data->rpc_done)
2295                 goto out_free;
2296         state = nfs4_opendata_to_nfs4_state(data);
2297         if (!IS_ERR(state))
2298                 nfs4_close_state(state, data->o_arg.fmode);
2299 out_free:
2300         nfs4_opendata_put(data);
2301 }
2302
2303 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2304         .rpc_call_prepare = nfs4_open_confirm_prepare,
2305         .rpc_call_done = nfs4_open_confirm_done,
2306         .rpc_release = nfs4_open_confirm_release,
2307 };
2308
2309 /*
2310  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2311  */
2312 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2313 {
2314         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2315         struct rpc_task *task;
2316         struct  rpc_message msg = {
2317                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2318                 .rpc_argp = &data->c_arg,
2319                 .rpc_resp = &data->c_res,
2320                 .rpc_cred = data->owner->so_cred,
2321         };
2322         struct rpc_task_setup task_setup_data = {
2323                 .rpc_client = server->client,
2324                 .rpc_message = &msg,
2325                 .callback_ops = &nfs4_open_confirm_ops,
2326                 .callback_data = data,
2327                 .workqueue = nfsiod_workqueue,
2328                 .flags = RPC_TASK_ASYNC,
2329         };
2330         int status;
2331
2332         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2333                                 data->is_recover);
2334         kref_get(&data->kref);
2335         data->rpc_done = false;
2336         data->rpc_status = 0;
2337         data->timestamp = jiffies;
2338         task = rpc_run_task(&task_setup_data);
2339         if (IS_ERR(task))
2340                 return PTR_ERR(task);
2341         status = rpc_wait_for_completion_task(task);
2342         if (status != 0) {
2343                 data->cancelled = true;
2344                 smp_wmb();
2345         } else
2346                 status = data->rpc_status;
2347         rpc_put_task(task);
2348         return status;
2349 }
2350
2351 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2352 {
2353         struct nfs4_opendata *data = calldata;
2354         struct nfs4_state_owner *sp = data->owner;
2355         struct nfs_client *clp = sp->so_server->nfs_client;
2356         enum open_claim_type4 claim = data->o_arg.claim;
2357
2358         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2359                 goto out_wait;
2360         /*
2361          * Check if we still need to send an OPEN call, or if we can use
2362          * a delegation instead.
2363          */
2364         if (data->state != NULL) {
2365                 struct nfs_delegation *delegation;
2366
2367                 if (can_open_cached(data->state, data->o_arg.fmode,
2368                                         data->o_arg.open_flags, claim))
2369                         goto out_no_action;
2370                 rcu_read_lock();
2371                 delegation = nfs4_get_valid_delegation(data->state->inode);
2372                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2373                         goto unlock_no_action;
2374                 rcu_read_unlock();
2375         }
2376         /* Update client id. */
2377         data->o_arg.clientid = clp->cl_clientid;
2378         switch (claim) {
2379         default:
2380                 break;
2381         case NFS4_OPEN_CLAIM_PREVIOUS:
2382         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2383         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2384                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2385                 /* Fall through */
2386         case NFS4_OPEN_CLAIM_FH:
2387                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2388         }
2389         data->timestamp = jiffies;
2390         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2391                                 &data->o_arg.seq_args,
2392                                 &data->o_res.seq_res,
2393                                 task) != 0)
2394                 nfs_release_seqid(data->o_arg.seqid);
2395
2396         /* Set the create mode (note dependency on the session type) */
2397         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2398         if (data->o_arg.open_flags & O_EXCL) {
2399                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2400                 if (nfs4_has_persistent_session(clp))
2401                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2402                 else if (clp->cl_mvops->minor_version > 0)
2403                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2404         }
2405         return;
2406 unlock_no_action:
2407         trace_nfs4_cached_open(data->state);
2408         rcu_read_unlock();
2409 out_no_action:
2410         task->tk_action = NULL;
2411 out_wait:
2412         nfs4_sequence_done(task, &data->o_res.seq_res);
2413 }
2414
2415 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2416 {
2417         struct nfs4_opendata *data = calldata;
2418
2419         data->rpc_status = task->tk_status;
2420
2421         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2422                 return;
2423
2424         if (task->tk_status == 0) {
2425                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2426                         switch (data->o_res.f_attr->mode & S_IFMT) {
2427                         case S_IFREG:
2428                                 break;
2429                         case S_IFLNK:
2430                                 data->rpc_status = -ELOOP;
2431                                 break;
2432                         case S_IFDIR:
2433                                 data->rpc_status = -EISDIR;
2434                                 break;
2435                         default:
2436                                 data->rpc_status = -ENOTDIR;
2437                         }
2438                 }
2439                 renew_lease(data->o_res.server, data->timestamp);
2440                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2441                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2442         }
2443         data->rpc_done = true;
2444 }
2445
2446 static void nfs4_open_release(void *calldata)
2447 {
2448         struct nfs4_opendata *data = calldata;
2449         struct nfs4_state *state = NULL;
2450
2451         /* If this request hasn't been cancelled, do nothing */
2452         if (!data->cancelled)
2453                 goto out_free;
2454         /* In case of error, no cleanup! */
2455         if (data->rpc_status != 0 || !data->rpc_done)
2456                 goto out_free;
2457         /* In case we need an open_confirm, no cleanup! */
2458         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2459                 goto out_free;
2460         state = nfs4_opendata_to_nfs4_state(data);
2461         if (!IS_ERR(state))
2462                 nfs4_close_state(state, data->o_arg.fmode);
2463 out_free:
2464         nfs4_opendata_put(data);
2465 }
2466
2467 static const struct rpc_call_ops nfs4_open_ops = {
2468         .rpc_call_prepare = nfs4_open_prepare,
2469         .rpc_call_done = nfs4_open_done,
2470         .rpc_release = nfs4_open_release,
2471 };
2472
2473 static int nfs4_run_open_task(struct nfs4_opendata *data,
2474                               struct nfs_open_context *ctx)
2475 {
2476         struct inode *dir = d_inode(data->dir);
2477         struct nfs_server *server = NFS_SERVER(dir);
2478         struct nfs_openargs *o_arg = &data->o_arg;
2479         struct nfs_openres *o_res = &data->o_res;
2480         struct rpc_task *task;
2481         struct rpc_message msg = {
2482                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2483                 .rpc_argp = o_arg,
2484                 .rpc_resp = o_res,
2485                 .rpc_cred = data->owner->so_cred,
2486         };
2487         struct rpc_task_setup task_setup_data = {
2488                 .rpc_client = server->client,
2489                 .rpc_message = &msg,
2490                 .callback_ops = &nfs4_open_ops,
2491                 .callback_data = data,
2492                 .workqueue = nfsiod_workqueue,
2493                 .flags = RPC_TASK_ASYNC,
2494         };
2495         int status;
2496
2497         kref_get(&data->kref);
2498         data->rpc_done = false;
2499         data->rpc_status = 0;
2500         data->cancelled = false;
2501         data->is_recover = false;
2502         if (!ctx) {
2503                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2504                 data->is_recover = true;
2505                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2506         } else {
2507                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2508                 pnfs_lgopen_prepare(data, ctx);
2509         }
2510         task = rpc_run_task(&task_setup_data);
2511         if (IS_ERR(task))
2512                 return PTR_ERR(task);
2513         status = rpc_wait_for_completion_task(task);
2514         if (status != 0) {
2515                 data->cancelled = true;
2516                 smp_wmb();
2517         } else
2518                 status = data->rpc_status;
2519         rpc_put_task(task);
2520
2521         return status;
2522 }
2523
2524 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2525 {
2526         struct inode *dir = d_inode(data->dir);
2527         struct nfs_openres *o_res = &data->o_res;
2528         int status;
2529
2530         status = nfs4_run_open_task(data, NULL);
2531         if (status != 0 || !data->rpc_done)
2532                 return status;
2533
2534         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2535
2536         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2537                 status = _nfs4_proc_open_confirm(data);
2538
2539         return status;
2540 }
2541
2542 /*
2543  * Additional permission checks in order to distinguish between an
2544  * open for read, and an open for execute. This works around the
2545  * fact that NFSv4 OPEN treats read and execute permissions as being
2546  * the same.
2547  * Note that in the non-execute case, we want to turn off permission
2548  * checking if we just created a new file (POSIX open() semantics).
2549  */
2550 static int nfs4_opendata_access(const struct cred *cred,
2551                                 struct nfs4_opendata *opendata,
2552                                 struct nfs4_state *state, fmode_t fmode,
2553                                 int openflags)
2554 {
2555         struct nfs_access_entry cache;
2556         u32 mask, flags;
2557
2558         /* access call failed or for some reason the server doesn't
2559          * support any access modes -- defer access call until later */
2560         if (opendata->o_res.access_supported == 0)
2561                 return 0;
2562
2563         mask = 0;
2564         /*
2565          * Use openflags to check for exec, because fmode won't
2566          * always have FMODE_EXEC set when file open for exec.
2567          */
2568         if (openflags & __FMODE_EXEC) {
2569                 /* ONLY check for exec rights */
2570                 if (S_ISDIR(state->inode->i_mode))
2571                         mask = NFS4_ACCESS_LOOKUP;
2572                 else
2573                         mask = NFS4_ACCESS_EXECUTE;
2574         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2575                 mask = NFS4_ACCESS_READ;
2576
2577         cache.cred = cred;
2578         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2579         nfs_access_add_cache(state->inode, &cache);
2580
2581         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2582         if ((mask & ~cache.mask & flags) == 0)
2583                 return 0;
2584
2585         return -EACCES;
2586 }
2587
2588 /*
2589  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2590  */
2591 static int _nfs4_proc_open(struct nfs4_opendata *data,
2592                            struct nfs_open_context *ctx)
2593 {
2594         struct inode *dir = d_inode(data->dir);
2595         struct nfs_server *server = NFS_SERVER(dir);
2596         struct nfs_openargs *o_arg = &data->o_arg;
2597         struct nfs_openres *o_res = &data->o_res;
2598         int status;
2599
2600         status = nfs4_run_open_task(data, ctx);
2601         if (!data->rpc_done)
2602                 return status;
2603         if (status != 0) {
2604                 if (status == -NFS4ERR_BADNAME &&
2605                                 !(o_arg->open_flags & O_CREAT))
2606                         return -ENOENT;
2607                 return status;
2608         }
2609
2610         nfs_fattr_map_and_free_names(server, &data->f_attr);
2611
2612         if (o_arg->open_flags & O_CREAT) {
2613                 if (o_arg->open_flags & O_EXCL)
2614                         data->file_created = true;
2615                 else if (o_res->cinfo.before != o_res->cinfo.after)
2616                         data->file_created = true;
2617                 if (data->file_created ||
2618                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2619                         update_changeattr(dir, &o_res->cinfo,
2620                                         o_res->f_attr->time_start, 0);
2621         }
2622         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2623                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2624         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2625                 status = _nfs4_proc_open_confirm(data);
2626                 if (status != 0)
2627                         return status;
2628         }
2629         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2630                 nfs4_sequence_free_slot(&o_res->seq_res);
2631                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2632                                 o_res->f_label, NULL);
2633         }
2634         return 0;
2635 }
2636
2637 /*
2638  * OPEN_EXPIRED:
2639  *      reclaim state on the server after a network partition.
2640  *      Assumes caller holds the appropriate lock
2641  */
2642 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2643 {
2644         struct nfs4_opendata *opendata;
2645         int ret;
2646
2647         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2648                         NFS4_OPEN_CLAIM_FH);
2649         if (IS_ERR(opendata))
2650                 return PTR_ERR(opendata);
2651         ret = nfs4_open_recover(opendata, state);
2652         if (ret == -ESTALE)
2653                 d_drop(ctx->dentry);
2654         nfs4_opendata_put(opendata);
2655         return ret;
2656 }
2657
2658 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2659 {
2660         struct nfs_server *server = NFS_SERVER(state->inode);
2661         struct nfs4_exception exception = { };
2662         int err;
2663
2664         do {
2665                 err = _nfs4_open_expired(ctx, state);
2666                 trace_nfs4_open_expired(ctx, 0, err);
2667                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2668                         continue;
2669                 switch (err) {
2670                 default:
2671                         goto out;
2672                 case -NFS4ERR_GRACE:
2673                 case -NFS4ERR_DELAY:
2674                         nfs4_handle_exception(server, err, &exception);
2675                         err = 0;
2676                 }
2677         } while (exception.retry);
2678 out:
2679         return err;
2680 }
2681
2682 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2683 {
2684         struct nfs_open_context *ctx;
2685         int ret;
2686
2687         ctx = nfs4_state_find_open_context(state);
2688         if (IS_ERR(ctx))
2689                 return -EAGAIN;
2690         ret = nfs4_do_open_expired(ctx, state);
2691         put_nfs_open_context(ctx);
2692         return ret;
2693 }
2694
2695 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2696                 const nfs4_stateid *stateid)
2697 {
2698         nfs_remove_bad_delegation(state->inode, stateid);
2699         nfs_state_clear_delegation(state);
2700 }
2701
2702 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2703 {
2704         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2705                 nfs_finish_clear_delegation_stateid(state, NULL);
2706 }
2707
2708 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2709 {
2710         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2711         nfs40_clear_delegation_stateid(state);
2712         nfs_state_clear_open_state_flags(state);
2713         return nfs4_open_expired(sp, state);
2714 }
2715
2716 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2717                 nfs4_stateid *stateid,
2718                 const struct cred *cred)
2719 {
2720         return -NFS4ERR_BAD_STATEID;
2721 }
2722
2723 #if defined(CONFIG_NFS_V4_1)
2724 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2725                 nfs4_stateid *stateid,
2726                 const struct cred *cred)
2727 {
2728         int status;
2729
2730         switch (stateid->type) {
2731         default:
2732                 break;
2733         case NFS4_INVALID_STATEID_TYPE:
2734         case NFS4_SPECIAL_STATEID_TYPE:
2735                 return -NFS4ERR_BAD_STATEID;
2736         case NFS4_REVOKED_STATEID_TYPE:
2737                 goto out_free;
2738         }
2739
2740         status = nfs41_test_stateid(server, stateid, cred);
2741         switch (status) {
2742         case -NFS4ERR_EXPIRED:
2743         case -NFS4ERR_ADMIN_REVOKED:
2744         case -NFS4ERR_DELEG_REVOKED:
2745                 break;
2746         default:
2747                 return status;
2748         }
2749 out_free:
2750         /* Ack the revoked state to the server */
2751         nfs41_free_stateid(server, stateid, cred, true);
2752         return -NFS4ERR_EXPIRED;
2753 }
2754
2755 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2756 {
2757         struct nfs_server *server = NFS_SERVER(state->inode);
2758         nfs4_stateid stateid;
2759         struct nfs_delegation *delegation;
2760         const struct cred *cred = NULL;
2761         int status, ret = NFS_OK;
2762
2763         /* Get the delegation credential for use by test/free_stateid */
2764         rcu_read_lock();
2765         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2766         if (delegation == NULL) {
2767                 rcu_read_unlock();
2768                 nfs_state_clear_delegation(state);
2769                 return NFS_OK;
2770         }
2771
2772         nfs4_stateid_copy(&stateid, &delegation->stateid);
2773
2774         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2775                                 &delegation->flags)) {
2776                 rcu_read_unlock();
2777                 return NFS_OK;
2778         }
2779
2780         if (delegation->cred)
2781                 cred = get_cred(delegation->cred);
2782         rcu_read_unlock();
2783         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2784         trace_nfs4_test_delegation_stateid(state, NULL, status);
2785         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2786                 nfs_finish_clear_delegation_stateid(state, &stateid);
2787         else
2788                 ret = status;
2789
2790         put_cred(cred);
2791         return ret;
2792 }
2793
2794 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2795 {
2796         nfs4_stateid tmp;
2797
2798         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2799             nfs4_copy_delegation_stateid(state->inode, state->state,
2800                                 &tmp, NULL) &&
2801             nfs4_stateid_match_other(&state->stateid, &tmp))
2802                 nfs_state_set_delegation(state, &tmp, state->state);
2803         else
2804                 nfs_state_clear_delegation(state);
2805 }
2806
2807 /**
2808  * nfs41_check_expired_locks - possibly free a lock stateid
2809  *
2810  * @state: NFSv4 state for an inode
2811  *
2812  * Returns NFS_OK if recovery for this stateid is now finished.
2813  * Otherwise a negative NFS4ERR value is returned.
2814  */
2815 static int nfs41_check_expired_locks(struct nfs4_state *state)
2816 {
2817         int status, ret = NFS_OK;
2818         struct nfs4_lock_state *lsp, *prev = NULL;
2819         struct nfs_server *server = NFS_SERVER(state->inode);
2820
2821         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2822                 goto out;
2823
2824         spin_lock(&state->state_lock);
2825         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2826                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2827                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2828
2829                         refcount_inc(&lsp->ls_count);
2830                         spin_unlock(&state->state_lock);
2831
2832                         nfs4_put_lock_state(prev);
2833                         prev = lsp;
2834
2835                         status = nfs41_test_and_free_expired_stateid(server,
2836                                         &lsp->ls_stateid,
2837                                         cred);
2838                         trace_nfs4_test_lock_stateid(state, lsp, status);
2839                         if (status == -NFS4ERR_EXPIRED ||
2840                             status == -NFS4ERR_BAD_STATEID) {
2841                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2842                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2843                                 if (!recover_lost_locks)
2844                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2845                         } else if (status != NFS_OK) {
2846                                 ret = status;
2847                                 nfs4_put_lock_state(prev);
2848                                 goto out;
2849                         }
2850                         spin_lock(&state->state_lock);
2851                 }
2852         }
2853         spin_unlock(&state->state_lock);
2854         nfs4_put_lock_state(prev);
2855 out:
2856         return ret;
2857 }
2858
2859 /**
2860  * nfs41_check_open_stateid - possibly free an open stateid
2861  *
2862  * @state: NFSv4 state for an inode
2863  *
2864  * Returns NFS_OK if recovery for this stateid is now finished.
2865  * Otherwise a negative NFS4ERR value is returned.
2866  */
2867 static int nfs41_check_open_stateid(struct nfs4_state *state)
2868 {
2869         struct nfs_server *server = NFS_SERVER(state->inode);
2870         nfs4_stateid *stateid = &state->open_stateid;
2871         const struct cred *cred = state->owner->so_cred;
2872         int status;
2873
2874         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2875                 return -NFS4ERR_BAD_STATEID;
2876         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2877         trace_nfs4_test_open_stateid(state, NULL, status);
2878         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2879                 nfs_state_clear_open_state_flags(state);
2880                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2881                 return status;
2882         }
2883         if (nfs_open_stateid_recover_openmode(state))
2884                 return -NFS4ERR_OPENMODE;
2885         return NFS_OK;
2886 }
2887
2888 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2889 {
2890         int status;
2891
2892         status = nfs41_check_delegation_stateid(state);
2893         if (status != NFS_OK)
2894                 return status;
2895         nfs41_delegation_recover_stateid(state);
2896
2897         status = nfs41_check_expired_locks(state);
2898         if (status != NFS_OK)
2899                 return status;
2900         status = nfs41_check_open_stateid(state);
2901         if (status != NFS_OK)
2902                 status = nfs4_open_expired(sp, state);
2903         return status;
2904 }
2905 #endif
2906
2907 /*
2908  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2909  * fields corresponding to attributes that were used to store the verifier.
2910  * Make sure we clobber those fields in the later setattr call
2911  */
2912 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2913                                 struct iattr *sattr, struct nfs4_label **label)
2914 {
2915         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2916         __u32 attrset[3];
2917         unsigned ret;
2918         unsigned i;
2919
2920         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2921                 attrset[i] = opendata->o_res.attrset[i];
2922                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2923                         attrset[i] &= ~bitmask[i];
2924         }
2925
2926         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2927                 sattr->ia_valid : 0;
2928
2929         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2930                 if (sattr->ia_valid & ATTR_ATIME_SET)
2931                         ret |= ATTR_ATIME_SET;
2932                 else
2933                         ret |= ATTR_ATIME;
2934         }
2935
2936         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2937                 if (sattr->ia_valid & ATTR_MTIME_SET)
2938                         ret |= ATTR_MTIME_SET;
2939                 else
2940                         ret |= ATTR_MTIME;
2941         }
2942
2943         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2944                 *label = NULL;
2945         return ret;
2946 }
2947
2948 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2949                 int flags, struct nfs_open_context *ctx)
2950 {
2951         struct nfs4_state_owner *sp = opendata->owner;
2952         struct nfs_server *server = sp->so_server;
2953         struct dentry *dentry;
2954         struct nfs4_state *state;
2955         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2956         unsigned int seq;
2957         int ret;
2958
2959         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2960
2961         ret = _nfs4_proc_open(opendata, ctx);
2962         if (ret != 0)
2963                 goto out;
2964
2965         state = _nfs4_opendata_to_nfs4_state(opendata);
2966         ret = PTR_ERR(state);
2967         if (IS_ERR(state))
2968                 goto out;
2969         ctx->state = state;
2970         if (server->caps & NFS_CAP_POSIX_LOCK)
2971                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2972         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2973                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2974
2975         dentry = opendata->dentry;
2976         if (d_really_is_negative(dentry)) {
2977                 struct dentry *alias;
2978                 d_drop(dentry);
2979                 alias = d_exact_alias(dentry, state->inode);
2980                 if (!alias)
2981                         alias = d_splice_alias(igrab(state->inode), dentry);
2982                 /* d_splice_alias() can't fail here - it's a non-directory */
2983                 if (alias) {
2984                         dput(ctx->dentry);
2985                         ctx->dentry = dentry = alias;
2986                 }
2987                 nfs_set_verifier(dentry,
2988                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2989         }
2990
2991         /* Parse layoutget results before we check for access */
2992         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2993
2994         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
2995                         acc_mode, flags);
2996         if (ret != 0)
2997                 goto out;
2998
2999         if (d_inode(dentry) == state->inode) {
3000                 nfs_inode_attach_open_context(ctx);
3001                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3002                         nfs4_schedule_stateid_recovery(server, state);
3003         }
3004
3005 out:
3006         if (!opendata->cancelled)
3007                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3008         return ret;
3009 }
3010
3011 /*
3012  * Returns a referenced nfs4_state
3013  */
3014 static int _nfs4_do_open(struct inode *dir,
3015                         struct nfs_open_context *ctx,
3016                         int flags,
3017                         const struct nfs4_open_createattrs *c,
3018                         int *opened)
3019 {
3020         struct nfs4_state_owner  *sp;
3021         struct nfs4_state     *state = NULL;
3022         struct nfs_server       *server = NFS_SERVER(dir);
3023         struct nfs4_opendata *opendata;
3024         struct dentry *dentry = ctx->dentry;
3025         const struct cred *cred = ctx->cred;
3026         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3027         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3028         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3029         struct iattr *sattr = c->sattr;
3030         struct nfs4_label *label = c->label;
3031         struct nfs4_label *olabel = NULL;
3032         int status;
3033
3034         /* Protect against reboot recovery conflicts */
3035         status = -ENOMEM;
3036         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3037         if (sp == NULL) {
3038                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3039                 goto out_err;
3040         }
3041         status = nfs4_client_recover_expired_lease(server->nfs_client);
3042         if (status != 0)
3043                 goto err_put_state_owner;
3044         if (d_really_is_positive(dentry))
3045                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3046         status = -ENOMEM;
3047         if (d_really_is_positive(dentry))
3048                 claim = NFS4_OPEN_CLAIM_FH;
3049         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3050                         c, claim, GFP_KERNEL);
3051         if (opendata == NULL)
3052                 goto err_put_state_owner;
3053
3054         if (label) {
3055                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3056                 if (IS_ERR(olabel)) {
3057                         status = PTR_ERR(olabel);
3058                         goto err_opendata_put;
3059                 }
3060         }
3061
3062         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3063                 if (!opendata->f_attr.mdsthreshold) {
3064                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3065                         if (!opendata->f_attr.mdsthreshold)
3066                                 goto err_free_label;
3067                 }
3068                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3069         }
3070         if (d_really_is_positive(dentry))
3071                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3072
3073         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3074         if (status != 0)
3075                 goto err_free_label;
3076         state = ctx->state;
3077
3078         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3079             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3080                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3081                 /*
3082                  * send create attributes which was not set by open
3083                  * with an extra setattr.
3084                  */
3085                 if (attrs || label) {
3086                         unsigned ia_old = sattr->ia_valid;
3087
3088                         sattr->ia_valid = attrs;
3089                         nfs_fattr_init(opendata->o_res.f_attr);
3090                         status = nfs4_do_setattr(state->inode, cred,
3091                                         opendata->o_res.f_attr, sattr,
3092                                         ctx, label, olabel);
3093                         if (status == 0) {
3094                                 nfs_setattr_update_inode(state->inode, sattr,
3095                                                 opendata->o_res.f_attr);
3096                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3097                         }
3098                         sattr->ia_valid = ia_old;
3099                 }
3100         }
3101         if (opened && opendata->file_created)
3102                 *opened = 1;
3103
3104         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3105                 *ctx_th = opendata->f_attr.mdsthreshold;
3106                 opendata->f_attr.mdsthreshold = NULL;
3107         }
3108
3109         nfs4_label_free(olabel);
3110
3111         nfs4_opendata_put(opendata);
3112         nfs4_put_state_owner(sp);
3113         return 0;
3114 err_free_label:
3115         nfs4_label_free(olabel);
3116 err_opendata_put:
3117         nfs4_opendata_put(opendata);
3118 err_put_state_owner:
3119         nfs4_put_state_owner(sp);
3120 out_err:
3121         return status;
3122 }
3123
3124
3125 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3126                                         struct nfs_open_context *ctx,
3127                                         int flags,
3128                                         struct iattr *sattr,
3129                                         struct nfs4_label *label,
3130                                         int *opened)
3131 {
3132         struct nfs_server *server = NFS_SERVER(dir);
3133         struct nfs4_exception exception = {
3134                 .interruptible = true,
3135         };
3136         struct nfs4_state *res;
3137         struct nfs4_open_createattrs c = {
3138                 .label = label,
3139                 .sattr = sattr,
3140                 .verf = {
3141                         [0] = (__u32)jiffies,
3142                         [1] = (__u32)current->pid,
3143                 },
3144         };
3145         int status;
3146
3147         do {
3148                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3149                 res = ctx->state;
3150                 trace_nfs4_open_file(ctx, flags, status);
3151                 if (status == 0)
3152                         break;
3153                 /* NOTE: BAD_SEQID means the server and client disagree about the
3154                  * book-keeping w.r.t. state-changing operations
3155                  * (OPEN/CLOSE/LOCK/LOCKU...)
3156                  * It is actually a sign of a bug on the client or on the server.
3157                  *
3158                  * If we receive a BAD_SEQID error in the particular case of
3159                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3160                  * have unhashed the old state_owner for us, and that we can
3161                  * therefore safely retry using a new one. We should still warn
3162                  * the user though...
3163                  */
3164                 if (status == -NFS4ERR_BAD_SEQID) {
3165                         pr_warn_ratelimited("NFS: v4 server %s "
3166                                         " returned a bad sequence-id error!\n",
3167                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3168                         exception.retry = 1;
3169                         continue;
3170                 }
3171                 /*
3172                  * BAD_STATEID on OPEN means that the server cancelled our
3173                  * state before it received the OPEN_CONFIRM.
3174                  * Recover by retrying the request as per the discussion
3175                  * on Page 181 of RFC3530.
