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 *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  95 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 101                 const struct cred *cred,
 102                 struct nfs4_slot *slot,
 103                 bool is_privileged);
 104 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 105                 const struct cred *);
 106 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 107                 const struct cred *, bool);
 108 #endif
 109
 110 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 111 static inline struct nfs4_label *
 112 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 113         struct iattr *sattr, struct nfs4_label *label)
 114 {
 115         int err;
 116
 117         if (label == NULL)
 118                 return NULL;
 119
 120         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 121                 return NULL;
 122
 123         err = security_dentry_init_security(dentry, sattr->ia_mode,
 124                                 &dentry->d_name, (void **)&label->label, &label->len);
 125         if (err == 0)
 126                 return label;
 127
 128         return NULL;
 129 }
 130 static inline void
 131 nfs4_label_release_security(struct nfs4_label *label)
 132 {
 133         if (label)
 134                 security_release_secctx(label->label, label->len);
 135 }
 136 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 137 {
 138         if (label)
 139                 return server->attr_bitmask;
 140
 141         return server->attr_bitmask_nl;
 142 }
 143 #else
 144 static inline struct nfs4_label *
 145 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 146         struct iattr *sattr, struct nfs4_label *l)
 147 { return NULL; }
 148 static inline void
 149 nfs4_label_release_security(struct nfs4_label *label)
 150 { return; }
 151 static inline u32 *
 152 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 153 { return server->attr_bitmask; }
 154 #endif
 155
 156 /* Prevent leaks of NFSv4 errors into userland */
 157 static int nfs4_map_errors(int err)
 158 {
 159         if (err >= -1000)
 160                 return err;
 161         switch (err) {
 162         case -NFS4ERR_RESOURCE:
 163         case -NFS4ERR_LAYOUTTRYLATER:
 164         case -NFS4ERR_RECALLCONFLICT:
 165                 return -EREMOTEIO;
 166         case -NFS4ERR_WRONGSEC:
 167         case -NFS4ERR_WRONG_CRED:
 168                 return -EPERM;
 169         case -NFS4ERR_BADOWNER:
 170         case -NFS4ERR_BADNAME:
 171                 return -EINVAL;
 172         case -NFS4ERR_SHARE_DENIED:
 173                 return -EACCES;
 174         case -NFS4ERR_MINOR_VERS_MISMATCH:
 175                 return -EPROTONOSUPPORT;
 176         case -NFS4ERR_FILE_OPEN:
 177                 return -EBUSY;
 178         default:
 179                 dprintk("%s could not handle NFSv4 error %d\n",
 180                                 __func__, -err);
 181                 break;
 182         }
 183         return -EIO;
 184 }
 185
 186 /*
 187  * This is our standard bitmap for GETATTR requests.
 188  */
 189 const u32 nfs4_fattr_bitmap[3] = {
 190         FATTR4_WORD0_TYPE
 191         | FATTR4_WORD0_CHANGE
 192         | FATTR4_WORD0_SIZE
 193         | FATTR4_WORD0_FSID
 194         | FATTR4_WORD0_FILEID,
 195         FATTR4_WORD1_MODE
 196         | FATTR4_WORD1_NUMLINKS
 197         | FATTR4_WORD1_OWNER
 198         | FATTR4_WORD1_OWNER_GROUP
 199         | FATTR4_WORD1_RAWDEV
 200         | FATTR4_WORD1_SPACE_USED
 201         | FATTR4_WORD1_TIME_ACCESS
 202         | FATTR4_WORD1_TIME_METADATA
 203         | FATTR4_WORD1_TIME_MODIFY
 204         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 205 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 206         FATTR4_WORD2_SECURITY_LABEL
 207 #endif
 208 };
 209
 210 static const u32 nfs4_pnfs_open_bitmap[3] = {
 211         FATTR4_WORD0_TYPE
 212         | FATTR4_WORD0_CHANGE
 213         | FATTR4_WORD0_SIZE
 214         | FATTR4_WORD0_FSID
 215         | FATTR4_WORD0_FILEID,
 216         FATTR4_WORD1_MODE
 217         | FATTR4_WORD1_NUMLINKS
 218         | FATTR4_WORD1_OWNER
 219         | FATTR4_WORD1_OWNER_GROUP
 220         | FATTR4_WORD1_RAWDEV
 221         | FATTR4_WORD1_SPACE_USED
 222         | FATTR4_WORD1_TIME_ACCESS
 223         | FATTR4_WORD1_TIME_METADATA
 224         | FATTR4_WORD1_TIME_MODIFY,
 225         FATTR4_WORD2_MDSTHRESHOLD
 226 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 227         | FATTR4_WORD2_SECURITY_LABEL
 228 #endif
 229 };
 230
 231 static const u32 nfs4_open_noattr_bitmap[3] = {
 232         FATTR4_WORD0_TYPE
 233         | FATTR4_WORD0_FILEID,
 234 };
 235
 236 const u32 nfs4_statfs_bitmap[3] = {
 237         FATTR4_WORD0_FILES_AVAIL
 238         | FATTR4_WORD0_FILES_FREE
 239         | FATTR4_WORD0_FILES_TOTAL,
 240         FATTR4_WORD1_SPACE_AVAIL
 241         | FATTR4_WORD1_SPACE_FREE
 242         | FATTR4_WORD1_SPACE_TOTAL
 243 };
 244
 245 const u32 nfs4_pathconf_bitmap[3] = {
 246         FATTR4_WORD0_MAXLINK
 247         | FATTR4_WORD0_MAXNAME,
 248         0
 249 };
 250
 251 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 252                         | FATTR4_WORD0_MAXREAD
 253                         | FATTR4_WORD0_MAXWRITE
 254                         | FATTR4_WORD0_LEASE_TIME,
 255                         FATTR4_WORD1_TIME_DELTA
 256                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 257                         FATTR4_WORD2_LAYOUT_BLKSIZE
 258                         | FATTR4_WORD2_CLONE_BLKSIZE
 259 };
 260
 261 const u32 nfs4_fs_locations_bitmap[3] = {
 262         FATTR4_WORD0_CHANGE
 263         | FATTR4_WORD0_SIZE
 264         | FATTR4_WORD0_FSID
 265         | FATTR4_WORD0_FILEID
 266         | FATTR4_WORD0_FS_LOCATIONS,
 267         FATTR4_WORD1_OWNER
 268         | FATTR4_WORD1_OWNER_GROUP
 269         | FATTR4_WORD1_RAWDEV
 270         | FATTR4_WORD1_SPACE_USED
 271         | FATTR4_WORD1_TIME_ACCESS
 272         | FATTR4_WORD1_TIME_METADATA
 273         | FATTR4_WORD1_TIME_MODIFY
 274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 275 };
 276
 277 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 278                 struct inode *inode)
 279 {
 280         unsigned long cache_validity;
 281
 282         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 283         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 284                 return;
 285
 286         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 287         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 288                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 289                                 | NFS_INO_INVALID_SIZE);
 290
 291         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 292                 dst[0] &= ~FATTR4_WORD0_SIZE;
 293
 294         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 295                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 296 }
 297
 298 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 299                 const __u32 *src, struct inode *inode)
 300 {
 301         nfs4_bitmap_copy_adjust(dst, src, inode);
 302 }
 303
 304 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 305                 struct nfs4_readdir_arg *readdir)
 306 {
 307         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 308         __be32 *start, *p;
 309
 310         if (cookie > 2) {
 311                 readdir->cookie = cookie;
 312                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 313                 return;
 314         }
 315
 316         readdir->cookie = 0;
 317         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 318         if (cookie == 2)
 319                 return;
 320
 321         /*
 322          * NFSv4 servers do not return entries for '.' and '..'
 323          * Therefore, we fake these entries here.  We let '.'
 324          * have cookie 0 and '..' have cookie 1.  Note that
 325          * when talking to the server, we always send cookie 0
 326          * instead of 1 or 2.
 327          */
 328         start = p = kmap_atomic(*readdir->pages);
 329
 330         if (cookie == 0) {
 331                 *p++ = xdr_one;                                  /* next */
 332                 *p++ = xdr_zero;                   /* cookie, first word */
 333                 *p++ = xdr_one;                   /* cookie, second word */
 334                 *p++ = xdr_one;                             /* entry len */
 335                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 336                 p++;
 337                 *p++ = xdr_one;                         /* bitmap length */
 338                 *p++ = htonl(attrs);                           /* bitmap */
 339                 *p++ = htonl(12);             /* attribute buffer length */
 340                 *p++ = htonl(NF4DIR);
 341                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 342         }
 343
 344         *p++ = xdr_one;                                  /* next */
 345         *p++ = xdr_zero;                   /* cookie, first word */
 346         *p++ = xdr_two;                   /* cookie, second word */
 347         *p++ = xdr_two;                             /* entry len */
 348         memcpy(p, "..\0\0", 4);                         /* entry */
 349         p++;
 350         *p++ = xdr_one;                         /* bitmap length */
 351         *p++ = htonl(attrs);                           /* bitmap */
 352         *p++ = htonl(12);             /* attribute buffer length */
 353         *p++ = htonl(NF4DIR);
 354         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 355
 356         readdir->pgbase = (char *)p - (char *)start;
 357         readdir->count -= readdir->pgbase;
 358         kunmap_atomic(start);
 359 }
 360
 361 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 362                 nfs4_stateid *stateid,
 363                 const struct cred *cred)
 364 {
 365         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 366
 367         ops->test_and_free_expired(server, stateid, cred);
 368 }
 369
 370 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 371                 nfs4_stateid *stateid,
 372                 const struct cred *cred)
 373 {
 374         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 375         nfs4_test_and_free_stateid(server, stateid, cred);
 376 }
 377
 378 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 379                 const nfs4_stateid *stateid,
 380                 const struct cred *cred)
 381 {
 382         nfs4_stateid tmp;
 383
 384         nfs4_stateid_copy(&tmp, stateid);
 385         __nfs4_free_revoked_stateid(server, &tmp, cred);
 386 }
 387
 388 static long nfs4_update_delay(long *timeout)
 389 {
 390         long ret;
 391         if (!timeout)
 392                 return NFS4_POLL_RETRY_MAX;
 393         if (*timeout <= 0)
 394                 *timeout = NFS4_POLL_RETRY_MIN;
 395         if (*timeout > NFS4_POLL_RETRY_MAX)
 396                 *timeout = NFS4_POLL_RETRY_MAX;
 397         ret = *timeout;
 398         *timeout <<= 1;
 399         return ret;
 400 }
 401
 402 static int nfs4_delay_killable(long *timeout)
 403 {
 404         might_sleep();
 405
 406         freezable_schedule_timeout_killable_unsafe(
 407                 nfs4_update_delay(timeout));
 408         if (!__fatal_signal_pending(current))
 409                 return 0;
 410         return -EINTR;
 411 }
 412
 413 static int nfs4_delay_interruptible(long *timeout)
 414 {
 415         might_sleep();
 416
 417         freezable_schedule_timeout_interruptible(nfs4_update_delay(timeout));
 418         if (!signal_pending(current))
 419                 return 0;
 420         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 421 }
 422
 423 static int nfs4_delay(long *timeout, bool interruptible)
 424 {
 425         if (interruptible)
 426                 return nfs4_delay_interruptible(timeout);
 427         return nfs4_delay_killable(timeout);
 428 }
 429
 430 static const nfs4_stateid *
 431 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 432 {
 433         if (!stateid)
 434                 return NULL;
 435         switch (stateid->type) {
 436         case NFS4_OPEN_STATEID_TYPE:
 437         case NFS4_LOCK_STATEID_TYPE:
 438         case NFS4_DELEGATION_STATEID_TYPE:
 439                 return stateid;
 440         default:
 441                 break;
 442         }
 443         return NULL;
 444 }
 445
 446 /* This is the error handling routine for processes that are allowed
 447  * to sleep.
 448  */
 449 static int nfs4_do_handle_exception(struct nfs_server *server,
 450                 int errorcode, struct nfs4_exception *exception)
 451 {
 452         struct nfs_client *clp = server->nfs_client;
 453         struct nfs4_state *state = exception->state;
 454         const nfs4_stateid *stateid;
 455         struct inode *inode = exception->inode;
 456         int ret = errorcode;
 457
 458         exception->delay = 0;
 459         exception->recovering = 0;
 460         exception->retry = 0;
 461
 462         stateid = nfs4_recoverable_stateid(exception->stateid);
 463         if (stateid == NULL && state != NULL)
 464                 stateid = nfs4_recoverable_stateid(&state->stateid);
 465
 466         switch(errorcode) {
 467                 case 0:
 468                         return 0;
 469                 case -NFS4ERR_BADHANDLE:
 470                 case -ESTALE:
 471                         if (inode != NULL && S_ISREG(inode->i_mode))
 472                                 pnfs_destroy_layout(NFS_I(inode));
 473                         break;
 474                 case -NFS4ERR_DELEG_REVOKED:
 475                 case -NFS4ERR_ADMIN_REVOKED:
 476                 case -NFS4ERR_EXPIRED:
 477                 case -NFS4ERR_BAD_STATEID:
 478                 case -NFS4ERR_PARTNER_NO_AUTH:
 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                         /* Handled in nfs41_sequence_process() */
 525                         goto wait_on_recovery;
 526 #endif /* defined(CONFIG_NFS_V4_1) */
 527                 case -NFS4ERR_FILE_OPEN:
 528                         if (exception->timeout > HZ) {
 529                                 /* We have retried a decent amount, time to
 530                                  * fail
 531                                  */
 532                                 ret = -EBUSY;
 533                                 break;
 534                         }
 535                         /* Fall through */
 536                 case -NFS4ERR_DELAY:
 537                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 538                         /* Fall through */
 539                 case -NFS4ERR_GRACE:
 540                 case -NFS4ERR_LAYOUTTRYLATER:
 541                 case -NFS4ERR_RECALLCONFLICT:
 542                         exception->delay = 1;
 543                         return 0;
 544
 545                 case -NFS4ERR_RETRY_UNCACHED_REP:
 546                 case -NFS4ERR_OLD_STATEID:
 547                         exception->retry = 1;
 548                         break;
 549                 case -NFS4ERR_BADOWNER:
 550                         /* The following works around a Linux server bug! */
 551                 case -NFS4ERR_BADNAME:
 552                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 553                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 554                                 exception->retry = 1;
 555                                 printk(KERN_WARNING "NFS: v4 server %s "
 556                                                 "does not accept raw "
 557                                                 "uid/gids. "
 558                                                 "Reenabling the idmapper.\n",
 559                                                 server->nfs_client->cl_hostname);
 560                         }
 561         }
 562         /* We failed to handle the error */
 563         return nfs4_map_errors(ret);
 564 wait_on_recovery:
 565         exception->recovering = 1;
 566         return 0;
 567 }
 568
 569 /* This is the error handling routine for processes that are allowed
 570  * to sleep.
 571  */
 572 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 573 {
 574         struct nfs_client *clp = server->nfs_client;
 575         int ret;
 576
 577         ret = nfs4_do_handle_exception(server, errorcode, exception);
 578         if (exception->delay) {
 579                 ret = nfs4_delay(&exception->timeout,
 580                                 exception->interruptible);
 581                 goto out_retry;
 582         }
 583         if (exception->recovering) {
 584                 ret = nfs4_wait_clnt_recover(clp);
 585                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 586                         return -EIO;
 587                 goto out_retry;
 588         }
 589         return ret;
 590 out_retry:
 591         if (ret == 0)
 592                 exception->retry = 1;
 593         return ret;
 594 }
 595
 596 static int
 597 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 598                 int errorcode, struct nfs4_exception *exception)
 599 {
 600         struct nfs_client *clp = server->nfs_client;
 601         int ret;
 602
 603         ret = nfs4_do_handle_exception(server, errorcode, exception);
 604         if (exception->delay) {
 605                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 606                 goto out_retry;
 607         }
 608         if (exception->recovering) {
 609                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 610                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 611                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 612                 goto out_retry;
 613         }
 614         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 615                 ret = -EIO;
 616         return ret;
 617 out_retry:
 618         if (ret == 0) {
 619                 exception->retry = 1;
 620                 /*
 621                  * For NFS4ERR_MOVED, the client transport will need to
 622                  * be recomputed after migration recovery has completed.
 623                  */
 624                 if (errorcode == -NFS4ERR_MOVED)
 625                         rpc_task_release_transport(task);
 626         }
 627         return ret;
 628 }
 629
 630 int
 631 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 632                         struct nfs4_state *state, long *timeout)
 633 {
 634         struct nfs4_exception exception = {
 635                 .state = state,
 636         };
 637
 638         if (task->tk_status >= 0)
 639                 return 0;
 640         if (timeout)
 641                 exception.timeout = *timeout;
 642         task->tk_status = nfs4_async_handle_exception(task, server,
 643                         task->tk_status,
 644                         &exception);
 645         if (exception.delay && timeout)
 646                 *timeout = exception.timeout;
 647         if (exception.retry)
 648                 return -EAGAIN;
 649         return 0;
 650 }
 651
 652 /*
 653  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 654  * or 'false' otherwise.
 655  */
 656 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 657 {
 658         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 659         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 660 }
 661
 662 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 663 {
 664         spin_lock(&clp->cl_lock);
 665         if (time_before(clp->cl_last_renewal,timestamp))
 666                 clp->cl_last_renewal = timestamp;
 667         spin_unlock(&clp->cl_lock);
 668 }
 669
 670 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 671 {
 672         struct nfs_client *clp = server->nfs_client;
 673
 674         if (!nfs4_has_session(clp))
 675                 do_renew_lease(clp, timestamp);
 676 }
 677
 678 struct nfs4_call_sync_data {
 679         const struct nfs_server *seq_server;
 680         struct nfs4_sequence_args *seq_args;
 681         struct nfs4_sequence_res *seq_res;
 682 };
 683
 684 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 685                         struct nfs4_sequence_res *res, int cache_reply,
 686                         int privileged)
 687 {
 688         args->sa_slot = NULL;
 689         args->sa_cache_this = cache_reply;
 690         args->sa_privileged = privileged;
 691
 692         res->sr_slot = NULL;
 693 }
 694
 695 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 696 {
 697         struct nfs4_slot *slot = res->sr_slot;
 698         struct nfs4_slot_table *tbl;
 699
 700         tbl = slot->table;
 701         spin_lock(&tbl->slot_tbl_lock);
 702         if (!nfs41_wake_and_assign_slot(tbl, slot))
 703                 nfs4_free_slot(tbl, slot);
 704         spin_unlock(&tbl->slot_tbl_lock);
 705
 706         res->sr_slot = NULL;
 707 }
 708
 709 static int nfs40_sequence_done(struct rpc_task *task,
 710                                struct nfs4_sequence_res *res)
 711 {
 712         if (res->sr_slot != NULL)
 713                 nfs40_sequence_free_slot(res);
 714         return 1;
 715 }
 716
 717 #if defined(CONFIG_NFS_V4_1)
 718
 719 static void nfs41_release_slot(struct nfs4_slot *slot)
 720 {
 721         struct nfs4_session *session;
 722         struct nfs4_slot_table *tbl;
 723         bool send_new_highest_used_slotid = false;
 724
 725         if (!slot)
 726                 return;
 727         tbl = slot->table;
 728         session = tbl->session;
 729
 730         /* Bump the slot sequence number */
 731         if (slot->seq_done)
 732                 slot->seq_nr++;
 733         slot->seq_done = 0;
 734
 735         spin_lock(&tbl->slot_tbl_lock);
 736         /* Be nice to the server: try to ensure that the last transmitted
 737          * value for highest_user_slotid <= target_highest_slotid
 738          */
 739         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 740                 send_new_highest_used_slotid = true;
 741
 742         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 743                 send_new_highest_used_slotid = false;
 744                 goto out_unlock;
 745         }
 746         nfs4_free_slot(tbl, slot);
 747
 748         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 749                 send_new_highest_used_slotid = false;
 750 out_unlock:
 751         spin_unlock(&tbl->slot_tbl_lock);
 752         if (send_new_highest_used_slotid)
 753                 nfs41_notify_server(session->clp);
 754         if (waitqueue_active(&tbl->slot_waitq))
 755                 wake_up_all(&tbl->slot_waitq);
 756 }
 757
 758 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 759 {
 760         nfs41_release_slot(res->sr_slot);
 761         res->sr_slot = NULL;
 762 }
 763
 764 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 765                 u32 seqnr)
 766 {
 767         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 768                 slot->seq_nr_highest_sent = seqnr;
 769 }
 770 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 771                 u32 seqnr)
 772 {
 773         slot->seq_nr_highest_sent = seqnr;
 774         slot->seq_nr_last_acked = seqnr;
 775 }
 776
 777 static int nfs41_sequence_process(struct rpc_task *task,
 778                 struct nfs4_sequence_res *res)
 779 {
 780         struct nfs4_session *session;
 781         struct nfs4_slot *slot = res->sr_slot;
 782         struct nfs_client *clp;
 783         int status;
 784         int ret = 1;
 785
 786         if (slot == NULL)
 787                 goto out_noaction;
 788         /* don't increment the sequence number if the task wasn't sent */
 789         if (!RPC_WAS_SENT(task) || slot->seq_done)
 790                 goto out;
 791
 792         session = slot->table->session;
 793
 794         trace_nfs4_sequence_done(session, res);
 795
 796         status = res->sr_status;
 797         if (task->tk_status == -NFS4ERR_DEADSESSION)
 798                 status = -NFS4ERR_DEADSESSION;
 799
 800         /* Check the SEQUENCE operation status */
 801         switch (status) {
 802         case 0:
 803                 /* Mark this sequence number as having been acked */
 804                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 805                 /* Update the slot's sequence and clientid lease timer */
 806                 slot->seq_done = 1;
 807                 clp = session->clp;
 808                 do_renew_lease(clp, res->sr_timestamp);
 809                 /* Check sequence flags */
 810                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 811                                 !!slot->privileged);
 812                 nfs41_update_target_slotid(slot->table, slot, res);
 813                 break;
 814         case 1:
 815                 /*
 816                  * sr_status remains 1 if an RPC level error occurred.