3176                  */
3177                 if (status == -NFS4ERR_BAD_STATEID) {
3178                         exception.retry = 1;
3179                         continue;
3180                 }
3181                 if (status == -EAGAIN) {
3182                         /* We must have found a delegation */
3183                         exception.retry = 1;
3184                         continue;
3185                 }
3186                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3187                         continue;
3188                 res = ERR_PTR(nfs4_handle_exception(server,
3189                                         status, &exception));
3190         } while (exception.retry);
3191         return res;
3192 }
3193
3194 static int _nfs4_do_setattr(struct inode *inode,
3195                             struct nfs_setattrargs *arg,
3196                             struct nfs_setattrres *res,
3197                             const struct cred *cred,
3198                             struct nfs_open_context *ctx)
3199 {
3200         struct nfs_server *server = NFS_SERVER(inode);
3201         struct rpc_message msg = {
3202                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3203                 .rpc_argp       = arg,
3204                 .rpc_resp       = res,
3205                 .rpc_cred       = cred,
3206         };
3207         const struct cred *delegation_cred = NULL;
3208         unsigned long timestamp = jiffies;
3209         bool truncate;
3210         int status;
3211
3212         nfs_fattr_init(res->fattr);
3213
3214         /* Servers should only apply open mode checks for file size changes */
3215         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3216         if (!truncate)
3217                 goto zero_stateid;
3218
3219         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3220                 /* Use that stateid */
3221         } else if (ctx != NULL && ctx->state) {
3222                 struct nfs_lock_context *l_ctx;
3223                 if (!nfs4_valid_open_stateid(ctx->state))
3224                         return -EBADF;
3225                 l_ctx = nfs_get_lock_context(ctx);
3226                 if (IS_ERR(l_ctx))
3227                         return PTR_ERR(l_ctx);
3228                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3229                                                 &arg->stateid, &delegation_cred);
3230                 nfs_put_lock_context(l_ctx);
3231                 if (status == -EIO)
3232                         return -EBADF;
3233         } else {
3234 zero_stateid:
3235                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3236         }
3237         if (delegation_cred)
3238                 msg.rpc_cred = delegation_cred;
3239
3240         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3241
3242         put_cred(delegation_cred);
3243         if (status == 0 && ctx != NULL)
3244                 renew_lease(server, timestamp);
3245         trace_nfs4_setattr(inode, &arg->stateid, status);
3246         return status;
3247 }
3248
3249 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3250                            struct nfs_fattr *fattr, struct iattr *sattr,
3251                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3252                            struct nfs4_label *olabel)
3253 {
3254         struct nfs_server *server = NFS_SERVER(inode);
3255         __u32 bitmask[NFS4_BITMASK_SZ];
3256         struct nfs4_state *state = ctx ? ctx->state : NULL;
3257         struct nfs_setattrargs  arg = {
3258                 .fh             = NFS_FH(inode),
3259                 .iap            = sattr,
3260                 .server         = server,
3261                 .bitmask = bitmask,
3262                 .label          = ilabel,
3263         };
3264         struct nfs_setattrres  res = {
3265                 .fattr          = fattr,
3266                 .label          = olabel,
3267                 .server         = server,
3268         };
3269         struct nfs4_exception exception = {
3270                 .state = state,
3271                 .inode = inode,
3272                 .stateid = &arg.stateid,
3273         };
3274         int err;
3275
3276         do {
3277                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3278                                 nfs4_bitmask(server, olabel),
3279                                 inode);
3280
3281                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3282                 switch (err) {
3283                 case -NFS4ERR_OPENMODE:
3284                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3285                                 pr_warn_once("NFSv4: server %s is incorrectly "
3286                                                 "applying open mode checks to "
3287                                                 "a SETATTR that is not "
3288                                                 "changing file size.\n",
3289                                                 server->nfs_client->cl_hostname);
3290                         }
3291                         if (state && !(state->state & FMODE_WRITE)) {
3292                                 err = -EBADF;
3293                                 if (sattr->ia_valid & ATTR_OPEN)
3294                                         err = -EACCES;
3295                                 goto out;
3296                         }
3297                 }
3298                 err = nfs4_handle_exception(server, err, &exception);
3299         } while (exception.retry);
3300 out:
3301         return err;
3302 }
3303
3304 static bool
3305 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3306 {
3307         if (inode == NULL || !nfs_have_layout(inode))
3308                 return false;
3309
3310         return pnfs_wait_on_layoutreturn(inode, task);
3311 }
3312
3313 /*
3314  * Update the seqid of an open stateid
3315  */
3316 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3317                 struct nfs4_state *state)
3318 {
3319         __be32 seqid_open;
3320         u32 dst_seqid;
3321         int seq;
3322
3323         for (;;) {
3324                 if (!nfs4_valid_open_stateid(state))
3325                         break;
3326                 seq = read_seqbegin(&state->seqlock);
3327                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3328                         nfs4_stateid_copy(dst, &state->open_stateid);
3329                         if (read_seqretry(&state->seqlock, seq))
3330                                 continue;
3331                         break;
3332                 }
3333                 seqid_open = state->open_stateid.seqid;
3334                 if (read_seqretry(&state->seqlock, seq))
3335                         continue;
3336
3337                 dst_seqid = be32_to_cpu(dst->seqid);
3338                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3339                         dst->seqid = seqid_open;
3340                 break;
3341         }
3342 }
3343
3344 /*
3345  * Update the seqid of an open stateid after receiving
3346  * NFS4ERR_OLD_STATEID
3347  */
3348 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3349                 struct nfs4_state *state)
3350 {
3351         __be32 seqid_open;
3352         u32 dst_seqid;
3353         bool ret;
3354         int seq;
3355
3356         for (;;) {
3357                 ret = false;
3358                 if (!nfs4_valid_open_stateid(state))
3359                         break;
3360                 seq = read_seqbegin(&state->seqlock);
3361                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3362                         if (read_seqretry(&state->seqlock, seq))
3363                                 continue;
3364                         break;
3365                 }
3366                 seqid_open = state->open_stateid.seqid;
3367                 if (read_seqretry(&state->seqlock, seq))
3368                         continue;
3369
3370                 dst_seqid = be32_to_cpu(dst->seqid);
3371                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) >= 0)
3372                         dst->seqid = cpu_to_be32(dst_seqid + 1);
3373                 else
3374                         dst->seqid = seqid_open;
3375                 ret = true;
3376                 break;
3377         }
3378
3379         return ret;
3380 }
3381
3382 struct nfs4_closedata {
3383         struct inode *inode;
3384         struct nfs4_state *state;
3385         struct nfs_closeargs arg;
3386         struct nfs_closeres res;
3387         struct {
3388                 struct nfs4_layoutreturn_args arg;
3389                 struct nfs4_layoutreturn_res res;
3390                 struct nfs4_xdr_opaque_data ld_private;
3391                 u32 roc_barrier;
3392                 bool roc;
3393         } lr;
3394         struct nfs_fattr fattr;
3395         unsigned long timestamp;
3396 };
3397
3398 static void nfs4_free_closedata(void *data)
3399 {
3400         struct nfs4_closedata *calldata = data;
3401         struct nfs4_state_owner *sp = calldata->state->owner;
3402         struct super_block *sb = calldata->state->inode->i_sb;
3403
3404         if (calldata->lr.roc)
3405                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3406                                 calldata->res.lr_ret);
3407         nfs4_put_open_state(calldata->state);
3408         nfs_free_seqid(calldata->arg.seqid);
3409         nfs4_put_state_owner(sp);
3410         nfs_sb_deactive(sb);
3411         kfree(calldata);
3412 }
3413
3414 static void nfs4_close_done(struct rpc_task *task, void *data)
3415 {
3416         struct nfs4_closedata *calldata = data;
3417         struct nfs4_state *state = calldata->state;
3418         struct nfs_server *server = NFS_SERVER(calldata->inode);
3419         nfs4_stateid *res_stateid = NULL;
3420         struct nfs4_exception exception = {
3421                 .state = state,
3422                 .inode = calldata->inode,
3423                 .stateid = &calldata->arg.stateid,
3424         };
3425
3426         dprintk("%s: begin!\n", __func__);
3427         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3428                 return;
3429         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3430
3431         /* Handle Layoutreturn errors */
3432         if (pnfs_roc_done(task, calldata->inode,
3433                                 &calldata->arg.lr_args,
3434                                 &calldata->res.lr_res,
3435                                 &calldata->res.lr_ret) == -EAGAIN)
3436                 goto out_restart;
3437
3438         /* hmm. we are done with the inode, and in the process of freeing
3439          * the state_owner. we keep this around to process errors
3440          */
3441         switch (task->tk_status) {
3442                 case 0:
3443                         res_stateid = &calldata->res.stateid;
3444                         renew_lease(server, calldata->timestamp);
3445                         break;
3446                 case -NFS4ERR_ACCESS:
3447                         if (calldata->arg.bitmask != NULL) {
3448                                 calldata->arg.bitmask = NULL;
3449                                 calldata->res.fattr = NULL;
3450                                 goto out_restart;
3451
3452                         }
3453                         break;
3454                 case -NFS4ERR_OLD_STATEID:
3455                         /* Did we race with OPEN? */
3456                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3457                                                 state))
3458                                 goto out_restart;
3459                         goto out_release;
3460                 case -NFS4ERR_ADMIN_REVOKED:
3461                 case -NFS4ERR_STALE_STATEID:
3462                 case -NFS4ERR_EXPIRED:
3463                         nfs4_free_revoked_stateid(server,
3464                                         &calldata->arg.stateid,
3465                                         task->tk_msg.rpc_cred);
3466                         /* Fallthrough */
3467                 case -NFS4ERR_BAD_STATEID:
3468                         if (calldata->arg.fmode == 0)
3469                                 break;
3470                         /* Fallthrough */
3471                 default:
3472                         task->tk_status = nfs4_async_handle_exception(task,
3473                                         server, task->tk_status, &exception);
3474                         if (exception.retry)
3475                                 goto out_restart;
3476         }
3477         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3478                         res_stateid, calldata->arg.fmode);
3479 out_release:
3480         task->tk_status = 0;
3481         nfs_release_seqid(calldata->arg.seqid);
3482         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3483         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3484         return;
3485 out_restart:
3486         task->tk_status = 0;
3487         rpc_restart_call_prepare(task);
3488         goto out_release;
3489 }
3490
3491 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3492 {
3493         struct nfs4_closedata *calldata = data;
3494         struct nfs4_state *state = calldata->state;
3495         struct inode *inode = calldata->inode;
3496         struct pnfs_layout_hdr *lo;
3497         bool is_rdonly, is_wronly, is_rdwr;
3498         int call_close = 0;
3499
3500         dprintk("%s: begin!\n", __func__);
3501         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3502                 goto out_wait;
3503
3504         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3505         spin_lock(&state->owner->so_lock);
3506         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3507         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3508         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3509         /* Calculate the change in open mode */
3510         calldata->arg.fmode = 0;
3511         if (state->n_rdwr == 0) {
3512                 if (state->n_rdonly == 0)
3513                         call_close |= is_rdonly;
3514                 else if (is_rdonly)
3515                         calldata->arg.fmode |= FMODE_READ;
3516                 if (state->n_wronly == 0)
3517                         call_close |= is_wronly;
3518                 else if (is_wronly)
3519                         calldata->arg.fmode |= FMODE_WRITE;
3520                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3521                         call_close |= is_rdwr;
3522         } else if (is_rdwr)
3523                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3524
3525         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3526         if (!nfs4_valid_open_stateid(state))
3527                 call_close = 0;
3528         spin_unlock(&state->owner->so_lock);
3529
3530         if (!call_close) {
3531                 /* Note: exit _without_ calling nfs4_close_done */
3532                 goto out_no_action;
3533         }
3534
3535         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3536                 nfs_release_seqid(calldata->arg.seqid);
3537                 goto out_wait;
3538         }
3539
3540         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3541         if (lo && !pnfs_layout_is_valid(lo)) {
3542                 calldata->arg.lr_args = NULL;
3543                 calldata->res.lr_res = NULL;
3544         }
3545
3546         if (calldata->arg.fmode == 0)
3547                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3548
3549         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3550                 /* Close-to-open cache consistency revalidation */
3551                 if (!nfs4_have_delegation(inode, FMODE_READ))
3552                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3553                 else
3554                         calldata->arg.bitmask = NULL;
3555         }
3556
3557         calldata->arg.share_access =
3558                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3559                                 calldata->arg.fmode, 0);
3560
3561         if (calldata->res.fattr == NULL)
3562                 calldata->arg.bitmask = NULL;
3563         else if (calldata->arg.bitmask == NULL)
3564                 calldata->res.fattr = NULL;
3565         calldata->timestamp = jiffies;
3566         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3567                                 &calldata->arg.seq_args,
3568                                 &calldata->res.seq_res,
3569                                 task) != 0)
3570                 nfs_release_seqid(calldata->arg.seqid);
3571         dprintk("%s: done!\n", __func__);
3572         return;
3573 out_no_action:
3574         task->tk_action = NULL;
3575 out_wait:
3576         nfs4_sequence_done(task, &calldata->res.seq_res);
3577 }
3578
3579 static const struct rpc_call_ops nfs4_close_ops = {
3580         .rpc_call_prepare = nfs4_close_prepare,
3581         .rpc_call_done = nfs4_close_done,
3582         .rpc_release = nfs4_free_closedata,
3583 };
3584
3585 /*
3586  * It is possible for data to be read/written from a mem-mapped file
3587  * after the sys_close call (which hits the vfs layer as a flush).
3588  * This means that we can't safely call nfsv4 close on a file until
3589  * the inode is cleared. This in turn means that we are not good
3590  * NFSv4 citizens - we do not indicate to the server to update the file's
3591  * share state even when we are done with one of the three share
3592  * stateid's in the inode.
3593  *
3594  * NOTE: Caller must be holding the sp->so_owner semaphore!
3595  */
3596 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3597 {
3598         struct nfs_server *server = NFS_SERVER(state->inode);
3599         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3600         struct nfs4_closedata *calldata;
3601         struct nfs4_state_owner *sp = state->owner;
3602         struct rpc_task *task;
3603         struct rpc_message msg = {
3604                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3605                 .rpc_cred = state->owner->so_cred,
3606         };
3607         struct rpc_task_setup task_setup_data = {
3608                 .rpc_client = server->client,
3609                 .rpc_message = &msg,
3610                 .callback_ops = &nfs4_close_ops,
3611                 .workqueue = nfsiod_workqueue,
3612                 .flags = RPC_TASK_ASYNC,
3613         };
3614         int status = -ENOMEM;
3615
3616         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3617                 &task_setup_data.rpc_client, &msg);
3618
3619         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3620         if (calldata == NULL)
3621                 goto out;
3622         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3623         calldata->inode = state->inode;
3624         calldata->state = state;
3625         calldata->arg.fh = NFS_FH(state->inode);
3626         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3627                 goto out_free_calldata;
3628         /* Serialization for the sequence id */
3629         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3630         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3631         if (IS_ERR(calldata->arg.seqid))
3632                 goto out_free_calldata;
3633         nfs_fattr_init(&calldata->fattr);
3634         calldata->arg.fmode = 0;
3635         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3636         calldata->res.fattr = &calldata->fattr;
3637         calldata->res.seqid = calldata->arg.seqid;
3638         calldata->res.server = server;
3639         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3640         calldata->lr.roc = pnfs_roc(state->inode,
3641                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3642         if (calldata->lr.roc) {
3643                 calldata->arg.lr_args = &calldata->lr.arg;
3644                 calldata->res.lr_res = &calldata->lr.res;
3645         }
3646         nfs_sb_active(calldata->inode->i_sb);
3647
3648         msg.rpc_argp = &calldata->arg;
3649         msg.rpc_resp = &calldata->res;
3650         task_setup_data.callback_data = calldata;
3651         task = rpc_run_task(&task_setup_data);
3652         if (IS_ERR(task))
3653                 return PTR_ERR(task);
3654         status = 0;
3655         if (wait)
3656                 status = rpc_wait_for_completion_task(task);
3657         rpc_put_task(task);
3658         return status;
3659 out_free_calldata:
3660         kfree(calldata);
3661 out:
3662         nfs4_put_open_state(state);
3663         nfs4_put_state_owner(sp);
3664         return status;
3665 }
3666
3667 static struct inode *
3668 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3669                 int open_flags, struct iattr *attr, int *opened)
3670 {
3671         struct nfs4_state *state;
3672         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3673
3674         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3675
3676         /* Protect against concurrent sillydeletes */
3677         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3678
3679         nfs4_label_release_security(label);
3680
3681         if (IS_ERR(state))
3682                 return ERR_CAST(state);
3683         return state->inode;
3684 }
3685
3686 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3687 {
3688         if (ctx->state == NULL)
3689                 return;
3690         if (is_sync)
3691                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3692         else
3693                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3694 }
3695
3696 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3697 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3698 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3699
3700 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3701 {
3702         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3703         struct nfs4_server_caps_arg args = {
3704                 .fhandle = fhandle,
3705                 .bitmask = bitmask,
3706         };
3707         struct nfs4_server_caps_res res = {};
3708         struct rpc_message msg = {
3709                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3710                 .rpc_argp = &args,
3711                 .rpc_resp = &res,
3712         };
3713         int status;
3714         int i;
3715
3716         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3717                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3718                      FATTR4_WORD0_LINK_SUPPORT |
3719                      FATTR4_WORD0_SYMLINK_SUPPORT |
3720                      FATTR4_WORD0_ACLSUPPORT;
3721         if (minorversion)
3722                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3723
3724         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3725         if (status == 0) {
3726                 /* Sanity check the server answers */
3727                 switch (minorversion) {
3728                 case 0:
3729                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3730                         res.attr_bitmask[2] = 0;
3731                         break;
3732                 case 1:
3733                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3734                         break;
3735                 case 2:
3736                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3737                 }
3738                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3739                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3740                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3741                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3742                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3743                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3744                                 NFS_CAP_SECURITY_LABEL);
3745                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3746                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3747                         server->caps |= NFS_CAP_ACLS;
3748                 if (res.has_links != 0)
3749                         server->caps |= NFS_CAP_HARDLINKS;
3750                 if (res.has_symlinks != 0)
3751                         server->caps |= NFS_CAP_SYMLINKS;
3752                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3753                         server->caps |= NFS_CAP_FILEID;
3754                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3755                         server->caps |= NFS_CAP_MODE;
3756                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3757                         server->caps |= NFS_CAP_NLINK;
3758                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3759                         server->caps |= NFS_CAP_OWNER;
3760                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3761                         server->caps |= NFS_CAP_OWNER_GROUP;
3762                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3763                         server->caps |= NFS_CAP_ATIME;
3764                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3765                         server->caps |= NFS_CAP_CTIME;
3766                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3767                         server->caps |= NFS_CAP_MTIME;
3768 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3769                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3770                         server->caps |= NFS_CAP_SECURITY_LABEL;
3771 #endif
3772                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3773                                 sizeof(server->attr_bitmask));
3774                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3775
3776                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3777                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3778                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3779                 server->cache_consistency_bitmask[2] = 0;
3780
3781                 /* Avoid a regression due to buggy server */
3782                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3783                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3784                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3785                         sizeof(server->exclcreat_bitmask));
3786
3787                 server->acl_bitmask = res.acl_bitmask;
3788                 server->fh_expire_type = res.fh_expire_type;
3789         }
3790
3791         return status;
3792 }
3793
3794 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3795 {
3796         struct nfs4_exception exception = {
3797                 .interruptible = true,
3798         };
3799         int err;
3800         do {
3801                 err = nfs4_handle_exception(server,
3802                                 _nfs4_server_capabilities(server, fhandle),
3803                                 &exception);
3804         } while (exception.retry);
3805         return err;
3806 }
3807
3808 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3809                 struct nfs_fsinfo *info)
3810 {
3811         u32 bitmask[3];
3812         struct nfs4_lookup_root_arg args = {
3813                 .bitmask = bitmask,
3814         };
3815         struct nfs4_lookup_res res = {
3816                 .server = server,
3817                 .fattr = info->fattr,
3818                 .fh = fhandle,
3819         };
3820         struct rpc_message msg = {
3821                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3822                 .rpc_argp = &args,
3823                 .rpc_resp = &res,
3824         };
3825
3826         bitmask[0] = nfs4_fattr_bitmap[0];
3827         bitmask[1] = nfs4_fattr_bitmap[1];
3828         /*
3829          * Process the label in the upcoming getfattr
3830          */
3831         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3832
3833         nfs_fattr_init(info->fattr);
3834         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3835 }
3836
3837 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3838                 struct nfs_fsinfo *info)
3839 {
3840         struct nfs4_exception exception = {
3841                 .interruptible = true,
3842         };
3843         int err;
3844         do {
3845                 err = _nfs4_lookup_root(server, fhandle, info);
3846                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3847                 switch (err) {
3848                 case 0:
3849                 case -NFS4ERR_WRONGSEC:
3850                         goto out;
3851                 default:
3852                         err = nfs4_handle_exception(server, err, &exception);
3853                 }
3854         } while (exception.retry);
3855 out:
3856         return err;
3857 }
3858
3859 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3860                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3861 {
3862         struct rpc_auth_create_args auth_args = {
3863                 .pseudoflavor = flavor,
3864         };
3865         struct rpc_auth *auth;
3866
3867         auth = rpcauth_create(&auth_args, server->client);
3868         if (IS_ERR(auth))
3869                 return -EACCES;
3870         return nfs4_lookup_root(server, fhandle, info);
3871 }
3872
3873 /*
3874  * Retry pseudoroot lookup with various security flavors.  We do this when:
3875  *
3876  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3877  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3878  *
3879  * Returns zero on success, or a negative NFS4ERR value, or a
3880  * negative errno value.
3881  */
3882 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3883                               struct nfs_fsinfo *info)
3884 {
3885         /* Per 3530bis 15.33.5 */
3886         static const rpc_authflavor_t flav_array[] = {
3887                 RPC_AUTH_GSS_KRB5P,
3888                 RPC_AUTH_GSS_KRB5I,
3889                 RPC_AUTH_GSS_KRB5,
3890                 RPC_AUTH_UNIX,                  /* courtesy */
3891                 RPC_AUTH_NULL,
3892         };
3893         int status = -EPERM;
3894         size_t i;
3895
3896         if (server->auth_info.flavor_len > 0) {
3897                 /* try each flavor specified by user */
3898                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3899                         status = nfs4_lookup_root_sec(server, fhandle, info,
3900                                                 server->auth_info.flavors[i]);
3901                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3902                                 continue;
3903                         break;
3904                 }
3905         } else {
3906                 /* no flavors specified by user, try default list */
3907                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3908                         status = nfs4_lookup_root_sec(server, fhandle, info,
3909                                                       flav_array[i]);
3910                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3911                                 continue;
3912                         break;
3913                 }
3914         }
3915
3916         /*
3917          * -EACCES could mean that the user doesn't have correct permissions
3918          * to access the mount.  It could also mean that we tried to mount
3919          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3920          * existing mount programs don't handle -EACCES very well so it should
3921          * be mapped to -EPERM instead.
3922          */
3923         if (status == -EACCES)
3924                 status = -EPERM;
3925         return status;
3926 }
3927
3928 /**
3929  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3930  * @server: initialized nfs_server handle
3931  * @fhandle: we fill in the pseudo-fs root file handle
3932  * @info: we fill in an FSINFO struct
3933  * @auth_probe: probe the auth flavours
3934  *
3935  * Returns zero on success, or a negative errno.
3936  */
3937 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3938                          struct nfs_fsinfo *info,
3939                          bool auth_probe)
3940 {
3941         int status = 0;
3942
3943         if (!auth_probe)
3944                 status = nfs4_lookup_root(server, fhandle, info);
3945
3946         if (auth_probe || status == NFS4ERR_WRONGSEC)
3947                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3948                                 fhandle, info);
3949
3950         if (status == 0)
3951                 status = nfs4_server_capabilities(server, fhandle);
3952         if (status == 0)
3953                 status = nfs4_do_fsinfo(server, fhandle, info);
3954
3955         return nfs4_map_errors(status);
3956 }
3957
3958 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3959                               struct nfs_fsinfo *info)
3960 {
3961         int error;
3962         struct nfs_fattr *fattr = info->fattr;
3963         struct nfs4_label *label = NULL;
3964
3965         error = nfs4_server_capabilities(server, mntfh);
3966         if (error < 0) {
3967                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3968                 return error;
3969         }
3970
3971         label = nfs4_label_alloc(server, GFP_KERNEL);
3972         if (IS_ERR(label))
3973                 return PTR_ERR(label);
3974
3975         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3976         if (error < 0) {
3977                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3978                 goto err_free_label;
3979         }
3980
3981         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3982             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3983                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3984
3985 err_free_label:
3986         nfs4_label_free(label);
3987
3988         return error;
3989 }
3990
3991 /*
3992  * Get locations and (maybe) other attributes of a referral.
3993  * Note that we'll actually follow the referral later when
3994  * we detect fsid mismatch in inode revalidation
3995  */
3996 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3997                              const struct qstr *name, struct nfs_fattr *fattr,
3998                              struct nfs_fh *fhandle)
3999 {
4000         int status = -ENOMEM;
4001         struct page *page = NULL;
4002         struct nfs4_fs_locations *locations = NULL;
4003
4004         page = alloc_page(GFP_KERNEL);
4005         if (page == NULL)
4006                 goto out;
4007         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4008         if (locations == NULL)
4009                 goto out;
4010
4011         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4012         if (status != 0)
4013                 goto out;
4014
4015         /*
4016          * If the fsid didn't change, this is a migration event, not a
4017          * referral.  Cause us to drop into the exception handler, which
4018          * will kick off migration recovery.
4019          */
4020         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4021                 dprintk("%s: server did not return a different fsid for"
4022                         " a referral at %s\n", __func__, name->name);
4023                 status = -NFS4ERR_MOVED;
4024                 goto out;
4025         }
4026         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4027         nfs_fixup_referral_attributes(&locations->fattr);
4028
4029         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4030         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4031         memset(fhandle, 0, sizeof(struct nfs_fh));
4032 out:
4033         if (page)
4034                 __free_page(page);
4035         kfree(locations);
4036         return status;
4037 }
4038
4039 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4040                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4041                                 struct inode *inode)
4042 {
4043         __u32 bitmask[NFS4_BITMASK_SZ];
4044         struct nfs4_getattr_arg args = {
4045                 .fh = fhandle,
4046                 .bitmask = bitmask,
4047         };
4048         struct nfs4_getattr_res res = {
4049                 .fattr = fattr,
4050                 .label = label,
4051                 .server = server,
4052         };
4053         struct rpc_message msg = {
4054                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4055                 .rpc_argp = &args,
4056                 .rpc_resp = &res,
4057         };
4058
4059         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4060
4061         nfs_fattr_init(fattr);
4062         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4063 }
4064
4065 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4066                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4067                                 struct inode *inode)
4068 {
4069         struct nfs4_exception exception = {
4070                 .interruptible = true,
4071         };
4072         int err;
4073         do {
4074                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4075                 trace_nfs4_getattr(server, fhandle, fattr, err);
4076                 err = nfs4_handle_exception(server, err,
4077                                 &exception);
4078         } while (exception.retry);
4079         return err;
4080 }
4081
4082 /*
4083  * The file is not closed if it is opened due to the a request to change
4084  * the size of the file. The open call will not be needed once the
4085  * VFS layer lookup-intents are implemented.
4086  *
4087  * Close is called when the inode is destroyed.
4088  * If we haven't opened the file for O_WRONLY, we
4089  * need to in the size_change case to obtain a stateid.
4090  *
4091  * Got race?
4092  * Because OPEN is always done by name in nfsv4, it is
4093  * possible that we opened a different file by the same
4094  * name.  We can recognize this race condition, but we
4095  * can't do anything about it besides returning an error.
4096  *
4097  * This will be fixed with VFS changes (lookup-intent).
4098  */
4099 static int
4100 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4101                   struct iattr *sattr)
4102 {
4103         struct inode *inode = d_inode(dentry);
4104         const struct cred *cred = NULL;
4105         struct nfs_open_context *ctx = NULL;
4106         struct nfs4_label *label = NULL;
4107         int status;
4108
4109         if (pnfs_ld_layoutret_on_setattr(inode) &&
4110             sattr->ia_valid & ATTR_SIZE &&
4111             sattr->ia_size < i_size_read(inode))
4112                 pnfs_commit_and_return_layout(inode);
4113
4114         nfs_fattr_init(fattr);
4115
4116         /* Deal with open(O_TRUNC) */
4117         if (sattr->ia_valid & ATTR_OPEN)
4118                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4119
4120         /* Optimization: if the end result is no change, don't RPC */
4121         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4122                 return 0;
4123
4124         /* Search for an existing open(O_WRITE) file */
4125         if (sattr->ia_valid & ATTR_FILE) {
4126
4127                 ctx = nfs_file_open_context(sattr->ia_file);
4128                 if (ctx)
4129                         cred = ctx->cred;
4130         }
4131
4132         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4133         if (IS_ERR(label))
4134                 return PTR_ERR(label);
4135
4136         /* Return any delegations if we're going to change ACLs */
4137         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4138                 nfs4_inode_make_writeable(inode);
4139
4140         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4141         if (status == 0) {
4142                 nfs_setattr_update_inode(inode, sattr, fattr);
4143                 nfs_setsecurity(inode, fattr, label);
4144         }
4145         nfs4_label_free(label);
4146         return status;
4147 }
4148
4149 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4150                 const struct qstr *name, struct nfs_fh *fhandle,
4151                 struct nfs_fattr *fattr, struct nfs4_label *label)
4152 {
4153         struct nfs_server *server = NFS_SERVER(dir);
4154         int                    status;
4155         struct nfs4_lookup_arg args = {
4156                 .bitmask = server->attr_bitmask,
4157                 .dir_fh = NFS_FH(dir),
4158                 .name = name,
4159         };
4160         struct nfs4_lookup_res res = {
4161                 .server = server,
4162                 .fattr = fattr,
4163                 .label = label,
4164                 .fh = fhandle,
4165         };
4166         struct rpc_message msg = {
4167                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4168                 .rpc_argp = &args,
4169                 .rpc_resp = &res,
4170         };
4171
4172         args.bitmask = nfs4_bitmask(server, label);
4173
4174         nfs_fattr_init(fattr);
4175
4176         dprintk("NFS call  lookup %s\n", name->name);
4177         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4178         dprintk("NFS reply lookup: %d\n", status);
4179         return status;
4180 }
4181
4182 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4183 {
4184         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4185                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4186         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4187         fattr->nlink = 2;
4188 }
4189
4190 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4191                                    const struct qstr *name, struct nfs_fh *fhandle,
4192                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4193 {
4194         struct nfs4_exception exception = {
4195                 .interruptible = true,
4196         };
4197         struct rpc_clnt *client = *clnt;
4198         int err;
4199         do {
4200                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4201                 trace_nfs4_lookup(dir, name, err);
4202                 switch (err) {
4203                 case -NFS4ERR_BADNAME:
4204                         err = -ENOENT;
4205                         goto out;
4206                 case -NFS4ERR_MOVED:
4207                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4208                         if (err == -NFS4ERR_MOVED)
4209                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4210                         goto out;
4211                 case -NFS4ERR_WRONGSEC:
4212                         err = -EPERM;
4213                         if (client != *clnt)
4214                                 goto out;
4215                         client = nfs4_negotiate_security(client, dir, name);
4216                         if (IS_ERR(client))
4217                                 return PTR_ERR(client);
4218
4219                         exception.retry = 1;
4220                         break;
4221                 default:
4222                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4223                 }
4224         } while (exception.retry);
4225
4226 out:
4227         if (err == 0)
4228                 *clnt = client;
4229         else if (client != *clnt)
4230                 rpc_shutdown_client(client);
4231
4232         return err;
4233 }
4234
4235 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4236                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4237                             struct nfs4_label *label)
4238 {
4239         int status;
4240         struct rpc_clnt *client = NFS_CLIENT(dir);
4241
4242         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4243         if (client != NFS_CLIENT(dir)) {
4244                 rpc_shutdown_client(client);
4245                 nfs_fixup_secinfo_attributes(fattr);
4246         }
4247         return status;
4248 }
4249
4250 struct rpc_clnt *
4251 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4252                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4253 {
4254         struct rpc_clnt *client = NFS_CLIENT(dir);
4255         int status;
4256
4257         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4258         if (status < 0)
4259                 return ERR_PTR(status);
4260         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4261 }
4262
4263 static int _nfs4_proc_lookupp(struct inode *inode,
4264                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4265                 struct nfs4_label *label)
4266 {
4267         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4268         struct nfs_server *server = NFS_SERVER(inode);
4269         int                    status;
4270         struct nfs4_lookupp_arg args = {
4271                 .bitmask = server->attr_bitmask,
4272                 .fh = NFS_FH(inode),
4273         };
4274         struct nfs4_lookupp_res res = {
4275                 .server = server,
4276                 .fattr = fattr,
4277                 .label = label,
4278                 .fh = fhandle,
4279         };
4280         struct rpc_message msg = {
4281                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4282                 .rpc_argp = &args,
4283                 .rpc_resp = &res,
4284         };
4285
4286         args.bitmask = nfs4_bitmask(server, label);
4287
4288         nfs_fattr_init(fattr);
4289
4290         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4291         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4292                                 &res.seq_res, 0);
4293         dprintk("NFS reply lookupp: %d\n", status);
4294         return status;
4295 }
4296
4297 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4298                              struct nfs_fattr *fattr, struct nfs4_label *label)
4299 {
4300         struct nfs4_exception exception = {
4301                 .interruptible = true,
4302         };
4303         int err;
4304         do {
4305                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4306                 trace_nfs4_lookupp(inode, err);
4307                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4308                                 &exception);
4309         } while (exception.retry);
4310         return err;
4311 }
4312
4313 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4314 {
4315         struct nfs_server *server = NFS_SERVER(inode);
4316         struct nfs4_accessargs args = {
4317                 .fh = NFS_FH(inode),
4318                 .access = entry->mask,
4319         };
4320         struct nfs4_accessres res = {
4321                 .server = server,
4322         };
4323         struct rpc_message msg = {
4324                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4325                 .rpc_argp = &args,
4326                 .rpc_resp = &res,
4327                 .rpc_cred = entry->cred,
4328         };
4329         int status = 0;
4330
4331         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4332                 res.fattr = nfs_alloc_fattr();
4333                 if (res.fattr == NULL)
4334                         return -ENOMEM;
4335                 args.bitmask = server->cache_consistency_bitmask;
4336         }
4337         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4338         if (!status) {
4339                 nfs_access_set_mask(entry, res.access);
4340                 if (res.fattr)
4341                         nfs_refresh_inode(inode, res.fattr);
4342         }
4343         nfs_free_fattr(res.fattr);
4344         return status;
4345 }
4346
4347 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4348 {
4349         struct nfs4_exception exception = {
4350                 .interruptible = true,
4351         };
4352         int err;
4353         do {
4354                 err = _nfs4_proc_access(inode, entry);
4355                 trace_nfs4_access(inode, err);
4356                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4357                                 &exception);
4358         } while (exception.retry);
4359         return err;
4360 }
4361
4362 /*
4363  * TODO: For the time being, we don't try to get any attributes
4364  * along with any of the zero-copy operations READ, READDIR,
4365  * READLINK, WRITE.
4366  *
4367  * In the case of the first three, we want to put the GETATTR
4368  * after the read-type operation -- this is because it is hard
4369  * to predict the length of a GETATTR response in v4, and thus
4370  * align the READ data correctly.  This means that the GETATTR
4371  * may end up partially falling into the page cache, and we should
4372  * shift it into the 'tail' of the xdr_buf before processing.
4373  * To do this efficiently, we need to know the total length
4374  * of data received, which doesn't seem to be available outside
4375  * of the RPC layer.
4376  *
4377  * In the case of WRITE, we also want to put the GETATTR after
4378  * the operation -- in this case because we want to make sure
4379  * we get the post-operation mtime and size.
4380  *
4381  * Both of these changes to the XDR layer would in fact be quite
4382  * minor, but I decided to leave them for a subsequent patch.
4383  */
4384 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4385                 unsigned int pgbase, unsigned int pglen)
4386 {
4387         struct nfs4_readlink args = {
4388                 .fh       = NFS_FH(inode),
4389                 .pgbase   = pgbase,
4390                 .pglen    = pglen,
4391                 .pages    = &page,
4392         };
4393         struct nfs4_readlink_res res;
4394         struct rpc_message msg = {
4395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4396                 .rpc_argp = &args,
4397                 .rpc_resp = &res,
4398         };
4399
4400         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4401 }
4402
4403 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4404                 unsigned int pgbase, unsigned int pglen)
4405 {
4406         struct nfs4_exception exception = {
4407                 .interruptible = true,
4408         };
4409         int err;
4410         do {
4411                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4412                 trace_nfs4_readlink(inode, err);
4413                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4414                                 &exception);
4415         } while (exception.retry);
4416         return err;
4417 }
4418
4419 /*
4420  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4421  */
4422 static int
4423 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4424                  int flags)
4425 {
4426         struct nfs_server *server = NFS_SERVER(dir);
4427         struct nfs4_label l, *ilabel = NULL;
4428         struct nfs_open_context *ctx;
4429         struct nfs4_state *state;
4430         int status = 0;
4431
4432         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4433         if (IS_ERR(ctx))
4434                 return PTR_ERR(ctx);
4435
4436         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4437
4438         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4439                 sattr->ia_mode &= ~current_umask();
4440         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4441         if (IS_ERR(state)) {
4442                 status = PTR_ERR(state);
4443                 goto out;
4444         }
4445 out:
4446         nfs4_label_release_security(ilabel);
4447         put_nfs_open_context(ctx);
4448         return status;
4449 }
4450
4451 static int
4452 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4453 {
4454         struct nfs_server *server = NFS_SERVER(dir);
4455         struct nfs_removeargs args = {
4456                 .fh = NFS_FH(dir),
4457                 .name = *name,
4458         };
4459         struct nfs_removeres res = {
4460                 .server = server,
4461         };
4462         struct rpc_message msg = {
4463                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4464                 .rpc_argp = &args,
4465                 .rpc_resp = &res,
4466         };
4467         unsigned long timestamp = jiffies;
4468         int status;
4469
4470         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4471         if (status == 0) {
4472                 spin_lock(&dir->i_lock);
4473                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4474                 /* Removing a directory decrements nlink in the parent */
4475                 if (ftype == NF4DIR && dir->i_nlink > 2)
4476                         nfs4_dec_nlink_locked(dir);
4477                 spin_unlock(&dir->i_lock);
4478         }
4479         return status;
4480 }
4481
4482 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4483 {
4484         struct nfs4_exception exception = {
4485                 .interruptible = true,
4486         };
4487         struct inode *inode = d_inode(dentry);
4488         int err;
4489
4490         if (inode) {
4491                 if (inode->i_nlink == 1)
4492                         nfs4_inode_return_delegation(inode);
4493                 else
4494                         nfs4_inode_make_writeable(inode);
4495         }
4496         do {
4497                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4498                 trace_nfs4_remove(dir, &dentry->d_name, err);
4499                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4500                                 &exception);
4501         } while (exception.retry);
4502         return err;
4503 }
4504
4505 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4506 {
4507         struct nfs4_exception exception = {
4508                 .interruptible = true,
4509         };
4510         int err;
4511
4512         do {
4513                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4514                 trace_nfs4_remove(dir, name, err);
4515                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4516                                 &exception);
4517         } while (exception.retry);
4518         return err;
4519 }
4520
4521 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4522                 struct dentry *dentry,
4523                 struct inode *inode)
4524 {
4525         struct nfs_removeargs *args = msg->rpc_argp;
4526         struct nfs_removeres *res = msg->rpc_resp;
4527
4528         res->server = NFS_SB(dentry->d_sb);
4529         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4530         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4531
4532         nfs_fattr_init(res->dir_attr);
4533
4534         if (inode)
4535                 nfs4_inode_return_delegation(inode);
4536 }
4537
4538 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4539 {
4540         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4541                         &data->args.seq_args,
4542                         &data->res.seq_res,
4543                         task);
4544 }
4545
4546 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4547 {
4548         struct nfs_unlinkdata *data = task->tk_calldata;
4549         struct nfs_removeres *res = &data->res;
4550
4551         if (!nfs4_sequence_done(task, &res->seq_res))
4552                 return 0;
4553         if (nfs4_async_handle_error(task, res->server, NULL,
4554                                     &data->timeout) == -EAGAIN)
4555                 return 0;
4556         if (task->tk_status == 0)
4557                 update_changeattr(dir, &res->cinfo,
4558                                 res->dir_attr->time_start, 0);
4559         return 1;
4560 }
4561
4562 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4563                 struct dentry *old_dentry,
4564                 struct dentry *new_dentry)
4565 {
4566         struct nfs_renameargs *arg = msg->rpc_argp;
4567         struct nfs_renameres *res = msg->rpc_resp;
4568         struct inode *old_inode = d_inode(old_dentry);
4569         struct inode *new_inode = d_inode(new_dentry);
4570
4571         if (old_inode)
4572                 nfs4_inode_make_writeable(old_inode);
4573         if (new_inode)
4574                 nfs4_inode_return_delegation(new_inode);
4575         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4576         res->server = NFS_SB(old_dentry->d_sb);
4577         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4578 }
4579
4580 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4581 {
4582         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4583                         &data->args.seq_args,
4584                         &data->res.seq_res,
4585                         task);
4586 }
4587
4588 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4589                                  struct inode *new_dir)
4590 {
4591         struct nfs_renamedata *data = task->tk_calldata;
4592         struct nfs_renameres *res = &data->res;
4593
4594         if (!nfs4_sequence_done(task, &res->seq_res))
4595                 return 0;
4596         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4597                 return 0;
4598
4599         if (task->tk_status == 0) {
4600                 if (new_dir != old_dir) {
4601                         /* Note: If we moved a directory, nlink will change */
4602                         update_changeattr(old_dir, &res->old_cinfo,
4603                                         res->old_fattr->time_start,
4604                                         NFS_INO_INVALID_OTHER);
4605                         update_changeattr(new_dir, &res->new_cinfo,
4606                                         res->new_fattr->time_start,
4607                                         NFS_INO_INVALID_OTHER);
4608                 } else
4609                         update_changeattr(old_dir, &res->old_cinfo,
4610                                         res->old_fattr->time_start,
4611                                         0);
4612         }
4613         return 1;
4614 }
4615
4616 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4617 {
4618         struct nfs_server *server = NFS_SERVER(inode);
4619         __u32 bitmask[NFS4_BITMASK_SZ];
4620         struct nfs4_link_arg arg = {
4621                 .fh     = NFS_FH(inode),
4622                 .dir_fh = NFS_FH(dir),
4623                 .name   = name,
4624                 .bitmask = bitmask,
4625         };
4626         struct nfs4_link_res res = {
4627                 .server = server,
4628                 .label = NULL,
4629         };
4630         struct rpc_message msg = {
4631                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4632                 .rpc_argp = &arg,
4633                 .rpc_resp = &res,
4634         };
4635         int status = -ENOMEM;
4636
4637         res.fattr = nfs_alloc_fattr();
4638         if (res.fattr == NULL)
4639                 goto out;
4640
4641         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4642         if (IS_ERR(res.label)) {
4643                 status = PTR_ERR(res.label);
4644                 goto out;
4645         }
4646
4647         nfs4_inode_make_writeable(inode);
4648         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4649
4650         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4651         if (!status) {
4652                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4653                 status = nfs_post_op_update_inode(inode, res.fattr);
4654                 if (!status)
4655                         nfs_setsecurity(inode, res.fattr, res.label);
4656         }
4657
4658
4659         nfs4_label_free(res.label);
4660
4661 out:
4662         nfs_free_fattr(res.fattr);
4663         return status;
4664 }
4665
4666 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4667 {
4668         struct nfs4_exception exception = {
4669                 .interruptible = true,
4670         };
4671         int err;
4672         do {
4673                 err = nfs4_handle_exception(NFS_SERVER(inode),
4674                                 _nfs4_proc_link(inode, dir, name),
4675                                 &exception);
4676         } while (exception.retry);
4677         return err;
4678 }
4679
4680 struct nfs4_createdata {
4681         struct rpc_message msg;
4682         struct nfs4_create_arg arg;
4683         struct nfs4_create_res res;
4684         struct nfs_fh fh;
4685         struct nfs_fattr fattr;
4686         struct nfs4_label *label;
4687 };
4688
4689 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4690                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4691 {
4692         struct nfs4_createdata *data;
4693
4694         data = kzalloc(sizeof(*data), GFP_KERNEL);
4695         if (data != NULL) {
4696                 struct nfs_server *server = NFS_SERVER(dir);
4697
4698                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4699                 if (IS_ERR(data->label))
4700                         goto out_free;
4701
4702                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4703                 data->msg.rpc_argp = &data->arg;
4704                 data->msg.rpc_resp = &data->res;
4705                 data->arg.dir_fh = NFS_FH(dir);
4706                 data->arg.server = server;
4707                 data->arg.name = name;
4708                 data->arg.attrs = sattr;
4709                 data->arg.ftype = ftype;
4710                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4711                 data->arg.umask = current_umask();
4712                 data->res.server = server;
4713                 data->res.fh = &data->fh;
4714                 data->res.fattr = &data->fattr;
4715                 data->res.label = data->label;
4716                 nfs_fattr_init(data->res.fattr);
4717         }
4718         return data;
4719 out_free:
4720         kfree(data);
4721         return NULL;
4722 }
4723
4724 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4725 {
4726         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4727                                     &data->arg.seq_args, &data->res.seq_res, 1);
4728         if (status == 0) {
4729                 spin_lock(&dir->i_lock);
4730                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4731                                 data->res.fattr->time_start, 0);
4732                 /* Creating a directory bumps nlink in the parent */
4733                 if (data->arg.ftype == NF4DIR)
4734                         nfs4_inc_nlink_locked(dir);
4735                 spin_unlock(&dir->i_lock);
4736                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4737         }
4738         return status;
4739 }
4740
4741 static void nfs4_free_createdata(struct nfs4_createdata *data)
4742 {
4743         nfs4_label_free(data->label);
4744         kfree(data);
4745 }
4746
4747 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4748                 struct page *page, unsigned int len, struct iattr *sattr,
4749                 struct nfs4_label *label)
4750 {
4751         struct nfs4_createdata *data;
4752         int status = -ENAMETOOLONG;
4753
4754         if (len > NFS4_MAXPATHLEN)
4755                 goto out;
4756
4757         status = -ENOMEM;
4758         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4759         if (data == NULL)
4760                 goto out;
4761
4762         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4763         data->arg.u.symlink.pages = &page;
4764         data->arg.u.symlink.len = len;
4765         data->arg.label = label;
4766
4767         status = nfs4_do_create(dir, dentry, data);
4768
4769         nfs4_free_createdata(data);
4770 out:
4771         return status;
4772 }
4773
4774 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4775                 struct page *page, unsigned int len, struct iattr *sattr)
4776 {
4777         struct nfs4_exception exception = {
4778                 .interruptible = true,
4779         };
4780         struct nfs4_label l, *label = NULL;
4781         int err;
4782
4783         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4784
4785         do {
4786                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4787                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4788                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4789                                 &exception);
4790         } while (exception.retry);
4791
4792         nfs4_label_release_security(label);
4793         return err;
4794 }
4795
4796 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4797                 struct iattr *sattr, struct nfs4_label *label)
4798 {
4799         struct nfs4_createdata *data;
4800         int status = -ENOMEM;
4801
4802         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4803         if (data == NULL)
4804                 goto out;
4805
4806         data->arg.label = label;
4807         status = nfs4_do_create(dir, dentry, data);
4808
4809         nfs4_free_createdata(data);
4810 out:
4811         return status;
4812 }
4813
4814 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4815                 struct iattr *sattr)
4816 {
4817         struct nfs_server *server = NFS_SERVER(dir);
4818         struct nfs4_exception exception = {
4819                 .interruptible = true,
4820         };
4821         struct nfs4_label l, *label = NULL;
4822         int err;
4823
4824         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4825
4826         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4827                 sattr->ia_mode &= ~current_umask();
4828         do {
4829                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4830                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4831                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4832                                 &exception);
4833         } while (exception.retry);
4834         nfs4_label_release_security(label);
4835
4836         return err;
4837 }
4838
4839 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4840                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4841 {
4842         struct inode            *dir = d_inode(dentry);
4843         struct nfs4_readdir_arg args = {
4844                 .fh = NFS_FH(dir),
4845                 .pages = pages,
4846                 .pgbase = 0,
4847                 .count = count,
4848                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4849                 .plus = plus,
4850         };
4851         struct nfs4_readdir_res res;
4852         struct rpc_message msg = {
4853                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4854                 .rpc_argp = &args,
4855                 .rpc_resp = &res,
4856                 .rpc_cred = cred,
4857         };
4858         int                     status;
4859
4860         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4861                         dentry,
4862                         (unsigned long long)cookie);
4863         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4864         res.pgbase = args.pgbase;
4865         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4866         if (status >= 0) {
4867                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4868                 status += args.pgbase;
4869         }
4870
4871         nfs_invalidate_atime(dir);
4872
4873         dprintk("%s: returns %d\n", __func__, status);
4874         return status;
4875 }
4876
4877 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4878                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4879 {
4880         struct nfs4_exception exception = {
4881                 .interruptible = true,
4882         };
4883         int err;
4884         do {
4885                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4886                                 pages, count, plus);
4887                 trace_nfs4_readdir(d_inode(dentry), err);
4888                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4889                                 &exception);
4890         } while (exception.retry);
4891         return err;
4892 }
4893
4894 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4895                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4896 {
4897         struct nfs4_createdata *data;
4898         int mode = sattr->ia_mode;
4899         int status = -ENOMEM;
4900
4901         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4902         if (data == NULL)
4903                 goto out;
4904
4905         if (S_ISFIFO(mode))
4906                 data->arg.ftype = NF4FIFO;
4907         else if (S_ISBLK(mode)) {
4908                 data->arg.ftype = NF4BLK;
4909                 data->arg.u.device.specdata1 = MAJOR(rdev);
4910                 data->arg.u.device.specdata2 = MINOR(rdev);
4911         }
4912         else if (S_ISCHR(mode)) {
4913                 data->arg.ftype = NF4CHR;
4914                 data->arg.u.device.specdata1 = MAJOR(rdev);
4915                 data->arg.u.device.specdata2 = MINOR(rdev);
4916         } else if (!S_ISSOCK(mode)) {
4917                 status = -EINVAL;
4918                 goto out_free;
4919         }
4920
4921         data->arg.label = label;
4922         status = nfs4_do_create(dir, dentry, data);
4923 out_free:
4924         nfs4_free_createdata(data);
4925 out:
4926         return status;
4927 }
4928
4929 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4930                 struct iattr *sattr, dev_t rdev)
4931 {
4932         struct nfs_server *server = NFS_SERVER(dir);
4933         struct nfs4_exception exception = {
4934                 .interruptible = true,
4935         };
4936         struct nfs4_label l, *label = NULL;
4937         int err;
4938
4939         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4940
4941         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4942                 sattr->ia_mode &= ~current_umask();
4943         do {
4944                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4945                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4946                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4947                                 &exception);
4948         } while (exception.retry);
4949
4950         nfs4_label_release_security(label);
4951
4952         return err;
4953 }
4954
4955 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4956                  struct nfs_fsstat *fsstat)
4957 {
4958         struct nfs4_statfs_arg args = {
4959                 .fh = fhandle,
4960                 .bitmask = server->attr_bitmask,
4961         };
4962         struct nfs4_statfs_res res = {
4963                 .fsstat = fsstat,
4964         };
4965         struct rpc_message msg = {
4966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4967                 .rpc_argp = &args,
4968                 .rpc_resp = &res,
4969         };
4970
4971         nfs_fattr_init(fsstat->fattr);
4972         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4973 }
4974
4975 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4976 {
4977         struct nfs4_exception exception = {
4978                 .interruptible = true,
4979         };
4980         int err;
4981         do {
4982                 err = nfs4_handle_exception(server,
4983                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4984                                 &exception);
4985         } while (exception.retry);
4986         return err;
4987 }
4988
4989 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4990                 struct nfs_fsinfo *fsinfo)
4991 {
4992         struct nfs4_fsinfo_arg args = {
4993                 .fh = fhandle,
4994                 .bitmask = server->attr_bitmask,
4995         };
4996         struct nfs4_fsinfo_res res = {
4997                 .fsinfo = fsinfo,
4998         };
4999         struct rpc_message msg = {
5000                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5001                 .rpc_argp = &args,
5002                 .rpc_resp = &res,
5003         };
5004
5005         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5006 }
5007
5008 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5009 {
5010         struct nfs4_exception exception = {
5011                 .interruptible = true,
5012         };
5013         unsigned long now = jiffies;
5014         int err;
5015
5016         do {
5017                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5018                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5019                 if (err == 0) {
5020                         nfs4_set_lease_period(server->nfs_client,
5021                                         fsinfo->lease_time * HZ,
5022                                         now);
5023                         break;
5024                 }
5025                 err = nfs4_handle_exception(server, err, &exception);
5026         } while (exception.retry);
5027         return err;
5028 }
5029
5030 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5031 {
5032         int error;
5033
5034         nfs_fattr_init(fsinfo->fattr);
5035         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5036         if (error == 0) {
5037                 /* block layout checks this! */
5038                 server->pnfs_blksize = fsinfo->blksize;
5039                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5040         }
5041
5042         return error;
5043 }
5044
5045 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5046                 struct nfs_pathconf *pathconf)
5047 {
5048         struct nfs4_pathconf_arg args = {
5049                 .fh = fhandle,
5050                 .bitmask = server->attr_bitmask,
5051         };
5052         struct nfs4_pathconf_res res = {
5053                 .pathconf = pathconf,
5054         };
5055         struct rpc_message msg = {
5056                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5057                 .rpc_argp = &args,
5058                 .rpc_resp = &res,
5059         };
5060
5061         /* None of the pathconf attributes are mandatory to implement */
5062         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5063                 memset(pathconf, 0, sizeof(*pathconf));
5064                 return 0;
5065         }
5066
5067         nfs_fattr_init(pathconf->fattr);
5068         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5069 }
5070
5071 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5072                 struct nfs_pathconf *pathconf)
5073 {
5074         struct nfs4_exception exception = {
5075                 .interruptible = true,
5076         };
5077         int err;
5078
5079         do {
5080                 err = nfs4_handle_exception(server,
5081                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5082                                 &exception);
5083         } while (exception.retry);
5084         return err;
5085 }
5086
5087 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5088                 const struct nfs_open_context *ctx,
5089                 const struct nfs_lock_context *l_ctx,
5090                 fmode_t fmode)
5091 {
5092         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5093 }
5094 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5095
5096 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5097                 const struct nfs_open_context *ctx,
5098                 const struct nfs_lock_context *l_ctx,
5099                 fmode_t fmode)
5100 {
5101         nfs4_stateid current_stateid;
5102
5103         /* If the current stateid represents a lost lock, then exit */
5104         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
5105                 return true;
5106         return nfs4_stateid_match(stateid, &current_stateid);
5107 }
5108
5109 static bool nfs4_error_stateid_expired(int err)
5110 {
5111         switch (err) {
5112         case -NFS4ERR_DELEG_REVOKED:
5113         case -NFS4ERR_ADMIN_REVOKED:
5114         case -NFS4ERR_BAD_STATEID:
5115         case -NFS4ERR_STALE_STATEID:
5116         case -NFS4ERR_OLD_STATEID:
5117         case -NFS4ERR_OPENMODE:
5118         case -NFS4ERR_EXPIRED:
5119                 return true;
5120         }
5121         return false;
5122 }
5123
5124 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5125 {
5126         struct nfs_server *server = NFS_SERVER(hdr->inode);
5127
5128         trace_nfs4_read(hdr, task->tk_status);
5129         if (task->tk_status < 0) {
5130                 struct nfs4_exception exception = {
5131                         .inode = hdr->inode,
5132                         .state = hdr->args.context->state,
5133                         .stateid = &hdr->args.stateid,
5134                 };
5135                 task->tk_status = nfs4_async_handle_exception(task,
5136                                 server, task->tk_status, &exception);
5137                 if (exception.retry) {
5138                         rpc_restart_call_prepare(task);
5139                         return -EAGAIN;
5140                 }
5141         }
5142
5143         if (task->tk_status > 0)
5144                 renew_lease(server, hdr->timestamp);
5145         return 0;
5146 }
5147
5148 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5149                 struct nfs_pgio_args *args)
5150 {
5151
5152         if (!nfs4_error_stateid_expired(task->tk_status) ||
5153                 nfs4_stateid_is_current(&args->stateid,
5154                                 args->context,
5155                                 args->lock_context,
5156                                 FMODE_READ))
5157                 return false;
5158         rpc_restart_call_prepare(task);
5159         return true;
5160 }
5161
5162 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5163 {
5164
5165         dprintk("--> %s\n", __func__);
5166
5167         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5168                 return -EAGAIN;
5169         if (nfs4_read_stateid_changed(task, &hdr->args))
5170                 return -EAGAIN;
5171         if (task->tk_status > 0)
5172                 nfs_invalidate_atime(hdr->inode);
5173         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5174                                     nfs4_read_done_cb(task, hdr);
5175 }
5176
5177 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5178                                  struct rpc_message *msg)
5179 {
5180         hdr->timestamp   = jiffies;
5181         if (!hdr->pgio_done_cb)
5182                 hdr->pgio_done_cb = nfs4_read_done_cb;
5183         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5184         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5185 }
5186
5187 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5188                                       struct nfs_pgio_header *hdr)
5189 {
5190         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5191                         &hdr->args.seq_args,
5192                         &hdr->res.seq_res,
5193                         task))
5194                 return 0;
5195         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5196                                 hdr->args.lock_context,
5197                                 hdr->rw_mode) == -EIO)
5198                 return -EIO;
5199         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5200                 return -EIO;
5201         return 0;
5202 }
5203
5204 static int nfs4_write_done_cb(struct rpc_task *task,
5205                               struct nfs_pgio_header *hdr)
5206 {
5207         struct inode *inode = hdr->inode;
5208
5209         trace_nfs4_write(hdr, task->tk_status);
5210         if (task->tk_status < 0) {
5211                 struct nfs4_exception exception = {
5212                         .inode = hdr->inode,
5213                         .state = hdr->args.context->state,
5214                         .stateid = &hdr->args.stateid,
5215                 };
5216                 task->tk_status = nfs4_async_handle_exception(task,
5217                                 NFS_SERVER(inode), task->tk_status,
5218                                 &exception);
5219                 if (exception.retry) {
5220                         rpc_restart_call_prepare(task);
5221                         return -EAGAIN;
5222                 }
5223         }
5224         if (task->tk_status >= 0) {
5225                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5226                 nfs_writeback_update_inode(hdr);
5227         }
5228         return 0;
5229 }
5230
5231 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5232                 struct nfs_pgio_args *args)
5233 {
5234
5235         if (!nfs4_error_stateid_expired(task->tk_status) ||
5236                 nfs4_stateid_is_current(&args->stateid,
5237                                 args->context,
5238                                 args->lock_context,
5239                                 FMODE_WRITE))
5240                 return false;
5241         rpc_restart_call_prepare(task);
5242         return true;
5243 }
5244
5245 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5246 {
5247         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5248                 return -EAGAIN;
5249         if (nfs4_write_stateid_changed(task, &hdr->args))
5250                 return -EAGAIN;
5251         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5252                 nfs4_write_done_cb(task, hdr);
5253 }
5254
5255 static
5256 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5257 {
5258         /* Don't request attributes for pNFS or O_DIRECT writes */
5259         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5260                 return false;
5261         /* Otherwise, request attributes if and only if we don't hold
5262          * a delegation
5263          */
5264         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5265 }
5266
5267 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5268                                   struct rpc_message *msg,
5269                                   struct rpc_clnt **clnt)
5270 {
5271         struct nfs_server *server = NFS_SERVER(hdr->inode);
5272
5273         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5274                 hdr->args.bitmask = NULL;
5275                 hdr->res.fattr = NULL;
5276         } else
5277                 hdr->args.bitmask = server->cache_consistency_bitmask;
5278
5279         if (!hdr->pgio_done_cb)
5280                 hdr->pgio_done_cb = nfs4_write_done_cb;
5281         hdr->res.server = server;
5282         hdr->timestamp   = jiffies;
5283
5284         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5285         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5286         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5287 }
5288
5289 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5290 {
5291         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5292                         &data->args.seq_args,
5293                         &data->res.seq_res,
5294                         task);
5295 }
5296
5297 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5298 {
5299         struct inode *inode = data->inode;
5300
5301         trace_nfs4_commit(data, task->tk_status);
5302         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5303                                     NULL, NULL) == -EAGAIN) {
5304                 rpc_restart_call_prepare(task);
5305                 return -EAGAIN;
5306         }
5307         return 0;
5308 }
5309
5310 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5311 {
5312         if (!nfs4_sequence_done(task, &data->res.seq_res))
5313                 return -EAGAIN;
5314         return data->commit_done_cb(task, data);
5315 }
5316
5317 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5318                                    struct rpc_clnt **clnt)
5319 {
5320         struct nfs_server *server = NFS_SERVER(data->inode);
5321
5322         if (data->commit_done_cb == NULL)
5323                 data->commit_done_cb = nfs4_commit_done_cb;
5324         data->res.server = server;
5325         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5326         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5327         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5328 }
5329
5330 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5331                                 struct nfs_commitres *res)
5332 {
5333         struct inode *dst_inode = file_inode(dst);
5334         struct nfs_server *server = NFS_SERVER(dst_inode);
5335         struct rpc_message msg = {
5336                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5337                 .rpc_argp = args,
5338                 .rpc_resp = res,
5339         };
5340
5341         args->fh = NFS_FH(dst_inode);
5342         return nfs4_call_sync(server->client, server, &msg,
5343                         &args->seq_args, &res->seq_res, 1);
5344 }
5345
5346 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5347 {
5348         struct nfs_commitargs args = {
5349                 .offset = offset,
5350                 .count = count,
5351         };
5352         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5353         struct nfs4_exception exception = { };
5354         int status;
5355
5356         do {
5357                 status = _nfs4_proc_commit(dst, &args, res);
5358                 status = nfs4_handle_exception(dst_server, status, &exception);
5359         } while (exception.retry);
5360
5361         return status;
5362 }
5363
5364 struct nfs4_renewdata {
5365         struct nfs_client       *client;
5366         unsigned long           timestamp;
5367 };
5368
5369 /*
5370  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5371  * standalone procedure for queueing an asynchronous RENEW.