 817                  * The server may or may not have processed the sequence
 818                  * operation..
 819                  */
 820                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 821                 slot->seq_done = 1;
 822                 goto out;
 823         case -NFS4ERR_DELAY:
 824                 /* The server detected a resend of the RPC call and
 825                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 826                  * of RFC5661.
 827                  */
 828                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 829                         __func__,
 830                         slot->slot_nr,
 831                         slot->seq_nr);
 832                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 833                 goto out_retry;
 834         case -NFS4ERR_RETRY_UNCACHED_REP:
 835         case -NFS4ERR_SEQ_FALSE_RETRY:
 836                 /*
 837                  * The server thinks we tried to replay a request.
 838                  * Retry the call after bumping the sequence ID.
 839                  */
 840                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 841                 goto retry_new_seq;
 842         case -NFS4ERR_BADSLOT:
 843                 /*
 844                  * The slot id we used was probably retired. Try again
 845                  * using a different slot id.
 846                  */
 847                 if (slot->slot_nr < slot->table->target_highest_slotid)
 848                         goto session_recover;
 849                 goto retry_nowait;
 850         case -NFS4ERR_SEQ_MISORDERED:
 851                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 852                 /*
 853                  * Were one or more calls using this slot interrupted?
 854                  * If the server never received the request, then our
 855                  * transmitted slot sequence number may be too high.
 856                  */
 857                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 858                         slot->seq_nr--;
 859                         goto retry_nowait;
 860                 }
 861                 /*
 862                  * RFC5661:
 863                  * A retry might be sent while the original request is
 864                  * still in progress on the replier. The replier SHOULD
 865                  * deal with the issue by returning NFS4ERR_DELAY as the
 866                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 867                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 868                  *
 869                  * Restart the search after a delay.
 870                  */
 871                 slot->seq_nr = slot->seq_nr_highest_sent;
 872                 goto out_retry;
 873         case -NFS4ERR_BADSESSION:
 874         case -NFS4ERR_DEADSESSION:
 875         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 876                 goto session_recover;
 877         default:
 878                 /* Just update the slot sequence no. */
 879                 slot->seq_done = 1;
 880         }
 881 out:
 882         /* The session may be reset by one of the error handlers. */
 883         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 884 out_noaction:
 885         return ret;
 886 session_recover:
 887         nfs4_schedule_session_recovery(session, status);
 888         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 889         nfs41_sequence_free_slot(res);
 890         goto out;
 891 retry_new_seq:
 892         ++slot->seq_nr;
 893 retry_nowait:
 894         if (rpc_restart_call_prepare(task)) {
 895                 nfs41_sequence_free_slot(res);
 896                 task->tk_status = 0;
 897                 ret = 0;
 898         }
 899         goto out;
 900 out_retry:
 901         if (!rpc_restart_call(task))
 902                 goto out;
 903         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 904         return 0;
 905 }
 906
 907 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 908 {
 909         if (!nfs41_sequence_process(task, res))
 910                 return 0;
 911         if (res->sr_slot != NULL)
 912                 nfs41_sequence_free_slot(res);
 913         return 1;
 914
 915 }
 916 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 917
 918 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 919 {
 920         if (res->sr_slot == NULL)
 921                 return 1;
 922         if (res->sr_slot->table->session != NULL)
 923                 return nfs41_sequence_process(task, res);
 924         return nfs40_sequence_done(task, res);
 925 }
 926
 927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot != NULL) {
 930                 if (res->sr_slot->table->session != NULL)
 931                         nfs41_sequence_free_slot(res);
 932                 else
 933                         nfs40_sequence_free_slot(res);
 934         }
 935 }
 936
 937 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 938 {
 939         if (res->sr_slot == NULL)
 940                 return 1;
 941         if (!res->sr_slot->table->session)
 942                 return nfs40_sequence_done(task, res);
 943         return nfs41_sequence_done(task, res);
 944 }
 945 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 946
 947 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950
 951         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 952
 953         nfs4_setup_sequence(data->seq_server->nfs_client,
 954                             data->seq_args, data->seq_res, task);
 955 }
 956
 957 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 958 {
 959         struct nfs4_call_sync_data *data = calldata;
 960
 961         nfs41_sequence_done(task, data->seq_res);
 962 }
 963
 964 static const struct rpc_call_ops nfs41_call_sync_ops = {
 965         .rpc_call_prepare = nfs41_call_sync_prepare,
 966         .rpc_call_done = nfs41_call_sync_done,
 967 };
 968
 969 #else   /* !CONFIG_NFS_V4_1 */
 970
 971 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 972 {
 973         return nfs40_sequence_done(task, res);
 974 }
 975
 976 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 977 {
 978         if (res->sr_slot != NULL)
 979                 nfs40_sequence_free_slot(res);
 980 }
 981
 982 int nfs4_sequence_done(struct rpc_task *task,
 983                        struct nfs4_sequence_res *res)
 984 {
 985         return nfs40_sequence_done(task, res);
 986 }
 987 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 988
 989 #endif  /* !CONFIG_NFS_V4_1 */
 990
 991 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
 992 {
 993         res->sr_timestamp = jiffies;
 994         res->sr_status_flags = 0;
 995         res->sr_status = 1;
 996 }
 997
 998 static
 999 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1000                 struct nfs4_sequence_res *res,
1001                 struct nfs4_slot *slot)
1002 {
1003         if (!slot)
1004                 return;
1005         slot->privileged = args->sa_privileged ? 1 : 0;
1006         args->sa_slot = slot;
1007
1008         res->sr_slot = slot;
1009 }
1010
1011 int nfs4_setup_sequence(struct nfs_client *client,
1012                         struct nfs4_sequence_args *args,
1013                         struct nfs4_sequence_res *res,
1014                         struct rpc_task *task)
1015 {
1016         struct nfs4_session *session = nfs4_get_session(client);
1017         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1018         struct nfs4_slot *slot;
1019
1020         /* slot already allocated? */
1021         if (res->sr_slot != NULL)
1022                 goto out_start;
1023
1024         if (session)
1025                 tbl = &session->fc_slot_table;
1026
1027         spin_lock(&tbl->slot_tbl_lock);
1028         /* The state manager will wait until the slot table is empty */
1029         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1030                 goto out_sleep;
1031
1032         slot = nfs4_alloc_slot(tbl);
1033         if (IS_ERR(slot)) {
1034                 if (slot == ERR_PTR(-ENOMEM))
1035                         goto out_sleep_timeout;
1036                 goto out_sleep;
1037         }
1038         spin_unlock(&tbl->slot_tbl_lock);
1039
1040         nfs4_sequence_attach_slot(args, res, slot);
1041
1042         trace_nfs4_setup_sequence(session, args);
1043 out_start:
1044         nfs41_sequence_res_init(res);
1045         rpc_call_start(task);
1046         return 0;
1047 out_sleep_timeout:
1048         /* Try again in 1/4 second */
1049         if (args->sa_privileged)
1050                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1051                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1052         else
1053                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1054                                 NULL, jiffies + (HZ >> 2));
1055         spin_unlock(&tbl->slot_tbl_lock);
1056         return -EAGAIN;
1057 out_sleep:
1058         if (args->sa_privileged)
1059                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1060                                 RPC_PRIORITY_PRIVILEGED);
1061         else
1062                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1063         spin_unlock(&tbl->slot_tbl_lock);
1064         return -EAGAIN;
1065 }
1066 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1067
1068 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1069 {
1070         struct nfs4_call_sync_data *data = calldata;
1071         nfs4_setup_sequence(data->seq_server->nfs_client,
1072                                 data->seq_args, data->seq_res, task);
1073 }
1074
1075 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1076 {
1077         struct nfs4_call_sync_data *data = calldata;
1078         nfs4_sequence_done(task, data->seq_res);
1079 }
1080
1081 static const struct rpc_call_ops nfs40_call_sync_ops = {
1082         .rpc_call_prepare = nfs40_call_sync_prepare,
1083         .rpc_call_done = nfs40_call_sync_done,
1084 };
1085
1086 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1087 {
1088         int ret;
1089         struct rpc_task *task;
1090
1091         task = rpc_run_task(task_setup);
1092         if (IS_ERR(task))
1093                 return PTR_ERR(task);
1094
1095         ret = task->tk_status;
1096         rpc_put_task(task);
1097         return ret;
1098 }
1099
1100 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1101                              struct nfs_server *server,
1102                              struct rpc_message *msg,
1103                              struct nfs4_sequence_args *args,
1104                              struct nfs4_sequence_res *res,
1105                              unsigned short task_flags)
1106 {
1107         struct nfs_client *clp = server->nfs_client;
1108         struct nfs4_call_sync_data data = {
1109                 .seq_server = server,
1110                 .seq_args = args,
1111                 .seq_res = res,
1112         };
1113         struct rpc_task_setup task_setup = {
1114                 .rpc_client = clnt,
1115                 .rpc_message = msg,
1116                 .callback_ops = clp->cl_mvops->call_sync_ops,
1117                 .callback_data = &data,
1118                 .flags = task_flags,
1119         };
1120
1121         return nfs4_call_sync_custom(&task_setup);
1122 }
1123
1124 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1125                                    struct nfs_server *server,
1126                                    struct rpc_message *msg,
1127                                    struct nfs4_sequence_args *args,
1128                                    struct nfs4_sequence_res *res)
1129 {
1130         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1131 }
1132
1133
1134 int nfs4_call_sync(struct rpc_clnt *clnt,
1135                    struct nfs_server *server,
1136                    struct rpc_message *msg,
1137                    struct nfs4_sequence_args *args,
1138                    struct nfs4_sequence_res *res,
1139                    int cache_reply)
1140 {
1141         nfs4_init_sequence(args, res, cache_reply, 0);
1142         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1143 }
1144
1145 static void
1146 nfs4_inc_nlink_locked(struct inode *inode)
1147 {
1148         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1149         inc_nlink(inode);
1150 }
1151
1152 static void
1153 nfs4_dec_nlink_locked(struct inode *inode)
1154 {
1155         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1156         drop_nlink(inode);
1157 }
1158
1159 static void
1160 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1161                 unsigned long timestamp, unsigned long cache_validity)
1162 {
1163         struct nfs_inode *nfsi = NFS_I(dir);
1164
1165         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1166                 | NFS_INO_INVALID_MTIME
1167                 | NFS_INO_INVALID_DATA
1168                 | cache_validity;
1169         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1170                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1171                 nfsi->attrtimeo_timestamp = jiffies;
1172         } else {
1173                 nfs_force_lookup_revalidate(dir);
1174                 if (cinfo->before != inode_peek_iversion_raw(dir))
1175                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1176                                 NFS_INO_INVALID_ACL;
1177         }
1178         inode_set_iversion_raw(dir, cinfo->after);
1179         nfsi->read_cache_jiffies = timestamp;
1180         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1181         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1182         nfs_fscache_invalidate(dir);
1183 }
1184
1185 static void
1186 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1187                 unsigned long timestamp, unsigned long cache_validity)
1188 {
1189         spin_lock(&dir->i_lock);
1190         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1191         spin_unlock(&dir->i_lock);
1192 }
1193
1194 struct nfs4_open_createattrs {
1195         struct nfs4_label *label;
1196         struct iattr *sattr;
1197         const __u32 verf[2];
1198 };
1199
1200 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1201                 int err, struct nfs4_exception *exception)
1202 {
1203         if (err != -EINVAL)
1204                 return false;
1205         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1206                 return false;
1207         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1208         exception->retry = 1;
1209         return true;
1210 }
1211
1212 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1213 {
1214          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1215 }
1216
1217 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1218 {
1219         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1220
1221         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1222 }
1223
1224 static u32
1225 nfs4_map_atomic_open_share(struct nfs_server *server,
1226                 fmode_t fmode, int openflags)
1227 {
1228         u32 res = 0;
1229
1230         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1231         case FMODE_READ:
1232                 res = NFS4_SHARE_ACCESS_READ;
1233                 break;
1234         case FMODE_WRITE:
1235                 res = NFS4_SHARE_ACCESS_WRITE;
1236                 break;
1237         case FMODE_READ|FMODE_WRITE:
1238                 res = NFS4_SHARE_ACCESS_BOTH;
1239         }
1240         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1241                 goto out;
1242         /* Want no delegation if we're using O_DIRECT */
1243         if (openflags & O_DIRECT)
1244                 res |= NFS4_SHARE_WANT_NO_DELEG;
1245 out:
1246         return res;
1247 }
1248
1249 static enum open_claim_type4
1250 nfs4_map_atomic_open_claim(struct nfs_server *server,
1251                 enum open_claim_type4 claim)
1252 {
1253         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1254                 return claim;
1255         switch (claim) {
1256         default:
1257                 return claim;
1258         case NFS4_OPEN_CLAIM_FH:
1259                 return NFS4_OPEN_CLAIM_NULL;
1260         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1261                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1262         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1263                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1264         }
1265 }
1266
1267 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1268 {
1269         p->o_res.f_attr = &p->f_attr;
1270         p->o_res.f_label = p->f_label;
1271         p->o_res.seqid = p->o_arg.seqid;
1272         p->c_res.seqid = p->c_arg.seqid;
1273         p->o_res.server = p->o_arg.server;
1274         p->o_res.access_request = p->o_arg.access;
1275         nfs_fattr_init(&p->f_attr);
1276         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1277 }
1278
1279 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1280                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1281                 const struct nfs4_open_createattrs *c,
1282                 enum open_claim_type4 claim,
1283                 gfp_t gfp_mask)
1284 {
1285         struct dentry *parent = dget_parent(dentry);
1286         struct inode *dir = d_inode(parent);
1287         struct nfs_server *server = NFS_SERVER(dir);
1288         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1289         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1290         struct nfs4_opendata *p;
1291
1292         p = kzalloc(sizeof(*p), gfp_mask);
1293         if (p == NULL)
1294                 goto err;
1295
1296         p->f_label = nfs4_label_alloc(server, gfp_mask);
1297         if (IS_ERR(p->f_label))
1298                 goto err_free_p;
1299
1300         p->a_label = nfs4_label_alloc(server, gfp_mask);
1301         if (IS_ERR(p->a_label))
1302                 goto err_free_f;
1303
1304         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1305         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1306         if (IS_ERR(p->o_arg.seqid))
1307                 goto err_free_label;
1308         nfs_sb_active(dentry->d_sb);
1309         p->dentry = dget(dentry);
1310         p->dir = parent;
1311         p->owner = sp;
1312         atomic_inc(&sp->so_count);
1313         p->o_arg.open_flags = flags;
1314         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1315         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1316         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1317                         fmode, flags);
1318         if (flags & O_CREAT) {
1319                 p->o_arg.umask = current_umask();
1320                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1321                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1322                         p->o_arg.u.attrs = &p->attrs;
1323                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1324
1325                         memcpy(p->o_arg.u.verifier.data, c->verf,
1326                                         sizeof(p->o_arg.u.verifier.data));
1327                 }
1328         }
1329         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1330          * will return permission denied for all bits until close */
1331         if (!(flags & O_EXCL)) {
1332                 /* ask server to check for all possible rights as results
1333                  * are cached */
1334                 switch (p->o_arg.claim) {
1335                 default:
1336                         break;
1337                 case NFS4_OPEN_CLAIM_NULL:
1338                 case NFS4_OPEN_CLAIM_FH:
1339                         p->o_arg.access = NFS4_ACCESS_READ |
1340                                 NFS4_ACCESS_MODIFY |
1341                                 NFS4_ACCESS_EXTEND |
1342                                 NFS4_ACCESS_EXECUTE;
1343                 }
1344         }
1345         p->o_arg.clientid = server->nfs_client->cl_clientid;
1346         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1347         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1348         p->o_arg.name = &dentry->d_name;
1349         p->o_arg.server = server;
1350         p->o_arg.bitmask = nfs4_bitmask(server, label);
1351         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1352         switch (p->o_arg.claim) {
1353         case NFS4_OPEN_CLAIM_NULL:
1354         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1355         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1356                 p->o_arg.fh = NFS_FH(dir);
1357                 break;
1358         case NFS4_OPEN_CLAIM_PREVIOUS:
1359         case NFS4_OPEN_CLAIM_FH:
1360         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1361         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1362                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1363         }
1364         p->c_arg.fh = &p->o_res.fh;
1365         p->c_arg.stateid = &p->o_res.stateid;
1366         p->c_arg.seqid = p->o_arg.seqid;
1367         nfs4_init_opendata_res(p);
1368         kref_init(&p->kref);
1369         return p;
1370
1371 err_free_label:
1372         nfs4_label_free(p->a_label);
1373 err_free_f:
1374         nfs4_label_free(p->f_label);
1375 err_free_p:
1376         kfree(p);
1377 err:
1378         dput(parent);
1379         return NULL;
1380 }
1381
1382 static void nfs4_opendata_free(struct kref *kref)
1383 {
1384         struct nfs4_opendata *p = container_of(kref,
1385                         struct nfs4_opendata, kref);
1386         struct super_block *sb = p->dentry->d_sb;
1387
1388         nfs4_lgopen_release(p->lgp);
1389         nfs_free_seqid(p->o_arg.seqid);
1390         nfs4_sequence_free_slot(&p->o_res.seq_res);
1391         if (p->state != NULL)
1392                 nfs4_put_open_state(p->state);
1393         nfs4_put_state_owner(p->owner);
1394
1395         nfs4_label_free(p->a_label);
1396         nfs4_label_free(p->f_label);
1397
1398         dput(p->dir);
1399         dput(p->dentry);
1400         nfs_sb_deactive(sb);
1401         nfs_fattr_free_names(&p->f_attr);
1402         kfree(p->f_attr.mdsthreshold);
1403         kfree(p);
1404 }
1405
1406 static void nfs4_opendata_put(struct nfs4_opendata *p)
1407 {
1408         if (p != NULL)
1409                 kref_put(&p->kref, nfs4_opendata_free);
1410 }
1411
1412 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1413                 fmode_t fmode)
1414 {
1415         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1416         case FMODE_READ|FMODE_WRITE:
1417                 return state->n_rdwr != 0;
1418         case FMODE_WRITE:
1419                 return state->n_wronly != 0;
1420         case FMODE_READ:
1421                 return state->n_rdonly != 0;
1422         }
1423         WARN_ON_ONCE(1);
1424         return false;
1425 }
1426
1427 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1428                 int open_mode, enum open_claim_type4 claim)
1429 {
1430         int ret = 0;
1431
1432         if (open_mode & (O_EXCL|O_TRUNC))
1433                 goto out;
1434         switch (claim) {
1435         case NFS4_OPEN_CLAIM_NULL:
1436         case NFS4_OPEN_CLAIM_FH:
1437                 goto out;
1438         default:
1439                 break;
1440         }
1441         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1442                 case FMODE_READ:
1443                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1444                                 && state->n_rdonly != 0;
1445                         break;
1446                 case FMODE_WRITE:
1447                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1448                                 && state->n_wronly != 0;
1449                         break;
1450                 case FMODE_READ|FMODE_WRITE:
1451                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1452                                 && state->n_rdwr != 0;
1453         }
1454 out:
1455         return ret;
1456 }
1457
1458 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1459                 enum open_claim_type4 claim)
1460 {
1461         if (delegation == NULL)
1462                 return 0;
1463         if ((delegation->type & fmode) != fmode)
1464                 return 0;
1465         switch (claim) {
1466         case NFS4_OPEN_CLAIM_NULL:
1467         case NFS4_OPEN_CLAIM_FH:
1468                 break;
1469         case NFS4_OPEN_CLAIM_PREVIOUS:
1470                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1471                         break;
1472                 /* Fall through */
1473         default:
1474                 return 0;
1475         }
1476         nfs_mark_delegation_referenced(delegation);
1477         return 1;
1478 }
1479
1480 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1481 {
1482         switch (fmode) {
1483                 case FMODE_WRITE:
1484                         state->n_wronly++;
1485                         break;
1486                 case FMODE_READ:
1487                         state->n_rdonly++;
1488                         break;
1489                 case FMODE_READ|FMODE_WRITE:
1490                         state->n_rdwr++;
1491         }
1492         nfs4_state_set_mode_locked(state, state->state | fmode);
1493 }
1494
1495 #ifdef CONFIG_NFS_V4_1
1496 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1497 {
1498         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1499                 return true;
1500         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1501                 return true;
1502         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1503                 return true;
1504         return false;
1505 }
1506 #endif /* CONFIG_NFS_V4_1 */
1507
1508 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1509 {
1510         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1511                 wake_up_all(&state->waitq);
1512 }
1513
1514 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1515                 const nfs4_stateid *stateid)
1516 {
1517         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1518         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1519
1520         if (stateid_seqid == state_seqid + 1U ||
1521             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1522                 nfs_state_log_update_open_stateid(state);
1523         else
1524                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1525 }
1526
1527 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1528 {
1529         struct nfs_client *clp = state->owner->so_server->nfs_client;
1530         bool need_recover = false;
1531
1532         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1533                 need_recover = true;
1534         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1535                 need_recover = true;
1536         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1537                 need_recover = true;
1538         if (need_recover)
1539                 nfs4_state_mark_reclaim_nograce(clp, state);
1540 }
1541
1542 /*
1543  * Check for whether or not the caller may update the open stateid
1544  * to the value passed in by stateid.