5372  */
5373 static void nfs4_renew_release(void *calldata)
5374 {
5375         struct nfs4_renewdata *data = calldata;
5376         struct nfs_client *clp = data->client;
5377
5378         if (refcount_read(&clp->cl_count) > 1)
5379                 nfs4_schedule_state_renewal(clp);
5380         nfs_put_client(clp);
5381         kfree(data);
5382 }
5383
5384 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5385 {
5386         struct nfs4_renewdata *data = calldata;
5387         struct nfs_client *clp = data->client;
5388         unsigned long timestamp = data->timestamp;
5389
5390         trace_nfs4_renew_async(clp, task->tk_status);
5391         switch (task->tk_status) {
5392         case 0:
5393                 break;
5394         case -NFS4ERR_LEASE_MOVED:
5395                 nfs4_schedule_lease_moved_recovery(clp);
5396                 break;
5397         default:
5398                 /* Unless we're shutting down, schedule state recovery! */
5399                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5400                         return;
5401                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5402                         nfs4_schedule_lease_recovery(clp);
5403                         return;
5404                 }
5405                 nfs4_schedule_path_down_recovery(clp);
5406         }
5407         do_renew_lease(clp, timestamp);
5408 }
5409
5410 static const struct rpc_call_ops nfs4_renew_ops = {
5411         .rpc_call_done = nfs4_renew_done,
5412         .rpc_release = nfs4_renew_release,
5413 };
5414
5415 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5416 {
5417         struct rpc_message msg = {
5418                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5419                 .rpc_argp       = clp,
5420                 .rpc_cred       = cred,
5421         };
5422         struct nfs4_renewdata *data;
5423
5424         if (renew_flags == 0)
5425                 return 0;
5426         if (!refcount_inc_not_zero(&clp->cl_count))
5427                 return -EIO;
5428         data = kmalloc(sizeof(*data), GFP_NOFS);
5429         if (data == NULL) {
5430                 nfs_put_client(clp);
5431                 return -ENOMEM;
5432         }
5433         data->client = clp;
5434         data->timestamp = jiffies;
5435         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5436                         &nfs4_renew_ops, data);
5437 }
5438
5439 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5440 {
5441         struct rpc_message msg = {
5442                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5443                 .rpc_argp       = clp,
5444                 .rpc_cred       = cred,
5445         };
5446         unsigned long now = jiffies;
5447         int status;
5448
5449         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5450         if (status < 0)
5451                 return status;
5452         do_renew_lease(clp, now);
5453         return 0;
5454 }
5455
5456 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5457 {
5458         return server->caps & NFS_CAP_ACLS;
5459 }
5460
5461 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5462  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5463  * the stack.
5464  */
5465 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5466
5467 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5468                 struct page **pages)
5469 {
5470         struct page *newpage, **spages;
5471         int rc = 0;
5472         size_t len;
5473         spages = pages;
5474
5475         do {
5476                 len = min_t(size_t, PAGE_SIZE, buflen);
5477                 newpage = alloc_page(GFP_KERNEL);
5478
5479                 if (newpage == NULL)
5480                         goto unwind;
5481                 memcpy(page_address(newpage), buf, len);
5482                 buf += len;
5483                 buflen -= len;
5484                 *pages++ = newpage;
5485                 rc++;
5486         } while (buflen != 0);
5487
5488         return rc;
5489
5490 unwind:
5491         for(; rc > 0; rc--)
5492                 __free_page(spages[rc-1]);
5493         return -ENOMEM;
5494 }
5495
5496 struct nfs4_cached_acl {
5497         int cached;
5498         size_t len;
5499         char data[0];
5500 };
5501
5502 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5503 {
5504         struct nfs_inode *nfsi = NFS_I(inode);
5505
5506         spin_lock(&inode->i_lock);
5507         kfree(nfsi->nfs4_acl);
5508         nfsi->nfs4_acl = acl;
5509         spin_unlock(&inode->i_lock);
5510 }
5511
5512 static void nfs4_zap_acl_attr(struct inode *inode)
5513 {
5514         nfs4_set_cached_acl(inode, NULL);
5515 }
5516
5517 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5518 {
5519         struct nfs_inode *nfsi = NFS_I(inode);
5520         struct nfs4_cached_acl *acl;
5521         int ret = -ENOENT;
5522
5523         spin_lock(&inode->i_lock);
5524         acl = nfsi->nfs4_acl;
5525         if (acl == NULL)
5526                 goto out;
5527         if (buf == NULL) /* user is just asking for length */
5528                 goto out_len;
5529         if (acl->cached == 0)
5530                 goto out;
5531         ret = -ERANGE; /* see getxattr(2) man page */
5532         if (acl->len > buflen)
5533                 goto out;
5534         memcpy(buf, acl->data, acl->len);
5535 out_len:
5536         ret = acl->len;
5537 out:
5538         spin_unlock(&inode->i_lock);
5539         return ret;
5540 }
5541
5542 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5543 {
5544         struct nfs4_cached_acl *acl;
5545         size_t buflen = sizeof(*acl) + acl_len;
5546
5547         if (buflen <= PAGE_SIZE) {
5548                 acl = kmalloc(buflen, GFP_KERNEL);
5549                 if (acl == NULL)
5550                         goto out;
5551                 acl->cached = 1;
5552                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5553         } else {
5554                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5555                 if (acl == NULL)
5556                         goto out;
5557                 acl->cached = 0;
5558         }
5559         acl->len = acl_len;
5560 out:
5561         nfs4_set_cached_acl(inode, acl);
5562 }
5563
5564 /*
5565  * The getxattr API returns the required buffer length when called with a
5566  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5567  * the required buf.  On a NULL buf, we send a page of data to the server
5568  * guessing that the ACL request can be serviced by a page. If so, we cache
5569  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5570  * the cache. If not so, we throw away the page, and cache the required
5571  * length. The next getxattr call will then produce another round trip to
5572  * the server, this time with the input buf of the required size.
5573  */
5574 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5575 {
5576         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5577         struct nfs_getaclargs args = {
5578                 .fh = NFS_FH(inode),
5579                 .acl_pages = pages,
5580                 .acl_len = buflen,
5581         };
5582         struct nfs_getaclres res = {
5583                 .acl_len = buflen,
5584         };
5585         struct rpc_message msg = {
5586                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5587                 .rpc_argp = &args,
5588                 .rpc_resp = &res,
5589         };
5590         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5591         int ret = -ENOMEM, i;
5592
5593         if (npages > ARRAY_SIZE(pages))
5594                 return -ERANGE;
5595
5596         for (i = 0; i < npages; i++) {
5597                 pages[i] = alloc_page(GFP_KERNEL);
5598                 if (!pages[i])
5599                         goto out_free;
5600         }
5601
5602         /* for decoding across pages */
5603         res.acl_scratch = alloc_page(GFP_KERNEL);
5604         if (!res.acl_scratch)
5605                 goto out_free;
5606
5607         args.acl_len = npages * PAGE_SIZE;
5608
5609         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5610                 __func__, buf, buflen, npages, args.acl_len);
5611         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5612                              &msg, &args.seq_args, &res.seq_res, 0);
5613         if (ret)
5614                 goto out_free;
5615
5616         /* Handle the case where the passed-in buffer is too short */
5617         if (res.acl_flags & NFS4_ACL_TRUNC) {
5618                 /* Did the user only issue a request for the acl length? */
5619                 if (buf == NULL)
5620                         goto out_ok;
5621                 ret = -ERANGE;
5622                 goto out_free;
5623         }
5624         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5625         if (buf) {
5626                 if (res.acl_len > buflen) {
5627                         ret = -ERANGE;
5628                         goto out_free;
5629                 }
5630                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5631         }
5632 out_ok:
5633         ret = res.acl_len;
5634 out_free:
5635         for (i = 0; i < npages; i++)
5636                 if (pages[i])
5637                         __free_page(pages[i]);
5638         if (res.acl_scratch)
5639                 __free_page(res.acl_scratch);
5640         return ret;
5641 }
5642
5643 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5644 {
5645         struct nfs4_exception exception = {
5646                 .interruptible = true,
5647         };
5648         ssize_t ret;
5649         do {
5650                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5651                 trace_nfs4_get_acl(inode, ret);
5652                 if (ret >= 0)
5653                         break;
5654                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5655         } while (exception.retry);
5656         return ret;
5657 }
5658
5659 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5660 {
5661         struct nfs_server *server = NFS_SERVER(inode);
5662         int ret;
5663
5664         if (!nfs4_server_supports_acls(server))
5665                 return -EOPNOTSUPP;
5666         ret = nfs_revalidate_inode(server, inode);
5667         if (ret < 0)
5668                 return ret;
5669         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5670                 nfs_zap_acl_cache(inode);
5671         ret = nfs4_read_cached_acl(inode, buf, buflen);
5672         if (ret != -ENOENT)
5673                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5674                  * but no cached acl data, just the acl length */
5675                 return ret;
5676         return nfs4_get_acl_uncached(inode, buf, buflen);
5677 }
5678
5679 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5680 {
5681         struct nfs_server *server = NFS_SERVER(inode);
5682         struct page *pages[NFS4ACL_MAXPAGES];
5683         struct nfs_setaclargs arg = {
5684                 .fh             = NFS_FH(inode),
5685                 .acl_pages      = pages,
5686                 .acl_len        = buflen,
5687         };
5688         struct nfs_setaclres res;
5689         struct rpc_message msg = {
5690                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5691                 .rpc_argp       = &arg,
5692                 .rpc_resp       = &res,
5693         };
5694         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5695         int ret, i;
5696
5697         if (!nfs4_server_supports_acls(server))
5698                 return -EOPNOTSUPP;
5699         if (npages > ARRAY_SIZE(pages))
5700                 return -ERANGE;
5701         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5702         if (i < 0)
5703                 return i;
5704         nfs4_inode_make_writeable(inode);
5705         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5706
5707         /*
5708          * Free each page after tx, so the only ref left is
5709          * held by the network stack
5710          */
5711         for (; i > 0; i--)
5712                 put_page(pages[i-1]);
5713
5714         /*
5715          * Acl update can result in inode attribute update.
5716          * so mark the attribute cache invalid.
5717          */
5718         spin_lock(&inode->i_lock);
5719         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5720                 | NFS_INO_INVALID_CTIME
5721                 | NFS_INO_REVAL_FORCED;
5722         spin_unlock(&inode->i_lock);
5723         nfs_access_zap_cache(inode);
5724         nfs_zap_acl_cache(inode);
5725         return ret;
5726 }
5727
5728 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5729 {
5730         struct nfs4_exception exception = { };
5731         int err;
5732         do {
5733                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5734                 trace_nfs4_set_acl(inode, err);
5735                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5736                                 &exception);
5737         } while (exception.retry);
5738         return err;
5739 }
5740
5741 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5742 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5743                                         size_t buflen)
5744 {
5745         struct nfs_server *server = NFS_SERVER(inode);
5746         struct nfs_fattr fattr;
5747         struct nfs4_label label = {0, 0, buflen, buf};
5748
5749         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5750         struct nfs4_getattr_arg arg = {
5751                 .fh             = NFS_FH(inode),
5752                 .bitmask        = bitmask,
5753         };
5754         struct nfs4_getattr_res res = {
5755                 .fattr          = &fattr,
5756                 .label          = &label,
5757                 .server         = server,
5758         };
5759         struct rpc_message msg = {
5760                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5761                 .rpc_argp       = &arg,
5762                 .rpc_resp       = &res,
5763         };
5764         int ret;
5765
5766         nfs_fattr_init(&fattr);
5767
5768         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5769         if (ret)
5770                 return ret;
5771         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5772                 return -ENOENT;
5773         if (buflen < label.len)
5774                 return -ERANGE;
5775         return 0;
5776 }
5777
5778 static int nfs4_get_security_label(struct inode *inode, void *buf,
5779                                         size_t buflen)
5780 {
5781         struct nfs4_exception exception = {
5782                 .interruptible = true,
5783         };
5784         int err;
5785
5786         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5787                 return -EOPNOTSUPP;
5788
5789         do {
5790                 err = _nfs4_get_security_label(inode, buf, buflen);
5791                 trace_nfs4_get_security_label(inode, err);
5792                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5793                                 &exception);
5794         } while (exception.retry);
5795         return err;
5796 }
5797
5798 static int _nfs4_do_set_security_label(struct inode *inode,
5799                 struct nfs4_label *ilabel,
5800                 struct nfs_fattr *fattr,
5801                 struct nfs4_label *olabel)
5802 {
5803
5804         struct iattr sattr = {0};
5805         struct nfs_server *server = NFS_SERVER(inode);
5806         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5807         struct nfs_setattrargs arg = {
5808                 .fh             = NFS_FH(inode),
5809                 .iap            = &sattr,
5810                 .server         = server,
5811                 .bitmask        = bitmask,
5812                 .label          = ilabel,
5813         };
5814         struct nfs_setattrres res = {
5815                 .fattr          = fattr,
5816                 .label          = olabel,
5817                 .server         = server,
5818         };
5819         struct rpc_message msg = {
5820                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5821                 .rpc_argp       = &arg,
5822                 .rpc_resp       = &res,
5823         };
5824         int status;
5825
5826         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5827
5828         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5829         if (status)
5830                 dprintk("%s failed: %d\n", __func__, status);
5831
5832         return status;
5833 }
5834
5835 static int nfs4_do_set_security_label(struct inode *inode,
5836                 struct nfs4_label *ilabel,
5837                 struct nfs_fattr *fattr,
5838                 struct nfs4_label *olabel)
5839 {
5840         struct nfs4_exception exception = { };
5841         int err;
5842
5843         do {
5844                 err = _nfs4_do_set_security_label(inode, ilabel,
5845                                 fattr, olabel);
5846                 trace_nfs4_set_security_label(inode, err);
5847                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5848                                 &exception);
5849         } while (exception.retry);
5850         return err;
5851 }
5852
5853 static int
5854 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5855 {
5856         struct nfs4_label ilabel, *olabel = NULL;
5857         struct nfs_fattr fattr;
5858         int status;
5859
5860         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5861                 return -EOPNOTSUPP;
5862
5863         nfs_fattr_init(&fattr);
5864
5865         ilabel.pi = 0;
5866         ilabel.lfs = 0;
5867         ilabel.label = (char *)buf;
5868         ilabel.len = buflen;
5869
5870         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5871         if (IS_ERR(olabel)) {
5872                 status = -PTR_ERR(olabel);
5873                 goto out;
5874         }
5875
5876         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5877         if (status == 0)
5878                 nfs_setsecurity(inode, &fattr, olabel);
5879
5880         nfs4_label_free(olabel);
5881 out:
5882         return status;
5883 }
5884 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5885
5886
5887 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5888                                     nfs4_verifier *bootverf)
5889 {
5890         __be32 verf[2];
5891
5892         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5893                 /* An impossible timestamp guarantees this value
5894                  * will never match a generated boot time. */
5895                 verf[0] = cpu_to_be32(U32_MAX);
5896                 verf[1] = cpu_to_be32(U32_MAX);
5897         } else {
5898                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5899                 u64 ns = ktime_to_ns(nn->boot_time);
5900
5901                 verf[0] = cpu_to_be32(ns >> 32);
5902                 verf[1] = cpu_to_be32(ns);
5903         }
5904         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5905 }
5906
5907 static int
5908 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5909 {
5910         size_t len;
5911         char *str;
5912
5913         if (clp->cl_owner_id != NULL)
5914                 return 0;
5915
5916         rcu_read_lock();
5917         len = 14 +
5918                 strlen(clp->cl_rpcclient->cl_nodename) +
5919                 1 +
5920                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5921                 1;
5922         rcu_read_unlock();
5923         if (nfs4_client_id_uniquifier[0] != '\0')
5924                 len += strlen(nfs4_client_id_uniquifier) + 1;
5925         if (len > NFS4_OPAQUE_LIMIT + 1)
5926                 return -EINVAL;
5927
5928         /*
5929          * Since this string is allocated at mount time, and held until the
5930          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5931          * about a memory-reclaim deadlock.
5932          */
5933         str = kmalloc(len, GFP_KERNEL);
5934         if (!str)
5935                 return -ENOMEM;
5936
5937         rcu_read_lock();
5938         if (nfs4_client_id_uniquifier[0] != '\0')
5939                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5940                           clp->cl_rpcclient->cl_nodename,
5941                           nfs4_client_id_uniquifier,
5942                           rpc_peeraddr2str(clp->cl_rpcclient,
5943                                            RPC_DISPLAY_ADDR));
5944         else
5945                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5946                           clp->cl_rpcclient->cl_nodename,
5947                           rpc_peeraddr2str(clp->cl_rpcclient,
5948                                            RPC_DISPLAY_ADDR));
5949         rcu_read_unlock();
5950
5951         clp->cl_owner_id = str;
5952         return 0;
5953 }
5954
5955 static int
5956 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5957 {
5958         size_t len;
5959         char *str;
5960
5961         len = 10 + 10 + 1 + 10 + 1 +
5962                 strlen(nfs4_client_id_uniquifier) + 1 +
5963                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5964
5965         if (len > NFS4_OPAQUE_LIMIT + 1)
5966                 return -EINVAL;
5967
5968         /*
5969          * Since this string is allocated at mount time, and held until the
5970          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5971          * about a memory-reclaim deadlock.
5972          */
5973         str = kmalloc(len, GFP_KERNEL);
5974         if (!str)
5975                 return -ENOMEM;
5976
5977         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5978                         clp->rpc_ops->version, clp->cl_minorversion,
5979                         nfs4_client_id_uniquifier,
5980                         clp->cl_rpcclient->cl_nodename);
5981         clp->cl_owner_id = str;
5982         return 0;
5983 }
5984
5985 static int
5986 nfs4_init_uniform_client_string(struct nfs_client *clp)
5987 {
5988         size_t len;
5989         char *str;
5990
5991         if (clp->cl_owner_id != NULL)
5992                 return 0;
5993
5994         if (nfs4_client_id_uniquifier[0] != '\0')
5995                 return nfs4_init_uniquifier_client_string(clp);
5996
5997         len = 10 + 10 + 1 + 10 + 1 +
5998                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5999
6000         if (len > NFS4_OPAQUE_LIMIT + 1)
6001                 return -EINVAL;
6002
6003         /*
6004          * Since this string is allocated at mount time, and held until the
6005          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6006          * about a memory-reclaim deadlock.
6007          */
6008         str = kmalloc(len, GFP_KERNEL);
6009         if (!str)
6010                 return -ENOMEM;
6011
6012         scnprintf(str, len, "Linux NFSv%u.%u %s",
6013                         clp->rpc_ops->version, clp->cl_minorversion,
6014                         clp->cl_rpcclient->cl_nodename);
6015         clp->cl_owner_id = str;
6016         return 0;
6017 }
6018
6019 /*
6020  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6021  * services.  Advertise one based on the address family of the
6022  * clientaddr.
6023  */
6024 static unsigned int
6025 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6026 {
6027         if (strchr(clp->cl_ipaddr, ':') != NULL)
6028                 return scnprintf(buf, len, "tcp6");
6029         else
6030                 return scnprintf(buf, len, "tcp");
6031 }
6032
6033 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6034 {
6035         struct nfs4_setclientid *sc = calldata;
6036
6037         if (task->tk_status == 0)
6038                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6039 }
6040
6041 static const struct rpc_call_ops nfs4_setclientid_ops = {
6042         .rpc_call_done = nfs4_setclientid_done,
6043 };
6044
6045 /**
6046  * nfs4_proc_setclientid - Negotiate client ID
6047  * @clp: state data structure
6048  * @program: RPC program for NFSv4 callback service
6049  * @port: IP port number for NFS4 callback service
6050  * @cred: credential to use for this call
6051  * @res: where to place the result
6052  *
6053  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6054  */
6055 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6056                 unsigned short port, const struct cred *cred,
6057                 struct nfs4_setclientid_res *res)
6058 {
6059         nfs4_verifier sc_verifier;
6060         struct nfs4_setclientid setclientid = {
6061                 .sc_verifier = &sc_verifier,
6062                 .sc_prog = program,
6063                 .sc_clnt = clp,
6064         };
6065         struct rpc_message msg = {
6066                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6067                 .rpc_argp = &setclientid,
6068                 .rpc_resp = res,
6069                 .rpc_cred = cred,
6070         };
6071         struct rpc_task_setup task_setup_data = {
6072                 .rpc_client = clp->cl_rpcclient,
6073                 .rpc_message = &msg,
6074                 .callback_ops = &nfs4_setclientid_ops,
6075                 .callback_data = &setclientid,
6076                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6077         };
6078         int status;
6079
6080         /* nfs_client_id4 */
6081         nfs4_init_boot_verifier(clp, &sc_verifier);
6082
6083         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6084                 status = nfs4_init_uniform_client_string(clp);
6085         else
6086                 status = nfs4_init_nonuniform_client_string(clp);
6087
6088         if (status)
6089                 goto out;
6090
6091         /* cb_client4 */
6092         setclientid.sc_netid_len =
6093                                 nfs4_init_callback_netid(clp,
6094                                                 setclientid.sc_netid,
6095                                                 sizeof(setclientid.sc_netid));
6096         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6097                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6098                                 clp->cl_ipaddr, port >> 8, port & 255);
6099
6100         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6101                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6102                 clp->cl_owner_id);
6103
6104         status = nfs4_call_sync_custom(&task_setup_data);
6105         if (setclientid.sc_cred) {
6106                 kfree(clp->cl_acceptor);
6107                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6108                 put_rpccred(setclientid.sc_cred);
6109         }
6110 out:
6111         trace_nfs4_setclientid(clp, status);
6112         dprintk("NFS reply setclientid: %d\n", status);
6113         return status;
6114 }
6115
6116 /**
6117  * nfs4_proc_setclientid_confirm - Confirm client ID
6118  * @clp: state data structure
6119  * @arg: result of a previous SETCLIENTID
6120  * @cred: credential to use for this call
6121  *
6122  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6123  */
6124 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6125                 struct nfs4_setclientid_res *arg,
6126                 const struct cred *cred)
6127 {
6128         struct rpc_message msg = {
6129                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6130                 .rpc_argp = arg,
6131                 .rpc_cred = cred,
6132         };
6133         int status;
6134
6135         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6136                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6137                 clp->cl_clientid);
6138         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6139                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6140         trace_nfs4_setclientid_confirm(clp, status);
6141         dprintk("NFS reply setclientid_confirm: %d\n", status);
6142         return status;
6143 }
6144
6145 struct nfs4_delegreturndata {
6146         struct nfs4_delegreturnargs args;
6147         struct nfs4_delegreturnres res;
6148         struct nfs_fh fh;
6149         nfs4_stateid stateid;
6150         unsigned long timestamp;
6151         struct {
6152                 struct nfs4_layoutreturn_args arg;
6153                 struct nfs4_layoutreturn_res res;
6154                 struct nfs4_xdr_opaque_data ld_private;
6155                 u32 roc_barrier;
6156                 bool roc;
6157         } lr;
6158         struct nfs_fattr fattr;
6159         int rpc_status;
6160         struct inode *inode;
6161 };
6162
6163 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6164 {
6165         struct nfs4_delegreturndata *data = calldata;
6166         struct nfs4_exception exception = {
6167                 .inode = data->inode,
6168                 .stateid = &data->stateid,
6169         };
6170
6171         if (!nfs4_sequence_done(task, &data->res.seq_res))
6172                 return;
6173
6174         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6175
6176         /* Handle Layoutreturn errors */
6177         if (pnfs_roc_done(task, data->inode,
6178                                 &data->args.lr_args,
6179                                 &data->res.lr_res,
6180                                 &data->res.lr_ret) == -EAGAIN)
6181                 goto out_restart;
6182
6183         switch (task->tk_status) {
6184         case 0:
6185                 renew_lease(data->res.server, data->timestamp);
6186                 break;
6187         case -NFS4ERR_ADMIN_REVOKED:
6188         case -NFS4ERR_DELEG_REVOKED:
6189         case -NFS4ERR_EXPIRED:
6190                 nfs4_free_revoked_stateid(data->res.server,
6191                                 data->args.stateid,
6192                                 task->tk_msg.rpc_cred);
6193                 /* Fallthrough */
6194         case -NFS4ERR_BAD_STATEID:
6195         case -NFS4ERR_STALE_STATEID:
6196                 task->tk_status = 0;
6197                 break;
6198         case -NFS4ERR_OLD_STATEID:
6199                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6200                         goto out_restart;
6201                 task->tk_status = 0;
6202                 break;
6203         case -NFS4ERR_ACCESS:
6204                 if (data->args.bitmask) {
6205                         data->args.bitmask = NULL;
6206                         data->res.fattr = NULL;
6207                         goto out_restart;
6208                 }
6209                 /* Fallthrough */
6210         default:
6211                 task->tk_status = nfs4_async_handle_exception(task,
6212                                 data->res.server, task->tk_status,
6213                                 &exception);
6214                 if (exception.retry)
6215                         goto out_restart;
6216         }
6217         data->rpc_status = task->tk_status;
6218         return;
6219 out_restart:
6220         task->tk_status = 0;
6221         rpc_restart_call_prepare(task);
6222 }
6223
6224 static void nfs4_delegreturn_release(void *calldata)
6225 {
6226         struct nfs4_delegreturndata *data = calldata;
6227         struct inode *inode = data->inode;
6228
6229         if (inode) {
6230                 if (data->lr.roc)
6231                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6232                                         data->res.lr_ret);
6233                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6234                 nfs_iput_and_deactive(inode);
6235         }
6236         kfree(calldata);
6237 }
6238
6239 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6240 {
6241         struct nfs4_delegreturndata *d_data;
6242         struct pnfs_layout_hdr *lo;
6243
6244         d_data = (struct nfs4_delegreturndata *)data;
6245
6246         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6247                 return;
6248
6249         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6250         if (lo && !pnfs_layout_is_valid(lo)) {
6251                 d_data->args.lr_args = NULL;
6252                 d_data->res.lr_res = NULL;
6253         }
6254
6255         nfs4_setup_sequence(d_data->res.server->nfs_client,
6256                         &d_data->args.seq_args,
6257                         &d_data->res.seq_res,
6258                         task);
6259 }
6260
6261 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6262         .rpc_call_prepare = nfs4_delegreturn_prepare,
6263         .rpc_call_done = nfs4_delegreturn_done,
6264         .rpc_release = nfs4_delegreturn_release,
6265 };
6266
6267 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6268 {
6269         struct nfs4_delegreturndata *data;
6270         struct nfs_server *server = NFS_SERVER(inode);
6271         struct rpc_task *task;
6272         struct rpc_message msg = {
6273                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6274                 .rpc_cred = cred,
6275         };
6276         struct rpc_task_setup task_setup_data = {
6277                 .rpc_client = server->client,
6278                 .rpc_message = &msg,
6279                 .callback_ops = &nfs4_delegreturn_ops,
6280                 .flags = RPC_TASK_ASYNC,
6281         };
6282         int status = 0;
6283
6284         data = kzalloc(sizeof(*data), GFP_NOFS);
6285         if (data == NULL)
6286                 return -ENOMEM;
6287         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6288
6289         nfs4_state_protect(server->nfs_client,
6290                         NFS_SP4_MACH_CRED_CLEANUP,
6291                         &task_setup_data.rpc_client, &msg);
6292
6293         data->args.fhandle = &data->fh;
6294         data->args.stateid = &data->stateid;
6295         data->args.bitmask = server->cache_consistency_bitmask;
6296         nfs_copy_fh(&data->fh, NFS_FH(inode));
6297         nfs4_stateid_copy(&data->stateid, stateid);
6298         data->res.fattr = &data->fattr;
6299         data->res.server = server;
6300         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6301         data->lr.arg.ld_private = &data->lr.ld_private;
6302         nfs_fattr_init(data->res.fattr);
6303         data->timestamp = jiffies;
6304         data->rpc_status = 0;
6305         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6306         data->inode = nfs_igrab_and_active(inode);
6307         if (data->inode) {
6308                 if (data->lr.roc) {
6309                         data->args.lr_args = &data->lr.arg;
6310                         data->res.lr_res = &data->lr.res;
6311                 }
6312         } else if (data->lr.roc) {
6313                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6314                 data->lr.roc = false;
6315         }
6316
6317         task_setup_data.callback_data = data;
6318         msg.rpc_argp = &data->args;
6319         msg.rpc_resp = &data->res;
6320         task = rpc_run_task(&task_setup_data);
6321         if (IS_ERR(task))
6322                 return PTR_ERR(task);
6323         if (!issync)
6324                 goto out;
6325         status = rpc_wait_for_completion_task(task);
6326         if (status != 0)
6327                 goto out;
6328         status = data->rpc_status;
6329 out:
6330         rpc_put_task(task);
6331         return status;
6332 }
6333
6334 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6335 {
6336         struct nfs_server *server = NFS_SERVER(inode);
6337         struct nfs4_exception exception = { };
6338         int err;
6339         do {
6340                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6341                 trace_nfs4_delegreturn(inode, stateid, err);
6342                 switch (err) {
6343                         case -NFS4ERR_STALE_STATEID:
6344                         case -NFS4ERR_EXPIRED:
6345                         case 0:
6346                                 return 0;
6347                 }
6348                 err = nfs4_handle_exception(server, err, &exception);
6349         } while (exception.retry);
6350         return err;
6351 }
6352
6353 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6354 {
6355         struct inode *inode = state->inode;
6356         struct nfs_server *server = NFS_SERVER(inode);
6357         struct nfs_client *clp = server->nfs_client;
6358         struct nfs_lockt_args arg = {
6359                 .fh = NFS_FH(inode),
6360                 .fl = request,
6361         };
6362         struct nfs_lockt_res res = {
6363                 .denied = request,
6364         };
6365         struct rpc_message msg = {
6366                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6367                 .rpc_argp       = &arg,
6368                 .rpc_resp       = &res,
6369                 .rpc_cred       = state->owner->so_cred,
6370         };
6371         struct nfs4_lock_state *lsp;
6372         int status;
6373
6374         arg.lock_owner.clientid = clp->cl_clientid;
6375         status = nfs4_set_lock_state(state, request);
6376         if (status != 0)
6377                 goto out;
6378         lsp = request->fl_u.nfs4_fl.owner;
6379         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6380         arg.lock_owner.s_dev = server->s_dev;
6381         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6382         switch (status) {
6383                 case 0:
6384                         request->fl_type = F_UNLCK;
6385                         break;
6386                 case -NFS4ERR_DENIED:
6387                         status = 0;
6388         }
6389         request->fl_ops->fl_release_private(request);
6390         request->fl_ops = NULL;
6391 out:
6392         return status;
6393 }
6394
6395 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6396 {
6397         struct nfs4_exception exception = {
6398                 .interruptible = true,
6399         };
6400         int err;
6401
6402         do {
6403                 err = _nfs4_proc_getlk(state, cmd, request);
6404                 trace_nfs4_get_lock(request, state, cmd, err);
6405                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6406                                 &exception);
6407         } while (exception.retry);
6408         return err;
6409 }
6410
6411 /*
6412  * Update the seqid of a lock stateid after receiving
6413  * NFS4ERR_OLD_STATEID
6414  */
6415 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6416                 struct nfs4_lock_state *lsp)
6417 {
6418         struct nfs4_state *state = lsp->ls_state;
6419         bool ret = false;
6420
6421         spin_lock(&state->state_lock);
6422         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6423                 goto out;
6424         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6425                 nfs4_stateid_seqid_inc(dst);
6426         else
6427                 dst->seqid = lsp->ls_stateid.seqid;
6428         ret = true;
6429 out:
6430         spin_unlock(&state->state_lock);
6431         return ret;
6432 }
6433
6434 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6435                 struct nfs4_lock_state *lsp)
6436 {
6437         struct nfs4_state *state = lsp->ls_state;
6438         bool ret;
6439
6440         spin_lock(&state->state_lock);
6441         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6442         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6443         spin_unlock(&state->state_lock);
6444         return ret;
6445 }
6446
6447 struct nfs4_unlockdata {
6448         struct nfs_locku_args arg;
6449         struct nfs_locku_res res;
6450         struct nfs4_lock_state *lsp;
6451         struct nfs_open_context *ctx;
6452         struct nfs_lock_context *l_ctx;
6453         struct file_lock fl;
6454         struct nfs_server *server;
6455         unsigned long timestamp;
6456 };
6457
6458 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6459                 struct nfs_open_context *ctx,
6460                 struct nfs4_lock_state *lsp,
6461                 struct nfs_seqid *seqid)
6462 {
6463         struct nfs4_unlockdata *p;
6464         struct nfs4_state *state = lsp->ls_state;
6465         struct inode *inode = state->inode;
6466
6467         p = kzalloc(sizeof(*p), GFP_NOFS);
6468         if (p == NULL)
6469                 return NULL;
6470         p->arg.fh = NFS_FH(inode);
6471         p->arg.fl = &p->fl;
6472         p->arg.seqid = seqid;
6473         p->res.seqid = seqid;
6474         p->lsp = lsp;
6475         /* Ensure we don't close file until we're done freeing locks! */
6476         p->ctx = get_nfs_open_context(ctx);
6477         p->l_ctx = nfs_get_lock_context(ctx);
6478         locks_init_lock(&p->fl);
6479         locks_copy_lock(&p->fl, fl);
6480         p->server = NFS_SERVER(inode);
6481         spin_lock(&state->state_lock);
6482         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6483         spin_unlock(&state->state_lock);
6484         return p;
6485 }
6486
6487 static void nfs4_locku_release_calldata(void *data)
6488 {
6489         struct nfs4_unlockdata *calldata = data;
6490         nfs_free_seqid(calldata->arg.seqid);
6491         nfs4_put_lock_state(calldata->lsp);
6492         nfs_put_lock_context(calldata->l_ctx);
6493         put_nfs_open_context(calldata->ctx);
6494         kfree(calldata);
6495 }
6496
6497 static void nfs4_locku_done(struct rpc_task *task, void *data)
6498 {
6499         struct nfs4_unlockdata *calldata = data;
6500         struct nfs4_exception exception = {
6501                 .inode = calldata->lsp->ls_state->inode,
6502                 .stateid = &calldata->arg.stateid,
6503         };
6504
6505         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6506                 return;
6507         switch (task->tk_status) {
6508                 case 0:
6509                         renew_lease(calldata->server, calldata->timestamp);
6510                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6511                         if (nfs4_update_lock_stateid(calldata->lsp,
6512                                         &calldata->res.stateid))
6513                                 break;
6514                         /* Fall through */
6515                 case -NFS4ERR_ADMIN_REVOKED:
6516                 case -NFS4ERR_EXPIRED:
6517                         nfs4_free_revoked_stateid(calldata->server,
6518                                         &calldata->arg.stateid,
6519                                         task->tk_msg.rpc_cred);
6520                         /* Fall through */
6521                 case -NFS4ERR_BAD_STATEID:
6522                 case -NFS4ERR_STALE_STATEID:
6523                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6524                                                 calldata->lsp))
6525                                 rpc_restart_call_prepare(task);
6526                         break;
6527                 case -NFS4ERR_OLD_STATEID:
6528                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6529                                                 calldata->lsp))
6530                                 rpc_restart_call_prepare(task);
6531                         break;
6532                 default:
6533                         task->tk_status = nfs4_async_handle_exception(task,
6534                                         calldata->server, task->tk_status,
6535                                         &exception);
6536                         if (exception.retry)
6537                                 rpc_restart_call_prepare(task);
6538         }
6539         nfs_release_seqid(calldata->arg.seqid);
6540 }
6541
6542 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6543 {
6544         struct nfs4_unlockdata *calldata = data;
6545
6546         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6547                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6548                 return;
6549
6550         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6551                 goto out_wait;
6552         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6553                 /* Note: exit _without_ running nfs4_locku_done */
6554                 goto out_no_action;
6555         }
6556         calldata->timestamp = jiffies;
6557         if (nfs4_setup_sequence(calldata->server->nfs_client,
6558                                 &calldata->arg.seq_args,
6559                                 &calldata->res.seq_res,
6560                                 task) != 0)
6561                 nfs_release_seqid(calldata->arg.seqid);
6562         return;
6563 out_no_action:
6564         task->tk_action = NULL;
6565 out_wait:
6566         nfs4_sequence_done(task, &calldata->res.seq_res);
6567 }
6568
6569 static const struct rpc_call_ops nfs4_locku_ops = {
6570         .rpc_call_prepare = nfs4_locku_prepare,
6571         .rpc_call_done = nfs4_locku_done,
6572         .rpc_release = nfs4_locku_release_calldata,
6573 };
6574
6575 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6576                 struct nfs_open_context *ctx,
6577                 struct nfs4_lock_state *lsp,
6578                 struct nfs_seqid *seqid)
6579 {
6580         struct nfs4_unlockdata *data;
6581         struct rpc_message msg = {
6582                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6583                 .rpc_cred = ctx->cred,
6584         };
6585         struct rpc_task_setup task_setup_data = {
6586                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6587                 .rpc_message = &msg,
6588                 .callback_ops = &nfs4_locku_ops,
6589                 .workqueue = nfsiod_workqueue,
6590                 .flags = RPC_TASK_ASYNC,
6591         };
6592
6593         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6594                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6595
6596         /* Ensure this is an unlock - when canceling a lock, the
6597          * canceled lock is passed in, and it won't be an unlock.
6598          */
6599         fl->fl_type = F_UNLCK;
6600         if (fl->fl_flags & FL_CLOSE)
6601                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6602
6603         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6604         if (data == NULL) {
6605                 nfs_free_seqid(seqid);
6606                 return ERR_PTR(-ENOMEM);
6607         }
6608
6609         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6610         msg.rpc_argp = &data->arg;
6611         msg.rpc_resp = &data->res;
6612         task_setup_data.callback_data = data;
6613         return rpc_run_task(&task_setup_data);
6614 }
6615
6616 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6617 {
6618         struct inode *inode = state->inode;
6619         struct nfs4_state_owner *sp = state->owner;
6620         struct nfs_inode *nfsi = NFS_I(inode);
6621         struct nfs_seqid *seqid;
6622         struct nfs4_lock_state *lsp;
6623         struct rpc_task *task;
6624         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6625         int status = 0;
6626         unsigned char fl_flags = request->fl_flags;
6627
6628         status = nfs4_set_lock_state(state, request);
6629         /* Unlock _before_ we do the RPC call */
6630         request->fl_flags |= FL_EXISTS;
6631         /* Exclude nfs_delegation_claim_locks() */
6632         mutex_lock(&sp->so_delegreturn_mutex);
6633         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6634         down_read(&nfsi->rwsem);
6635         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6636                 up_read(&nfsi->rwsem);
6637                 mutex_unlock(&sp->so_delegreturn_mutex);
6638                 goto out;
6639         }
6640         up_read(&nfsi->rwsem);
6641         mutex_unlock(&sp->so_delegreturn_mutex);
6642         if (status != 0)
6643                 goto out;
6644         /* Is this a delegated lock? */
6645         lsp = request->fl_u.nfs4_fl.owner;
6646         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6647                 goto out;
6648         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6649         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6650         status = -ENOMEM;
6651         if (IS_ERR(seqid))
6652                 goto out;
6653         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6654         status = PTR_ERR(task);
6655         if (IS_ERR(task))
6656                 goto out;
6657         status = rpc_wait_for_completion_task(task);
6658         rpc_put_task(task);
6659 out:
6660         request->fl_flags = fl_flags;
6661         trace_nfs4_unlock(request, state, F_SETLK, status);
6662         return status;
6663 }
6664
6665 struct nfs4_lockdata {
6666         struct nfs_lock_args arg;
6667         struct nfs_lock_res res;
6668         struct nfs4_lock_state *lsp;
6669         struct nfs_open_context *ctx;
6670         struct file_lock fl;
6671         unsigned long timestamp;
6672         int rpc_status;
6673         int cancelled;
6674         struct nfs_server *server;
6675 };
6676
6677 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6678                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6679                 gfp_t gfp_mask)
6680 {
6681         struct nfs4_lockdata *p;
6682         struct inode *inode = lsp->ls_state->inode;
6683         struct nfs_server *server = NFS_SERVER(inode);
6684         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6685
6686         p = kzalloc(sizeof(*p), gfp_mask);
6687         if (p == NULL)
6688                 return NULL;
6689
6690         p->arg.fh = NFS_FH(inode);
6691         p->arg.fl = &p->fl;
6692         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6693         if (IS_ERR(p->arg.open_seqid))
6694                 goto out_free;
6695         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6696         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6697         if (IS_ERR(p->arg.lock_seqid))
6698                 goto out_free_seqid;
6699         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6700         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6701         p->arg.lock_owner.s_dev = server->s_dev;
6702         p->res.lock_seqid = p->arg.lock_seqid;
6703         p->lsp = lsp;
6704         p->server = server;
6705         p->ctx = get_nfs_open_context(ctx);
6706         locks_init_lock(&p->fl);
6707         locks_copy_lock(&p->fl, fl);
6708         return p;
6709 out_free_seqid:
6710         nfs_free_seqid(p->arg.open_seqid);
6711 out_free:
6712         kfree(p);
6713         return NULL;
6714 }
6715
6716 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6717 {
6718         struct nfs4_lockdata *data = calldata;
6719         struct nfs4_state *state = data->lsp->ls_state;
6720
6721         dprintk("%s: begin!\n", __func__);
6722         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6723                 goto out_wait;
6724         /* Do we need to do an open_to_lock_owner? */
6725         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6726                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6727                         goto out_release_lock_seqid;
6728                 }
6729                 nfs4_stateid_copy(&data->arg.open_stateid,
6730                                 &state->open_stateid);
6731                 data->arg.new_lock_owner = 1;
6732                 data->res.open_seqid = data->arg.open_seqid;
6733         } else {
6734                 data->arg.new_lock_owner = 0;
6735                 nfs4_stateid_copy(&data->arg.lock_stateid,
6736                                 &data->lsp->ls_stateid);
6737         }
6738         if (!nfs4_valid_open_stateid(state)) {
6739                 data->rpc_status = -EBADF;
6740                 task->tk_action = NULL;
6741                 goto out_release_open_seqid;
6742         }
6743         data->timestamp = jiffies;
6744         if (nfs4_setup_sequence(data->server->nfs_client,
6745                                 &data->arg.seq_args,
6746                                 &data->res.seq_res,
6747                                 task) == 0)
6748                 return;
6749 out_release_open_seqid:
6750         nfs_release_seqid(data->arg.open_seqid);
6751 out_release_lock_seqid:
6752         nfs_release_seqid(data->arg.lock_seqid);
6753 out_wait:
6754         nfs4_sequence_done(task, &data->res.seq_res);
6755         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6756 }
6757
6758 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6759 {
6760         struct nfs4_lockdata *data = calldata;
6761         struct nfs4_lock_state *lsp = data->lsp;
6762
6763         dprintk("%s: begin!\n", __func__);
6764
6765         if (!nfs4_sequence_done(task, &data->res.seq_res))
6766                 return;
6767
6768         data->rpc_status = task->tk_status;
6769         switch (task->tk_status) {
6770         case 0:
6771                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6772                                 data->timestamp);
6773                 if (data->arg.new_lock && !data->cancelled) {
6774                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6775                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6776                                 goto out_restart;
6777                 }
6778                 if (data->arg.new_lock_owner != 0) {
6779                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6780                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6781                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6782                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6783                         goto out_restart;
6784                 break;
6785         case -NFS4ERR_BAD_STATEID:
6786         case -NFS4ERR_OLD_STATEID:
6787         case -NFS4ERR_STALE_STATEID:
6788         case -NFS4ERR_EXPIRED:
6789                 if (data->arg.new_lock_owner != 0) {
6790                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6791                                                 &lsp->ls_state->open_stateid))
6792                                 goto out_restart;
6793                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6794                                                 &lsp->ls_stateid))
6795                                 goto out_restart;
6796         }
6797 out_done:
6798         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6799         return;
6800 out_restart:
6801         if (!data->cancelled)
6802                 rpc_restart_call_prepare(task);
6803         goto out_done;
6804 }
6805
6806 static void nfs4_lock_release(void *calldata)
6807 {
6808         struct nfs4_lockdata *data = calldata;
6809
6810         dprintk("%s: begin!\n", __func__);
6811         nfs_free_seqid(data->arg.open_seqid);
6812         if (data->cancelled && data->rpc_status == 0) {
6813                 struct rpc_task *task;
6814                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6815                                 data->arg.lock_seqid);
6816                 if (!IS_ERR(task))
6817                         rpc_put_task_async(task);
6818                 dprintk("%s: cancelling lock!\n", __func__);
6819         } else
6820                 nfs_free_seqid(data->arg.lock_seqid);
6821         nfs4_put_lock_state(data->lsp);
6822         put_nfs_open_context(data->ctx);
6823         kfree(data);
6824         dprintk("%s: done!\n", __func__);
6825 }
6826
6827 static const struct rpc_call_ops nfs4_lock_ops = {
6828         .rpc_call_prepare = nfs4_lock_prepare,
6829         .rpc_call_done = nfs4_lock_done,
6830         .rpc_release = nfs4_lock_release,
6831 };
6832
6833 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6834 {
6835         switch (error) {
6836         case -NFS4ERR_ADMIN_REVOKED:
6837         case -NFS4ERR_EXPIRED:
6838         case -NFS4ERR_BAD_STATEID:
6839                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6840                 if (new_lock_owner != 0 ||
6841                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6842                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6843                 break;
6844         case -NFS4ERR_STALE_STATEID:
6845                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6846                 nfs4_schedule_lease_recovery(server->nfs_client);
6847         };
6848 }
6849
6850 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6851 {
6852         struct nfs4_lockdata *data;
6853         struct rpc_task *task;
6854         struct rpc_message msg = {
6855                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6856                 .rpc_cred = state->owner->so_cred,
6857         };
6858         struct rpc_task_setup task_setup_data = {
6859                 .rpc_client = NFS_CLIENT(state->inode),
6860                 .rpc_message = &msg,
6861                 .callback_ops = &nfs4_lock_ops,
6862                 .workqueue = nfsiod_workqueue,
6863                 .flags = RPC_TASK_ASYNC,
6864         };
6865         int ret;
6866
6867         dprintk("%s: begin!\n", __func__);
6868         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6869                         fl->fl_u.nfs4_fl.owner,
6870                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6871         if (data == NULL)
6872                 return -ENOMEM;
6873         if (IS_SETLKW(cmd))
6874                 data->arg.block = 1;
6875         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6876                                 recovery_type > NFS_LOCK_NEW);
6877         msg.rpc_argp = &data->arg;
6878         msg.rpc_resp = &data->res;
6879         task_setup_data.callback_data = data;
6880         if (recovery_type > NFS_LOCK_NEW) {
6881                 if (recovery_type == NFS_LOCK_RECLAIM)
6882                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6883         } else
6884                 data->arg.new_lock = 1;
6885         task = rpc_run_task(&task_setup_data);
6886         if (IS_ERR(task))
6887                 return PTR_ERR(task);
6888         ret = rpc_wait_for_completion_task(task);
6889         if (ret == 0) {
6890                 ret = data->rpc_status;
6891                 if (ret)
6892                         nfs4_handle_setlk_error(data->server, data->lsp,
6893                                         data->arg.new_lock_owner, ret);
6894         } else
6895                 data->cancelled = true;
6896         rpc_put_task(task);
6897         dprintk("%s: done, ret = %d!\n", __func__, ret);
6898         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6899         return ret;
6900 }
6901
6902 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6903 {
6904         struct nfs_server *server = NFS_SERVER(state->inode);
6905         struct nfs4_exception exception = {
6906                 .inode = state->inode,
6907         };
6908         int err;
6909
6910         do {
6911                 /* Cache the lock if possible... */
6912                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6913                         return 0;
6914                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6915                 if (err != -NFS4ERR_DELAY)
6916                         break;
6917                 nfs4_handle_exception(server, err, &exception);
6918         } while (exception.retry);
6919         return err;
6920 }
6921
6922 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6923 {
6924         struct nfs_server *server = NFS_SERVER(state->inode);
6925         struct nfs4_exception exception = {
6926                 .inode = state->inode,
6927         };
6928         int err;
6929
6930         err = nfs4_set_lock_state(state, request);
6931         if (err != 0)
6932                 return err;
6933         if (!recover_lost_locks) {
6934                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6935                 return 0;
6936         }
6937         do {
6938                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6939                         return 0;
6940                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6941                 switch (err) {
6942                 default:
6943                         goto out;
6944                 case -NFS4ERR_GRACE:
6945                 case -NFS4ERR_DELAY:
6946                         nfs4_handle_exception(server, err, &exception);
6947                         err = 0;
6948                 }
6949         } while (exception.retry);
6950 out:
6951         return err;
6952 }
6953
6954 #if defined(CONFIG_NFS_V4_1)
6955 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6956 {
6957         struct nfs4_lock_state *lsp;
6958         int status;
6959
6960         status = nfs4_set_lock_state(state, request);
6961         if (status != 0)
6962                 return status;
6963         lsp = request->fl_u.nfs4_fl.owner;
6964         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6965             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6966                 return 0;
6967         return nfs4_lock_expired(state, request);
6968 }
6969 #endif
6970
6971 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6972 {
6973         struct nfs_inode *nfsi = NFS_I(state->inode);
6974         struct nfs4_state_owner *sp = state->owner;
6975         unsigned char fl_flags = request->fl_flags;
6976         int status;
6977
6978         request->fl_flags |= FL_ACCESS;
6979         status = locks_lock_inode_wait(state->inode, request);
6980         if (status < 0)
6981                 goto out;
6982         mutex_lock(&sp->so_delegreturn_mutex);
6983         down_read(&nfsi->rwsem);
6984         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6985                 /* Yes: cache locks! */
6986                 /* ...but avoid races with delegation recall... */
6987                 request->fl_flags = fl_flags & ~FL_SLEEP;
6988                 status = locks_lock_inode_wait(state->inode, request);
6989                 up_read(&nfsi->rwsem);
6990                 mutex_unlock(&sp->so_delegreturn_mutex);
6991                 goto out;
6992         }
6993         up_read(&nfsi->rwsem);
6994         mutex_unlock(&sp->so_delegreturn_mutex);
6995         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6996 out:
6997         request->fl_flags = fl_flags;
6998         return status;
6999 }
7000
7001 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7002 {
7003         struct nfs4_exception exception = {
7004                 .state = state,
7005                 .inode = state->inode,
7006                 .interruptible = true,
7007         };
7008         int err;
7009
7010         do {
7011                 err = _nfs4_proc_setlk(state, cmd, request);
7012                 if (err == -NFS4ERR_DENIED)
7013                         err = -EAGAIN;
7014                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7015                                 err, &exception);
7016         } while (exception.retry);
7017         return err;
7018 }
7019
7020 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7021 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7022
7023 static int
7024 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7025                         struct file_lock *request)
7026 {
7027         int             status = -ERESTARTSYS;
7028         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7029
7030         while(!signalled()) {
7031                 status = nfs4_proc_setlk(state, cmd, request);
7032                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7033                         break;
7034                 freezable_schedule_timeout_interruptible(timeout);
7035                 timeout *= 2;
7036                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7037                 status = -ERESTARTSYS;
7038         }
7039         return status;
7040 }
7041
7042 #ifdef CONFIG_NFS_V4_1
7043 struct nfs4_lock_waiter {
7044         struct task_struct      *task;
7045         struct inode            *inode;
7046         struct nfs_lowner       *owner;
7047 };
7048
7049 static int
7050 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7051 {
7052         int ret;
7053         struct nfs4_lock_waiter *waiter = wait->private;
7054
7055         /* NULL key means to wake up everyone */
7056         if (key) {
7057                 struct cb_notify_lock_args      *cbnl = key;
7058                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7059                                                 *wowner = waiter->owner;
7060
7061                 /* Only wake if the callback was for the same owner. */
7062                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7063                         return 0;
7064
7065                 /* Make sure it's for the right inode */
7066                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7067                         return 0;
7068         }
7069
7070         /* override "private" so we can use default_wake_function */
7071         wait->private = waiter->task;
7072         ret = woken_wake_function(wait, mode, flags, key);
7073         if (ret)
7074                 list_del_init(&wait->entry);
7075         wait->private = waiter;
7076         return ret;
7077 }
7078
7079 static int
7080 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7081 {
7082         int status = -ERESTARTSYS;
7083         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7084         struct nfs_server *server = NFS_SERVER(state->inode);
7085         struct nfs_client *clp = server->nfs_client;
7086         wait_queue_head_t *q = &clp->cl_lock_waitq;
7087         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7088                                     .id = lsp->ls_seqid.owner_id,
7089                                     .s_dev = server->s_dev };
7090         struct nfs4_lock_waiter waiter = { .task  = current,
7091                                            .inode = state->inode,
7092                                            .owner = &owner};
7093         wait_queue_entry_t wait;
7094
7095         /* Don't bother with waitqueue if we don't expect a callback */
7096         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7097                 return nfs4_retry_setlk_simple(state, cmd, request);
7098
7099         init_wait(&wait);
7100         wait.private = &waiter;
7101         wait.func = nfs4_wake_lock_waiter;
7102
7103         while(!signalled()) {
7104                 add_wait_queue(q, &wait);
7105                 status = nfs4_proc_setlk(state, cmd, request);
7106                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7107                         finish_wait(q, &wait);
7108                         break;
7109                 }
7110
7111                 status = -ERESTARTSYS;
7112                 freezer_do_not_count();
7113                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7114                 freezer_count();
7115                 finish_wait(q, &wait);
7116         }
7117
7118         return status;
7119 }
7120 #else /* !CONFIG_NFS_V4_1 */
7121 static inline int
7122 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7123 {
7124         return nfs4_retry_setlk_simple(state, cmd, request);
7125 }
7126 #endif
7127
7128 static int
7129 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7130 {
7131         struct nfs_open_context *ctx;
7132         struct nfs4_state *state;
7133         int status;
7134
7135         /* verify open state */
7136         ctx = nfs_file_open_context(filp);
7137         state = ctx->state;
7138
7139         if (IS_GETLK(cmd)) {
7140                 if (state != NULL)
7141                         return nfs4_proc_getlk(state, F_GETLK, request);
7142                 return 0;
7143         }
7144
7145         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7146                 return -EINVAL;
7147
7148         if (request->fl_type == F_UNLCK) {
7149                 if (state != NULL)
7150                         return nfs4_proc_unlck(state, cmd, request);
7151                 return 0;
7152         }
7153
7154         if (state == NULL)
7155                 return -ENOLCK;
7156
7157         if ((request->fl_flags & FL_POSIX) &&
7158             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7159                 return -ENOLCK;
7160
7161         /*
7162          * Don't rely on the VFS having checked the file open mode,
7163          * since it won't do this for flock() locks.