1545  *
1546  * Note: This function relies heavily on the server implementing
1547  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1548  * correctly.
1549  * i.e. The stateid seqids have to be initialised to 1, and
1550  * are then incremented on every state transition.
1551  */
1552 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1553                 const nfs4_stateid *stateid)
1554 {
1555         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1556             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1557                 if (stateid->seqid == cpu_to_be32(1))
1558                         nfs_state_log_update_open_stateid(state);
1559                 else
1560                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1561                 return true;
1562         }
1563
1564         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1565                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1566                 return true;
1567         }
1568         return false;
1569 }
1570
1571 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1572 {
1573         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1574                 return;
1575         if (state->n_wronly)
1576                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1577         if (state->n_rdonly)
1578                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1579         if (state->n_rdwr)
1580                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1581         set_bit(NFS_OPEN_STATE, &state->flags);
1582 }
1583
1584 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1585                 nfs4_stateid *stateid, fmode_t fmode)
1586 {
1587         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1588         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1589         case FMODE_WRITE:
1590                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1591                 break;
1592         case FMODE_READ:
1593                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1594                 break;
1595         case 0:
1596                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1597                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1598                 clear_bit(NFS_OPEN_STATE, &state->flags);
1599         }
1600         if (stateid == NULL)
1601                 return;
1602         /* Handle OPEN+OPEN_DOWNGRADE races */
1603         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1604             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1605                 nfs_resync_open_stateid_locked(state);
1606                 goto out;
1607         }
1608         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1609                 nfs4_stateid_copy(&state->stateid, stateid);
1610         nfs4_stateid_copy(&state->open_stateid, stateid);
1611         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1612 out:
1613         nfs_state_log_update_open_stateid(state);
1614 }
1615
1616 static void nfs_clear_open_stateid(struct nfs4_state *state,
1617         nfs4_stateid *arg_stateid,
1618         nfs4_stateid *stateid, fmode_t fmode)
1619 {
1620         write_seqlock(&state->seqlock);
1621         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1622         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1623                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1624         write_sequnlock(&state->seqlock);
1625         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1626                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1627 }
1628
1629 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1630                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1631         __must_hold(&state->owner->so_lock)
1632         __must_hold(&state->seqlock)
1633         __must_hold(RCU)
1634
1635 {
1636         DEFINE_WAIT(wait);
1637         int status = 0;
1638         for (;;) {
1639
1640                 if (!nfs_need_update_open_stateid(state, stateid))
1641                         return;
1642                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1643                         break;
1644                 if (status)
1645                         break;
1646                 /* Rely on seqids for serialisation with NFSv4.0 */
1647                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1648                         break;
1649
1650                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1651                 /*
1652                  * Ensure we process the state changes in the same order
1653                  * in which the server processed them by delaying the
1654                  * update of the stateid until we are in sequence.
1655                  */
1656                 write_sequnlock(&state->seqlock);
1657                 spin_unlock(&state->owner->so_lock);
1658                 rcu_read_unlock();
1659                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1660                 if (!signal_pending(current)) {
1661                         if (schedule_timeout(5*HZ) == 0)
1662                                 status = -EAGAIN;
1663                         else
1664                                 status = 0;
1665                 } else
1666                         status = -EINTR;
1667                 finish_wait(&state->waitq, &wait);
1668                 rcu_read_lock();
1669                 spin_lock(&state->owner->so_lock);
1670                 write_seqlock(&state->seqlock);
1671         }
1672
1673         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1674             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1675                 nfs4_stateid_copy(freeme, &state->open_stateid);
1676                 nfs_test_and_clear_all_open_stateid(state);
1677         }
1678
1679         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1680                 nfs4_stateid_copy(&state->stateid, stateid);
1681         nfs4_stateid_copy(&state->open_stateid, stateid);
1682         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1683         nfs_state_log_update_open_stateid(state);
1684 }
1685
1686 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1687                 const nfs4_stateid *open_stateid,
1688                 fmode_t fmode,
1689                 nfs4_stateid *freeme)
1690 {
1691         /*
1692          * Protect the call to nfs4_state_set_mode_locked and
1693          * serialise the stateid update
1694          */
1695         write_seqlock(&state->seqlock);
1696         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1697         switch (fmode) {
1698         case FMODE_READ:
1699                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1700                 break;
1701         case FMODE_WRITE:
1702                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1703                 break;
1704         case FMODE_READ|FMODE_WRITE:
1705                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1706         }
1707         set_bit(NFS_OPEN_STATE, &state->flags);
1708         write_sequnlock(&state->seqlock);
1709 }
1710
1711 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1712 {
1713         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1714         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1715         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1716         clear_bit(NFS_OPEN_STATE, &state->flags);
1717 }
1718
1719 static void nfs_state_set_delegation(struct nfs4_state *state,
1720                 const nfs4_stateid *deleg_stateid,
1721                 fmode_t fmode)
1722 {
1723         /*
1724          * Protect the call to nfs4_state_set_mode_locked and
1725          * serialise the stateid update
1726          */
1727         write_seqlock(&state->seqlock);
1728         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1729         set_bit(NFS_DELEGATED_STATE, &state->flags);
1730         write_sequnlock(&state->seqlock);
1731 }
1732
1733 static void nfs_state_clear_delegation(struct nfs4_state *state)
1734 {
1735         write_seqlock(&state->seqlock);
1736         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1737         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1738         write_sequnlock(&state->seqlock);
1739 }
1740
1741 int update_open_stateid(struct nfs4_state *state,
1742                 const nfs4_stateid *open_stateid,
1743                 const nfs4_stateid *delegation,
1744                 fmode_t fmode)
1745 {
1746         struct nfs_server *server = NFS_SERVER(state->inode);
1747         struct nfs_client *clp = server->nfs_client;
1748         struct nfs_inode *nfsi = NFS_I(state->inode);
1749         struct nfs_delegation *deleg_cur;
1750         nfs4_stateid freeme = { };
1751         int ret = 0;
1752
1753         fmode &= (FMODE_READ|FMODE_WRITE);
1754
1755         rcu_read_lock();
1756         spin_lock(&state->owner->so_lock);
1757         if (open_stateid != NULL) {
1758                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1759                 ret = 1;
1760         }
1761
1762         deleg_cur = nfs4_get_valid_delegation(state->inode);
1763         if (deleg_cur == NULL)
1764                 goto no_delegation;
1765
1766         spin_lock(&deleg_cur->lock);
1767         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1768            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1769             (deleg_cur->type & fmode) != fmode)
1770                 goto no_delegation_unlock;
1771
1772         if (delegation == NULL)
1773                 delegation = &deleg_cur->stateid;
1774         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1775                 goto no_delegation_unlock;
1776
1777         nfs_mark_delegation_referenced(deleg_cur);
1778         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1779         ret = 1;
1780 no_delegation_unlock:
1781         spin_unlock(&deleg_cur->lock);
1782 no_delegation:
1783         if (ret)
1784                 update_open_stateflags(state, fmode);
1785         spin_unlock(&state->owner->so_lock);
1786         rcu_read_unlock();
1787
1788         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1789                 nfs4_schedule_state_manager(clp);
1790         if (freeme.type != 0)
1791                 nfs4_test_and_free_stateid(server, &freeme,
1792                                 state->owner->so_cred);
1793
1794         return ret;
1795 }
1796
1797 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1798                 const nfs4_stateid *stateid)
1799 {
1800         struct nfs4_state *state = lsp->ls_state;
1801         bool ret = false;
1802
1803         spin_lock(&state->state_lock);
1804         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1805                 goto out_noupdate;
1806         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1807                 goto out_noupdate;
1808         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1809         ret = true;
1810 out_noupdate:
1811         spin_unlock(&state->state_lock);
1812         return ret;
1813 }
1814
1815 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1816 {
1817         struct nfs_delegation *delegation;
1818
1819         fmode &= FMODE_READ|FMODE_WRITE;
1820         rcu_read_lock();
1821         delegation = nfs4_get_valid_delegation(inode);
1822         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1823                 rcu_read_unlock();
1824                 return;
1825         }
1826         rcu_read_unlock();
1827         nfs4_inode_return_delegation(inode);
1828 }
1829
1830 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1831 {
1832         struct nfs4_state *state = opendata->state;
1833         struct nfs_delegation *delegation;
1834         int open_mode = opendata->o_arg.open_flags;
1835         fmode_t fmode = opendata->o_arg.fmode;
1836         enum open_claim_type4 claim = opendata->o_arg.claim;
1837         nfs4_stateid stateid;
1838         int ret = -EAGAIN;
1839
1840         for (;;) {
1841                 spin_lock(&state->owner->so_lock);
1842                 if (can_open_cached(state, fmode, open_mode, claim)) {
1843                         update_open_stateflags(state, fmode);
1844                         spin_unlock(&state->owner->so_lock);
1845                         goto out_return_state;
1846                 }
1847                 spin_unlock(&state->owner->so_lock);
1848                 rcu_read_lock();
1849                 delegation = nfs4_get_valid_delegation(state->inode);
1850                 if (!can_open_delegated(delegation, fmode, claim)) {
1851                         rcu_read_unlock();
1852                         break;
1853                 }
1854                 /* Save the delegation */
1855                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1856                 rcu_read_unlock();
1857                 nfs_release_seqid(opendata->o_arg.seqid);
1858                 if (!opendata->is_recover) {
1859                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1860                         if (ret != 0)
1861                                 goto out;
1862                 }
1863                 ret = -EAGAIN;
1864
1865                 /* Try to update the stateid using the delegation */
1866                 if (update_open_stateid(state, NULL, &stateid, fmode))
1867                         goto out_return_state;
1868         }
1869 out:
1870         return ERR_PTR(ret);
1871 out_return_state:
1872         refcount_inc(&state->count);
1873         return state;
1874 }
1875
1876 static void
1877 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1878 {
1879         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1880         struct nfs_delegation *delegation;
1881         int delegation_flags = 0;
1882
1883         rcu_read_lock();
1884         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1885         if (delegation)
1886                 delegation_flags = delegation->flags;
1887         rcu_read_unlock();
1888         switch (data->o_arg.claim) {
1889         default:
1890                 break;
1891         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1892         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1893                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1894                                    "returning a delegation for "
1895                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1896                                    clp->cl_hostname);
1897                 return;
1898         }
1899         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1900                 nfs_inode_set_delegation(state->inode,
1901                                 data->owner->so_cred,
1902                                 data->o_res.delegation_type,
1903                                 &data->o_res.delegation,
1904                                 data->o_res.pagemod_limit);
1905         else
1906                 nfs_inode_reclaim_delegation(state->inode,
1907                                 data->owner->so_cred,
1908                                 data->o_res.delegation_type,
1909                                 &data->o_res.delegation,
1910                                 data->o_res.pagemod_limit);
1911
1912         if (data->o_res.do_recall)
1913                 nfs_async_inode_return_delegation(state->inode,
1914                                                   &data->o_res.delegation);
1915 }
1916
1917 /*
1918  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1919  * and update the nfs4_state.
1920  */
1921 static struct nfs4_state *
1922 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1923 {
1924         struct inode *inode = data->state->inode;
1925         struct nfs4_state *state = data->state;
1926         int ret;
1927
1928         if (!data->rpc_done) {
1929                 if (data->rpc_status)
1930                         return ERR_PTR(data->rpc_status);
1931                 /* cached opens have already been processed */
1932                 goto update;
1933         }
1934
1935         ret = nfs_refresh_inode(inode, &data->f_attr);
1936         if (ret)
1937                 return ERR_PTR(ret);
1938
1939         if (data->o_res.delegation_type != 0)
1940                 nfs4_opendata_check_deleg(data, state);
1941 update:
1942         if (!update_open_stateid(state, &data->o_res.stateid,
1943                                 NULL, data->o_arg.fmode))
1944                 return ERR_PTR(-EAGAIN);
1945         refcount_inc(&state->count);
1946
1947         return state;
1948 }
1949
1950 static struct inode *
1951 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1952 {
1953         struct inode *inode;
1954
1955         switch (data->o_arg.claim) {
1956         case NFS4_OPEN_CLAIM_NULL:
1957         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1958         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1959                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1960                         return ERR_PTR(-EAGAIN);
1961                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1962                                 &data->f_attr, data->f_label);
1963                 break;
1964         default:
1965                 inode = d_inode(data->dentry);
1966                 ihold(inode);
1967                 nfs_refresh_inode(inode, &data->f_attr);
1968         }
1969         return inode;
1970 }
1971
1972 static struct nfs4_state *
1973 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1974 {
1975         struct nfs4_state *state;
1976         struct inode *inode;
1977
1978         inode = nfs4_opendata_get_inode(data);
1979         if (IS_ERR(inode))
1980                 return ERR_CAST(inode);
1981         if (data->state != NULL && data->state->inode == inode) {
1982                 state = data->state;
1983                 refcount_inc(&state->count);
1984         } else
1985                 state = nfs4_get_open_state(inode, data->owner);
1986         iput(inode);
1987         if (state == NULL)
1988                 state = ERR_PTR(-ENOMEM);
1989         return state;
1990 }
1991
1992 static struct nfs4_state *
1993 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1994 {
1995         struct nfs4_state *state;
1996
1997         if (!data->rpc_done) {
1998                 state = nfs4_try_open_cached(data);
1999                 trace_nfs4_cached_open(data->state);
2000                 goto out;
2001         }
2002
2003         state = nfs4_opendata_find_nfs4_state(data);
2004         if (IS_ERR(state))
2005                 goto out;
2006
2007         if (data->o_res.delegation_type != 0)
2008                 nfs4_opendata_check_deleg(data, state);
2009         if (!update_open_stateid(state, &data->o_res.stateid,
2010                                 NULL, data->o_arg.fmode)) {
2011                 nfs4_put_open_state(state);
2012                 state = ERR_PTR(-EAGAIN);
2013         }
2014 out:
2015         nfs_release_seqid(data->o_arg.seqid);
2016         return state;
2017 }
2018
2019 static struct nfs4_state *
2020 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2021 {
2022         struct nfs4_state *ret;
2023
2024         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2025                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2026         else
2027                 ret = _nfs4_opendata_to_nfs4_state(data);
2028         nfs4_sequence_free_slot(&data->o_res.seq_res);
2029         return ret;
2030 }
2031
2032 static struct nfs_open_context *
2033 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2034 {
2035         struct nfs_inode *nfsi = NFS_I(state->inode);
2036         struct nfs_open_context *ctx;
2037
2038         rcu_read_lock();
2039         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2040                 if (ctx->state != state)
2041                         continue;
2042                 if ((ctx->mode & mode) != mode)
2043                         continue;
2044                 if (!get_nfs_open_context(ctx))
2045                         continue;
2046                 rcu_read_unlock();
2047                 return ctx;
2048         }
2049         rcu_read_unlock();
2050         return ERR_PTR(-ENOENT);
2051 }
2052
2053 static struct nfs_open_context *
2054 nfs4_state_find_open_context(struct nfs4_state *state)
2055 {
2056         struct nfs_open_context *ctx;
2057
2058         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2059         if (!IS_ERR(ctx))
2060                 return ctx;
2061         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2062         if (!IS_ERR(ctx))
2063                 return ctx;
2064         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2065 }
2066
2067 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2068                 struct nfs4_state *state, enum open_claim_type4 claim)
2069 {
2070         struct nfs4_opendata *opendata;
2071
2072         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2073                         NULL, claim, GFP_NOFS);
2074         if (opendata == NULL)
2075                 return ERR_PTR(-ENOMEM);
2076         opendata->state = state;
2077         refcount_inc(&state->count);
2078         return opendata;
2079 }
2080
2081 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2082                 fmode_t fmode)
2083 {
2084         struct nfs4_state *newstate;
2085         int ret;
2086
2087         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2088                 return 0;
2089         opendata->o_arg.open_flags = 0;
2090         opendata->o_arg.fmode = fmode;
2091         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2092                         NFS_SB(opendata->dentry->d_sb),
2093                         fmode, 0);
2094         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2095         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2096         nfs4_init_opendata_res(opendata);
2097         ret = _nfs4_recover_proc_open(opendata);
2098         if (ret != 0)
2099                 return ret;
2100         newstate = nfs4_opendata_to_nfs4_state(opendata);
2101         if (IS_ERR(newstate))
2102                 return PTR_ERR(newstate);
2103         if (newstate != opendata->state)
2104                 ret = -ESTALE;
2105         nfs4_close_state(newstate, fmode);
2106         return ret;
2107 }
2108
2109 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2110 {
2111         int ret;
2112
2113         /* memory barrier prior to reading state->n_* */
2114         smp_rmb();
2115         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2116         if (ret != 0)
2117                 return ret;
2118         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2119         if (ret != 0)
2120                 return ret;
2121         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2122         if (ret != 0)
2123                 return ret;
2124         /*
2125          * We may have performed cached opens for all three recoveries.
2126          * Check if we need to update the current stateid.
2127          */
2128         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2129             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2130                 write_seqlock(&state->seqlock);
2131                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2132                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2133                 write_sequnlock(&state->seqlock);
2134         }
2135         return 0;
2136 }
2137
2138 /*
2139  * OPEN_RECLAIM:
2140  *      reclaim state on the server after a reboot.
2141  */
2142 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2143 {
2144         struct nfs_delegation *delegation;
2145         struct nfs4_opendata *opendata;
2146         fmode_t delegation_type = 0;
2147         int status;
2148
2149         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2150                         NFS4_OPEN_CLAIM_PREVIOUS);
2151         if (IS_ERR(opendata))
2152                 return PTR_ERR(opendata);
2153         rcu_read_lock();
2154         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2155         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2156                 delegation_type = delegation->type;
2157         rcu_read_unlock();
2158         opendata->o_arg.u.delegation_type = delegation_type;
2159         status = nfs4_open_recover(opendata, state);
2160         nfs4_opendata_put(opendata);
2161         return status;
2162 }
2163
2164 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2165 {
2166         struct nfs_server *server = NFS_SERVER(state->inode);
2167         struct nfs4_exception exception = { };
2168         int err;
2169         do {
2170                 err = _nfs4_do_open_reclaim(ctx, state);
2171                 trace_nfs4_open_reclaim(ctx, 0, err);
2172                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2173                         continue;
2174                 if (err != -NFS4ERR_DELAY)
2175                         break;
2176                 nfs4_handle_exception(server, err, &exception);
2177         } while (exception.retry);
2178         return err;
2179 }
2180
2181 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2182 {
2183         struct nfs_open_context *ctx;
2184         int ret;
2185
2186         ctx = nfs4_state_find_open_context(state);
2187         if (IS_ERR(ctx))
2188                 return -EAGAIN;
2189         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2190         nfs_state_clear_open_state_flags(state);
2191         ret = nfs4_do_open_reclaim(ctx, state);
2192         put_nfs_open_context(ctx);
2193         return ret;
2194 }
2195
2196 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)
2197 {
2198         switch (err) {
2199                 default:
2200                         printk(KERN_ERR "NFS: %s: unhandled error "
2201                                         "%d.\n", __func__, err);
2202                 case 0:
2203                 case -ENOENT:
2204                 case -EAGAIN:
2205                 case -ESTALE:
2206                 case -ETIMEDOUT:
2207                         break;
2208                 case -NFS4ERR_BADSESSION:
2209                 case -NFS4ERR_BADSLOT:
2210                 case -NFS4ERR_BAD_HIGH_SLOT:
2211                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2212                 case -NFS4ERR_DEADSESSION:
2213                         return -EAGAIN;
2214                 case -NFS4ERR_STALE_CLIENTID:
2215                 case -NFS4ERR_STALE_STATEID:
2216                         /* Don't recall a delegation if it was lost */
2217                         nfs4_schedule_lease_recovery(server->nfs_client);
2218                         return -EAGAIN;
2219                 case -NFS4ERR_MOVED:
2220                         nfs4_schedule_migration_recovery(server);
2221                         return -EAGAIN;
2222                 case -NFS4ERR_LEASE_MOVED:
2223                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2224                         return -EAGAIN;
2225                 case -NFS4ERR_DELEG_REVOKED:
2226                 case -NFS4ERR_ADMIN_REVOKED:
2227                 case -NFS4ERR_EXPIRED:
2228                 case -NFS4ERR_BAD_STATEID:
2229                 case -NFS4ERR_OPENMODE:
2230                         nfs_inode_find_state_and_recover(state->inode,
2231                                         stateid);
2232                         nfs4_schedule_stateid_recovery(server, state);
2233                         return -EAGAIN;
2234                 case -NFS4ERR_DELAY:
2235                 case -NFS4ERR_GRACE:
2236                         ssleep(1);
2237                         return -EAGAIN;
2238                 case -ENOMEM:
2239                 case -NFS4ERR_DENIED:
2240                         if (fl) {
2241                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2242                                 if (lsp)
2243                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2244                         }
2245                         return 0;
2246         }
2247         return err;
2248 }
2249
2250 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2251                 struct nfs4_state *state, const nfs4_stateid *stateid)
2252 {
2253         struct nfs_server *server = NFS_SERVER(state->inode);
2254         struct nfs4_opendata *opendata;
2255         int err = 0;
2256
2257         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2258                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2259         if (IS_ERR(opendata))
2260                 return PTR_ERR(opendata);
2261         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2262         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2263                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2264                 if (err)
2265                         goto out;
2266         }
2267         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2268                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2269                 if (err)
2270                         goto out;
2271         }
2272         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2273                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2274                 if (err)
2275                         goto out;
2276         }
2277         nfs_state_clear_delegation(state);
2278 out:
2279         nfs4_opendata_put(opendata);
2280         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2281 }
2282
2283 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2284 {
2285         struct nfs4_opendata *data = calldata;
2286
2287         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2288                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2289 }
2290
2291 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2292 {
2293         struct nfs4_opendata *data = calldata;
2294
2295         nfs40_sequence_done(task, &data->c_res.seq_res);
2296
2297         data->rpc_status = task->tk_status;
2298         if (data->rpc_status == 0) {
2299                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2300                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2301                 renew_lease(data->o_res.server, data->timestamp);
2302                 data->rpc_done = true;
2303         }
2304 }
2305
2306 static void nfs4_open_confirm_release(void *calldata)
2307 {
2308         struct nfs4_opendata *data = calldata;
2309         struct nfs4_state *state = NULL;
2310
2311         /* If this request hasn't been cancelled, do nothing */
2312         if (!data->cancelled)
2313                 goto out_free;
2314         /* In case of error, no cleanup! */
2315         if (!data->rpc_done)
2316                 goto out_free;
2317         state = nfs4_opendata_to_nfs4_state(data);
2318         if (!IS_ERR(state))
2319                 nfs4_close_state(state, data->o_arg.fmode);
2320 out_free:
2321         nfs4_opendata_put(data);
2322 }
2323
2324 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2325         .rpc_call_prepare = nfs4_open_confirm_prepare,
2326         .rpc_call_done = nfs4_open_confirm_done,
2327         .rpc_release = nfs4_open_confirm_release,
2328 };
2329
2330 /*
2331  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2332  */
2333 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2334 {
2335         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2336         struct rpc_task *task;
2337         struct  rpc_message msg = {
2338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2339                 .rpc_argp = &data->c_arg,
2340                 .rpc_resp = &data->c_res,
2341                 .rpc_cred = data->owner->so_cred,
2342         };
2343         struct rpc_task_setup task_setup_data = {
2344                 .rpc_client = server->client,
2345                 .rpc_message = &msg,
2346                 .callback_ops = &nfs4_open_confirm_ops,
2347                 .callback_data = data,
2348                 .workqueue = nfsiod_workqueue,
2349                 .flags = RPC_TASK_ASYNC,
2350         };
2351         int status;
2352
2353         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2354                                 data->is_recover);
2355         kref_get(&data->kref);
2356         data->rpc_done = false;
2357         data->rpc_status = 0;
2358         data->timestamp = jiffies;
2359         task = rpc_run_task(&task_setup_data);
2360         if (IS_ERR(task))
2361                 return PTR_ERR(task);
2362         status = rpc_wait_for_completion_task(task);
2363         if (status != 0) {
2364                 data->cancelled = true;
2365                 smp_wmb();
2366         } else
2367                 status = data->rpc_status;
2368         rpc_put_task(task);
2369         return status;
2370 }
2371
2372 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2373 {
2374         struct nfs4_opendata *data = calldata;
2375         struct nfs4_state_owner *sp = data->owner;
2376         struct nfs_client *clp = sp->so_server->nfs_client;
2377         enum open_claim_type4 claim = data->o_arg.claim;
2378
2379         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2380                 goto out_wait;
2381         /*
2382          * Check if we still need to send an OPEN call, or if we can use
2383          * a delegation instead.