7164          */
7165         switch (request->fl_type) {
7166         case F_RDLCK:
7167                 if (!(filp->f_mode & FMODE_READ))
7168                         return -EBADF;
7169                 break;
7170         case F_WRLCK:
7171                 if (!(filp->f_mode & FMODE_WRITE))
7172                         return -EBADF;
7173         }
7174
7175         status = nfs4_set_lock_state(state, request);
7176         if (status != 0)
7177                 return status;
7178
7179         return nfs4_retry_setlk(state, cmd, request);
7180 }
7181
7182 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7183 {
7184         struct nfs_server *server = NFS_SERVER(state->inode);
7185         int err;
7186
7187         err = nfs4_set_lock_state(state, fl);
7188         if (err != 0)
7189                 return err;
7190         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7191         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7192 }
7193
7194 struct nfs_release_lockowner_data {
7195         struct nfs4_lock_state *lsp;
7196         struct nfs_server *server;
7197         struct nfs_release_lockowner_args args;
7198         struct nfs_release_lockowner_res res;
7199         unsigned long timestamp;
7200 };
7201
7202 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7203 {
7204         struct nfs_release_lockowner_data *data = calldata;
7205         struct nfs_server *server = data->server;
7206         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7207                            &data->res.seq_res, task);
7208         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7209         data->timestamp = jiffies;
7210 }
7211
7212 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7213 {
7214         struct nfs_release_lockowner_data *data = calldata;
7215         struct nfs_server *server = data->server;
7216
7217         nfs40_sequence_done(task, &data->res.seq_res);
7218
7219         switch (task->tk_status) {
7220         case 0:
7221                 renew_lease(server, data->timestamp);
7222                 break;
7223         case -NFS4ERR_STALE_CLIENTID:
7224         case -NFS4ERR_EXPIRED:
7225                 nfs4_schedule_lease_recovery(server->nfs_client);
7226                 break;
7227         case -NFS4ERR_LEASE_MOVED:
7228         case -NFS4ERR_DELAY:
7229                 if (nfs4_async_handle_error(task, server,
7230                                             NULL, NULL) == -EAGAIN)
7231                         rpc_restart_call_prepare(task);
7232         }
7233 }
7234
7235 static void nfs4_release_lockowner_release(void *calldata)
7236 {
7237         struct nfs_release_lockowner_data *data = calldata;
7238         nfs4_free_lock_state(data->server, data->lsp);
7239         kfree(calldata);
7240 }
7241
7242 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7243         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7244         .rpc_call_done = nfs4_release_lockowner_done,
7245         .rpc_release = nfs4_release_lockowner_release,
7246 };
7247
7248 static void
7249 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7250 {
7251         struct nfs_release_lockowner_data *data;
7252         struct rpc_message msg = {
7253                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7254         };
7255
7256         if (server->nfs_client->cl_mvops->minor_version != 0)
7257                 return;
7258
7259         data = kmalloc(sizeof(*data), GFP_NOFS);
7260         if (!data)
7261                 return;
7262         data->lsp = lsp;
7263         data->server = server;
7264         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7265         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7266         data->args.lock_owner.s_dev = server->s_dev;
7267
7268         msg.rpc_argp = &data->args;
7269         msg.rpc_resp = &data->res;
7270         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7271         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7272 }
7273
7274 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7275
7276 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7277                                    struct dentry *unused, struct inode *inode,
7278                                    const char *key, const void *buf,
7279                                    size_t buflen, int flags)
7280 {
7281         return nfs4_proc_set_acl(inode, buf, buflen);
7282 }
7283
7284 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7285                                    struct dentry *unused, struct inode *inode,
7286                                    const char *key, void *buf, size_t buflen)
7287 {
7288         return nfs4_proc_get_acl(inode, buf, buflen);
7289 }
7290
7291 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7292 {
7293         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7294 }
7295
7296 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7297
7298 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7299                                      struct dentry *unused, struct inode *inode,
7300                                      const char *key, const void *buf,
7301                                      size_t buflen, int flags)
7302 {
7303         if (security_ismaclabel(key))
7304                 return nfs4_set_security_label(inode, buf, buflen);
7305
7306         return -EOPNOTSUPP;
7307 }
7308
7309 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7310                                      struct dentry *unused, struct inode *inode,
7311                                      const char *key, void *buf, size_t buflen)
7312 {
7313         if (security_ismaclabel(key))
7314                 return nfs4_get_security_label(inode, buf, buflen);
7315         return -EOPNOTSUPP;
7316 }
7317
7318 static ssize_t
7319 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7320 {
7321         int len = 0;
7322
7323         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7324                 len = security_inode_listsecurity(inode, list, list_len);
7325                 if (list_len && len > list_len)
7326                         return -ERANGE;
7327         }
7328         return len;
7329 }
7330
7331 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7332         .prefix = XATTR_SECURITY_PREFIX,
7333         .get    = nfs4_xattr_get_nfs4_label,
7334         .set    = nfs4_xattr_set_nfs4_label,
7335 };
7336
7337 #else
7338
7339 static ssize_t
7340 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7341 {
7342         return 0;
7343 }
7344
7345 #endif
7346
7347 /*
7348  * nfs_fhget will use either the mounted_on_fileid or the fileid
7349  */
7350 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7351 {
7352         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7353                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7354               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7355               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7356                 return;
7357
7358         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7359                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7360         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7361         fattr->nlink = 2;
7362 }
7363
7364 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7365                                    const struct qstr *name,
7366                                    struct nfs4_fs_locations *fs_locations,
7367                                    struct page *page)
7368 {
7369         struct nfs_server *server = NFS_SERVER(dir);
7370         u32 bitmask[3];
7371         struct nfs4_fs_locations_arg args = {
7372                 .dir_fh = NFS_FH(dir),
7373                 .name = name,
7374                 .page = page,
7375                 .bitmask = bitmask,
7376         };
7377         struct nfs4_fs_locations_res res = {
7378                 .fs_locations = fs_locations,
7379         };
7380         struct rpc_message msg = {
7381                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7382                 .rpc_argp = &args,
7383                 .rpc_resp = &res,
7384         };
7385         int status;
7386
7387         dprintk("%s: start\n", __func__);
7388
7389         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7390         bitmask[1] = nfs4_fattr_bitmap[1];
7391
7392         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7393          * is not supported */
7394         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7395                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7396         else
7397                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7398
7399         nfs_fattr_init(&fs_locations->fattr);
7400         fs_locations->server = server;
7401         fs_locations->nlocations = 0;
7402         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7403         dprintk("%s: returned status = %d\n", __func__, status);
7404         return status;
7405 }
7406
7407 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7408                            const struct qstr *name,
7409                            struct nfs4_fs_locations *fs_locations,
7410                            struct page *page)
7411 {
7412         struct nfs4_exception exception = {
7413                 .interruptible = true,
7414         };
7415         int err;
7416         do {
7417                 err = _nfs4_proc_fs_locations(client, dir, name,
7418                                 fs_locations, page);
7419                 trace_nfs4_get_fs_locations(dir, name, err);
7420                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7421                                 &exception);
7422         } while (exception.retry);
7423         return err;
7424 }
7425
7426 /*
7427  * This operation also signals the server that this client is
7428  * performing migration recovery.  The server can stop returning
7429  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7430  * appended to this compound to identify the client ID which is
7431  * performing recovery.
7432  */
7433 static int _nfs40_proc_get_locations(struct inode *inode,
7434                                      struct nfs4_fs_locations *locations,
7435                                      struct page *page, const struct cred *cred)
7436 {
7437         struct nfs_server *server = NFS_SERVER(inode);
7438         struct rpc_clnt *clnt = server->client;
7439         u32 bitmask[2] = {
7440                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7441         };
7442         struct nfs4_fs_locations_arg args = {
7443                 .clientid       = server->nfs_client->cl_clientid,
7444                 .fh             = NFS_FH(inode),
7445                 .page           = page,
7446                 .bitmask        = bitmask,
7447                 .migration      = 1,            /* skip LOOKUP */
7448                 .renew          = 1,            /* append RENEW */
7449         };
7450         struct nfs4_fs_locations_res res = {
7451                 .fs_locations   = locations,
7452                 .migration      = 1,
7453                 .renew          = 1,
7454         };
7455         struct rpc_message msg = {
7456                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7457                 .rpc_argp       = &args,
7458                 .rpc_resp       = &res,
7459                 .rpc_cred       = cred,
7460         };
7461         unsigned long now = jiffies;
7462         int status;
7463
7464         nfs_fattr_init(&locations->fattr);
7465         locations->server = server;
7466         locations->nlocations = 0;
7467
7468         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7469         status = nfs4_call_sync_sequence(clnt, server, &msg,
7470                                         &args.seq_args, &res.seq_res);
7471         if (status)
7472                 return status;
7473
7474         renew_lease(server, now);
7475         return 0;
7476 }
7477
7478 #ifdef CONFIG_NFS_V4_1
7479
7480 /*
7481  * This operation also signals the server that this client is
7482  * performing migration recovery.  The server can stop asserting
7483  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7484  * performing this operation is identified in the SEQUENCE
7485  * operation in this compound.
7486  *
7487  * When the client supports GETATTR(fs_locations_info), it can
7488  * be plumbed in here.
7489  */
7490 static int _nfs41_proc_get_locations(struct inode *inode,
7491                                      struct nfs4_fs_locations *locations,
7492                                      struct page *page, const struct cred *cred)
7493 {
7494         struct nfs_server *server = NFS_SERVER(inode);
7495         struct rpc_clnt *clnt = server->client;
7496         u32 bitmask[2] = {
7497                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7498         };
7499         struct nfs4_fs_locations_arg args = {
7500                 .fh             = NFS_FH(inode),
7501                 .page           = page,
7502                 .bitmask        = bitmask,
7503                 .migration      = 1,            /* skip LOOKUP */
7504         };
7505         struct nfs4_fs_locations_res res = {
7506                 .fs_locations   = locations,
7507                 .migration      = 1,
7508         };
7509         struct rpc_message msg = {
7510                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7511                 .rpc_argp       = &args,
7512                 .rpc_resp       = &res,
7513                 .rpc_cred       = cred,
7514         };
7515         int status;
7516
7517         nfs_fattr_init(&locations->fattr);
7518         locations->server = server;
7519         locations->nlocations = 0;
7520
7521         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7522         status = nfs4_call_sync_sequence(clnt, server, &msg,
7523                                         &args.seq_args, &res.seq_res);
7524         if (status == NFS4_OK &&
7525             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7526                 status = -NFS4ERR_LEASE_MOVED;
7527         return status;
7528 }
7529
7530 #endif  /* CONFIG_NFS_V4_1 */
7531
7532 /**
7533  * nfs4_proc_get_locations - discover locations for a migrated FSID
7534  * @inode: inode on FSID that is migrating
7535  * @locations: result of query
7536  * @page: buffer
7537  * @cred: credential to use for this operation
7538  *
7539  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7540  * operation failed, or a negative errno if a local error occurred.
7541  *
7542  * On success, "locations" is filled in, but if the server has
7543  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7544  * asserted.
7545  *
7546  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7547  * from this client that require migration recovery.
7548  */
7549 int nfs4_proc_get_locations(struct inode *inode,
7550                             struct nfs4_fs_locations *locations,
7551                             struct page *page, const struct cred *cred)
7552 {
7553         struct nfs_server *server = NFS_SERVER(inode);
7554         struct nfs_client *clp = server->nfs_client;
7555         const struct nfs4_mig_recovery_ops *ops =
7556                                         clp->cl_mvops->mig_recovery_ops;
7557         struct nfs4_exception exception = {
7558                 .interruptible = true,
7559         };
7560         int status;
7561
7562         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7563                 (unsigned long long)server->fsid.major,
7564                 (unsigned long long)server->fsid.minor,
7565                 clp->cl_hostname);
7566         nfs_display_fhandle(NFS_FH(inode), __func__);
7567
7568         do {
7569                 status = ops->get_locations(inode, locations, page, cred);
7570                 if (status != -NFS4ERR_DELAY)
7571                         break;
7572                 nfs4_handle_exception(server, status, &exception);
7573         } while (exception.retry);
7574         return status;
7575 }
7576
7577 /*
7578  * This operation also signals the server that this client is
7579  * performing "lease moved" recovery.  The server can stop
7580  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7581  * is appended to this compound to identify the client ID which is
7582  * performing recovery.
7583  */
7584 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7585 {
7586         struct nfs_server *server = NFS_SERVER(inode);
7587         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7588         struct rpc_clnt *clnt = server->client;
7589         struct nfs4_fsid_present_arg args = {
7590                 .fh             = NFS_FH(inode),
7591                 .clientid       = clp->cl_clientid,
7592                 .renew          = 1,            /* append RENEW */
7593         };
7594         struct nfs4_fsid_present_res res = {
7595                 .renew          = 1,
7596         };
7597         struct rpc_message msg = {
7598                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7599                 .rpc_argp       = &args,
7600                 .rpc_resp       = &res,
7601                 .rpc_cred       = cred,
7602         };
7603         unsigned long now = jiffies;
7604         int status;
7605
7606         res.fh = nfs_alloc_fhandle();
7607         if (res.fh == NULL)
7608                 return -ENOMEM;
7609
7610         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7611         status = nfs4_call_sync_sequence(clnt, server, &msg,
7612                                                 &args.seq_args, &res.seq_res);
7613         nfs_free_fhandle(res.fh);
7614         if (status)
7615                 return status;
7616
7617         do_renew_lease(clp, now);
7618         return 0;
7619 }
7620
7621 #ifdef CONFIG_NFS_V4_1
7622
7623 /*
7624  * This operation also signals the server that this client is
7625  * performing "lease moved" recovery.  The server can stop asserting
7626  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7627  * this operation is identified in the SEQUENCE operation in this
7628  * compound.
7629  */
7630 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7631 {
7632         struct nfs_server *server = NFS_SERVER(inode);
7633         struct rpc_clnt *clnt = server->client;
7634         struct nfs4_fsid_present_arg args = {
7635                 .fh             = NFS_FH(inode),
7636         };
7637         struct nfs4_fsid_present_res res = {
7638         };
7639         struct rpc_message msg = {
7640                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7641                 .rpc_argp       = &args,
7642                 .rpc_resp       = &res,
7643                 .rpc_cred       = cred,
7644         };
7645         int status;
7646
7647         res.fh = nfs_alloc_fhandle();
7648         if (res.fh == NULL)
7649                 return -ENOMEM;
7650
7651         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7652         status = nfs4_call_sync_sequence(clnt, server, &msg,
7653                                                 &args.seq_args, &res.seq_res);
7654         nfs_free_fhandle(res.fh);
7655         if (status == NFS4_OK &&
7656             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7657                 status = -NFS4ERR_LEASE_MOVED;
7658         return status;
7659 }
7660
7661 #endif  /* CONFIG_NFS_V4_1 */
7662
7663 /**
7664  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7665  * @inode: inode on FSID to check
7666  * @cred: credential to use for this operation
7667  *
7668  * Server indicates whether the FSID is present, moved, or not
7669  * recognized.  This operation is necessary to clear a LEASE_MOVED
7670  * condition for this client ID.
7671  *
7672  * Returns NFS4_OK if the FSID is present on this server,
7673  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7674  *  NFS4ERR code if some error occurred on the server, or a
7675  *  negative errno if a local failure occurred.
7676  */
7677 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7678 {
7679         struct nfs_server *server = NFS_SERVER(inode);
7680         struct nfs_client *clp = server->nfs_client;
7681         const struct nfs4_mig_recovery_ops *ops =
7682                                         clp->cl_mvops->mig_recovery_ops;
7683         struct nfs4_exception exception = {
7684                 .interruptible = true,
7685         };
7686         int status;
7687
7688         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7689                 (unsigned long long)server->fsid.major,
7690                 (unsigned long long)server->fsid.minor,
7691                 clp->cl_hostname);
7692         nfs_display_fhandle(NFS_FH(inode), __func__);
7693
7694         do {
7695                 status = ops->fsid_present(inode, cred);
7696                 if (status != -NFS4ERR_DELAY)
7697                         break;
7698                 nfs4_handle_exception(server, status, &exception);
7699         } while (exception.retry);
7700         return status;
7701 }
7702
7703 /*
7704  * If 'use_integrity' is true and the state managment nfs_client
7705  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7706  * and the machine credential as per RFC3530bis and RFC5661 Security
7707  * Considerations sections. Otherwise, just use the user cred with the
7708  * filesystem's rpc_client.
7709  */
7710 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7711 {
7712         int status;
7713         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7714         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7715         struct nfs4_secinfo_arg args = {
7716                 .dir_fh = NFS_FH(dir),
7717                 .name   = name,
7718         };
7719         struct nfs4_secinfo_res res = {
7720                 .flavors     = flavors,
7721         };
7722         struct rpc_message msg = {
7723                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7724                 .rpc_argp = &args,
7725                 .rpc_resp = &res,
7726         };
7727         struct nfs4_call_sync_data data = {
7728                 .seq_server = NFS_SERVER(dir),
7729                 .seq_args = &args.seq_args,
7730                 .seq_res = &res.seq_res,
7731         };
7732         struct rpc_task_setup task_setup = {
7733                 .rpc_client = clnt,
7734                 .rpc_message = &msg,
7735                 .callback_ops = clp->cl_mvops->call_sync_ops,
7736                 .callback_data = &data,
7737                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7738         };
7739         const struct cred *cred = NULL;
7740
7741         if (use_integrity) {
7742                 clnt = clp->cl_rpcclient;
7743                 task_setup.rpc_client = clnt;
7744
7745                 cred = nfs4_get_clid_cred(clp);
7746                 msg.rpc_cred = cred;
7747         }
7748
7749         dprintk("NFS call  secinfo %s\n", name->name);
7750
7751         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7752         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7753         status = nfs4_call_sync_custom(&task_setup);
7754
7755         dprintk("NFS reply  secinfo: %d\n", status);
7756
7757         put_cred(cred);
7758         return status;
7759 }
7760
7761 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7762                       struct nfs4_secinfo_flavors *flavors)
7763 {
7764         struct nfs4_exception exception = {
7765                 .interruptible = true,
7766         };
7767         int err;
7768         do {
7769                 err = -NFS4ERR_WRONGSEC;
7770
7771                 /* try to use integrity protection with machine cred */
7772                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7773                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7774
7775                 /*
7776                  * if unable to use integrity protection, or SECINFO with
7777                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7778                  * disallowed by spec, but exists in deployed servers) use
7779                  * the current filesystem's rpc_client and the user cred.
7780                  */
7781                 if (err == -NFS4ERR_WRONGSEC)
7782                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7783
7784                 trace_nfs4_secinfo(dir, name, err);
7785                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7786                                 &exception);
7787         } while (exception.retry);
7788         return err;
7789 }
7790
7791 #ifdef CONFIG_NFS_V4_1
7792 /*
7793  * Check the exchange flags returned by the server for invalid flags, having
7794  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7795  * DS flags set.
7796  */
7797 static int nfs4_check_cl_exchange_flags(u32 flags)
7798 {
7799         if (flags & ~EXCHGID4_FLAG_MASK_R)
7800                 goto out_inval;
7801         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7802             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7803                 goto out_inval;
7804         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7805                 goto out_inval;
7806         return NFS_OK;
7807 out_inval:
7808         return -NFS4ERR_INVAL;
7809 }
7810
7811 static bool
7812 nfs41_same_server_scope(struct nfs41_server_scope *a,
7813                         struct nfs41_server_scope *b)
7814 {
7815         if (a->server_scope_sz != b->server_scope_sz)
7816                 return false;
7817         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7818 }
7819
7820 static void
7821 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7822 {
7823 }
7824
7825 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7826         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7827 };
7828
7829 /*
7830  * nfs4_proc_bind_one_conn_to_session()
7831  *
7832  * The 4.1 client currently uses the same TCP connection for the
7833  * fore and backchannel.
7834  */
7835 static
7836 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7837                 struct rpc_xprt *xprt,
7838                 struct nfs_client *clp,
7839                 const struct cred *cred)
7840 {
7841         int status;
7842         struct nfs41_bind_conn_to_session_args args = {
7843                 .client = clp,
7844                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7845         };
7846         struct nfs41_bind_conn_to_session_res res;
7847         struct rpc_message msg = {
7848                 .rpc_proc =
7849                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7850                 .rpc_argp = &args,
7851                 .rpc_resp = &res,
7852                 .rpc_cred = cred,
7853         };
7854         struct rpc_task_setup task_setup_data = {
7855                 .rpc_client = clnt,
7856                 .rpc_xprt = xprt,
7857                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7858                 .rpc_message = &msg,
7859                 .flags = RPC_TASK_TIMEOUT,
7860         };
7861         struct rpc_task *task;
7862
7863         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7864         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7865                 args.dir = NFS4_CDFC4_FORE;
7866
7867         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7868         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7869                 args.dir = NFS4_CDFC4_FORE;
7870
7871         task = rpc_run_task(&task_setup_data);
7872         if (!IS_ERR(task)) {
7873                 status = task->tk_status;
7874                 rpc_put_task(task);
7875         } else
7876                 status = PTR_ERR(task);
7877         trace_nfs4_bind_conn_to_session(clp, status);
7878         if (status == 0) {
7879                 if (memcmp(res.sessionid.data,
7880                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7881                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7882                         return -EIO;
7883                 }
7884                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7885                         dprintk("NFS: %s: Unexpected direction from server\n",
7886                                 __func__);
7887                         return -EIO;
7888                 }
7889                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7890                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7891                                 __func__);
7892                         return -EIO;
7893                 }
7894         }
7895
7896         return status;
7897 }
7898
7899 struct rpc_bind_conn_calldata {
7900         struct nfs_client *clp;
7901         const struct cred *cred;
7902 };
7903
7904 static int
7905 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7906                 struct rpc_xprt *xprt,
7907                 void *calldata)
7908 {
7909         struct rpc_bind_conn_calldata *p = calldata;
7910
7911         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7912 }
7913
7914 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7915 {
7916         struct rpc_bind_conn_calldata data = {
7917                 .clp = clp,
7918                 .cred = cred,
7919         };
7920         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7921                         nfs4_proc_bind_conn_to_session_callback, &data);
7922 }
7923
7924 /*
7925  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7926  * and operations we'd like to see to enable certain features in the allow map
7927  */
7928 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7929         .how = SP4_MACH_CRED,
7930         .enforce.u.words = {
7931                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7932                       1 << (OP_EXCHANGE_ID - 32) |
7933                       1 << (OP_CREATE_SESSION - 32) |
7934                       1 << (OP_DESTROY_SESSION - 32) |
7935                       1 << (OP_DESTROY_CLIENTID - 32)
7936         },
7937         .allow.u.words = {
7938                 [0] = 1 << (OP_CLOSE) |
7939                       1 << (OP_OPEN_DOWNGRADE) |
7940                       1 << (OP_LOCKU) |
7941                       1 << (OP_DELEGRETURN) |
7942                       1 << (OP_COMMIT),
7943                 [1] = 1 << (OP_SECINFO - 32) |
7944                       1 << (OP_SECINFO_NO_NAME - 32) |
7945                       1 << (OP_LAYOUTRETURN - 32) |
7946                       1 << (OP_TEST_STATEID - 32) |
7947                       1 << (OP_FREE_STATEID - 32) |
7948                       1 << (OP_WRITE - 32)
7949         }
7950 };
7951
7952 /*
7953  * Select the state protection mode for client `clp' given the server results
7954  * from exchange_id in `sp'.
7955  *
7956  * Returns 0 on success, negative errno otherwise.
7957  */
7958 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7959                                  struct nfs41_state_protection *sp)
7960 {
7961         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7962                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7963                       1 << (OP_EXCHANGE_ID - 32) |
7964                       1 << (OP_CREATE_SESSION - 32) |
7965                       1 << (OP_DESTROY_SESSION - 32) |
7966                       1 << (OP_DESTROY_CLIENTID - 32)
7967         };
7968         unsigned long flags = 0;
7969         unsigned int i;
7970         int ret = 0;
7971
7972         if (sp->how == SP4_MACH_CRED) {
7973                 /* Print state protect result */
7974                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7975                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7976                         if (test_bit(i, sp->enforce.u.longs))
7977                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7978                         if (test_bit(i, sp->allow.u.longs))
7979                                 dfprintk(MOUNT, "  allow op %d\n", i);
7980                 }
7981
7982                 /* make sure nothing is on enforce list that isn't supported */
7983                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7984                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7985                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7986                                 ret = -EINVAL;
7987                                 goto out;
7988                         }
7989                 }
7990
7991                 /*
7992                  * Minimal mode - state operations are allowed to use machine
7993                  * credential.  Note this already happens by default, so the
7994                  * client doesn't have to do anything more than the negotiation.
7995                  *
7996                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7997                  *       we're already using the machine cred for exchange_id
7998                  *       and will never use a different cred.
7999                  */
8000                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8001                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8002                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8003                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8004                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8005                         dfprintk(MOUNT, "  minimal mode enabled\n");
8006                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8007                 } else {
8008                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8009                         ret = -EINVAL;
8010                         goto out;
8011                 }
8012
8013                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8014                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8015                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8016                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8017                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8018                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8019                 }
8020
8021                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8022                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8023                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8024                 }
8025
8026                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8027                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8028                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8029                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8030                 }
8031
8032                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8033                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8034                         dfprintk(MOUNT, "  stateid mode enabled\n");
8035                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8036                 }
8037
8038                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8039                         dfprintk(MOUNT, "  write mode enabled\n");
8040                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8041                 }
8042
8043                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8044                         dfprintk(MOUNT, "  commit mode enabled\n");
8045                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8046                 }
8047         }
8048 out:
8049         clp->cl_sp4_flags = flags;
8050         return ret;
8051 }
8052
8053 struct nfs41_exchange_id_data {
8054         struct nfs41_exchange_id_res res;
8055         struct nfs41_exchange_id_args args;
8056 };
8057
8058 static void nfs4_exchange_id_release(void *data)
8059 {
8060         struct nfs41_exchange_id_data *cdata =
8061                                         (struct nfs41_exchange_id_data *)data;
8062
8063         nfs_put_client(cdata->args.client);
8064         kfree(cdata->res.impl_id);
8065         kfree(cdata->res.server_scope);
8066         kfree(cdata->res.server_owner);
8067         kfree(cdata);
8068 }
8069
8070 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8071         .rpc_release = nfs4_exchange_id_release,
8072 };
8073
8074 /*
8075  * _nfs4_proc_exchange_id()
8076  *
8077  * Wrapper for EXCHANGE_ID operation.
8078  */
8079 static struct rpc_task *
8080 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8081                         u32 sp4_how, struct rpc_xprt *xprt)
8082 {
8083         struct rpc_message msg = {
8084                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8085                 .rpc_cred = cred,
8086         };
8087         struct rpc_task_setup task_setup_data = {
8088                 .rpc_client = clp->cl_rpcclient,
8089                 .callback_ops = &nfs4_exchange_id_call_ops,
8090                 .rpc_message = &msg,
8091                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8092         };
8093         struct nfs41_exchange_id_data *calldata;
8094         int status;
8095
8096         if (!refcount_inc_not_zero(&clp->cl_count))
8097                 return ERR_PTR(-EIO);
8098
8099         status = -ENOMEM;
8100         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8101         if (!calldata)
8102                 goto out;
8103
8104         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8105
8106         status = nfs4_init_uniform_client_string(clp);
8107         if (status)
8108                 goto out_calldata;
8109
8110         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8111                                                 GFP_NOFS);
8112         status = -ENOMEM;
8113         if (unlikely(calldata->res.server_owner == NULL))
8114                 goto out_calldata;
8115
8116         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8117                                         GFP_NOFS);
8118         if (unlikely(calldata->res.server_scope == NULL))
8119                 goto out_server_owner;
8120
8121         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8122         if (unlikely(calldata->res.impl_id == NULL))
8123                 goto out_server_scope;
8124
8125         switch (sp4_how) {
8126         case SP4_NONE:
8127                 calldata->args.state_protect.how = SP4_NONE;
8128                 break;
8129
8130         case SP4_MACH_CRED:
8131                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8132                 break;
8133
8134         default:
8135                 /* unsupported! */
8136                 WARN_ON_ONCE(1);
8137                 status = -EINVAL;
8138                 goto out_impl_id;
8139         }
8140         if (xprt) {
8141                 task_setup_data.rpc_xprt = xprt;
8142                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8143                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8144                                 sizeof(calldata->args.verifier.data));
8145         }
8146         calldata->args.client = clp;
8147         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8148         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8149 #ifdef CONFIG_NFS_V4_1_MIGRATION
8150         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8151 #endif
8152         msg.rpc_argp = &calldata->args;
8153         msg.rpc_resp = &calldata->res;
8154         task_setup_data.callback_data = calldata;
8155
8156         return rpc_run_task(&task_setup_data);
8157
8158 out_impl_id:
8159         kfree(calldata->res.impl_id);
8160 out_server_scope:
8161         kfree(calldata->res.server_scope);
8162 out_server_owner:
8163         kfree(calldata->res.server_owner);
8164 out_calldata:
8165         kfree(calldata);
8166 out:
8167         nfs_put_client(clp);
8168         return ERR_PTR(status);
8169 }
8170
8171 /*
8172  * _nfs4_proc_exchange_id()
8173  *
8174  * Wrapper for EXCHANGE_ID operation.