2384          */
2385         if (data->state != NULL) {
2386                 struct nfs_delegation *delegation;
2387
2388                 if (can_open_cached(data->state, data->o_arg.fmode,
2389                                         data->o_arg.open_flags, claim))
2390                         goto out_no_action;
2391                 rcu_read_lock();
2392                 delegation = nfs4_get_valid_delegation(data->state->inode);
2393                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2394                         goto unlock_no_action;
2395                 rcu_read_unlock();
2396         }
2397         /* Update client id. */
2398         data->o_arg.clientid = clp->cl_clientid;
2399         switch (claim) {
2400         default:
2401                 break;
2402         case NFS4_OPEN_CLAIM_PREVIOUS:
2403         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2404         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2405                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2406                 /* Fall through */
2407         case NFS4_OPEN_CLAIM_FH:
2408                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2409         }
2410         data->timestamp = jiffies;
2411         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2412                                 &data->o_arg.seq_args,
2413                                 &data->o_res.seq_res,
2414                                 task) != 0)
2415                 nfs_release_seqid(data->o_arg.seqid);
2416
2417         /* Set the create mode (note dependency on the session type) */
2418         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2419         if (data->o_arg.open_flags & O_EXCL) {
2420                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2421                 if (nfs4_has_persistent_session(clp))
2422                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2423                 else if (clp->cl_mvops->minor_version > 0)
2424                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2425         }
2426         return;
2427 unlock_no_action:
2428         trace_nfs4_cached_open(data->state);
2429         rcu_read_unlock();
2430 out_no_action:
2431         task->tk_action = NULL;
2432 out_wait:
2433         nfs4_sequence_done(task, &data->o_res.seq_res);
2434 }
2435
2436 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2437 {
2438         struct nfs4_opendata *data = calldata;
2439
2440         data->rpc_status = task->tk_status;
2441
2442         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2443                 return;
2444
2445         if (task->tk_status == 0) {
2446                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2447                         switch (data->o_res.f_attr->mode & S_IFMT) {
2448                         case S_IFREG:
2449                                 break;
2450                         case S_IFLNK:
2451                                 data->rpc_status = -ELOOP;
2452                                 break;
2453                         case S_IFDIR:
2454                                 data->rpc_status = -EISDIR;
2455                                 break;
2456                         default:
2457                                 data->rpc_status = -ENOTDIR;
2458                         }
2459                 }
2460                 renew_lease(data->o_res.server, data->timestamp);
2461                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2462                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2463         }
2464         data->rpc_done = true;
2465 }
2466
2467 static void nfs4_open_release(void *calldata)
2468 {
2469         struct nfs4_opendata *data = calldata;
2470         struct nfs4_state *state = NULL;
2471
2472         /* If this request hasn't been cancelled, do nothing */
2473         if (!data->cancelled)
2474                 goto out_free;
2475         /* In case of error, no cleanup! */
2476         if (data->rpc_status != 0 || !data->rpc_done)
2477                 goto out_free;
2478         /* In case we need an open_confirm, no cleanup! */
2479         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2480                 goto out_free;
2481         state = nfs4_opendata_to_nfs4_state(data);
2482         if (!IS_ERR(state))
2483                 nfs4_close_state(state, data->o_arg.fmode);
2484 out_free:
2485         nfs4_opendata_put(data);
2486 }
2487
2488 static const struct rpc_call_ops nfs4_open_ops = {
2489         .rpc_call_prepare = nfs4_open_prepare,
2490         .rpc_call_done = nfs4_open_done,
2491         .rpc_release = nfs4_open_release,
2492 };
2493
2494 static int nfs4_run_open_task(struct nfs4_opendata *data,
2495                               struct nfs_open_context *ctx)
2496 {
2497         struct inode *dir = d_inode(data->dir);
2498         struct nfs_server *server = NFS_SERVER(dir);
2499         struct nfs_openargs *o_arg = &data->o_arg;
2500         struct nfs_openres *o_res = &data->o_res;
2501         struct rpc_task *task;
2502         struct rpc_message msg = {
2503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2504                 .rpc_argp = o_arg,
2505                 .rpc_resp = o_res,
2506                 .rpc_cred = data->owner->so_cred,
2507         };
2508         struct rpc_task_setup task_setup_data = {
2509                 .rpc_client = server->client,
2510                 .rpc_message = &msg,
2511                 .callback_ops = &nfs4_open_ops,
2512                 .callback_data = data,
2513                 .workqueue = nfsiod_workqueue,
2514                 .flags = RPC_TASK_ASYNC,
2515         };
2516         int status;
2517
2518         kref_get(&data->kref);
2519         data->rpc_done = false;
2520         data->rpc_status = 0;
2521         data->cancelled = false;
2522         data->is_recover = false;
2523         if (!ctx) {
2524                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2525                 data->is_recover = true;
2526                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2527         } else {
2528                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2529                 pnfs_lgopen_prepare(data, ctx);
2530         }
2531         task = rpc_run_task(&task_setup_data);
2532         if (IS_ERR(task))
2533                 return PTR_ERR(task);
2534         status = rpc_wait_for_completion_task(task);
2535         if (status != 0) {
2536                 data->cancelled = true;
2537                 smp_wmb();
2538         } else
2539                 status = data->rpc_status;
2540         rpc_put_task(task);
2541
2542         return status;
2543 }
2544
2545 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2546 {
2547         struct inode *dir = d_inode(data->dir);
2548         struct nfs_openres *o_res = &data->o_res;
2549         int status;
2550
2551         status = nfs4_run_open_task(data, NULL);
2552         if (status != 0 || !data->rpc_done)
2553                 return status;
2554
2555         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2556
2557         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2558                 status = _nfs4_proc_open_confirm(data);
2559
2560         return status;
2561 }
2562
2563 /*
2564  * Additional permission checks in order to distinguish between an
2565  * open for read, and an open for execute. This works around the
2566  * fact that NFSv4 OPEN treats read and execute permissions as being
2567  * the same.
2568  * Note that in the non-execute case, we want to turn off permission
2569  * checking if we just created a new file (POSIX open() semantics).
2570  */
2571 static int nfs4_opendata_access(const struct cred *cred,
2572                                 struct nfs4_opendata *opendata,
2573                                 struct nfs4_state *state, fmode_t fmode,
2574                                 int openflags)
2575 {
2576         struct nfs_access_entry cache;
2577         u32 mask, flags;
2578
2579         /* access call failed or for some reason the server doesn't
2580          * support any access modes -- defer access call until later */
2581         if (opendata->o_res.access_supported == 0)
2582                 return 0;
2583
2584         mask = 0;
2585         /*
2586          * Use openflags to check for exec, because fmode won't
2587          * always have FMODE_EXEC set when file open for exec.
2588          */
2589         if (openflags & __FMODE_EXEC) {
2590                 /* ONLY check for exec rights */
2591                 if (S_ISDIR(state->inode->i_mode))
2592                         mask = NFS4_ACCESS_LOOKUP;
2593                 else
2594                         mask = NFS4_ACCESS_EXECUTE;
2595         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2596                 mask = NFS4_ACCESS_READ;
2597
2598         cache.cred = cred;
2599         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2600         nfs_access_add_cache(state->inode, &cache);
2601
2602         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2603         if ((mask & ~cache.mask & flags) == 0)
2604                 return 0;
2605
2606         return -EACCES;
2607 }
2608
2609 /*
2610  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2611  */
2612 static int _nfs4_proc_open(struct nfs4_opendata *data,
2613                            struct nfs_open_context *ctx)
2614 {
2615         struct inode *dir = d_inode(data->dir);
2616         struct nfs_server *server = NFS_SERVER(dir);
2617         struct nfs_openargs *o_arg = &data->o_arg;
2618         struct nfs_openres *o_res = &data->o_res;
2619         int status;
2620
2621         status = nfs4_run_open_task(data, ctx);
2622         if (!data->rpc_done)
2623                 return status;
2624         if (status != 0) {
2625                 if (status == -NFS4ERR_BADNAME &&
2626                                 !(o_arg->open_flags & O_CREAT))
2627                         return -ENOENT;
2628                 return status;
2629         }
2630
2631         nfs_fattr_map_and_free_names(server, &data->f_attr);
2632
2633         if (o_arg->open_flags & O_CREAT) {
2634                 if (o_arg->open_flags & O_EXCL)
2635                         data->file_created = true;
2636                 else if (o_res->cinfo.before != o_res->cinfo.after)
2637                         data->file_created = true;
2638                 if (data->file_created ||
2639                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2640                         update_changeattr(dir, &o_res->cinfo,
2641                                         o_res->f_attr->time_start, 0);
2642         }
2643         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2644                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2645         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2646                 status = _nfs4_proc_open_confirm(data);
2647                 if (status != 0)
2648                         return status;
2649         }
2650         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2651                 nfs4_sequence_free_slot(&o_res->seq_res);
2652                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2653                                 o_res->f_label, NULL);
2654         }
2655         return 0;
2656 }
2657
2658 /*
2659  * OPEN_EXPIRED:
2660  *      reclaim state on the server after a network partition.
2661  *      Assumes caller holds the appropriate lock
2662  */
2663 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2664 {
2665         struct nfs4_opendata *opendata;
2666         int ret;
2667
2668         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2669                         NFS4_OPEN_CLAIM_FH);
2670         if (IS_ERR(opendata))
2671                 return PTR_ERR(opendata);
2672         ret = nfs4_open_recover(opendata, state);
2673         if (ret == -ESTALE)
2674                 d_drop(ctx->dentry);
2675         nfs4_opendata_put(opendata);
2676         return ret;
2677 }
2678
2679 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2680 {
2681         struct nfs_server *server = NFS_SERVER(state->inode);
2682         struct nfs4_exception exception = { };
2683         int err;
2684
2685         do {
2686                 err = _nfs4_open_expired(ctx, state);
2687                 trace_nfs4_open_expired(ctx, 0, err);
2688                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2689                         continue;
2690                 switch (err) {
2691                 default:
2692                         goto out;
2693                 case -NFS4ERR_GRACE:
2694                 case -NFS4ERR_DELAY:
2695                         nfs4_handle_exception(server, err, &exception);
2696                         err = 0;
2697                 }
2698         } while (exception.retry);
2699 out:
2700         return err;
2701 }
2702
2703 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2704 {
2705         struct nfs_open_context *ctx;
2706         int ret;
2707
2708         ctx = nfs4_state_find_open_context(state);
2709         if (IS_ERR(ctx))
2710                 return -EAGAIN;
2711         ret = nfs4_do_open_expired(ctx, state);
2712         put_nfs_open_context(ctx);
2713         return ret;
2714 }
2715
2716 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2717                 const nfs4_stateid *stateid)
2718 {
2719         nfs_remove_bad_delegation(state->inode, stateid);
2720         nfs_state_clear_delegation(state);
2721 }
2722
2723 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2724 {
2725         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2726                 nfs_finish_clear_delegation_stateid(state, NULL);
2727 }
2728
2729 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2730 {
2731         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2732         nfs40_clear_delegation_stateid(state);
2733         nfs_state_clear_open_state_flags(state);
2734         return nfs4_open_expired(sp, state);
2735 }
2736
2737 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2738                 nfs4_stateid *stateid,
2739                 const struct cred *cred)
2740 {
2741         return -NFS4ERR_BAD_STATEID;
2742 }
2743
2744 #if defined(CONFIG_NFS_V4_1)
2745 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2746                 nfs4_stateid *stateid,
2747                 const struct cred *cred)
2748 {
2749         int status;
2750
2751         switch (stateid->type) {
2752         default:
2753                 break;
2754         case NFS4_INVALID_STATEID_TYPE:
2755         case NFS4_SPECIAL_STATEID_TYPE:
2756                 return -NFS4ERR_BAD_STATEID;
2757         case NFS4_REVOKED_STATEID_TYPE:
2758                 goto out_free;
2759         }
2760
2761         status = nfs41_test_stateid(server, stateid, cred);
2762         switch (status) {
2763         case -NFS4ERR_EXPIRED:
2764         case -NFS4ERR_ADMIN_REVOKED:
2765         case -NFS4ERR_DELEG_REVOKED:
2766                 break;
2767         default:
2768                 return status;
2769         }
2770 out_free:
2771         /* Ack the revoked state to the server */
2772         nfs41_free_stateid(server, stateid, cred, true);
2773         return -NFS4ERR_EXPIRED;
2774 }
2775
2776 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2777 {
2778         struct nfs_server *server = NFS_SERVER(state->inode);
2779         nfs4_stateid stateid;
2780         struct nfs_delegation *delegation;
2781         const struct cred *cred = NULL;
2782         int status, ret = NFS_OK;
2783
2784         /* Get the delegation credential for use by test/free_stateid */
2785         rcu_read_lock();
2786         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2787         if (delegation == NULL) {
2788                 rcu_read_unlock();
2789                 nfs_state_clear_delegation(state);
2790                 return NFS_OK;
2791         }
2792
2793         nfs4_stateid_copy(&stateid, &delegation->stateid);
2794
2795         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2796                                 &delegation->flags)) {
2797                 rcu_read_unlock();
2798                 return NFS_OK;
2799         }
2800
2801         if (delegation->cred)
2802                 cred = get_cred(delegation->cred);
2803         rcu_read_unlock();
2804         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2805         trace_nfs4_test_delegation_stateid(state, NULL, status);
2806         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2807                 nfs_finish_clear_delegation_stateid(state, &stateid);
2808         else
2809                 ret = status;
2810
2811         put_cred(cred);
2812         return ret;
2813 }
2814
2815 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2816 {
2817         nfs4_stateid tmp;
2818
2819         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2820             nfs4_copy_delegation_stateid(state->inode, state->state,
2821                                 &tmp, NULL) &&
2822             nfs4_stateid_match_other(&state->stateid, &tmp))
2823                 nfs_state_set_delegation(state, &tmp, state->state);
2824         else
2825                 nfs_state_clear_delegation(state);
2826 }
2827
2828 /**
2829  * nfs41_check_expired_locks - possibly free a lock stateid
2830  *
2831  * @state: NFSv4 state for an inode
2832  *
2833  * Returns NFS_OK if recovery for this stateid is now finished.
2834  * Otherwise a negative NFS4ERR value is returned.
2835  */
2836 static int nfs41_check_expired_locks(struct nfs4_state *state)
2837 {
2838         int status, ret = NFS_OK;
2839         struct nfs4_lock_state *lsp, *prev = NULL;
2840         struct nfs_server *server = NFS_SERVER(state->inode);
2841
2842         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2843                 goto out;
2844
2845         spin_lock(&state->state_lock);
2846         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2847                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2848                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2849
2850                         refcount_inc(&lsp->ls_count);
2851                         spin_unlock(&state->state_lock);
2852
2853                         nfs4_put_lock_state(prev);
2854                         prev = lsp;
2855
2856                         status = nfs41_test_and_free_expired_stateid(server,
2857                                         &lsp->ls_stateid,
2858                                         cred);
2859                         trace_nfs4_test_lock_stateid(state, lsp, status);
2860                         if (status == -NFS4ERR_EXPIRED ||
2861                             status == -NFS4ERR_BAD_STATEID) {
2862                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2863                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2864                                 if (!recover_lost_locks)
2865                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2866                         } else if (status != NFS_OK) {
2867                                 ret = status;
2868                                 nfs4_put_lock_state(prev);
2869                                 goto out;
2870                         }
2871                         spin_lock(&state->state_lock);
2872                 }
2873         }
2874         spin_unlock(&state->state_lock);
2875         nfs4_put_lock_state(prev);
2876 out:
2877         return ret;
2878 }
2879
2880 /**
2881  * nfs41_check_open_stateid - possibly free an open stateid
2882  *
2883  * @state: NFSv4 state for an inode
2884  *
2885  * Returns NFS_OK if recovery for this stateid is now finished.
2886  * Otherwise a negative NFS4ERR value is returned.
2887  */
2888 static int nfs41_check_open_stateid(struct nfs4_state *state)
2889 {
2890         struct nfs_server *server = NFS_SERVER(state->inode);
2891         nfs4_stateid *stateid = &state->open_stateid;
2892         const struct cred *cred = state->owner->so_cred;
2893         int status;
2894
2895         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2896                 return -NFS4ERR_BAD_STATEID;
2897         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2898         trace_nfs4_test_open_stateid(state, NULL, status);
2899         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2900                 nfs_state_clear_open_state_flags(state);
2901                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2902                 return status;
2903         }
2904         if (nfs_open_stateid_recover_openmode(state))
2905                 return -NFS4ERR_OPENMODE;
2906         return NFS_OK;
2907 }
2908
2909 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2910 {
2911         int status;
2912
2913         status = nfs41_check_delegation_stateid(state);
2914         if (status != NFS_OK)
2915                 return status;
2916         nfs41_delegation_recover_stateid(state);
2917
2918         status = nfs41_check_expired_locks(state);
2919         if (status != NFS_OK)
2920                 return status;
2921         status = nfs41_check_open_stateid(state);
2922         if (status != NFS_OK)
2923                 status = nfs4_open_expired(sp, state);
2924         return status;
2925 }
2926 #endif
2927
2928 /*
2929  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2930  * fields corresponding to attributes that were used to store the verifier.