8175  */
8176 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8177                         u32 sp4_how)
8178 {
8179         struct rpc_task *task;
8180         struct nfs41_exchange_id_args *argp;
8181         struct nfs41_exchange_id_res *resp;
8182         int status;
8183
8184         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8185         if (IS_ERR(task))
8186                 return PTR_ERR(task);
8187
8188         argp = task->tk_msg.rpc_argp;
8189         resp = task->tk_msg.rpc_resp;
8190         status = task->tk_status;
8191         if (status  != 0)
8192                 goto out;
8193
8194         status = nfs4_check_cl_exchange_flags(resp->flags);
8195         if (status  != 0)
8196                 goto out;
8197
8198         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8199         if (status != 0)
8200                 goto out;
8201
8202         clp->cl_clientid = resp->clientid;
8203         clp->cl_exchange_flags = resp->flags;
8204         clp->cl_seqid = resp->seqid;
8205         /* Client ID is not confirmed */
8206         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8207                 clear_bit(NFS4_SESSION_ESTABLISHED,
8208                           &clp->cl_session->session_state);
8209
8210         if (clp->cl_serverscope != NULL &&
8211             !nfs41_same_server_scope(clp->cl_serverscope,
8212                                 resp->server_scope)) {
8213                 dprintk("%s: server_scope mismatch detected\n",
8214                         __func__);
8215                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8216         }
8217
8218         swap(clp->cl_serverowner, resp->server_owner);
8219         swap(clp->cl_serverscope, resp->server_scope);
8220         swap(clp->cl_implid, resp->impl_id);
8221
8222         /* Save the EXCHANGE_ID verifier session trunk tests */
8223         memcpy(clp->cl_confirm.data, argp->verifier.data,
8224                sizeof(clp->cl_confirm.data));
8225 out:
8226         trace_nfs4_exchange_id(clp, status);
8227         rpc_put_task(task);
8228         return status;
8229 }
8230
8231 /*
8232  * nfs4_proc_exchange_id()
8233  *
8234  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8235  *
8236  * Since the clientid has expired, all compounds using sessions
8237  * associated with the stale clientid will be returning
8238  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8239  * be in some phase of session reset.
8240  *
8241  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8242  */
8243 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8244 {
8245         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8246         int status;
8247
8248         /* try SP4_MACH_CRED if krb5i/p */
8249         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8250             authflavor == RPC_AUTH_GSS_KRB5P) {
8251                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8252                 if (!status)
8253                         return 0;
8254         }
8255
8256         /* try SP4_NONE */
8257         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8258 }
8259
8260 /**
8261  * nfs4_test_session_trunk
8262  *
8263  * This is an add_xprt_test() test function called from
8264  * rpc_clnt_setup_test_and_add_xprt.
8265  *
8266  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8267  * and is dereferrenced in nfs4_exchange_id_release
8268  *
8269  * Upon success, add the new transport to the rpc_clnt
8270  *
8271  * @clnt: struct rpc_clnt to get new transport
8272  * @xprt: the rpc_xprt to test
8273  * @data: call data for _nfs4_proc_exchange_id.
8274  */
8275 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8276                             void *data)
8277 {
8278         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8279         struct rpc_task *task;
8280         int status;
8281
8282         u32 sp4_how;
8283
8284         dprintk("--> %s try %s\n", __func__,
8285                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8286
8287         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8288
8289         /* Test connection for session trunking. Async exchange_id call */
8290         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8291         if (IS_ERR(task))
8292                 return;
8293
8294         status = task->tk_status;
8295         if (status == 0)
8296                 status = nfs4_detect_session_trunking(adata->clp,
8297                                 task->tk_msg.rpc_resp, xprt);
8298
8299         if (status == 0)
8300                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8301
8302         rpc_put_task(task);
8303 }
8304 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8305
8306 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8307                 const struct cred *cred)
8308 {
8309         struct rpc_message msg = {
8310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8311                 .rpc_argp = clp,
8312                 .rpc_cred = cred,
8313         };
8314         int status;
8315
8316         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8317                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8318         trace_nfs4_destroy_clientid(clp, status);
8319         if (status)
8320                 dprintk("NFS: Got error %d from the server %s on "
8321                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8322         return status;
8323 }
8324
8325 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8326                 const struct cred *cred)
8327 {
8328         unsigned int loop;
8329         int ret;
8330
8331         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8332                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8333                 switch (ret) {
8334                 case -NFS4ERR_DELAY:
8335                 case -NFS4ERR_CLIENTID_BUSY:
8336                         ssleep(1);
8337                         break;
8338                 default:
8339                         return ret;
8340                 }
8341         }
8342         return 0;
8343 }
8344
8345 int nfs4_destroy_clientid(struct nfs_client *clp)
8346 {
8347         const struct cred *cred;
8348         int ret = 0;
8349
8350         if (clp->cl_mvops->minor_version < 1)
8351                 goto out;
8352         if (clp->cl_exchange_flags == 0)
8353                 goto out;
8354         if (clp->cl_preserve_clid)
8355                 goto out;
8356         cred = nfs4_get_clid_cred(clp);
8357         ret = nfs4_proc_destroy_clientid(clp, cred);
8358         put_cred(cred);
8359         switch (ret) {
8360         case 0:
8361         case -NFS4ERR_STALE_CLIENTID:
8362                 clp->cl_exchange_flags = 0;
8363         }
8364 out:
8365         return ret;
8366 }
8367
8368 #endif /* CONFIG_NFS_V4_1 */
8369
8370 struct nfs4_get_lease_time_data {
8371         struct nfs4_get_lease_time_args *args;
8372         struct nfs4_get_lease_time_res *res;
8373         struct nfs_client *clp;
8374 };
8375
8376 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8377                                         void *calldata)
8378 {
8379         struct nfs4_get_lease_time_data *data =
8380                         (struct nfs4_get_lease_time_data *)calldata;
8381
8382         dprintk("--> %s\n", __func__);
8383         /* just setup sequence, do not trigger session recovery
8384            since we're invoked within one */
8385         nfs4_setup_sequence(data->clp,
8386                         &data->args->la_seq_args,
8387                         &data->res->lr_seq_res,
8388                         task);
8389         dprintk("<-- %s\n", __func__);
8390 }
8391
8392 /*
8393  * Called from nfs4_state_manager thread for session setup, so don't recover
8394  * from sequence operation or clientid errors.
8395  */
8396 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8397 {
8398         struct nfs4_get_lease_time_data *data =
8399                         (struct nfs4_get_lease_time_data *)calldata;
8400
8401         dprintk("--> %s\n", __func__);
8402         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8403                 return;
8404         switch (task->tk_status) {
8405         case -NFS4ERR_DELAY:
8406         case -NFS4ERR_GRACE:
8407                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8408                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8409                 task->tk_status = 0;
8410                 /* fall through */
8411         case -NFS4ERR_RETRY_UNCACHED_REP:
8412                 rpc_restart_call_prepare(task);
8413                 return;
8414         }
8415         dprintk("<-- %s\n", __func__);
8416 }
8417
8418 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8419         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8420         .rpc_call_done = nfs4_get_lease_time_done,
8421 };
8422
8423 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8424 {
8425         struct nfs4_get_lease_time_args args;
8426         struct nfs4_get_lease_time_res res = {
8427                 .lr_fsinfo = fsinfo,
8428         };
8429         struct nfs4_get_lease_time_data data = {
8430                 .args = &args,
8431                 .res = &res,
8432                 .clp = clp,
8433         };
8434         struct rpc_message msg = {
8435                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8436                 .rpc_argp = &args,
8437                 .rpc_resp = &res,
8438         };
8439         struct rpc_task_setup task_setup = {
8440                 .rpc_client = clp->cl_rpcclient,
8441                 .rpc_message = &msg,
8442                 .callback_ops = &nfs4_get_lease_time_ops,
8443                 .callback_data = &data,
8444                 .flags = RPC_TASK_TIMEOUT,
8445         };
8446
8447         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8448         return nfs4_call_sync_custom(&task_setup);
8449 }
8450
8451 #ifdef CONFIG_NFS_V4_1
8452
8453 /*
8454  * Initialize the values to be used by the client in CREATE_SESSION
8455  * If nfs4_init_session set the fore channel request and response sizes,
8456  * use them.
8457  *
8458  * Set the back channel max_resp_sz_cached to zero to force the client to
8459  * always set csa_cachethis to FALSE because the current implementation
8460  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8461  */
8462 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8463                                     struct rpc_clnt *clnt)
8464 {
8465         unsigned int max_rqst_sz, max_resp_sz;
8466         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8467         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8468
8469         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8470         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8471
8472         /* Fore channel attributes */
8473         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8474         args->fc_attrs.max_resp_sz = max_resp_sz;
8475         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8476         args->fc_attrs.max_reqs = max_session_slots;
8477
8478         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8479                 "max_ops=%u max_reqs=%u\n",
8480                 __func__,
8481                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8482                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8483
8484         /* Back channel attributes */
8485         args->bc_attrs.max_rqst_sz = max_bc_payload;
8486         args->bc_attrs.max_resp_sz = max_bc_payload;
8487         args->bc_attrs.max_resp_sz_cached = 0;
8488         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8489         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8490         if (args->bc_attrs.max_reqs > max_bc_slots)
8491                 args->bc_attrs.max_reqs = max_bc_slots;
8492
8493         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8494                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8495                 __func__,
8496                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8497                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8498                 args->bc_attrs.max_reqs);
8499 }
8500
8501 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8502                 struct nfs41_create_session_res *res)
8503 {
8504         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8505         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8506
8507         if (rcvd->max_resp_sz > sent->max_resp_sz)
8508                 return -EINVAL;
8509         /*
8510          * Our requested max_ops is the minimum we need; we're not
8511          * prepared to break up compounds into smaller pieces than that.
8512          * So, no point even trying to continue if the server won't
8513          * cooperate:
8514          */
8515         if (rcvd->max_ops < sent->max_ops)
8516                 return -EINVAL;
8517         if (rcvd->max_reqs == 0)
8518                 return -EINVAL;
8519         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8520                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8521         return 0;
8522 }
8523
8524 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8525                 struct nfs41_create_session_res *res)
8526 {
8527         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8528         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8529
8530         if (!(res->flags & SESSION4_BACK_CHAN))
8531                 goto out;
8532         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8533                 return -EINVAL;
8534         if (rcvd->max_resp_sz < sent->max_resp_sz)
8535                 return -EINVAL;
8536         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8537                 return -EINVAL;
8538         if (rcvd->max_ops > sent->max_ops)
8539                 return -EINVAL;
8540         if (rcvd->max_reqs > sent->max_reqs)
8541                 return -EINVAL;
8542 out:
8543         return 0;
8544 }
8545
8546 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8547                                      struct nfs41_create_session_res *res)
8548 {
8549         int ret;
8550
8551         ret = nfs4_verify_fore_channel_attrs(args, res);
8552         if (ret)
8553                 return ret;
8554         return nfs4_verify_back_channel_attrs(args, res);
8555 }
8556
8557 static void nfs4_update_session(struct nfs4_session *session,
8558                 struct nfs41_create_session_res *res)
8559 {
8560         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8561         /* Mark client id and session as being confirmed */
8562         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8563         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8564         session->flags = res->flags;
8565         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8566         if (res->flags & SESSION4_BACK_CHAN)
8567                 memcpy(&session->bc_attrs, &res->bc_attrs,
8568                                 sizeof(session->bc_attrs));
8569 }
8570
8571 static int _nfs4_proc_create_session(struct nfs_client *clp,
8572                 const struct cred *cred)
8573 {
8574         struct nfs4_session *session = clp->cl_session;
8575         struct nfs41_create_session_args args = {
8576                 .client = clp,
8577                 .clientid = clp->cl_clientid,
8578                 .seqid = clp->cl_seqid,
8579                 .cb_program = NFS4_CALLBACK,
8580         };
8581         struct nfs41_create_session_res res;
8582
8583         struct rpc_message msg = {
8584                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8585                 .rpc_argp = &args,
8586                 .rpc_resp = &res,
8587                 .rpc_cred = cred,
8588         };
8589         int status;
8590
8591         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8592         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8593
8594         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8595                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8596         trace_nfs4_create_session(clp, status);
8597
8598         switch (status) {
8599         case -NFS4ERR_STALE_CLIENTID:
8600         case -NFS4ERR_DELAY:
8601         case -ETIMEDOUT:
8602         case -EACCES:
8603         case -EAGAIN:
8604                 goto out;
8605         };
8606
8607         clp->cl_seqid++;
8608         if (!status) {
8609                 /* Verify the session's negotiated channel_attrs values */
8610                 status = nfs4_verify_channel_attrs(&args, &res);
8611                 /* Increment the clientid slot sequence id */
8612                 if (status)
8613                         goto out;
8614                 nfs4_update_session(session, &res);
8615         }
8616 out:
8617         return status;
8618 }
8619
8620 /*
8621  * Issues a CREATE_SESSION operation to the server.
8622  * It is the responsibility of the caller to verify the session is
8623  * expired before calling this routine.
8624  */
8625 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8626 {
8627         int status;
8628         unsigned *ptr;
8629         struct nfs4_session *session = clp->cl_session;
8630
8631         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8632
8633         status = _nfs4_proc_create_session(clp, cred);
8634         if (status)
8635                 goto out;
8636
8637         /* Init or reset the session slot tables */
8638         status = nfs4_setup_session_slot_tables(session);
8639         dprintk("slot table setup returned %d\n", status);
8640         if (status)
8641                 goto out;
8642
8643         ptr = (unsigned *)&session->sess_id.data[0];
8644         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8645                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8646 out:
8647         dprintk("<-- %s\n", __func__);
8648         return status;
8649 }
8650
8651 /*
8652  * Issue the over-the-wire RPC DESTROY_SESSION.
8653  * The caller must serialize access to this routine.
8654  */
8655 int nfs4_proc_destroy_session(struct nfs4_session *session,
8656                 const struct cred *cred)
8657 {
8658         struct rpc_message msg = {
8659                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8660                 .rpc_argp = session,
8661                 .rpc_cred = cred,
8662         };
8663         int status = 0;
8664
8665         dprintk("--> nfs4_proc_destroy_session\n");
8666
8667         /* session is still being setup */
8668         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8669                 return 0;
8670
8671         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8672                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8673         trace_nfs4_destroy_session(session->clp, status);
8674
8675         if (status)
8676                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8677                         "Session has been destroyed regardless...\n", status);
8678
8679         dprintk("<-- nfs4_proc_destroy_session\n");
8680         return status;
8681 }
8682
8683 /*
8684  * Renew the cl_session lease.
8685  */
8686 struct nfs4_sequence_data {
8687         struct nfs_client *clp;
8688         struct nfs4_sequence_args args;
8689         struct nfs4_sequence_res res;
8690 };
8691
8692 static void nfs41_sequence_release(void *data)
8693 {
8694         struct nfs4_sequence_data *calldata = data;
8695         struct nfs_client *clp = calldata->clp;
8696
8697         if (refcount_read(&clp->cl_count) > 1)
8698                 nfs4_schedule_state_renewal(clp);
8699         nfs_put_client(clp);
8700         kfree(calldata);
8701 }
8702
8703 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8704 {
8705         switch(task->tk_status) {
8706         case -NFS4ERR_DELAY:
8707                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8708                 return -EAGAIN;
8709         default:
8710                 nfs4_schedule_lease_recovery(clp);
8711         }
8712         return 0;
8713 }
8714
8715 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8716 {
8717         struct nfs4_sequence_data *calldata = data;
8718         struct nfs_client *clp = calldata->clp;
8719
8720         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8721                 return;
8722
8723         trace_nfs4_sequence(clp, task->tk_status);
8724         if (task->tk_status < 0) {
8725                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8726                 if (refcount_read(&clp->cl_count) == 1)
8727                         goto out;
8728
8729                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8730                         rpc_restart_call_prepare(task);
8731                         return;
8732                 }
8733         }
8734         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8735 out:
8736         dprintk("<-- %s\n", __func__);
8737 }
8738
8739 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8740 {
8741         struct nfs4_sequence_data *calldata = data;
8742         struct nfs_client *clp = calldata->clp;
8743         struct nfs4_sequence_args *args;
8744         struct nfs4_sequence_res *res;
8745
8746         args = task->tk_msg.rpc_argp;
8747         res = task->tk_msg.rpc_resp;
8748
8749         nfs4_setup_sequence(clp, args, res, task);
8750 }
8751
8752 static const struct rpc_call_ops nfs41_sequence_ops = {
8753         .rpc_call_done = nfs41_sequence_call_done,
8754         .rpc_call_prepare = nfs41_sequence_prepare,
8755         .rpc_release = nfs41_sequence_release,
8756 };
8757
8758 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8759                 const struct cred *cred,
8760                 struct nfs4_slot *slot,
8761                 bool is_privileged)
8762 {
8763         struct nfs4_sequence_data *calldata;
8764         struct rpc_message msg = {
8765                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8766                 .rpc_cred = cred,
8767         };
8768         struct rpc_task_setup task_setup_data = {
8769                 .rpc_client = clp->cl_rpcclient,
8770                 .rpc_message = &msg,
8771                 .callback_ops = &nfs41_sequence_ops,
8772                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8773         };
8774         struct rpc_task *ret;
8775
8776         ret = ERR_PTR(-EIO);
8777         if (!refcount_inc_not_zero(&clp->cl_count))
8778                 goto out_err;
8779
8780         ret = ERR_PTR(-ENOMEM);
8781         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8782         if (calldata == NULL)
8783                 goto out_put_clp;
8784         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8785         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8786         msg.rpc_argp = &calldata->args;
8787         msg.rpc_resp = &calldata->res;
8788         calldata->clp = clp;
8789         task_setup_data.callback_data = calldata;
8790
8791         ret = rpc_run_task(&task_setup_data);
8792         if (IS_ERR(ret))
8793                 goto out_err;
8794         return ret;
8795 out_put_clp:
8796         nfs_put_client(clp);
8797 out_err:
8798         nfs41_release_slot(slot);
8799         return ret;
8800 }
8801
8802 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8803 {
8804         struct rpc_task *task;
8805         int ret = 0;
8806
8807         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8808                 return -EAGAIN;
8809         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8810         if (IS_ERR(task))
8811                 ret = PTR_ERR(task);
8812         else
8813                 rpc_put_task_async(task);
8814         dprintk("<-- %s status=%d\n", __func__, ret);
8815         return ret;
8816 }
8817
8818 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8819 {
8820         struct rpc_task *task;
8821         int ret;
8822
8823         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8824         if (IS_ERR(task)) {
8825                 ret = PTR_ERR(task);
8826                 goto out;
8827         }
8828         ret = rpc_wait_for_completion_task(task);
8829         if (!ret)
8830                 ret = task->tk_status;
8831         rpc_put_task(task);
8832 out:
8833         dprintk("<-- %s status=%d\n", __func__, ret);
8834         return ret;
8835 }
8836
8837 struct nfs4_reclaim_complete_data {
8838         struct nfs_client *clp;
8839         struct nfs41_reclaim_complete_args arg;
8840         struct nfs41_reclaim_complete_res res;
8841 };
8842
8843 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8844 {
8845         struct nfs4_reclaim_complete_data *calldata = data;
8846
8847         nfs4_setup_sequence(calldata->clp,
8848                         &calldata->arg.seq_args,
8849                         &calldata->res.seq_res,
8850                         task);
8851 }
8852
8853 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8854 {
8855         switch(task->tk_status) {
8856         case 0:
8857                 wake_up_all(&clp->cl_lock_waitq);
8858                 /* Fallthrough */
8859         case -NFS4ERR_COMPLETE_ALREADY:
8860         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8861                 break;
8862         case -NFS4ERR_DELAY:
8863                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8864                 /* fall through */
8865         case -NFS4ERR_RETRY_UNCACHED_REP:
8866                 return -EAGAIN;
8867         case -NFS4ERR_BADSESSION:
8868         case -NFS4ERR_DEADSESSION:
8869         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8870                 nfs4_schedule_session_recovery(clp->cl_session,
8871                                 task->tk_status);
8872                 break;
8873         default:
8874                 nfs4_schedule_lease_recovery(clp);
8875         }
8876         return 0;
8877 }
8878
8879 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8880 {
8881         struct nfs4_reclaim_complete_data *calldata = data;
8882         struct nfs_client *clp = calldata->clp;
8883         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8884
8885         dprintk("--> %s\n", __func__);
8886         if (!nfs41_sequence_done(task, res))
8887                 return;
8888
8889         trace_nfs4_reclaim_complete(clp, task->tk_status);
8890         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8891                 rpc_restart_call_prepare(task);
8892                 return;
8893         }
8894         dprintk("<-- %s\n", __func__);
8895 }
8896
8897 static void nfs4_free_reclaim_complete_data(void *data)
8898 {
8899         struct nfs4_reclaim_complete_data *calldata = data;
8900
8901         kfree(calldata);
8902 }
8903
8904 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8905         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8906         .rpc_call_done = nfs4_reclaim_complete_done,
8907         .rpc_release = nfs4_free_reclaim_complete_data,
8908 };
8909
8910 /*
8911  * Issue a global reclaim complete.
8912  */
8913 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8914                 const struct cred *cred)
8915 {
8916         struct nfs4_reclaim_complete_data *calldata;
8917         struct rpc_message msg = {
8918                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8919                 .rpc_cred = cred,
8920         };
8921         struct rpc_task_setup task_setup_data = {
8922                 .rpc_client = clp->cl_rpcclient,
8923                 .rpc_message = &msg,
8924                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8925                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8926         };
8927         int status = -ENOMEM;
8928
8929         dprintk("--> %s\n", __func__);
8930         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8931         if (calldata == NULL)
8932                 goto out;
8933         calldata->clp = clp;
8934         calldata->arg.one_fs = 0;
8935
8936         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8937         msg.rpc_argp = &calldata->arg;
8938         msg.rpc_resp = &calldata->res;
8939         task_setup_data.callback_data = calldata;
8940         status = nfs4_call_sync_custom(&task_setup_data);
8941 out:
8942         dprintk("<-- %s status=%d\n", __func__, status);
8943         return status;
8944 }
8945
8946 static void
8947 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8948 {
8949         struct nfs4_layoutget *lgp = calldata;
8950         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8951
8952         dprintk("--> %s\n", __func__);
8953         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8954                                 &lgp->res.seq_res, task);
8955         dprintk("<-- %s\n", __func__);
8956 }
8957
8958 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8959 {
8960         struct nfs4_layoutget *lgp = calldata;
8961
8962         dprintk("--> %s\n", __func__);
8963         nfs41_sequence_process(task, &lgp->res.seq_res);
8964         dprintk("<-- %s\n", __func__);
8965 }
8966
8967 static int
8968 nfs4_layoutget_handle_exception(struct rpc_task *task,
8969                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8970 {
8971         struct inode *inode = lgp->args.inode;
8972         struct nfs_server *server = NFS_SERVER(inode);
8973         struct pnfs_layout_hdr *lo;
8974         int nfs4err = task->tk_status;
8975         int err, status = 0;
8976         LIST_HEAD(head);
8977
8978         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8979
8980         nfs4_sequence_free_slot(&lgp->res.seq_res);
8981
8982         switch (nfs4err) {
8983         case 0:
8984                 goto out;
8985
8986         /*
8987          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8988          * on the file. set tk_status to -ENODATA to tell upper layer to
8989          * retry go inband.
8990          */
8991         case -NFS4ERR_LAYOUTUNAVAILABLE:
8992                 status = -ENODATA;
8993                 goto out;
8994         /*
8995          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8996          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8997          */
8998         case -NFS4ERR_BADLAYOUT:
8999                 status = -EOVERFLOW;
9000                 goto out;
9001         /*
9002          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9003          * (or clients) writing to the same RAID stripe except when
9004          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9005          *
9006          * Treat it like we would RECALLCONFLICT -- we retry for a little
9007          * while, and then eventually give up.