2931  * Make sure we clobber those fields in the later setattr call
2932  */
2933 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2934                                 struct iattr *sattr, struct nfs4_label **label)
2935 {
2936         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2937         __u32 attrset[3];
2938         unsigned ret;
2939         unsigned i;
2940
2941         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2942                 attrset[i] = opendata->o_res.attrset[i];
2943                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2944                         attrset[i] &= ~bitmask[i];
2945         }
2946
2947         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2948                 sattr->ia_valid : 0;
2949
2950         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2951                 if (sattr->ia_valid & ATTR_ATIME_SET)
2952                         ret |= ATTR_ATIME_SET;
2953                 else
2954                         ret |= ATTR_ATIME;
2955         }
2956
2957         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2958                 if (sattr->ia_valid & ATTR_MTIME_SET)
2959                         ret |= ATTR_MTIME_SET;
2960                 else
2961                         ret |= ATTR_MTIME;
2962         }
2963
2964         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2965                 *label = NULL;
2966         return ret;
2967 }
2968
2969 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2970                 int flags, struct nfs_open_context *ctx)
2971 {
2972         struct nfs4_state_owner *sp = opendata->owner;
2973         struct nfs_server *server = sp->so_server;
2974         struct dentry *dentry;
2975         struct nfs4_state *state;
2976         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2977         struct inode *dir = d_inode(opendata->dir);
2978         unsigned long dir_verifier;
2979         unsigned int seq;
2980         int ret;
2981
2982         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2983         dir_verifier = nfs_save_change_attribute(dir);
2984
2985         ret = _nfs4_proc_open(opendata, ctx);
2986         if (ret != 0)
2987                 goto out;
2988
2989         state = _nfs4_opendata_to_nfs4_state(opendata);
2990         ret = PTR_ERR(state);
2991         if (IS_ERR(state))
2992                 goto out;
2993         ctx->state = state;
2994         if (server->caps & NFS_CAP_POSIX_LOCK)
2995                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2996         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2997                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2998
2999         dentry = opendata->dentry;
3000         if (d_really_is_negative(dentry)) {
3001                 struct dentry *alias;
3002                 d_drop(dentry);
3003                 alias = d_exact_alias(dentry, state->inode);
3004                 if (!alias)
3005                         alias = d_splice_alias(igrab(state->inode), dentry);
3006                 /* d_splice_alias() can't fail here - it's a non-directory */
3007                 if (alias) {
3008                         dput(ctx->dentry);
3009                         ctx->dentry = dentry = alias;
3010                 }
3011         }
3012
3013         switch(opendata->o_arg.claim) {
3014         default:
3015                 break;
3016         case NFS4_OPEN_CLAIM_NULL:
3017         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3018         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3019                 if (!opendata->rpc_done)
3020                         break;
3021                 if (opendata->o_res.delegation_type != 0)
3022                         dir_verifier = nfs_save_change_attribute(dir);
3023                 nfs_set_verifier(dentry, dir_verifier);
3024         }
3025
3026         /* Parse layoutget results before we check for access */
3027         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3028
3029         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3030                         acc_mode, flags);
3031         if (ret != 0)
3032                 goto out;
3033
3034         if (d_inode(dentry) == state->inode) {
3035                 nfs_inode_attach_open_context(ctx);
3036                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3037                         nfs4_schedule_stateid_recovery(server, state);
3038         }
3039
3040 out:
3041         if (!opendata->cancelled)
3042                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3043         return ret;
3044 }
3045
3046 /*
3047  * Returns a referenced nfs4_state
3048  */
3049 static int _nfs4_do_open(struct inode *dir,
3050                         struct nfs_open_context *ctx,
3051                         int flags,
3052                         const struct nfs4_open_createattrs *c,
3053                         int *opened)
3054 {
3055         struct nfs4_state_owner  *sp;
3056         struct nfs4_state     *state = NULL;
3057         struct nfs_server       *server = NFS_SERVER(dir);
3058         struct nfs4_opendata *opendata;
3059         struct dentry *dentry = ctx->dentry;
3060         const struct cred *cred = ctx->cred;
3061         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3062         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3063         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3064         struct iattr *sattr = c->sattr;
3065         struct nfs4_label *label = c->label;
3066         struct nfs4_label *olabel = NULL;
3067         int status;
3068
3069         /* Protect against reboot recovery conflicts */
3070         status = -ENOMEM;
3071         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3072         if (sp == NULL) {
3073                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3074                 goto out_err;
3075         }
3076         status = nfs4_client_recover_expired_lease(server->nfs_client);
3077         if (status != 0)
3078                 goto err_put_state_owner;
3079         if (d_really_is_positive(dentry))
3080                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3081         status = -ENOMEM;
3082         if (d_really_is_positive(dentry))
3083                 claim = NFS4_OPEN_CLAIM_FH;
3084         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3085                         c, claim, GFP_KERNEL);
3086         if (opendata == NULL)
3087                 goto err_put_state_owner;
3088
3089         if (label) {
3090                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3091                 if (IS_ERR(olabel)) {
3092                         status = PTR_ERR(olabel);
3093                         goto err_opendata_put;
3094                 }
3095         }
3096
3097         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3098                 if (!opendata->f_attr.mdsthreshold) {
3099                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3100                         if (!opendata->f_attr.mdsthreshold)
3101                                 goto err_free_label;
3102                 }
3103                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3104         }
3105         if (d_really_is_positive(dentry))
3106                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3107
3108         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3109         if (status != 0)
3110                 goto err_free_label;
3111         state = ctx->state;
3112
3113         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3114             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3115                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3116                 /*
3117                  * send create attributes which was not set by open
3118                  * with an extra setattr.
3119                  */
3120                 if (attrs || label) {
3121                         unsigned ia_old = sattr->ia_valid;
3122
3123                         sattr->ia_valid = attrs;
3124                         nfs_fattr_init(opendata->o_res.f_attr);
3125                         status = nfs4_do_setattr(state->inode, cred,
3126                                         opendata->o_res.f_attr, sattr,
3127                                         ctx, label, olabel);
3128                         if (status == 0) {
3129                                 nfs_setattr_update_inode(state->inode, sattr,
3130                                                 opendata->o_res.f_attr);
3131                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3132                         }
3133                         sattr->ia_valid = ia_old;
3134                 }
3135         }
3136         if (opened && opendata->file_created)
3137                 *opened = 1;
3138
3139         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3140                 *ctx_th = opendata->f_attr.mdsthreshold;
3141                 opendata->f_attr.mdsthreshold = NULL;
3142         }
3143
3144         nfs4_label_free(olabel);
3145
3146         nfs4_opendata_put(opendata);
3147         nfs4_put_state_owner(sp);
3148         return 0;
3149 err_free_label:
3150         nfs4_label_free(olabel);
3151 err_opendata_put:
3152         nfs4_opendata_put(opendata);
3153 err_put_state_owner:
3154         nfs4_put_state_owner(sp);
3155 out_err:
3156         return status;
3157 }
3158
3159
3160 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3161                                         struct nfs_open_context *ctx,
3162                                         int flags,
3163                                         struct iattr *sattr,
3164                                         struct nfs4_label *label,
3165                                         int *opened)
3166 {
3167         struct nfs_server *server = NFS_SERVER(dir);
3168         struct nfs4_exception exception = {
3169                 .interruptible = true,
3170         };
3171         struct nfs4_state *res;
3172         struct nfs4_open_createattrs c = {
3173                 .label = label,
3174                 .sattr = sattr,
3175                 .verf = {
3176                         [0] = (__u32)jiffies,
3177                         [1] = (__u32)current->pid,
3178                 },
3179         };
3180         int status;
3181
3182         do {
3183                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3184                 res = ctx->state;
3185                 trace_nfs4_open_file(ctx, flags, status);
3186                 if (status == 0)
3187                         break;
3188                 /* NOTE: BAD_SEQID means the server and client disagree about the
3189                  * book-keeping w.r.t. state-changing operations
3190                  * (OPEN/CLOSE/LOCK/LOCKU...)
3191                  * It is actually a sign of a bug on the client or on the server.
3192                  *
3193                  * If we receive a BAD_SEQID error in the particular case of
3194                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3195                  * have unhashed the old state_owner for us, and that we can
3196                  * therefore safely retry using a new one. We should still warn
3197                  * the user though...
3198                  */
3199                 if (status == -NFS4ERR_BAD_SEQID) {
3200                         pr_warn_ratelimited("NFS: v4 server %s "
3201                                         " returned a bad sequence-id error!\n",
3202                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3203                         exception.retry = 1;
3204                         continue;
3205                 }
3206                 /*
3207                  * BAD_STATEID on OPEN means that the server cancelled our
3208                  * state before it received the OPEN_CONFIRM.
3209                  * Recover by retrying the request as per the discussion
3210                  * on Page 181 of RFC3530.
3211                  */
3212                 if (status == -NFS4ERR_BAD_STATEID) {
3213                         exception.retry = 1;
3214                         continue;
3215                 }
3216                 if (status == -NFS4ERR_EXPIRED) {
3217                         nfs4_schedule_lease_recovery(server->nfs_client);
3218                         exception.retry = 1;
3219                         continue;
3220                 }
3221                 if (status == -EAGAIN) {
3222                         /* We must have found a delegation */
3223                         exception.retry = 1;
3224                         continue;
3225                 }
3226                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3227                         continue;
3228                 res = ERR_PTR(nfs4_handle_exception(server,
3229                                         status, &exception));
3230         } while (exception.retry);
3231         return res;
3232 }
3233
3234 static int _nfs4_do_setattr(struct inode *inode,
3235                             struct nfs_setattrargs *arg,
3236                             struct nfs_setattrres *res,
3237                             const struct cred *cred,
3238                             struct nfs_open_context *ctx)
3239 {
3240         struct nfs_server *server = NFS_SERVER(inode);
3241         struct rpc_message msg = {
3242                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3243                 .rpc_argp       = arg,
3244                 .rpc_resp       = res,
3245                 .rpc_cred       = cred,
3246         };
3247         const struct cred *delegation_cred = NULL;
3248         unsigned long timestamp = jiffies;
3249         bool truncate;
3250         int status;
3251
3252         nfs_fattr_init(res->fattr);
3253
3254         /* Servers should only apply open mode checks for file size changes */
3255         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3256         if (!truncate)
3257                 goto zero_stateid;
3258
3259         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3260                 /* Use that stateid */
3261         } else if (ctx != NULL && ctx->state) {
3262                 struct nfs_lock_context *l_ctx;
3263                 if (!nfs4_valid_open_stateid(ctx->state))
3264                         return -EBADF;
3265                 l_ctx = nfs_get_lock_context(ctx);
3266                 if (IS_ERR(l_ctx))
3267                         return PTR_ERR(l_ctx);
3268                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3269                                                 &arg->stateid, &delegation_cred);
3270                 nfs_put_lock_context(l_ctx);
3271                 if (status == -EIO)
3272                         return -EBADF;
3273                 else if (status == -EAGAIN)
3274                         goto zero_stateid;
3275         } else {
3276 zero_stateid:
3277                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3278         }
3279         if (delegation_cred)
3280                 msg.rpc_cred = delegation_cred;
3281
3282         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3283
3284         put_cred(delegation_cred);
3285         if (status == 0 && ctx != NULL)
3286                 renew_lease(server, timestamp);
3287         trace_nfs4_setattr(inode, &arg->stateid, status);
3288         return status;
3289 }
3290
3291 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3292                            struct nfs_fattr *fattr, struct iattr *sattr,
3293                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3294                            struct nfs4_label *olabel)
3295 {
3296         struct nfs_server *server = NFS_SERVER(inode);
3297         __u32 bitmask[NFS4_BITMASK_SZ];
3298         struct nfs4_state *state = ctx ? ctx->state : NULL;
3299         struct nfs_setattrargs  arg = {
3300                 .fh             = NFS_FH(inode),
3301                 .iap            = sattr,
3302                 .server         = server,
3303                 .bitmask = bitmask,
3304                 .label          = ilabel,
3305         };
3306         struct nfs_setattrres  res = {
3307                 .fattr          = fattr,
3308                 .label          = olabel,
3309                 .server         = server,
3310         };
3311         struct nfs4_exception exception = {
3312                 .state = state,
3313                 .inode = inode,
3314                 .stateid = &arg.stateid,
3315         };
3316         int err;
3317
3318         do {
3319                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3320                                 nfs4_bitmask(server, olabel),
3321                                 inode);
3322
3323                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3324                 switch (err) {
3325                 case -NFS4ERR_OPENMODE:
3326                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3327                                 pr_warn_once("NFSv4: server %s is incorrectly "
3328                                                 "applying open mode checks to "
3329                                                 "a SETATTR that is not "
3330                                                 "changing file size.\n",
3331                                                 server->nfs_client->cl_hostname);
3332                         }
3333                         if (state && !(state->state & FMODE_WRITE)) {
3334                                 err = -EBADF;
3335                                 if (sattr->ia_valid & ATTR_OPEN)
3336                                         err = -EACCES;
3337                                 goto out;
3338                         }
3339                 }
3340                 err = nfs4_handle_exception(server, err, &exception);
3341         } while (exception.retry);
3342 out:
3343         return err;
3344 }
3345
3346 static bool
3347 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3348 {
3349         if (inode == NULL || !nfs_have_layout(inode))
3350                 return false;
3351
3352         return pnfs_wait_on_layoutreturn(inode, task);
3353 }
3354
3355 /*
3356  * Update the seqid of an open stateid
3357  */
3358 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3359                 struct nfs4_state *state)
3360 {
3361         __be32 seqid_open;
3362         u32 dst_seqid;
3363         int seq;
3364
3365         for (;;) {
3366                 if (!nfs4_valid_open_stateid(state))
3367                         break;
3368                 seq = read_seqbegin(&state->seqlock);
3369                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3370                         nfs4_stateid_copy(dst, &state->open_stateid);
3371                         if (read_seqretry(&state->seqlock, seq))
3372                                 continue;
3373                         break;
3374                 }
3375                 seqid_open = state->open_stateid.seqid;
3376                 if (read_seqretry(&state->seqlock, seq))
3377                         continue;
3378
3379                 dst_seqid = be32_to_cpu(dst->seqid);
3380                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3381                         dst->seqid = seqid_open;
3382                 break;
3383         }
3384 }
3385
3386 /*
3387  * Update the seqid of an open stateid after receiving
3388  * NFS4ERR_OLD_STATEID
3389  */
3390 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3391                 struct nfs4_state *state)
3392 {
3393         __be32 seqid_open;
3394         u32 dst_seqid;
3395         bool ret;
3396         int seq;
3397
3398         for (;;) {
3399                 ret = false;
3400                 if (!nfs4_valid_open_stateid(state))
3401                         break;
3402                 seq = read_seqbegin(&state->seqlock);
3403                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3404                         if (read_seqretry(&state->seqlock, seq))
3405                                 continue;
3406                         break;
3407                 }
3408                 seqid_open = state->open_stateid.seqid;
3409                 if (read_seqretry(&state->seqlock, seq))
3410                         continue;
3411
3412                 dst_seqid = be32_to_cpu(dst->seqid);
3413                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) >= 0)
3414                         dst->seqid = cpu_to_be32(dst_seqid + 1);
3415                 else
3416                         dst->seqid = seqid_open;
3417                 ret = true;
3418                 break;
3419         }
3420
3421         return ret;
3422 }
3423
3424 struct nfs4_closedata {
3425         struct inode *inode;
3426         struct nfs4_state *state;
3427         struct nfs_closeargs arg;
3428         struct nfs_closeres res;
3429         struct {
3430                 struct nfs4_layoutreturn_args arg;
3431                 struct nfs4_layoutreturn_res res;
3432                 struct nfs4_xdr_opaque_data ld_private;
3433                 u32 roc_barrier;
3434                 bool roc;
3435         } lr;
3436         struct nfs_fattr fattr;
3437         unsigned long timestamp;
3438 };
3439
3440 static void nfs4_free_closedata(void *data)
3441 {
3442         struct nfs4_closedata *calldata = data;
3443         struct nfs4_state_owner *sp = calldata->state->owner;
3444         struct super_block *sb = calldata->state->inode->i_sb;
3445
3446         if (calldata->lr.roc)
3447                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3448                                 calldata->res.lr_ret);
3449         nfs4_put_open_state(calldata->state);
3450         nfs_free_seqid(calldata->arg.seqid);
3451         nfs4_put_state_owner(sp);
3452         nfs_sb_deactive(sb);
3453         kfree(calldata);
3454 }
3455
3456 static void nfs4_close_done(struct rpc_task *task, void *data)
3457 {
3458         struct nfs4_closedata *calldata = data;
3459         struct nfs4_state *state = calldata->state;
3460         struct nfs_server *server = NFS_SERVER(calldata->inode);
3461         nfs4_stateid *res_stateid = NULL;
3462         struct nfs4_exception exception = {
3463                 .state = state,
3464                 .inode = calldata->inode,
3465                 .stateid = &calldata->arg.stateid,
3466         };
3467
3468         dprintk("%s: begin!\n", __func__);
3469         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3470                 return;
3471         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3472
3473         /* Handle Layoutreturn errors */
3474         if (pnfs_roc_done(task, calldata->inode,
3475                                 &calldata->arg.lr_args,
3476                                 &calldata->res.lr_res,
3477                                 &calldata->res.lr_ret) == -EAGAIN)
3478                 goto out_restart;
3479
3480         /* hmm. we are done with the inode, and in the process of freeing
3481          * the state_owner. we keep this around to process errors
3482          */
3483         switch (task->tk_status) {
3484                 case 0:
3485                         res_stateid = &calldata->res.stateid;
3486                         renew_lease(server, calldata->timestamp);
3487                         break;
3488                 case -NFS4ERR_ACCESS:
3489                         if (calldata->arg.bitmask != NULL) {
3490                                 calldata->arg.bitmask = NULL;
3491                                 calldata->res.fattr = NULL;
3492                                 goto out_restart;
3493
3494                         }
3495                         break;
3496                 case -NFS4ERR_OLD_STATEID:
3497                         /* Did we race with OPEN? */
3498                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3499                                                 state))
3500                                 goto out_restart;
3501                         goto out_release;
3502                 case -NFS4ERR_ADMIN_REVOKED:
3503                 case -NFS4ERR_STALE_STATEID:
3504                 case -NFS4ERR_EXPIRED:
3505                         nfs4_free_revoked_stateid(server,
3506                                         &calldata->arg.stateid,
3507                                         task->tk_msg.rpc_cred);
3508                         /* Fallthrough */
3509                 case -NFS4ERR_BAD_STATEID:
3510                         if (calldata->arg.fmode == 0)
3511                                 break;
3512                         /* Fallthrough */
3513                 default:
3514                         task->tk_status = nfs4_async_handle_exception(task,
3515                                         server, task->tk_status, &exception);
3516                         if (exception.retry)
3517                                 goto out_restart;
3518         }
3519         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3520                         res_stateid, calldata->arg.fmode);
3521 out_release:
3522         task->tk_status = 0;
3523         nfs_release_seqid(calldata->arg.seqid);
3524         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3525         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3526         return;
3527 out_restart:
3528         task->tk_status = 0;
3529         rpc_restart_call_prepare(task);
3530         goto out_release;
3531 }
3532
3533 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3534 {
3535         struct nfs4_closedata *calldata = data;
3536         struct nfs4_state *state = calldata->state;
3537         struct inode *inode = calldata->inode;
3538         struct pnfs_layout_hdr *lo;
3539         bool is_rdonly, is_wronly, is_rdwr;
3540         int call_close = 0;
3541
3542         dprintk("%s: begin!\n", __func__);
3543         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3544                 goto out_wait;
3545
3546         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3547         spin_lock(&state->owner->so_lock);
3548         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3549         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3550         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3551         /* Calculate the change in open mode */
3552         calldata->arg.fmode = 0;
3553         if (state->n_rdwr == 0) {
3554                 if (state->n_rdonly == 0)
3555                         call_close |= is_rdonly;
3556                 else if (is_rdonly)
3557                         calldata->arg.fmode |= FMODE_READ;
3558                 if (state->n_wronly == 0)
3559                         call_close |= is_wronly;
3560                 else if (is_wronly)
3561                         calldata->arg.fmode |= FMODE_WRITE;
3562                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3563                         call_close |= is_rdwr;
3564         } else if (is_rdwr)
3565                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3566
3567         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3568         if (!nfs4_valid_open_stateid(state))
3569                 call_close = 0;
3570         spin_unlock(&state->owner->so_lock);
3571
3572         if (!call_close) {
3573                 /* Note: exit _without_ calling nfs4_close_done */
3574                 goto out_no_action;
3575         }
3576
3577         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3578                 nfs_release_seqid(calldata->arg.seqid);
3579                 goto out_wait;
3580         }
3581
3582         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3583         if (lo && !pnfs_layout_is_valid(lo)) {
3584                 calldata->arg.lr_args = NULL;
3585                 calldata->res.lr_res = NULL;
3586         }
3587
3588         if (calldata->arg.fmode == 0)
3589                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3590
3591         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3592                 /* Close-to-open cache consistency revalidation */
3593                 if (!nfs4_have_delegation(inode, FMODE_READ))
3594                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3595                 else
3596                         calldata->arg.bitmask = NULL;
3597         }
3598
3599         calldata->arg.share_access =
3600                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3601                                 calldata->arg.fmode, 0);
3602
3603         if (calldata->res.fattr == NULL)
3604                 calldata->arg.bitmask = NULL;
3605         else if (calldata->arg.bitmask == NULL)
3606                 calldata->res.fattr = NULL;
3607         calldata->timestamp = jiffies;
3608         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3609                                 &calldata->arg.seq_args,
3610                                 &calldata->res.seq_res,
3611                                 task) != 0)
3612                 nfs_release_seqid(calldata->arg.seqid);
3613         dprintk("%s: done!\n", __func__);
3614         return;
3615 out_no_action:
3616         task->tk_action = NULL;
3617 out_wait:
3618         nfs4_sequence_done(task, &calldata->res.seq_res);
3619 }
3620
3621 static const struct rpc_call_ops nfs4_close_ops = {
3622         .rpc_call_prepare = nfs4_close_prepare,
3623         .rpc_call_done = nfs4_close_done,
3624         .rpc_release = nfs4_free_closedata,
3625 };
3626
3627 /*
3628  * It is possible for data to be read/written from a mem-mapped file
3629  * after the sys_close call (which hits the vfs layer as a flush).
3630  * This means that we can't safely call nfsv4 close on a file until
3631  * the inode is cleared. This in turn means that we are not good
3632  * NFSv4 citizens - we do not indicate to the server to update the file's
3633  * share state even when we are done with one of the three share
3634  * stateid's in the inode.
3635  *
3636  * NOTE: Caller must be holding the sp->so_owner semaphore!