9008          */
9009         case -NFS4ERR_LAYOUTTRYLATER:
9010                 if (lgp->args.minlength == 0) {
9011                         status = -EOVERFLOW;
9012                         goto out;
9013                 }
9014                 status = -EBUSY;
9015                 break;
9016         case -NFS4ERR_RECALLCONFLICT:
9017                 status = -ERECALLCONFLICT;
9018                 break;
9019         case -NFS4ERR_DELEG_REVOKED:
9020         case -NFS4ERR_ADMIN_REVOKED:
9021         case -NFS4ERR_EXPIRED:
9022         case -NFS4ERR_BAD_STATEID:
9023                 exception->timeout = 0;
9024                 spin_lock(&inode->i_lock);
9025                 lo = NFS_I(inode)->layout;
9026                 /* If the open stateid was bad, then recover it. */
9027                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9028                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9029                         spin_unlock(&inode->i_lock);
9030                         exception->state = lgp->args.ctx->state;
9031                         exception->stateid = &lgp->args.stateid;
9032                         break;
9033                 }
9034
9035                 /*
9036                  * Mark the bad layout state as invalid, then retry
9037                  */
9038                 pnfs_mark_layout_stateid_invalid(lo, &head);
9039                 spin_unlock(&inode->i_lock);
9040                 nfs_commit_inode(inode, 0);
9041                 pnfs_free_lseg_list(&head);
9042                 status = -EAGAIN;
9043                 goto out;
9044         }
9045
9046         err = nfs4_handle_exception(server, nfs4err, exception);
9047         if (!status) {
9048                 if (exception->retry)
9049                         status = -EAGAIN;
9050                 else
9051                         status = err;
9052         }
9053 out:
9054         dprintk("<-- %s\n", __func__);
9055         return status;
9056 }
9057
9058 size_t max_response_pages(struct nfs_server *server)
9059 {
9060         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9061         return nfs_page_array_len(0, max_resp_sz);
9062 }
9063
9064 static void nfs4_layoutget_release(void *calldata)
9065 {
9066         struct nfs4_layoutget *lgp = calldata;
9067
9068         dprintk("--> %s\n", __func__);
9069         nfs4_sequence_free_slot(&lgp->res.seq_res);
9070         pnfs_layoutget_free(lgp);
9071         dprintk("<-- %s\n", __func__);
9072 }
9073
9074 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9075         .rpc_call_prepare = nfs4_layoutget_prepare,
9076         .rpc_call_done = nfs4_layoutget_done,
9077         .rpc_release = nfs4_layoutget_release,
9078 };
9079
9080 struct pnfs_layout_segment *
9081 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9082 {
9083         struct inode *inode = lgp->args.inode;
9084         struct nfs_server *server = NFS_SERVER(inode);
9085         struct rpc_task *task;
9086         struct rpc_message msg = {
9087                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9088                 .rpc_argp = &lgp->args,
9089                 .rpc_resp = &lgp->res,
9090                 .rpc_cred = lgp->cred,
9091         };
9092         struct rpc_task_setup task_setup_data = {
9093                 .rpc_client = server->client,
9094                 .rpc_message = &msg,
9095                 .callback_ops = &nfs4_layoutget_call_ops,
9096                 .callback_data = lgp,
9097                 .flags = RPC_TASK_ASYNC,
9098         };
9099         struct pnfs_layout_segment *lseg = NULL;
9100         struct nfs4_exception exception = {
9101                 .inode = inode,
9102                 .timeout = *timeout,
9103         };
9104         int status = 0;
9105
9106         dprintk("--> %s\n", __func__);
9107
9108         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9109         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9110
9111         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9112
9113         task = rpc_run_task(&task_setup_data);
9114         if (IS_ERR(task))
9115                 return ERR_CAST(task);
9116         status = rpc_wait_for_completion_task(task);
9117         if (status != 0)
9118                 goto out;
9119
9120         if (task->tk_status < 0) {
9121                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9122                 *timeout = exception.timeout;
9123         } else if (lgp->res.layoutp->len == 0) {
9124                 status = -EAGAIN;
9125                 *timeout = nfs4_update_delay(&exception.timeout);
9126         } else
9127                 lseg = pnfs_layout_process(lgp);
9128 out:
9129         trace_nfs4_layoutget(lgp->args.ctx,
9130                         &lgp->args.range,
9131                         &lgp->res.range,
9132                         &lgp->res.stateid,
9133                         status);
9134
9135         rpc_put_task(task);
9136         dprintk("<-- %s status=%d\n", __func__, status);
9137         if (status)
9138                 return ERR_PTR(status);
9139         return lseg;
9140 }
9141
9142 static void
9143 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9144 {
9145         struct nfs4_layoutreturn *lrp = calldata;
9146
9147         dprintk("--> %s\n", __func__);
9148         nfs4_setup_sequence(lrp->clp,
9149                         &lrp->args.seq_args,
9150                         &lrp->res.seq_res,
9151                         task);
9152         if (!pnfs_layout_is_valid(lrp->args.layout))
9153                 rpc_exit(task, 0);
9154 }
9155
9156 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9157 {
9158         struct nfs4_layoutreturn *lrp = calldata;
9159         struct nfs_server *server;
9160
9161         dprintk("--> %s\n", __func__);
9162
9163         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9164                 return;
9165
9166         /*
9167          * Was there an RPC level error? Assume the call succeeded,
9168          * and that we need to release the layout
9169          */
9170         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9171                 lrp->res.lrs_present = 0;
9172                 return;
9173         }
9174
9175         server = NFS_SERVER(lrp->args.inode);
9176         switch (task->tk_status) {
9177         case -NFS4ERR_OLD_STATEID:
9178                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9179                                         &lrp->args.range,
9180                                         lrp->args.inode))
9181                         goto out_restart;
9182                 /* Fallthrough */
9183         default:
9184                 task->tk_status = 0;
9185                 /* Fallthrough */
9186         case 0:
9187                 break;
9188         case -NFS4ERR_DELAY:
9189                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9190                         break;
9191                 goto out_restart;
9192         }
9193         dprintk("<-- %s\n", __func__);
9194         return;
9195 out_restart:
9196         task->tk_status = 0;
9197         nfs4_sequence_free_slot(&lrp->res.seq_res);
9198         rpc_restart_call_prepare(task);
9199 }
9200
9201 static void nfs4_layoutreturn_release(void *calldata)
9202 {
9203         struct nfs4_layoutreturn *lrp = calldata;
9204         struct pnfs_layout_hdr *lo = lrp->args.layout;
9205
9206         dprintk("--> %s\n", __func__);
9207         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9208                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9209         nfs4_sequence_free_slot(&lrp->res.seq_res);
9210         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9211                 lrp->ld_private.ops->free(&lrp->ld_private);
9212         pnfs_put_layout_hdr(lrp->args.layout);
9213         nfs_iput_and_deactive(lrp->inode);
9214         kfree(calldata);
9215         dprintk("<-- %s\n", __func__);
9216 }
9217
9218 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9219         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9220         .rpc_call_done = nfs4_layoutreturn_done,
9221         .rpc_release = nfs4_layoutreturn_release,
9222 };
9223
9224 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9225 {
9226         struct rpc_task *task;
9227         struct rpc_message msg = {
9228                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9229                 .rpc_argp = &lrp->args,
9230                 .rpc_resp = &lrp->res,
9231                 .rpc_cred = lrp->cred,
9232         };
9233         struct rpc_task_setup task_setup_data = {
9234                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9235                 .rpc_message = &msg,
9236                 .callback_ops = &nfs4_layoutreturn_call_ops,
9237                 .callback_data = lrp,
9238         };
9239         int status = 0;
9240
9241         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9242                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9243                         &task_setup_data.rpc_client, &msg);
9244
9245         dprintk("--> %s\n", __func__);
9246         if (!sync) {
9247                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9248                 if (!lrp->inode) {
9249                         nfs4_layoutreturn_release(lrp);
9250                         return -EAGAIN;
9251                 }
9252                 task_setup_data.flags |= RPC_TASK_ASYNC;
9253         }
9254         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9255         task = rpc_run_task(&task_setup_data);
9256         if (IS_ERR(task))
9257                 return PTR_ERR(task);
9258         if (sync)
9259                 status = task->tk_status;
9260         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9261         dprintk("<-- %s status=%d\n", __func__, status);
9262         rpc_put_task(task);
9263         return status;
9264 }
9265
9266 static int
9267 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9268                 struct pnfs_device *pdev,
9269                 const struct cred *cred)
9270 {
9271         struct nfs4_getdeviceinfo_args args = {
9272                 .pdev = pdev,
9273                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9274                         NOTIFY_DEVICEID4_DELETE,
9275         };
9276         struct nfs4_getdeviceinfo_res res = {
9277                 .pdev = pdev,
9278         };
9279         struct rpc_message msg = {
9280                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9281                 .rpc_argp = &args,
9282                 .rpc_resp = &res,
9283                 .rpc_cred = cred,
9284         };
9285         int status;
9286
9287         dprintk("--> %s\n", __func__);
9288         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9289         if (res.notification & ~args.notify_types)
9290                 dprintk("%s: unsupported notification\n", __func__);
9291         if (res.notification != args.notify_types)
9292                 pdev->nocache = 1;
9293
9294         dprintk("<-- %s status=%d\n", __func__, status);
9295
9296         return status;
9297 }
9298
9299 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9300                 struct pnfs_device *pdev,
9301                 const struct cred *cred)
9302 {
9303         struct nfs4_exception exception = { };
9304         int err;
9305
9306         do {
9307                 err = nfs4_handle_exception(server,
9308                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9309                                         &exception);
9310         } while (exception.retry);
9311         return err;
9312 }
9313 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9314
9315 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9316 {
9317         struct nfs4_layoutcommit_data *data = calldata;
9318         struct nfs_server *server = NFS_SERVER(data->args.inode);
9319
9320         nfs4_setup_sequence(server->nfs_client,
9321                         &data->args.seq_args,
9322                         &data->res.seq_res,
9323                         task);
9324 }
9325
9326 static void
9327 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9328 {
9329         struct nfs4_layoutcommit_data *data = calldata;
9330         struct nfs_server *server = NFS_SERVER(data->args.inode);
9331
9332         if (!nfs41_sequence_done(task, &data->res.seq_res))
9333                 return;
9334
9335         switch (task->tk_status) { /* Just ignore these failures */
9336         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9337         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9338         case -NFS4ERR_BADLAYOUT:     /* no layout */
9339         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9340                 task->tk_status = 0;
9341         case 0:
9342                 break;
9343         default:
9344                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9345                         rpc_restart_call_prepare(task);
9346                         return;
9347                 }
9348         }
9349 }
9350
9351 static void nfs4_layoutcommit_release(void *calldata)
9352 {
9353         struct nfs4_layoutcommit_data *data = calldata;
9354
9355         pnfs_cleanup_layoutcommit(data);
9356         nfs_post_op_update_inode_force_wcc(data->args.inode,
9357                                            data->res.fattr);
9358         put_cred(data->cred);
9359         nfs_iput_and_deactive(data->inode);
9360         kfree(data);
9361 }
9362
9363 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9364         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9365         .rpc_call_done = nfs4_layoutcommit_done,
9366         .rpc_release = nfs4_layoutcommit_release,
9367 };
9368
9369 int
9370 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9371 {
9372         struct rpc_message msg = {
9373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9374                 .rpc_argp = &data->args,
9375                 .rpc_resp = &data->res,
9376                 .rpc_cred = data->cred,
9377         };
9378         struct rpc_task_setup task_setup_data = {
9379                 .task = &data->task,
9380                 .rpc_client = NFS_CLIENT(data->args.inode),
9381                 .rpc_message = &msg,
9382                 .callback_ops = &nfs4_layoutcommit_ops,
9383                 .callback_data = data,
9384         };
9385         struct rpc_task *task;
9386         int status = 0;
9387
9388         dprintk("NFS: initiating layoutcommit call. sync %d "
9389                 "lbw: %llu inode %lu\n", sync,
9390                 data->args.lastbytewritten,
9391                 data->args.inode->i_ino);
9392
9393         if (!sync) {
9394                 data->inode = nfs_igrab_and_active(data->args.inode);
9395                 if (data->inode == NULL) {
9396                         nfs4_layoutcommit_release(data);
9397                         return -EAGAIN;
9398                 }
9399                 task_setup_data.flags = RPC_TASK_ASYNC;
9400         }
9401         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9402         task = rpc_run_task(&task_setup_data);
9403         if (IS_ERR(task))
9404                 return PTR_ERR(task);
9405         if (sync)
9406                 status = task->tk_status;
9407         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9408         dprintk("%s: status %d\n", __func__, status);
9409         rpc_put_task(task);
9410         return status;
9411 }
9412
9413 /*
9414  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9415  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9416  */
9417 static int
9418 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9419                     struct nfs_fsinfo *info,
9420                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9421 {
9422         struct nfs41_secinfo_no_name_args args = {
9423                 .style = SECINFO_STYLE_CURRENT_FH,
9424         };
9425         struct nfs4_secinfo_res res = {
9426                 .flavors = flavors,
9427         };
9428         struct rpc_message msg = {
9429                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9430                 .rpc_argp = &args,
9431                 .rpc_resp = &res,
9432         };
9433         struct rpc_clnt *clnt = server->client;
9434         struct nfs4_call_sync_data data = {
9435                 .seq_server = server,
9436                 .seq_args = &args.seq_args,
9437                 .seq_res = &res.seq_res,
9438         };
9439         struct rpc_task_setup task_setup = {
9440                 .rpc_client = server->client,
9441                 .rpc_message = &msg,
9442                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9443                 .callback_data = &data,
9444                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9445         };
9446         const struct cred *cred = NULL;
9447         int status;
9448
9449         if (use_integrity) {
9450                 clnt = server->nfs_client->cl_rpcclient;
9451                 task_setup.rpc_client = clnt;
9452
9453                 cred = nfs4_get_clid_cred(server->nfs_client);
9454                 msg.rpc_cred = cred;
9455         }
9456
9457         dprintk("--> %s\n", __func__);
9458         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9459         status = nfs4_call_sync_custom(&task_setup);
9460         dprintk("<-- %s status=%d\n", __func__, status);
9461
9462         put_cred(cred);
9463
9464         return status;
9465 }
9466
9467 static int
9468 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9469                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9470 {
9471         struct nfs4_exception exception = {
9472                 .interruptible = true,
9473         };
9474         int err;
9475         do {
9476                 /* first try using integrity protection */
9477                 err = -NFS4ERR_WRONGSEC;
9478
9479                 /* try to use integrity protection with machine cred */
9480                 if (_nfs4_is_integrity_protected(server->nfs_client))
9481                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9482                                                           flavors, true);
9483
9484                 /*
9485                  * if unable to use integrity protection, or SECINFO with
9486                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9487                  * disallowed by spec, but exists in deployed servers) use
9488                  * the current filesystem's rpc_client and the user cred.
9489                  */
9490                 if (err == -NFS4ERR_WRONGSEC)
9491                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9492                                                           flavors, false);
9493
9494                 switch (err) {
9495                 case 0:
9496                 case -NFS4ERR_WRONGSEC:
9497                 case -ENOTSUPP:
9498                         goto out;
9499                 default:
9500                         err = nfs4_handle_exception(server, err, &exception);
9501                 }
9502         } while (exception.retry);
9503 out:
9504         return err;
9505 }
9506
9507 static int
9508 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9509                     struct nfs_fsinfo *info)
9510 {
9511         int err;
9512         struct page *page;
9513         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9514         struct nfs4_secinfo_flavors *flavors;
9515         struct nfs4_secinfo4 *secinfo;
9516         int i;
9517
9518         page = alloc_page(GFP_KERNEL);
9519         if (!page) {
9520                 err = -ENOMEM;
9521                 goto out;
9522         }
9523
9524         flavors = page_address(page);
9525         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9526
9527         /*
9528          * Fall back on "guess and check" method if
9529          * the server doesn't support SECINFO_NO_NAME
9530          */
9531         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9532                 err = nfs4_find_root_sec(server, fhandle, info);
9533                 goto out_freepage;
9534         }
9535         if (err)
9536                 goto out_freepage;
9537
9538         for (i = 0; i < flavors->num_flavors; i++) {
9539                 secinfo = &flavors->flavors[i];
9540
9541                 switch (secinfo->flavor) {
9542                 case RPC_AUTH_NULL:
9543                 case RPC_AUTH_UNIX:
9544                 case RPC_AUTH_GSS:
9545                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9546                                         &secinfo->flavor_info);
9547                         break;
9548                 default:
9549                         flavor = RPC_AUTH_MAXFLAVOR;
9550                         break;
9551                 }
9552
9553                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9554                         flavor = RPC_AUTH_MAXFLAVOR;
9555
9556                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9557                         err = nfs4_lookup_root_sec(server, fhandle,
9558                                                    info, flavor);
9559                         if (!err)
9560                                 break;
9561                 }
9562         }
9563
9564         if (flavor == RPC_AUTH_MAXFLAVOR)
9565                 err = -EPERM;
9566
9567 out_freepage:
9568         put_page(page);
9569         if (err == -EACCES)
9570                 return -EPERM;
9571 out:
9572         return err;
9573 }
9574
9575 static int _nfs41_test_stateid(struct nfs_server *server,
9576                 nfs4_stateid *stateid,
9577                 const struct cred *cred)
9578 {
9579         int status;
9580         struct nfs41_test_stateid_args args = {
9581                 .stateid = stateid,
9582         };
9583         struct nfs41_test_stateid_res res;
9584         struct rpc_message msg = {
9585                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9586                 .rpc_argp = &args,
9587                 .rpc_resp = &res,
9588                 .rpc_cred = cred,
9589         };
9590         struct rpc_clnt *rpc_client = server->client;
9591
9592         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9593                 &rpc_client, &msg);
9594
9595         dprintk("NFS call  test_stateid %p\n", stateid);
9596         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9597         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9598                         &args.seq_args, &res.seq_res);
9599         if (status != NFS_OK) {
9600                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9601                 return status;
9602         }
9603         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9604         return -res.status;
9605 }
9606
9607 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9608                 int err, struct nfs4_exception *exception)
9609 {
9610         exception->retry = 0;
9611         switch(err) {
9612         case -NFS4ERR_DELAY:
9613         case -NFS4ERR_RETRY_UNCACHED_REP:
9614                 nfs4_handle_exception(server, err, exception);
9615                 break;
9616         case -NFS4ERR_BADSESSION:
9617         case -NFS4ERR_BADSLOT:
9618         case -NFS4ERR_BAD_HIGH_SLOT:
9619         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9620         case -NFS4ERR_DEADSESSION:
9621                 nfs4_do_handle_exception(server, err, exception);
9622         }
9623 }
9624
9625 /**
9626  * nfs41_test_stateid - perform a TEST_STATEID operation
9627  *
9628  * @server: server / transport on which to perform the operation
9629  * @stateid: state ID to test
9630  * @cred: credential
9631  *
9632  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9633  * Otherwise a negative NFS4ERR value is returned if the operation
9634  * failed or the state ID is not currently valid.
9635  */
9636 static int nfs41_test_stateid(struct nfs_server *server,
9637                 nfs4_stateid *stateid,
9638                 const struct cred *cred)
9639 {
9640         struct nfs4_exception exception = {
9641                 .interruptible = true,
9642         };
9643         int err;
9644         do {
9645                 err = _nfs41_test_stateid(server, stateid, cred);
9646                 nfs4_handle_delay_or_session_error(server, err, &exception);
9647         } while (exception.retry);
9648         return err;
9649 }
9650
9651 struct nfs_free_stateid_data {
9652         struct nfs_server *server;
9653         struct nfs41_free_stateid_args args;
9654         struct nfs41_free_stateid_res res;
9655 };
9656
9657 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9658 {
9659         struct nfs_free_stateid_data *data = calldata;
9660         nfs4_setup_sequence(data->server->nfs_client,
9661                         &data->args.seq_args,
9662                         &data->res.seq_res,
9663                         task);
9664 }
9665
9666 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9667 {
9668         struct nfs_free_stateid_data *data = calldata;
9669
9670         nfs41_sequence_done(task, &data->res.seq_res);
9671
9672         switch (task->tk_status) {
9673         case -NFS4ERR_DELAY:
9674                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9675                         rpc_restart_call_prepare(task);
9676         }
9677 }
9678
9679 static void nfs41_free_stateid_release(void *calldata)
9680 {
9681         kfree(calldata);
9682 }
9683
9684 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9685         .rpc_call_prepare = nfs41_free_stateid_prepare,
9686         .rpc_call_done = nfs41_free_stateid_done,
9687         .rpc_release = nfs41_free_stateid_release,
9688 };
9689
9690 /**
9691  * nfs41_free_stateid - perform a FREE_STATEID operation
9692  *
9693  * @server: server / transport on which to perform the operation
9694  * @stateid: state ID to release
9695  * @cred: credential
9696  * @privileged: set to true if this call needs to be privileged
9697  *
9698  * Note: this function is always asynchronous.
9699  */
9700 static int nfs41_free_stateid(struct nfs_server *server,
9701                 const nfs4_stateid *stateid,
9702                 const struct cred *cred,
9703                 bool privileged)
9704 {
9705         struct rpc_message msg = {
9706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9707                 .rpc_cred = cred,
9708         };
9709         struct rpc_task_setup task_setup = {
9710                 .rpc_client = server->client,
9711                 .rpc_message = &msg,
9712                 .callback_ops = &nfs41_free_stateid_ops,
9713                 .flags = RPC_TASK_ASYNC,
9714         };
9715         struct nfs_free_stateid_data *data;
9716         struct rpc_task *task;
9717
9718         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9719                 &task_setup.rpc_client, &msg);
9720
9721         dprintk("NFS call  free_stateid %p\n", stateid);
9722         data = kmalloc(sizeof(*data), GFP_NOFS);
9723         if (!data)
9724                 return -ENOMEM;
9725         data->server = server;
9726         nfs4_stateid_copy(&data->args.stateid, stateid);
9727
9728         task_setup.callback_data = data;
9729
9730         msg.rpc_argp = &data->args;
9731         msg.rpc_resp = &data->res;
9732         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9733         task = rpc_run_task(&task_setup);
9734         if (IS_ERR(task))
9735                 return PTR_ERR(task);
9736         rpc_put_task(task);
9737         return 0;
9738 }
9739
9740 static void
9741 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9742 {
9743         const struct cred *cred = lsp->ls_state->owner->so_cred;
9744
9745         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9746         nfs4_free_lock_state(server, lsp);
9747 }
9748
9749 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9750                 const nfs4_stateid *s2)
9751 {
9752         if (s1->type != s2->type)
9753                 return false;
9754
9755         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9756                 return false;
9757
9758         if (s1->seqid == s2->seqid)
9759                 return true;
9760
9761         return s1->seqid == 0 || s2->seqid == 0;
9762 }
9763
9764 #endif /* CONFIG_NFS_V4_1 */
9765
9766 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9767                 const nfs4_stateid *s2)
9768 {
9769         return nfs4_stateid_match(s1, s2);
9770 }
9771
9772
9773 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9774         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9775         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9776         .recover_open   = nfs4_open_reclaim,
9777         .recover_lock   = nfs4_lock_reclaim,
9778         .establish_clid = nfs4_init_clientid,
9779         .detect_trunking = nfs40_discover_server_trunking,
9780 };
9781
9782 #if defined(CONFIG_NFS_V4_1)
9783 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9784         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9785         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9786         .recover_open   = nfs4_open_reclaim,
9787         .recover_lock   = nfs4_lock_reclaim,
9788         .establish_clid = nfs41_init_clientid,
9789         .reclaim_complete = nfs41_proc_reclaim_complete,
9790         .detect_trunking = nfs41_discover_server_trunking,
9791 };
9792 #endif /* CONFIG_NFS_V4_1 */
9793
9794 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9795         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9796         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9797         .recover_open   = nfs40_open_expired,
9798         .recover_lock   = nfs4_lock_expired,
9799         .establish_clid = nfs4_init_clientid,
9800 };
9801
9802 #if defined(CONFIG_NFS_V4_1)
9803 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9804         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9805         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9806         .recover_open   = nfs41_open_expired,
9807         .recover_lock   = nfs41_lock_expired,
9808         .establish_clid = nfs41_init_clientid,
9809 };
9810 #endif /* CONFIG_NFS_V4_1 */
9811
9812 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9813         .sched_state_renewal = nfs4_proc_async_renew,
9814         .get_state_renewal_cred = nfs4_get_renew_cred,
9815         .renew_lease = nfs4_proc_renew,
9816 };
9817
9818 #if defined(CONFIG_NFS_V4_1)
9819 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9820         .sched_state_renewal = nfs41_proc_async_sequence,
9821         .get_state_renewal_cred = nfs4_get_machine_cred,
9822         .renew_lease = nfs4_proc_sequence,
9823 };
9824 #endif
9825
9826 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9827         .get_locations = _nfs40_proc_get_locations,
9828         .fsid_present = _nfs40_proc_fsid_present,
9829 };
9830
9831 #if defined(CONFIG_NFS_V4_1)
9832 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9833         .get_locations = _nfs41_proc_get_locations,
9834         .fsid_present = _nfs41_proc_fsid_present,
9835 };
9836 #endif  /* CONFIG_NFS_V4_1 */
9837
9838 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9839         .minor_version = 0,
9840         .init_caps = NFS_CAP_READDIRPLUS
9841                 | NFS_CAP_ATOMIC_OPEN
9842                 | NFS_CAP_POSIX_LOCK,
9843         .init_client = nfs40_init_client,
9844         .shutdown_client = nfs40_shutdown_client,
9845         .match_stateid = nfs4_match_stateid,
9846         .find_root_sec = nfs4_find_root_sec,
9847         .free_lock_state = nfs4_release_lockowner,
9848         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9849         .alloc_seqid = nfs_alloc_seqid,
9850         .call_sync_ops = &nfs40_call_sync_ops,
9851         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9852         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9853         .state_renewal_ops = &nfs40_state_renewal_ops,
9854         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9855 };
9856
9857 #if defined(CONFIG_NFS_V4_1)
9858 static struct nfs_seqid *
9859 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9860 {
9861         return NULL;
9862 }
9863
9864 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9865         .minor_version = 1,
9866         .init_caps = NFS_CAP_READDIRPLUS
9867                 | NFS_CAP_ATOMIC_OPEN
9868                 | NFS_CAP_POSIX_LOCK
9869                 | NFS_CAP_STATEID_NFSV41
9870                 | NFS_CAP_ATOMIC_OPEN_V1
9871                 | NFS_CAP_LGOPEN,
9872         .init_client = nfs41_init_client,
9873         .shutdown_client = nfs41_shutdown_client,
9874         .match_stateid = nfs41_match_stateid,
9875         .find_root_sec = nfs41_find_root_sec,
9876         .free_lock_state = nfs41_free_lock_state,
9877         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9878         .alloc_seqid = nfs_alloc_no_seqid,
9879         .session_trunk = nfs4_test_session_trunk,
9880         .call_sync_ops = &nfs41_call_sync_ops,
9881         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9882         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9883         .state_renewal_ops = &nfs41_state_renewal_ops,
9884         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9885 };
9886 #endif
9887
9888 #if defined(CONFIG_NFS_V4_2)
9889 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9890         .minor_version = 2,
9891         .init_caps = NFS_CAP_READDIRPLUS
9892                 | NFS_CAP_ATOMIC_OPEN
9893                 | NFS_CAP_POSIX_LOCK
9894                 | NFS_CAP_STATEID_NFSV41
9895                 | NFS_CAP_ATOMIC_OPEN_V1
9896                 | NFS_CAP_LGOPEN
9897                 | NFS_CAP_ALLOCATE
9898                 | NFS_CAP_COPY
9899                 | NFS_CAP_OFFLOAD_CANCEL
9900                 | NFS_CAP_DEALLOCATE
9901                 | NFS_CAP_SEEK
9902                 | NFS_CAP_LAYOUTSTATS
9903                 | NFS_CAP_CLONE
9904                 | NFS_CAP_LAYOUTERROR,
9905         .init_client = nfs41_init_client,
9906         .shutdown_client = nfs41_shutdown_client,
9907         .match_stateid = nfs41_match_stateid,
9908         .find_root_sec = nfs41_find_root_sec,
9909         .free_lock_state = nfs41_free_lock_state,
9910         .call_sync_ops = &nfs41_call_sync_ops,
9911         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9912         .alloc_seqid = nfs_alloc_no_seqid,
9913         .session_trunk = nfs4_test_session_trunk,
9914         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9915         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9916         .state_renewal_ops = &nfs41_state_renewal_ops,
9917         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9918 };
9919 #endif
9920
9921 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9922         [0] = &nfs_v4_0_minor_ops,
9923 #if defined(CONFIG_NFS_V4_1)
9924         [1] = &nfs_v4_1_minor_ops,
9925 #endif
9926 #if defined(CONFIG_NFS_V4_2)
9927         [2] = &nfs_v4_2_minor_ops,
9928 #endif
9929 };
9930
9931 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9932 {
9933         ssize_t error, error2;
9934
9935         error = generic_listxattr(dentry, list, size);
9936         if (error < 0)
9937                 return error;
9938         if (list) {
9939                 list += error;
9940                 size -= error;
9941         }
9942
9943         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9944         if (error2 < 0)
9945                 return error2;
9946         return error + error2;
9947 }
9948
9949 static const struct inode_operations nfs4_dir_inode_operations = {
9950         .create         = nfs_create,
9951         .lookup         = nfs_lookup,
9952         .atomic_open    = nfs_atomic_open,
9953         .link           = nfs_link,
9954         .unlink         = nfs_unlink,
9955         .symlink        = nfs_symlink,
9956         .mkdir          = nfs_mkdir,
9957         .rmdir          = nfs_rmdir,
9958         .mknod          = nfs_mknod,
9959         .rename         = nfs_rename,
9960         .permission     = nfs_permission,
9961         .getattr        = nfs_getattr,
9962         .setattr        = nfs_setattr,
9963         .listxattr      = nfs4_listxattr,
9964 };
9965
9966 static const struct inode_operations nfs4_file_inode_operations = {
9967         .permission     = nfs_permission,
9968         .getattr        = nfs_getattr,
9969         .setattr        = nfs_setattr,
9970         .listxattr      = nfs4_listxattr,
9971 };
9972
9973 const struct nfs_rpc_ops nfs_v4_clientops = {
9974         .version        = 4,                    /* protocol version */
9975         .dentry_ops     = &nfs4_dentry_operations,
9976         .dir_inode_ops  = &nfs4_dir_inode_operations,
9977         .file_inode_ops = &nfs4_file_inode_operations,
9978         .file_ops       = &nfs4_file_operations,
9979         .getroot        = nfs4_proc_get_root,
9980         .submount       = nfs4_submount,
9981         .try_mount      = nfs4_try_mount,
9982         .getattr        = nfs4_proc_getattr,
9983         .setattr        = nfs4_proc_setattr,
9984         .lookup         = nfs4_proc_lookup,
9985         .lookupp        = nfs4_proc_lookupp,
9986         .access         = nfs4_proc_access,
9987         .readlink       = nfs4_proc_readlink,
9988         .create         = nfs4_proc_create,
9989         .remove         = nfs4_proc_remove,
9990         .unlink_setup   = nfs4_proc_unlink_setup,
9991         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9992         .unlink_done    = nfs4_proc_unlink_done,
9993         .rename_setup   = nfs4_proc_rename_setup,
9994         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9995         .rename_done    = nfs4_proc_rename_done,
9996         .link           = nfs4_proc_link,
9997         .symlink        = nfs4_proc_symlink,
9998         .mkdir          = nfs4_proc_mkdir,
9999         .rmdir          = nfs4_proc_rmdir,
10000         .readdir        = nfs4_proc_readdir,
10001         .mknod          = nfs4_proc_mknod,
10002         .statfs         = nfs4_proc_statfs,
10003         .fsinfo         = nfs4_proc_fsinfo,
10004         .pathconf       = nfs4_proc_pathconf,
10005         .set_capabilities = nfs4_server_capabilities,
10006         .decode_dirent  = nfs4_decode_dirent,
10007         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10008         .read_setup     = nfs4_proc_read_setup,
10009         .read_done      = nfs4_read_done,
10010         .write_setup    = nfs4_proc_write_setup,
10011         .write_done     = nfs4_write_done,
10012         .commit_setup   = nfs4_proc_commit_setup,
10013         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10014         .commit_done    = nfs4_commit_done,
10015         .lock           = nfs4_proc_lock,
10016         .clear_acl_cache = nfs4_zap_acl_attr,
10017         .close_context  = nfs4_close_context,
10018         .open_context   = nfs4_atomic_open,
10019         .have_delegation = nfs4_have_delegation,
10020         .alloc_client   = nfs4_alloc_client,
10021         .init_client    = nfs4_init_client,
10022         .free_client    = nfs4_free_client,
10023         .create_server  = nfs4_create_server,
10024         .clone_server   = nfs_clone_server,
10025 };
10026
10027 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10028         .name   = XATTR_NAME_NFSV4_ACL,
10029         .list   = nfs4_xattr_list_nfs4_acl,
10030         .get    = nfs4_xattr_get_nfs4_acl,
10031         .set    = nfs4_xattr_set_nfs4_acl,
10032 };
10033
10034 const struct xattr_handler *nfs4_xattr_handlers[] = {
10035         &nfs4_xattr_nfs4_acl_handler,
10036 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10037         &nfs4_xattr_nfs4_label_handler,
10038 #endif
10039         NULL
10040 };
10041
10042 /*
10043  * Local variables:
10044  *  c-basic-offset: 8
10045  * End:
10046  */