3637  */
3638 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3639 {
3640         struct nfs_server *server = NFS_SERVER(state->inode);
3641         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3642         struct nfs4_closedata *calldata;
3643         struct nfs4_state_owner *sp = state->owner;
3644         struct rpc_task *task;
3645         struct rpc_message msg = {
3646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3647                 .rpc_cred = state->owner->so_cred,
3648         };
3649         struct rpc_task_setup task_setup_data = {
3650                 .rpc_client = server->client,
3651                 .rpc_message = &msg,
3652                 .callback_ops = &nfs4_close_ops,
3653                 .workqueue = nfsiod_workqueue,
3654                 .flags = RPC_TASK_ASYNC,
3655         };
3656         int status = -ENOMEM;
3657
3658         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3659                 &task_setup_data.rpc_client, &msg);
3660
3661         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3662         if (calldata == NULL)
3663                 goto out;
3664         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3665         calldata->inode = state->inode;
3666         calldata->state = state;
3667         calldata->arg.fh = NFS_FH(state->inode);
3668         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3669                 goto out_free_calldata;
3670         /* Serialization for the sequence id */
3671         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3672         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3673         if (IS_ERR(calldata->arg.seqid))
3674                 goto out_free_calldata;
3675         nfs_fattr_init(&calldata->fattr);
3676         calldata->arg.fmode = 0;
3677         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3678         calldata->res.fattr = &calldata->fattr;
3679         calldata->res.seqid = calldata->arg.seqid;
3680         calldata->res.server = server;
3681         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3682         calldata->lr.roc = pnfs_roc(state->inode,
3683                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3684         if (calldata->lr.roc) {
3685                 calldata->arg.lr_args = &calldata->lr.arg;
3686                 calldata->res.lr_res = &calldata->lr.res;
3687         }
3688         nfs_sb_active(calldata->inode->i_sb);
3689
3690         msg.rpc_argp = &calldata->arg;
3691         msg.rpc_resp = &calldata->res;
3692         task_setup_data.callback_data = calldata;
3693         task = rpc_run_task(&task_setup_data);
3694         if (IS_ERR(task))
3695                 return PTR_ERR(task);
3696         status = 0;
3697         if (wait)
3698                 status = rpc_wait_for_completion_task(task);
3699         rpc_put_task(task);
3700         return status;
3701 out_free_calldata:
3702         kfree(calldata);
3703 out:
3704         nfs4_put_open_state(state);
3705         nfs4_put_state_owner(sp);
3706         return status;
3707 }
3708
3709 static struct inode *
3710 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3711                 int open_flags, struct iattr *attr, int *opened)
3712 {
3713         struct nfs4_state *state;
3714         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3715
3716         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3717
3718         /* Protect against concurrent sillydeletes */
3719         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3720
3721         nfs4_label_release_security(label);
3722
3723         if (IS_ERR(state))
3724                 return ERR_CAST(state);
3725         return state->inode;
3726 }
3727
3728 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3729 {
3730         if (ctx->state == NULL)
3731                 return;
3732         if (is_sync)
3733                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3734         else
3735                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3736 }
3737
3738 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3739 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3740 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3741
3742 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3743 {
3744         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3745         struct nfs4_server_caps_arg args = {
3746                 .fhandle = fhandle,
3747                 .bitmask = bitmask,
3748         };
3749         struct nfs4_server_caps_res res = {};
3750         struct rpc_message msg = {
3751                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3752                 .rpc_argp = &args,
3753                 .rpc_resp = &res,
3754         };
3755         int status;
3756         int i;
3757
3758         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3759                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3760                      FATTR4_WORD0_LINK_SUPPORT |
3761                      FATTR4_WORD0_SYMLINK_SUPPORT |
3762                      FATTR4_WORD0_ACLSUPPORT;
3763         if (minorversion)
3764                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3765
3766         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3767         if (status == 0) {
3768                 /* Sanity check the server answers */
3769                 switch (minorversion) {
3770                 case 0:
3771                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3772                         res.attr_bitmask[2] = 0;
3773                         break;
3774                 case 1:
3775                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3776                         break;
3777                 case 2:
3778                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3779                 }
3780                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3781                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3782                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3783                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3784                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3785                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3786                                 NFS_CAP_SECURITY_LABEL);
3787                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3788                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3789                         server->caps |= NFS_CAP_ACLS;
3790                 if (res.has_links != 0)
3791                         server->caps |= NFS_CAP_HARDLINKS;
3792                 if (res.has_symlinks != 0)
3793                         server->caps |= NFS_CAP_SYMLINKS;
3794                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3795                         server->caps |= NFS_CAP_FILEID;
3796                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3797                         server->caps |= NFS_CAP_MODE;
3798                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3799                         server->caps |= NFS_CAP_NLINK;
3800                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3801                         server->caps |= NFS_CAP_OWNER;
3802                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3803                         server->caps |= NFS_CAP_OWNER_GROUP;
3804                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3805                         server->caps |= NFS_CAP_ATIME;
3806                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3807                         server->caps |= NFS_CAP_CTIME;
3808                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3809                         server->caps |= NFS_CAP_MTIME;
3810 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3811                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3812                         server->caps |= NFS_CAP_SECURITY_LABEL;
3813 #endif
3814                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3815                                 sizeof(server->attr_bitmask));
3816                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3817
3818                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3819                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3820                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3821                 server->cache_consistency_bitmask[2] = 0;
3822
3823                 /* Avoid a regression due to buggy server */
3824                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3825                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3826                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3827                         sizeof(server->exclcreat_bitmask));
3828
3829                 server->acl_bitmask = res.acl_bitmask;
3830                 server->fh_expire_type = res.fh_expire_type;
3831         }
3832
3833         return status;
3834 }
3835
3836 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3837 {
3838         struct nfs4_exception exception = {
3839                 .interruptible = true,
3840         };
3841         int err;
3842         do {
3843                 err = nfs4_handle_exception(server,
3844                                 _nfs4_server_capabilities(server, fhandle),
3845                                 &exception);
3846         } while (exception.retry);
3847         return err;
3848 }
3849
3850 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3851                 struct nfs_fsinfo *info)
3852 {
3853         u32 bitmask[3];
3854         struct nfs4_lookup_root_arg args = {
3855                 .bitmask = bitmask,
3856         };
3857         struct nfs4_lookup_res res = {
3858                 .server = server,
3859                 .fattr = info->fattr,
3860                 .fh = fhandle,
3861         };
3862         struct rpc_message msg = {
3863                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3864                 .rpc_argp = &args,
3865                 .rpc_resp = &res,
3866         };
3867
3868         bitmask[0] = nfs4_fattr_bitmap[0];
3869         bitmask[1] = nfs4_fattr_bitmap[1];
3870         /*
3871          * Process the label in the upcoming getfattr
3872          */
3873         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3874
3875         nfs_fattr_init(info->fattr);
3876         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3877 }
3878
3879 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3880                 struct nfs_fsinfo *info)
3881 {
3882         struct nfs4_exception exception = {
3883                 .interruptible = true,
3884         };
3885         int err;
3886         do {
3887                 err = _nfs4_lookup_root(server, fhandle, info);
3888                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3889                 switch (err) {
3890                 case 0:
3891                 case -NFS4ERR_WRONGSEC:
3892                         goto out;
3893                 default:
3894                         err = nfs4_handle_exception(server, err, &exception);
3895                 }
3896         } while (exception.retry);
3897 out:
3898         return err;
3899 }
3900
3901 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3902                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3903 {
3904         struct rpc_auth_create_args auth_args = {
3905                 .pseudoflavor = flavor,
3906         };
3907         struct rpc_auth *auth;
3908
3909         auth = rpcauth_create(&auth_args, server->client);
3910         if (IS_ERR(auth))
3911                 return -EACCES;
3912         return nfs4_lookup_root(server, fhandle, info);
3913 }
3914
3915 /*
3916  * Retry pseudoroot lookup with various security flavors.  We do this when:
3917  *
3918  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3919  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3920  *
3921  * Returns zero on success, or a negative NFS4ERR value, or a
3922  * negative errno value.
3923  */
3924 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3925                               struct nfs_fsinfo *info)
3926 {
3927         /* Per 3530bis 15.33.5 */
3928         static const rpc_authflavor_t flav_array[] = {
3929                 RPC_AUTH_GSS_KRB5P,
3930                 RPC_AUTH_GSS_KRB5I,
3931                 RPC_AUTH_GSS_KRB5,
3932                 RPC_AUTH_UNIX,                  /* courtesy */
3933                 RPC_AUTH_NULL,
3934         };
3935         int status = -EPERM;
3936         size_t i;
3937
3938         if (server->auth_info.flavor_len > 0) {
3939                 /* try each flavor specified by user */
3940                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3941                         status = nfs4_lookup_root_sec(server, fhandle, info,
3942                                                 server->auth_info.flavors[i]);
3943                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3944                                 continue;
3945                         break;
3946                 }
3947         } else {
3948                 /* no flavors specified by user, try default list */
3949                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3950                         status = nfs4_lookup_root_sec(server, fhandle, info,
3951                                                       flav_array[i]);
3952                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3953                                 continue;
3954                         break;
3955                 }
3956         }
3957
3958         /*
3959          * -EACCES could mean that the user doesn't have correct permissions
3960          * to access the mount.  It could also mean that we tried to mount
3961          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3962          * existing mount programs don't handle -EACCES very well so it should
3963          * be mapped to -EPERM instead.
3964          */
3965         if (status == -EACCES)
3966                 status = -EPERM;
3967         return status;
3968 }
3969
3970 /**
3971  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3972  * @server: initialized nfs_server handle
3973  * @fhandle: we fill in the pseudo-fs root file handle
3974  * @info: we fill in an FSINFO struct
3975  * @auth_probe: probe the auth flavours
3976  *
3977  * Returns zero on success, or a negative errno.
3978  */
3979 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3980                          struct nfs_fsinfo *info,
3981                          bool auth_probe)
3982 {
3983         int status = 0;
3984
3985         if (!auth_probe)
3986                 status = nfs4_lookup_root(server, fhandle, info);
3987
3988         if (auth_probe || status == NFS4ERR_WRONGSEC)
3989                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3990                                 fhandle, info);
3991
3992         if (status == 0)
3993                 status = nfs4_server_capabilities(server, fhandle);
3994         if (status == 0)
3995                 status = nfs4_do_fsinfo(server, fhandle, info);
3996
3997         return nfs4_map_errors(status);
3998 }
3999
4000 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4001                               struct nfs_fsinfo *info)
4002 {
4003         int error;
4004         struct nfs_fattr *fattr = info->fattr;
4005         struct nfs4_label *label = fattr->label;
4006
4007         error = nfs4_server_capabilities(server, mntfh);
4008         if (error < 0) {
4009                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4010                 return error;
4011         }
4012
4013         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4014         if (error < 0) {
4015                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4016                 goto out;
4017         }
4018
4019         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4020             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4021                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4022
4023 out:
4024         return error;
4025 }
4026
4027 /*
4028  * Get locations and (maybe) other attributes of a referral.
4029  * Note that we'll actually follow the referral later when
4030  * we detect fsid mismatch in inode revalidation
4031  */
4032 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4033                              const struct qstr *name, struct nfs_fattr *fattr,
4034                              struct nfs_fh *fhandle)
4035 {
4036         int status = -ENOMEM;
4037         struct page *page = NULL;
4038         struct nfs4_fs_locations *locations = NULL;
4039
4040         page = alloc_page(GFP_KERNEL);
4041         if (page == NULL)
4042                 goto out;
4043         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4044         if (locations == NULL)
4045                 goto out;
4046
4047         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4048         if (status != 0)
4049                 goto out;
4050
4051         /*
4052          * If the fsid didn't change, this is a migration event, not a
4053          * referral.  Cause us to drop into the exception handler, which
4054          * will kick off migration recovery.
4055          */
4056         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4057                 dprintk("%s: server did not return a different fsid for"
4058                         " a referral at %s\n", __func__, name->name);
4059                 status = -NFS4ERR_MOVED;
4060                 goto out;
4061         }
4062         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4063         nfs_fixup_referral_attributes(&locations->fattr);
4064
4065         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4066         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4067         memset(fhandle, 0, sizeof(struct nfs_fh));
4068 out:
4069         if (page)
4070                 __free_page(page);
4071         kfree(locations);
4072         return status;
4073 }
4074
4075 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4076                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4077                                 struct inode *inode)
4078 {
4079         __u32 bitmask[NFS4_BITMASK_SZ];
4080         struct nfs4_getattr_arg args = {
4081                 .fh = fhandle,
4082                 .bitmask = bitmask,
4083         };
4084         struct nfs4_getattr_res res = {
4085                 .fattr = fattr,
4086                 .label = label,
4087                 .server = server,
4088         };
4089         struct rpc_message msg = {
4090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4091                 .rpc_argp = &args,
4092                 .rpc_resp = &res,
4093         };
4094         unsigned short task_flags = 0;
4095
4096         /* Is this is an attribute revalidation, subject to softreval? */
4097         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4098                 task_flags |= RPC_TASK_TIMEOUT;
4099
4100         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4101
4102         nfs_fattr_init(fattr);
4103         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4104         return nfs4_do_call_sync(server->client, server, &msg,
4105                         &args.seq_args, &res.seq_res, task_flags);
4106 }
4107
4108 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4109                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4110                                 struct inode *inode)
4111 {
4112         struct nfs4_exception exception = {
4113                 .interruptible = true,
4114         };
4115         int err;
4116         do {
4117                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4118                 trace_nfs4_getattr(server, fhandle, fattr, err);
4119                 err = nfs4_handle_exception(server, err,
4120                                 &exception);
4121         } while (exception.retry);
4122         return err;
4123 }
4124
4125 /*
4126  * The file is not closed if it is opened due to the a request to change
4127  * the size of the file. The open call will not be needed once the
4128  * VFS layer lookup-intents are implemented.
4129  *
4130  * Close is called when the inode is destroyed.
4131  * If we haven't opened the file for O_WRONLY, we
4132  * need to in the size_change case to obtain a stateid.
4133  *
4134  * Got race?
4135  * Because OPEN is always done by name in nfsv4, it is
4136  * possible that we opened a different file by the same
4137  * name.  We can recognize this race condition, but we
4138  * can't do anything about it besides returning an error.
4139  *
4140  * This will be fixed with VFS changes (lookup-intent).
4141  */
4142 static int
4143 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4144                   struct iattr *sattr)
4145 {
4146         struct inode *inode = d_inode(dentry);
4147         const struct cred *cred = NULL;
4148         struct nfs_open_context *ctx = NULL;
4149         struct nfs4_label *label = NULL;
4150         int status;
4151
4152         if (pnfs_ld_layoutret_on_setattr(inode) &&
4153             sattr->ia_valid & ATTR_SIZE &&
4154             sattr->ia_size < i_size_read(inode))
4155                 pnfs_commit_and_return_layout(inode);
4156
4157         nfs_fattr_init(fattr);
4158
4159         /* Deal with open(O_TRUNC) */
4160         if (sattr->ia_valid & ATTR_OPEN)
4161                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4162
4163         /* Optimization: if the end result is no change, don't RPC */
4164         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4165                 return 0;
4166
4167         /* Search for an existing open(O_WRITE) file */
4168         if (sattr->ia_valid & ATTR_FILE) {
4169
4170                 ctx = nfs_file_open_context(sattr->ia_file);
4171                 if (ctx)
4172                         cred = ctx->cred;
4173         }
4174
4175         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4176         if (IS_ERR(label))
4177                 return PTR_ERR(label);
4178
4179         /* Return any delegations if we're going to change ACLs */
4180         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4181                 nfs4_inode_make_writeable(inode);
4182
4183         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4184         if (status == 0) {
4185                 nfs_setattr_update_inode(inode, sattr, fattr);
4186                 nfs_setsecurity(inode, fattr, label);
4187         }
4188         nfs4_label_free(label);
4189         return status;
4190 }
4191
4192 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4193                 struct dentry *dentry, struct nfs_fh *fhandle,
4194                 struct nfs_fattr *fattr, struct nfs4_label *label)
4195 {
4196         struct nfs_server *server = NFS_SERVER(dir);
4197         int                    status;
4198         struct nfs4_lookup_arg args = {
4199                 .bitmask = server->attr_bitmask,
4200                 .dir_fh = NFS_FH(dir),
4201                 .name = &dentry->d_name,
4202         };
4203         struct nfs4_lookup_res res = {
4204                 .server = server,
4205                 .fattr = fattr,
4206                 .label = label,
4207                 .fh = fhandle,
4208         };
4209         struct rpc_message msg = {
4210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4211                 .rpc_argp = &args,
4212                 .rpc_resp = &res,
4213         };
4214         unsigned short task_flags = 0;
4215
4216         /* Is this is an attribute revalidation, subject to softreval? */
4217         if (nfs_lookup_is_soft_revalidate(dentry))
4218                 task_flags |= RPC_TASK_TIMEOUT;
4219
4220         args.bitmask = nfs4_bitmask(server, label);
4221
4222         nfs_fattr_init(fattr);
4223
4224         dprintk("NFS call  lookup %pd2\n", dentry);
4225         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4226         status = nfs4_do_call_sync(clnt, server, &msg,
4227                         &args.seq_args, &res.seq_res, task_flags);
4228         dprintk("NFS reply lookup: %d\n", status);
4229         return status;
4230 }
4231
4232 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4233 {
4234         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4235                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4236         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4237         fattr->nlink = 2;
4238 }
4239
4240 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4241                                    struct dentry *dentry, struct nfs_fh *fhandle,
4242                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4243 {
4244         struct nfs4_exception exception = {
4245                 .interruptible = true,
4246         };
4247         struct rpc_clnt *client = *clnt;
4248         const struct qstr *name = &dentry->d_name;
4249         int err;
4250         do {
4251                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4252                 trace_nfs4_lookup(dir, name, err);
4253                 switch (err) {
4254                 case -NFS4ERR_BADNAME:
4255                         err = -ENOENT;
4256                         goto out;
4257                 case -NFS4ERR_MOVED:
4258                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4259                         if (err == -NFS4ERR_MOVED)
4260                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4261                         goto out;
4262                 case -NFS4ERR_WRONGSEC:
4263                         err = -EPERM;
4264                         if (client != *clnt)
4265                                 goto out;
4266                         client = nfs4_negotiate_security(client, dir, name);
4267                         if (IS_ERR(client))
4268                                 return PTR_ERR(client);
4269
4270                         exception.retry = 1;
4271                         break;
4272                 default:
4273                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4274                 }
4275         } while (exception.retry);
4276
4277 out:
4278         if (err == 0)
4279                 *clnt = client;
4280         else if (client != *clnt)
4281                 rpc_shutdown_client(client);
4282
4283         return err;
4284 }
4285
4286 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4287                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4288                             struct nfs4_label *label)
4289 {
4290         int status;
4291         struct rpc_clnt *client = NFS_CLIENT(dir);
4292
4293         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4294         if (client != NFS_CLIENT(dir)) {
4295                 rpc_shutdown_client(client);
4296                 nfs_fixup_secinfo_attributes(fattr);
4297         }
4298         return status;
4299 }
4300
4301 struct rpc_clnt *
4302 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4303                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4304 {
4305         struct rpc_clnt *client = NFS_CLIENT(dir);
4306         int status;
4307
4308         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4309         if (status < 0)
4310                 return ERR_PTR(status);
4311         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4312 }
4313
4314 static int _nfs4_proc_lookupp(struct inode *inode,
4315                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4316                 struct nfs4_label *label)
4317 {
4318         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4319         struct nfs_server *server = NFS_SERVER(inode);
4320         int                    status;
4321         struct nfs4_lookupp_arg args = {
4322                 .bitmask = server->attr_bitmask,
4323                 .fh = NFS_FH(inode),
4324         };
4325         struct nfs4_lookupp_res res = {
4326                 .server = server,
4327                 .fattr = fattr,
4328                 .label = label,
4329                 .fh = fhandle,
4330         };
4331         struct rpc_message msg = {
4332                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4333                 .rpc_argp = &args,
4334                 .rpc_resp = &res,
4335         };
4336
4337         args.bitmask = nfs4_bitmask(server, label);
4338
4339         nfs_fattr_init(fattr);
4340
4341         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4342         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4343                                 &res.seq_res, 0);
4344         dprintk("NFS reply lookupp: %d\n", status);
4345         return status;
4346 }
4347
4348 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4349                              struct nfs_fattr *fattr, struct nfs4_label *label)
4350 {
4351         struct nfs4_exception exception = {
4352                 .interruptible = true,
4353         };
4354         int err;
4355         do {
4356                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4357                 trace_nfs4_lookupp(inode, err);
4358                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4359                                 &exception);
4360         } while (exception.retry);
4361         return err;
4362 }
4363
4364 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4365 {
4366         struct nfs_server *server = NFS_SERVER(inode);
4367         struct nfs4_accessargs args = {
4368                 .fh = NFS_FH(inode),
4369                 .access = entry->mask,
4370         };
4371         struct nfs4_accessres res = {
4372                 .server = server,
4373         };
4374         struct rpc_message msg = {
4375                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4376                 .rpc_argp = &args,
4377                 .rpc_resp = &res,
4378                 .rpc_cred = entry->cred,
4379         };
4380         int status = 0;
4381
4382         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4383                 res.fattr = nfs_alloc_fattr();
4384                 if (res.fattr == NULL)
4385                         return -ENOMEM;
4386                 args.bitmask = server->cache_consistency_bitmask;
4387         }
4388         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4389         if (!status) {
4390                 nfs_access_set_mask(entry, res.access);
4391                 if (res.fattr)
4392                         nfs_refresh_inode(inode, res.fattr);
4393         }
4394         nfs_free_fattr(res.fattr);
4395         return status;
4396 }
4397
4398 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4399 {
4400         struct nfs4_exception exception = {
4401                 .interruptible = true,
4402         };
4403         int err;
4404         do {
4405                 err = _nfs4_proc_access(inode, entry);
4406                 trace_nfs4_access(inode, err);
4407                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4408                                 &exception);
4409         } while (exception.retry);
4410         return err;
4411 }
4412
4413 /*
4414  * TODO: For the time being, we don't try to get any attributes
4415  * along with any of the zero-copy operations READ, READDIR,
4416  * READLINK, WRITE.
4417  *
4418  * In the case of the first three, we want to put the GETATTR
4419  * after the read-type operation -- this is because it is hard
4420  * to predict the length of a GETATTR response in v4, and thus
4421  * align the READ data correctly.  This means that the GETATTR
4422  * may end up partially falling into the page cache, and we should
4423  * shift it into the 'tail' of the xdr_buf before processing.
4424  * To do this efficiently, we need to know the total length
4425  * of data received, which doesn't seem to be available outside
4426  * of the RPC layer.
4427  *
4428  * In the case of WRITE, we also want to put the GETATTR after
4429  * the operation -- in this case because we want to make sure
4430  * we get the post-operation mtime and size.
4431  *
4432  * Both of these changes to the XDR layer would in fact be quite
4433  * minor, but I decided to leave them for a subsequent patch.
4434  */
4435 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4436                 unsigned int pgbase, unsigned int pglen)
4437 {
4438         struct nfs4_readlink args = {
4439                 .fh       = NFS_FH(inode),
4440                 .pgbase   = pgbase,
4441                 .pglen    = pglen,
4442                 .pages    = &page,
4443         };
4444         struct nfs4_readlink_res res;
4445         struct rpc_message msg = {
4446                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4447                 .rpc_argp = &args,
4448                 .rpc_resp = &res,
4449         };
4450
4451         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4452 }
4453
4454 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4455                 unsigned int pgbase, unsigned int pglen)
4456 {
4457         struct nfs4_exception exception = {
4458                 .interruptible = true,
4459         };
4460         int err;
4461         do {
4462                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4463                 trace_nfs4_readlink(inode, err);
4464                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4465                                 &exception);
4466         } while (exception.retry);
4467         return err;
4468 }
4469
4470 /*
4471  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4472  */
4473 static int
4474 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4475                  int flags)
4476 {
4477         struct nfs_server *server = NFS_SERVER(dir);
4478         struct nfs4_label l, *ilabel = NULL;
4479         struct nfs_open_context *ctx;
4480         struct nfs4_state *state;
4481         int status = 0;
4482
4483         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4484         if (IS_ERR(ctx))
4485                 return PTR_ERR(ctx);
4486
4487         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4488
4489         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4490                 sattr->ia_mode &= ~current_umask();
4491         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4492         if (IS_ERR(state)) {
4493                 status = PTR_ERR(state);
4494                 goto out;
4495         }
4496 out:
4497         nfs4_label_release_security(ilabel);
4498         put_nfs_open_context(ctx);
4499         return status;
4500 }
4501
4502 static int
4503 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4504 {
4505         struct nfs_server *server = NFS_SERVER(dir);
4506         struct nfs_removeargs args = {
4507                 .fh = NFS_FH(dir),
4508                 .name = *name,
4509         };
4510         struct nfs_removeres res = {
4511                 .server = server,
4512         };
4513         struct rpc_message msg = {
4514                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4515                 .rpc_argp = &args,
4516                 .rpc_resp = &res,
4517         };
4518         unsigned long timestamp = jiffies;
4519         int status;
4520
4521         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4522         if (status == 0) {
4523                 spin_lock(&dir->i_lock);
4524                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4525                 /* Removing a directory decrements nlink in the parent */
4526                 if (ftype == NF4DIR && dir->i_nlink > 2)
4527                         nfs4_dec_nlink_locked(dir);
4528                 spin_unlock(&dir->i_lock);
4529         }
4530         return status;
4531 }
4532
4533 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4534 {
4535         struct nfs4_exception exception = {
4536                 .interruptible = true,
4537         };
4538         struct inode *inode = d_inode(dentry);
4539         int err;
4540
4541         if (inode) {
4542                 if (inode->i_nlink == 1)
4543                         nfs4_inode_return_delegation(inode);
4544                 else
4545                         nfs4_inode_make_writeable(inode);
4546         }
4547         do {
4548                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4549                 trace_nfs4_remove(dir, &dentry->d_name, err);
4550                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4551                                 &exception);
4552         } while (exception.retry);
4553         return err;
4554 }
4555
4556 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4557 {
4558         struct nfs4_exception exception = {
4559                 .interruptible = true,
4560         };
4561         int err;
4562
4563         do {
4564                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4565                 trace_nfs4_remove(dir, name, err);
4566                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4567                                 &exception);
4568         } while (exception.retry);
4569         return err;
4570 }
4571
4572 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4573                 struct dentry *dentry,
4574                 struct inode *inode)
4575 {
4576         struct nfs_removeargs *args = msg->rpc_argp;
4577         struct nfs_removeres *res = msg->rpc_resp;
4578
4579         res->server = NFS_SB(dentry->d_sb);
4580         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4581         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4582
4583         nfs_fattr_init(res->dir_attr);
4584
4585         if (inode)
4586                 nfs4_inode_return_delegation(inode);
4587 }
4588
4589 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4590 {
4591         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4592                         &data->args.seq_args,
4593                         &data->res.seq_res,
4594                         task);
4595 }
4596
4597 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4598 {
4599         struct nfs_unlinkdata *data = task->tk_calldata;
4600         struct nfs_removeres *res = &data->res;
4601
4602         if (!nfs4_sequence_done(task, &res->seq_res))
4603                 return 0;
4604         if (nfs4_async_handle_error(task, res->server, NULL,
4605                                     &data->timeout) == -EAGAIN)
4606                 return 0;
4607         if (task->tk_status == 0)
4608                 update_changeattr(dir, &res->cinfo,
4609                                 res->dir_attr->time_start, 0);
4610         return 1;
4611 }
4612
4613 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4614                 struct dentry *old_dentry,
4615                 struct dentry *new_dentry)
4616 {
4617         struct nfs_renameargs *arg = msg->rpc_argp;
4618         struct nfs_renameres *res = msg->rpc_resp;
4619         struct inode *old_inode = d_inode(old_dentry);
4620         struct inode *new_inode = d_inode(new_dentry);
4621
4622         if (old_inode)
4623                 nfs4_inode_make_writeable(old_inode);
4624         if (new_inode)
4625                 nfs4_inode_return_delegation(new_inode);
4626         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4627         res->server = NFS_SB(old_dentry->d_sb);
4628         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4629 }
4630
4631 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4632 {
4633         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4634                         &data->args.seq_args,
4635                         &data->res.seq_res,
4636                         task);
4637 }
4638
4639 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4640                                  struct inode *new_dir)
4641 {
4642         struct nfs_renamedata *data = task->tk_calldata;
4643         struct nfs_renameres *res = &data->res;
4644
4645         if (!nfs4_sequence_done(task, &res->seq_res))
4646                 return 0;
4647         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4648                 return 0;
4649
4650         if (task->tk_status == 0) {
4651                 if (new_dir != old_dir) {
4652                         /* Note: If we moved a directory, nlink will change */
4653                         update_changeattr(old_dir, &res->old_cinfo,
4654                                         res->old_fattr->time_start,
4655                                         NFS_INO_INVALID_OTHER);
4656                         update_changeattr(new_dir, &res->new_cinfo,
4657                                         res->new_fattr->time_start,
4658                                         NFS_INO_INVALID_OTHER);
4659                 } else
4660                         update_changeattr(old_dir, &res->old_cinfo,
4661                                         res->old_fattr->time_start,
4662                                         0);
4663         }
4664         return 1;
4665 }
4666
4667 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4668 {
4669         struct nfs_server *server = NFS_SERVER(inode);
4670         __u32 bitmask[NFS4_BITMASK_SZ];
4671         struct nfs4_link_arg arg = {
4672                 .fh     = NFS_FH(inode),
4673                 .dir_fh = NFS_FH(dir),
4674                 .name   = name,
4675                 .bitmask = bitmask,
4676         };
4677         struct nfs4_link_res res = {
4678                 .server = server,
4679                 .label = NULL,
4680         };
4681         struct rpc_message msg = {
4682                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4683                 .rpc_argp = &arg,
4684                 .rpc_resp = &res,
4685         };
4686         int status = -ENOMEM;
4687
4688         res.fattr = nfs_alloc_fattr();
4689         if (res.fattr == NULL)
4690                 goto out;
4691
4692         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4693         if (IS_ERR(res.label)) {
4694                 status = PTR_ERR(res.label);
4695                 goto out;
4696         }
4697
4698         nfs4_inode_make_writeable(inode);
4699         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4700
4701         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4702         if (!status) {
4703                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4704                 status = nfs_post_op_update_inode(inode, res.fattr);
4705                 if (!status)
4706                         nfs_setsecurity(inode, res.fattr, res.label);
4707         }
4708
4709
4710         nfs4_label_free(res.label);
4711
4712 out:
4713         nfs_free_fattr(res.fattr);
4714         return status;
4715 }
4716
4717 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4718 {
4719         struct nfs4_exception exception = {
4720                 .interruptible = true,
4721         };
4722         int err;
4723         do {
4724                 err = nfs4_handle_exception(NFS_SERVER(inode),
4725                                 _nfs4_proc_link(inode, dir, name),
4726                                 &exception);
4727         } while (exception.retry);
4728         return err;
4729 }
4730
4731 struct nfs4_createdata {
4732         struct rpc_message msg;
4733         struct nfs4_create_arg arg;
4734         struct nfs4_create_res res;
4735         struct nfs_fh fh;
4736         struct nfs_fattr fattr;
4737         struct nfs4_label *label;
4738 };
4739
4740 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4741                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4742 {
4743         struct nfs4_createdata *data;
4744
4745         data = kzalloc(sizeof(*data), GFP_KERNEL);
4746         if (data != NULL) {
4747                 struct nfs_server *server = NFS_SERVER(dir);
4748
4749                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4750                 if (IS_ERR(data->label))
4751                         goto out_free;
4752
4753                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4754                 data->msg.rpc_argp = &data->arg;
4755                 data->msg.rpc_resp = &data->res;
4756                 data->arg.dir_fh = NFS_FH(dir);
4757                 data->arg.server = server;
4758                 data->arg.name = name;
4759                 data->arg.attrs = sattr;
4760                 data->arg.ftype = ftype;
4761                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4762                 data->arg.umask = current_umask();
4763                 data->res.server = server;
4764                 data->res.fh = &data->fh;
4765                 data->res.fattr = &data->fattr;
4766                 data->res.label = data->label;
4767                 nfs_fattr_init(data->res.fattr);
4768         }
4769         return data;
4770 out_free:
4771         kfree(data);
4772         return NULL;
4773 }
4774
4775 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4776 {
4777         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4778                                     &data->arg.seq_args, &data->res.seq_res, 1);
4779         if (status == 0) {
4780                 spin_lock(&dir->i_lock);
4781                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4782                                 data->res.fattr->time_start, 0);
4783                 /* Creating a directory bumps nlink in the parent */
4784                 if (data->arg.ftype == NF4DIR)
4785                         nfs4_inc_nlink_locked(dir);
4786                 spin_unlock(&dir->i_lock);
4787                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4788         }
4789         return status;
4790 }
4791
4792 static void nfs4_free_createdata(struct nfs4_createdata *data)
4793 {
4794         nfs4_label_free(data->label);
4795         kfree(data);
4796 }
4797
4798 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4799                 struct page *page, unsigned int len, struct iattr *sattr,
4800                 struct nfs4_label *label)
4801 {
4802         struct nfs4_createdata *data;
4803         int status = -ENAMETOOLONG;
4804
4805         if (len > NFS4_MAXPATHLEN)
4806                 goto out;
4807
4808         status = -ENOMEM;
4809         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4810         if (data == NULL)
4811                 goto out;
4812
4813         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4814         data->arg.u.symlink.pages = &page;
4815         data->arg.u.symlink.len = len;
4816         data->arg.label = label;
4817
4818         status = nfs4_do_create(dir, dentry, data);
4819
4820         nfs4_free_createdata(data);
4821 out:
4822         return status;
4823 }
4824
4825 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4826                 struct page *page, unsigned int len, struct iattr *sattr)
4827 {
4828         struct nfs4_exception exception = {
4829                 .interruptible = true,
4830         };
4831         struct nfs4_label l, *label = NULL;
4832         int err;
4833
4834         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4835
4836         do {
4837                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4838                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4839                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4840                                 &exception);
4841         } while (exception.retry);
4842
4843         nfs4_label_release_security(label);
4844         return err;
4845 }
4846
4847 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4848                 struct iattr *sattr, struct nfs4_label *label)
4849 {
4850         struct nfs4_createdata *data;
4851         int status = -ENOMEM;
4852
4853         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4854         if (data == NULL)
4855                 goto out;
4856
4857         data->arg.label = label;
4858         status = nfs4_do_create(dir, dentry, data);
4859
4860         nfs4_free_createdata(data);
4861 out:
4862         return status;
4863 }
4864
4865 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4866                 struct iattr *sattr)
4867 {
4868         struct nfs_server *server = NFS_SERVER(dir);
4869         struct nfs4_exception exception = {
4870                 .interruptible = true,
4871         };
4872         struct nfs4_label l, *label = NULL;
4873         int err;
4874
4875         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4876
4877         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4878                 sattr->ia_mode &= ~current_umask();
4879         do {
4880                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4881                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4882                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4883                                 &exception);
4884         } while (exception.retry);
4885         nfs4_label_release_security(label);
4886
4887         return err;
4888 }
4889
4890 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4891                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4892 {
4893         struct inode            *dir = d_inode(dentry);
4894         struct nfs4_readdir_arg args = {
4895                 .fh = NFS_FH(dir),
4896                 .pages = pages,
4897                 .pgbase = 0,
4898                 .count = count,
4899                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4900                 .plus = plus,
4901         };
4902         struct nfs4_readdir_res res;
4903         struct rpc_message msg = {
4904                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4905                 .rpc_argp = &args,
4906                 .rpc_resp = &res,
4907                 .rpc_cred = cred,
4908         };
4909         int                     status;
4910
4911         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4912                         dentry,
4913                         (unsigned long long)cookie);
4914         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4915         res.pgbase = args.pgbase;
4916         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4917         if (status >= 0) {
4918                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4919                 status += args.pgbase;
4920         }
4921
4922         nfs_invalidate_atime(dir);
4923
4924         dprintk("%s: returns %d\n", __func__, status);
4925         return status;
4926 }
4927
4928 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4929                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4930 {
4931         struct nfs4_exception exception = {
4932                 .interruptible = true,
4933         };
4934         int err;
4935         do {
4936                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4937                                 pages, count, plus);
4938                 trace_nfs4_readdir(d_inode(dentry), err);
4939                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4940                                 &exception);
4941         } while (exception.retry);
4942         return err;
4943 }
4944
4945 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4946                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4947 {
4948         struct nfs4_createdata *data;
4949         int mode = sattr->ia_mode;
4950         int status = -ENOMEM;
4951
4952         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4953         if (data == NULL)
4954                 goto out;
4955
4956         if (S_ISFIFO(mode))
4957                 data->arg.ftype = NF4FIFO;
4958         else if (S_ISBLK(mode)) {
4959                 data->arg.ftype = NF4BLK;
4960                 data->arg.u.device.specdata1 = MAJOR(rdev);
4961                 data->arg.u.device.specdata2 = MINOR(rdev);
4962         }
4963         else if (S_ISCHR(mode)) {
4964                 data->arg.ftype = NF4CHR;
4965                 data->arg.u.device.specdata1 = MAJOR(rdev);
4966                 data->arg.u.device.specdata2 = MINOR(rdev);
4967         } else if (!S_ISSOCK(mode)) {
4968                 status = -EINVAL;
4969                 goto out_free;
4970         }
4971
4972         data->arg.label = label;
4973         status = nfs4_do_create(dir, dentry, data);
4974 out_free:
4975         nfs4_free_createdata(data);
4976 out:
4977         return status;
4978 }
4979
4980 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4981                 struct iattr *sattr, dev_t rdev)
4982 {
4983         struct nfs_server *server = NFS_SERVER(dir);
4984         struct nfs4_exception exception = {
4985                 .interruptible = true,
4986         };
4987         struct nfs4_label l, *label = NULL;
4988         int err;
4989
4990         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4991
4992         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4993                 sattr->ia_mode &= ~current_umask();
4994         do {
4995                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4996                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4997                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4998                                 &exception);
4999         } while (exception.retry);
5000
5001         nfs4_label_release_security(label);
5002
5003         return err;
5004 }
5005
5006 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5007                  struct nfs_fsstat *fsstat)
5008 {
5009         struct nfs4_statfs_arg args = {
5010                 .fh = fhandle,
5011                 .bitmask = server->attr_bitmask,
5012         };
5013         struct nfs4_statfs_res res = {
5014                 .fsstat = fsstat,
5015         };
5016         struct rpc_message msg = {
5017                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5018                 .rpc_argp = &args,
5019                 .rpc_resp = &res,
5020         };
5021
5022         nfs_fattr_init(fsstat->fattr);
5023         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5024 }
5025
5026 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5027 {
5028         struct nfs4_exception exception = {
5029                 .interruptible = true,
5030         };
5031         int err;
5032         do {
5033                 err = nfs4_handle_exception(server,
5034                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5035                                 &exception);
5036         } while (exception.retry);
5037         return err;
5038 }
5039
5040 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5041                 struct nfs_fsinfo *fsinfo)
5042 {
5043         struct nfs4_fsinfo_arg args = {
5044                 .fh = fhandle,
5045                 .bitmask = server->attr_bitmask,
5046         };
5047         struct nfs4_fsinfo_res res = {
5048                 .fsinfo = fsinfo,
5049         };
5050         struct rpc_message msg = {
5051                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5052                 .rpc_argp = &args,
5053                 .rpc_resp = &res,
5054         };
5055
5056         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5057 }
5058
5059 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5060 {
5061         struct nfs4_exception exception = {
5062                 .interruptible = true,
5063         };
5064         int err;
5065
5066         do {
5067                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5068                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5069                 if (err == 0) {
5070                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5071                         break;
5072                 }
5073                 err = nfs4_handle_exception(server, err, &exception);
5074         } while (exception.retry);
5075         return err;
5076 }
5077
5078 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5079 {
5080         int error;
5081
5082         nfs_fattr_init(fsinfo->fattr);
5083         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5084         if (error == 0) {
5085                 /* block layout checks this! */
5086                 server->pnfs_blksize = fsinfo->blksize;
5087                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5088         }
5089
5090         return error;
5091 }
5092
5093 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5094                 struct nfs_pathconf *pathconf)
5095 {
5096         struct nfs4_pathconf_arg args = {
5097                 .fh = fhandle,
5098                 .bitmask = server->attr_bitmask,
5099         };
5100         struct nfs4_pathconf_res res = {
5101                 .pathconf = pathconf,
5102         };
5103         struct rpc_message msg = {
5104                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5105                 .rpc_argp = &args,
5106                 .rpc_resp = &res,
5107         };
5108
5109         /* None of the pathconf attributes are mandatory to implement */
5110         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5111                 memset(pathconf, 0, sizeof(*pathconf));
5112                 return 0;
5113         }
5114
5115         nfs_fattr_init(pathconf->fattr);
5116         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5117 }
5118
5119 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5120                 struct nfs_pathconf *pathconf)
5121 {
5122         struct nfs4_exception exception = {
5123                 .interruptible = true,
5124         };
5125         int err;
5126
5127         do {
5128                 err = nfs4_handle_exception(server,
5129                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5130                                 &exception);
5131         } while (exception.retry);
5132         return err;
5133 }
5134
5135 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5136                 const struct nfs_open_context *ctx,
5137                 const struct nfs_lock_context *l_ctx,
5138                 fmode_t fmode)
5139 {
5140         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5141 }
5142 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5143
5144 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5145                 const struct nfs_open_context *ctx,
5146                 const struct nfs_lock_context *l_ctx,
5147                 fmode_t fmode)
5148 {
5149         nfs4_stateid _current_stateid;
5150
5151         /* If the current stateid represents a lost lock, then exit */
5152         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5153                 return true;
5154         return nfs4_stateid_match(stateid, &_current_stateid);
5155 }
5156
5157 static bool nfs4_error_stateid_expired(int err)
5158 {
5159         switch (err) {
5160         case -NFS4ERR_DELEG_REVOKED:
5161         case -NFS4ERR_ADMIN_REVOKED:
5162         case -NFS4ERR_BAD_STATEID:
5163         case -NFS4ERR_STALE_STATEID:
5164         case -NFS4ERR_OLD_STATEID:
5165         case -NFS4ERR_OPENMODE:
5166         case -NFS4ERR_EXPIRED:
5167                 return true;
5168         }
5169         return false;
5170 }
5171
5172 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5173 {
5174         struct nfs_server *server = NFS_SERVER(hdr->inode);
5175
5176         trace_nfs4_read(hdr, task->tk_status);
5177         if (task->tk_status < 0) {
5178                 struct nfs4_exception exception = {
5179                         .inode = hdr->inode,
5180                         .state = hdr->args.context->state,
5181                         .stateid = &hdr->args.stateid,
5182                 };
5183                 task->tk_status = nfs4_async_handle_exception(task,
5184                                 server, task->tk_status, &exception);
5185                 if (exception.retry) {
5186                         rpc_restart_call_prepare(task);
5187                         return -EAGAIN;
5188                 }
5189         }
5190
5191         if (task->tk_status > 0)
5192                 renew_lease(server, hdr->timestamp);
5193         return 0;
5194 }
5195
5196 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5197                 struct nfs_pgio_args *args)
5198 {
5199
5200         if (!nfs4_error_stateid_expired(task->tk_status) ||
5201                 nfs4_stateid_is_current(&args->stateid,
5202                                 args->context,
5203                                 args->lock_context,
5204                                 FMODE_READ))
5205                 return false;
5206         rpc_restart_call_prepare(task);
5207         return true;
5208 }
5209
5210 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5211 {
5212
5213         dprintk("--> %s\n", __func__);
5214
5215         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5216                 return -EAGAIN;
5217         if (nfs4_read_stateid_changed(task, &hdr->args))
5218                 return -EAGAIN;
5219         if (task->tk_status > 0)
5220                 nfs_invalidate_atime(hdr->inode);
5221         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5222                                     nfs4_read_done_cb(task, hdr);
5223 }
5224
5225 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5226                                  struct rpc_message *msg)
5227 {
5228         hdr->timestamp   = jiffies;
5229         if (!hdr->pgio_done_cb)
5230                 hdr->pgio_done_cb = nfs4_read_done_cb;
5231         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5232         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5233 }
5234
5235 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5236                                       struct nfs_pgio_header *hdr)
5237 {
5238         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5239                         &hdr->args.seq_args,
5240                         &hdr->res.seq_res,
5241                         task))
5242                 return 0;
5243         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5244                                 hdr->args.lock_context,
5245                                 hdr->rw_mode) == -EIO)
5246                 return -EIO;
5247         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5248                 return -EIO;
5249         return 0;
5250 }
5251
5252 static int nfs4_write_done_cb(struct rpc_task *task,
5253                               struct nfs_pgio_header *hdr)
5254 {
5255         struct inode *inode = hdr->inode;
5256
5257         trace_nfs4_write(hdr, task->tk_status);
5258         if (task->tk_status < 0) {
5259                 struct nfs4_exception exception = {
5260                         .inode = hdr->inode,
5261                         .state = hdr->args.context->state,
5262                         .stateid = &hdr->args.stateid,
5263                 };
5264                 task->tk_status = nfs4_async_handle_exception(task,
5265                                 NFS_SERVER(inode), task->tk_status,
5266                                 &exception);
5267                 if (exception.retry) {
5268                         rpc_restart_call_prepare(task);
5269                         return -EAGAIN;
5270                 }
5271         }
5272         if (task->tk_status >= 0) {
5273                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5274                 nfs_writeback_update_inode(hdr);
5275         }
5276         return 0;
5277 }
5278
5279 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5280                 struct nfs_pgio_args *args)
5281 {
5282
5283         if (!nfs4_error_stateid_expired(task->tk_status) ||
5284                 nfs4_stateid_is_current(&args->stateid,
5285                                 args->context,
5286                                 args->lock_context,
5287                                 FMODE_WRITE))
5288                 return false;
5289         rpc_restart_call_prepare(task);
5290         return true;
5291 }
5292
5293 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5294 {
5295         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5296                 return -EAGAIN;
5297         if (nfs4_write_stateid_changed(task, &hdr->args))
5298                 return -EAGAIN;
5299         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5300                 nfs4_write_done_cb(task, hdr);
5301 }
5302
5303 static
5304 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5305 {
5306         /* Don't request attributes for pNFS or O_DIRECT writes */
5307         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5308                 return false;
5309         /* Otherwise, request attributes if and only if we don't hold
5310          * a delegation
5311          */
5312         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5313 }
5314
5315 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5316                                   struct rpc_message *msg,
5317                                   struct rpc_clnt **clnt)
5318 {
5319         struct nfs_server *server = NFS_SERVER(hdr->inode);
5320
5321         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5322                 hdr->args.bitmask = NULL;
5323                 hdr->res.fattr = NULL;
5324         } else
5325                 hdr->args.bitmask = server->cache_consistency_bitmask;
5326
5327         if (!hdr->pgio_done_cb)
5328                 hdr->pgio_done_cb = nfs4_write_done_cb;
5329         hdr->res.server = server;
5330         hdr->timestamp   = jiffies;
5331
5332         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5333         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5334         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5335 }
5336
5337 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5338 {
5339         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5340                         &data->args.seq_args,
5341                         &data->res.seq_res,
5342                         task);
5343 }
5344
5345 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5346 {
5347         struct inode *inode = data->inode;
5348
5349         trace_nfs4_commit(data, task->tk_status);
5350         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5351                                     NULL, NULL) == -EAGAIN) {
5352                 rpc_restart_call_prepare(task);
5353                 return -EAGAIN;
5354         }
5355         return 0;
5356 }
5357
5358 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5359 {
5360         if (!nfs4_sequence_done(task, &data->res.seq_res))
5361                 return -EAGAIN;
5362         return data->commit_done_cb(task, data);
5363 }
5364
5365 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5366                                    struct rpc_clnt **clnt)
5367 {
5368         struct nfs_server *server = NFS_SERVER(data->inode);
5369
5370         if (data->commit_done_cb == NULL)
5371                 data->commit_done_cb = nfs4_commit_done_cb;
5372         data->res.server = server;
5373         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5374         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5375         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5376 }
5377
5378 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5379                                 struct nfs_commitres *res)
5380 {
5381         struct inode *dst_inode = file_inode(dst);
5382         struct nfs_server *server = NFS_SERVER(dst_inode);
5383         struct rpc_message msg = {
5384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5385                 .rpc_argp = args,
5386                 .rpc_resp = res,
5387         };
5388
5389         args->fh = NFS_FH(dst_inode);
5390         return nfs4_call_sync(server->client, server, &msg,
5391                         &args->seq_args, &res->seq_res, 1);
5392 }
5393
5394 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5395 {
5396         struct nfs_commitargs args = {
5397                 .offset = offset,
5398                 .count = count,
5399         };
5400         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5401         struct nfs4_exception exception = { };
5402         int status;
5403
5404         do {
5405                 status = _nfs4_proc_commit(dst, &args, res);
5406                 status = nfs4_handle_exception(dst_server, status, &exception);
5407         } while (exception.retry);
5408
5409         return status;
5410 }
5411
5412 struct nfs4_renewdata {
5413         struct nfs_client       *client;
5414         unsigned long           timestamp;
5415 };
5416
5417 /*
5418  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5419  * standalone procedure for queueing an asynchronous RENEW.
5420  */
5421 static void nfs4_renew_release(void *calldata)
5422 {
5423         struct nfs4_renewdata *data = calldata;
5424         struct nfs_client *clp = data->client;
5425
5426         if (refcount_read(&clp->cl_count) > 1)
5427                 nfs4_schedule_state_renewal(clp);
5428         nfs_put_client(clp);
5429         kfree(data);
5430 }
5431
5432 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5433 {
5434         struct nfs4_renewdata *data = calldata;
5435         struct nfs_client *clp = data->client;
5436         unsigned long timestamp = data->timestamp;
5437
5438         trace_nfs4_renew_async(clp, task->tk_status);
5439         switch (task->tk_status) {
5440         case 0:
5441                 break;
5442         case -NFS4ERR_LEASE_MOVED:
5443                 nfs4_schedule_lease_moved_recovery(clp);
5444                 break;
5445         default:
5446                 /* Unless we're shutting down, schedule state recovery! */
5447                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5448                         return;
5449                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5450                         nfs4_schedule_lease_recovery(clp);
5451                         return;
5452                 }
5453                 nfs4_schedule_path_down_recovery(clp);
5454         }
5455         do_renew_lease(clp, timestamp);
5456 }
5457
5458 static const struct rpc_call_ops nfs4_renew_ops = {
5459         .rpc_call_done = nfs4_renew_done,
5460         .rpc_release = nfs4_renew_release,
5461 };
5462
5463 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5464 {
5465         struct rpc_message msg = {
5466                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5467                 .rpc_argp       = clp,
5468                 .rpc_cred       = cred,
5469         };
5470         struct nfs4_renewdata *data;
5471
5472         if (renew_flags == 0)
5473                 return 0;
5474         if (!refcount_inc_not_zero(&clp->cl_count))
5475                 return -EIO;
5476         data = kmalloc(sizeof(*data), GFP_NOFS);
5477         if (data == NULL) {
5478                 nfs_put_client(clp);
5479                 return -ENOMEM;
5480         }
5481         data->client = clp;
5482         data->timestamp = jiffies;
5483         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5484                         &nfs4_renew_ops, data);
5485 }
5486
5487 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5488 {
5489         struct rpc_message msg = {
5490                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5491                 .rpc_argp       = clp,
5492                 .rpc_cred       = cred,
5493         };
5494         unsigned long now = jiffies;
5495         int status;
5496
5497         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5498         if (status < 0)
5499                 return status;
5500         do_renew_lease(clp, now);
5501         return 0;
5502 }
5503
5504 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5505 {
5506         return server->caps & NFS_CAP_ACLS;
5507 }
5508
5509 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5510  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5511  * the stack.
5512  */
5513 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5514
5515 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5516                 struct page **pages)
5517 {
5518         struct page *newpage, **spages;
5519         int rc = 0;
5520         size_t len;
5521         spages = pages;
5522
5523         do {
5524                 len = min_t(size_t, PAGE_SIZE, buflen);
5525                 newpage = alloc_page(GFP_KERNEL);
5526
5527                 if (newpage == NULL)
5528                         goto unwind;
5529                 memcpy(page_address(newpage), buf, len);
5530                 buf += len;
5531                 buflen -= len;
5532                 *pages++ = newpage;
5533                 rc++;
5534         } while (buflen != 0);
5535
5536         return rc;
5537
5538 unwind:
5539         for(; rc > 0; rc--)
5540                 __free_page(spages[rc-1]);
5541         return -ENOMEM;
5542 }
5543
5544 struct nfs4_cached_acl {
5545         int cached;
5546         size_t len;
5547         char data[0];
5548 };
5549
5550 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5551 {
5552         struct nfs_inode *nfsi = NFS_I(inode);
5553
5554         spin_lock(&inode->i_lock);
5555         kfree(nfsi->nfs4_acl);
5556         nfsi->nfs4_acl = acl;
5557         spin_unlock(&inode->i_lock);
5558 }
5559
5560 static void nfs4_zap_acl_attr(struct inode *inode)
5561 {
5562         nfs4_set_cached_acl(inode, NULL);
5563 }
5564
5565 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5566 {
5567         struct nfs_inode *nfsi = NFS_I(inode);
5568         struct nfs4_cached_acl *acl;
5569         int ret = -ENOENT;
5570
5571         spin_lock(&inode->i_lock);
5572         acl = nfsi->nfs4_acl;
5573         if (acl == NULL)
5574                 goto out;
5575         if (buf == NULL) /* user is just asking for length */
5576                 goto out_len;
5577         if (acl->cached == 0)
5578                 goto out;
5579         ret = -ERANGE; /* see getxattr(2) man page */
5580         if (acl->len > buflen)
5581                 goto out;
5582         memcpy(buf, acl->data, acl->len);
5583 out_len:
5584         ret = acl->len;
5585 out:
5586         spin_unlock(&inode->i_lock);
5587         return ret;
5588 }
5589
5590 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5591 {
5592         struct nfs4_cached_acl *acl;
5593         size_t buflen = sizeof(*acl) + acl_len;
5594
5595         if (buflen <= PAGE_SIZE) {
5596                 acl = kmalloc(buflen, GFP_KERNEL);
5597                 if (acl == NULL)
5598                         goto out;
5599                 acl->cached = 1;
5600                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5601         } else {
5602                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5603                 if (acl == NULL)
5604                         goto out;
5605                 acl->cached = 0;
5606         }
5607         acl->len = acl_len;
5608 out:
5609         nfs4_set_cached_acl(inode, acl);
5610 }
5611
5612 /*
5613  * The getxattr API returns the required buffer length when called with a
5614  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5615  * the required buf.  On a NULL buf, we send a page of data to the server
5616  * guessing that the ACL request can be serviced by a page. If so, we cache
5617  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5618  * the cache. If not so, we throw away the page, and cache the required
5619  * length. The next getxattr call will then produce another round trip to
5620  * the server, this time with the input buf of the required size.
5621  */
5622 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5623 {
5624         struct page **pages;
5625         struct nfs_getaclargs args = {
5626                 .fh = NFS_FH(inode),
5627                 .acl_len = buflen,
5628         };
5629         struct nfs_getaclres res = {
5630                 .acl_len = buflen,
5631         };
5632         struct rpc_message msg = {
5633                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5634                 .rpc_argp = &args,
5635                 .rpc_resp = &res,
5636         };
5637         unsigned int npages;
5638         int ret = -ENOMEM, i;
5639         struct nfs_server *server = NFS_SERVER(inode);
5640
5641         if (buflen == 0)
5642                 buflen = server->rsize;
5643
5644         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5645         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5646         if (!pages)
5647                 return -ENOMEM;
5648
5649         args.acl_pages = pages;
5650
5651         for (i = 0; i < npages; i++) {
5652                 pages[i] = alloc_page(GFP_KERNEL);
5653                 if (!pages[i])
5654                         goto out_free;
5655         }
5656
5657         /* for decoding across pages */
5658         res.acl_scratch = alloc_page(GFP_KERNEL);
5659         if (!res.acl_scratch)
5660                 goto out_free;
5661
5662         args.acl_len = npages * PAGE_SIZE;
5663
5664         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5665                 __func__, buf, buflen, npages, args.acl_len);
5666         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5667                              &msg, &args.seq_args, &res.seq_res, 0);
5668         if (ret)
5669                 goto out_free;
5670
5671         /* Handle the case where the passed-in buffer is too short */
5672         if (res.acl_flags & NFS4_ACL_TRUNC) {
5673                 /* Did the user only issue a request for the acl length? */
5674                 if (buf == NULL)
5675                         goto out_ok;
5676                 ret = -ERANGE;
5677                 goto out_free;
5678         }
5679         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5680         if (buf) {
5681                 if (res.acl_len > buflen) {
5682                         ret = -ERANGE;
5683                         goto out_free;
5684                 }
5685                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5686         }
5687 out_ok:
5688         ret = res.acl_len;
5689 out_free:
5690         for (i = 0; i < npages; i++)
5691                 if (pages[i])
5692                         __free_page(pages[i]);
5693         if (res.acl_scratch)
5694                 __free_page(res.acl_scratch);
5695         kfree(pages);
5696         return ret;
5697 }
5698
5699 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5700 {
5701         struct nfs4_exception exception = {
5702                 .interruptible = true,
5703         };
5704         ssize_t ret;
5705         do {
5706                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5707                 trace_nfs4_get_acl(inode, ret);
5708                 if (ret >= 0)
5709                         break;
5710                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5711         } while (exception.retry);
5712         return ret;
5713 }
5714
5715 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5716 {
5717         struct nfs_server *server = NFS_SERVER(inode);
5718         int ret;
5719
5720         if (!nfs4_server_supports_acls(server))
5721                 return -EOPNOTSUPP;
5722         ret = nfs_revalidate_inode(server, inode);
5723         if (ret < 0)
5724                 return ret;
5725         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5726                 nfs_zap_acl_cache(inode);
5727         ret = nfs4_read_cached_acl(inode, buf, buflen);
5728         if (ret != -ENOENT)
5729                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5730                  * but no cached acl data, just the acl length */
5731                 return ret;
5732         return nfs4_get_acl_uncached(inode, buf, buflen);
5733 }
5734
5735 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5736 {
5737         struct nfs_server *server = NFS_SERVER(inode);
5738         struct page *pages[NFS4ACL_MAXPAGES];
5739         struct nfs_setaclargs arg = {
5740                 .fh             = NFS_FH(inode),
5741                 .acl_pages      = pages,
5742                 .acl_len        = buflen,
5743         };
5744         struct nfs_setaclres res;
5745         struct rpc_message msg = {
5746                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5747                 .rpc_argp       = &arg,
5748                 .rpc_resp       = &res,
5749         };
5750         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5751         int ret, i;
5752
5753         if (!nfs4_server_supports_acls(server))
5754                 return -EOPNOTSUPP;
5755         if (npages > ARRAY_SIZE(pages))
5756                 return -ERANGE;
5757         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5758         if (i < 0)
5759                 return i;
5760         nfs4_inode_make_writeable(inode);
5761         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5762
5763         /*
5764          * Free each page after tx, so the only ref left is
5765          * held by the network stack
5766          */
5767         for (; i > 0; i--)
5768                 put_page(pages[i-1]);
5769
5770         /*
5771          * Acl update can result in inode attribute update.
5772          * so mark the attribute cache invalid.
5773          */
5774         spin_lock(&inode->i_lock);
5775         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5776                 | NFS_INO_INVALID_CTIME
5777                 | NFS_INO_REVAL_FORCED;
5778         spin_unlock(&inode->i_lock);
5779         nfs_access_zap_cache(inode);
5780         nfs_zap_acl_cache(inode);
5781         return ret;
5782 }
5783
5784 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5785 {
5786         struct nfs4_exception exception = { };
5787         int err;
5788         do {
5789                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5790                 trace_nfs4_set_acl(inode, err);
5791                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5792                                 &exception);
5793         } while (exception.retry);
5794         return err;
5795 }
5796
5797 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5798 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5799                                         size_t buflen)
5800 {
5801         struct nfs_server *server = NFS_SERVER(inode);
5802         struct nfs_fattr fattr;
5803         struct nfs4_label label = {0, 0, buflen, buf};
5804
5805         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5806         struct nfs4_getattr_arg arg = {
5807                 .fh             = NFS_FH(inode),
5808                 .bitmask        = bitmask,
5809         };
5810         struct nfs4_getattr_res res = {
5811                 .fattr          = &fattr,
5812                 .label          = &label,
5813                 .server         = server,
5814         };
5815         struct rpc_message msg = {
5816                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5817                 .rpc_argp       = &arg,
5818                 .rpc_resp       = &res,
5819         };
5820         int ret;
5821
5822         nfs_fattr_init(&fattr);
5823
5824         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5825         if (ret)
5826                 return ret;
5827         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5828                 return -ENOENT;
5829         if (buflen < label.len)
5830                 return -ERANGE;
5831         return 0;
5832 }
5833
5834 static int nfs4_get_security_label(struct inode *inode, void *buf,
5835                                         size_t buflen)
5836 {
5837         struct nfs4_exception exception = {
5838                 .interruptible = true,
5839         };
5840         int err;
5841
5842         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5843                 return -EOPNOTSUPP;
5844
5845         do {
5846                 err = _nfs4_get_security_label(inode, buf, buflen);
5847                 trace_nfs4_get_security_label(inode, err);
5848                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5849                                 &exception);
5850         } while (exception.retry);
5851         return err;
5852 }
5853
5854 static int _nfs4_do_set_security_label(struct inode *inode,
5855                 struct nfs4_label *ilabel,
5856                 struct nfs_fattr *fattr,
5857                 struct nfs4_label *olabel)
5858 {
5859
5860         struct iattr sattr = {0};
5861         struct nfs_server *server = NFS_SERVER(inode);
5862         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5863         struct nfs_setattrargs arg = {
5864                 .fh             = NFS_FH(inode),
5865                 .iap            = &sattr,
5866                 .server         = server,
5867                 .bitmask        = bitmask,
5868                 .label          = ilabel,
5869         };
5870         struct nfs_setattrres res = {
5871                 .fattr          = fattr,
5872                 .label          = olabel,
5873                 .server         = server,
5874         };
5875         struct rpc_message msg = {
5876                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5877                 .rpc_argp       = &arg,
5878                 .rpc_resp       = &res,
5879         };
5880         int status;
5881
5882         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5883
5884         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5885         if (status)
5886                 dprintk("%s failed: %d\n", __func__, status);
5887
5888         return status;
5889 }
5890
5891 static int nfs4_do_set_security_label(struct inode *inode,
5892                 struct nfs4_label *ilabel,
5893                 struct nfs_fattr *fattr,
5894                 struct nfs4_label *olabel)
5895 {
5896         struct nfs4_exception exception = { };
5897         int err;
5898
5899         do {
5900                 err = _nfs4_do_set_security_label(inode, ilabel,
5901                                 fattr, olabel);
5902                 trace_nfs4_set_security_label(inode, err);
5903                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5904                                 &exception);
5905         } while (exception.retry);
5906         return err;
5907 }
5908
5909 static int
5910 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5911 {
5912         struct nfs4_label ilabel, *olabel = NULL;
5913         struct nfs_fattr fattr;
5914         int status;
5915
5916         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5917                 return -EOPNOTSUPP;
5918
5919         nfs_fattr_init(&fattr);
5920
5921         ilabel.pi = 0;
5922         ilabel.lfs = 0;
5923         ilabel.label = (char *)buf;
5924         ilabel.len = buflen;
5925
5926         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5927         if (IS_ERR(olabel)) {
5928                 status = -PTR_ERR(olabel);
5929                 goto out;
5930         }
5931
5932         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5933         if (status == 0)
5934                 nfs_setsecurity(inode, &fattr, olabel);
5935
5936         nfs4_label_free(olabel);
5937 out:
5938         return status;
5939 }
5940 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5941
5942
5943 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5944                                     nfs4_verifier *bootverf)
5945 {
5946         __be32 verf[2];
5947
5948         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5949                 /* An impossible timestamp guarantees this value
5950                  * will never match a generated boot time. */
5951                 verf[0] = cpu_to_be32(U32_MAX);
5952                 verf[1] = cpu_to_be32(U32_MAX);
5953         } else {
5954                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5955                 u64 ns = ktime_to_ns(nn->boot_time);
5956
5957                 verf[0] = cpu_to_be32(ns >> 32);
5958                 verf[1] = cpu_to_be32(ns);
5959         }
5960         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5961 }
5962
5963 static int
5964 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5965 {
5966         size_t len;
5967         char *str;
5968
5969         if (clp->cl_owner_id != NULL)
5970                 return 0;
5971
5972         rcu_read_lock();
5973         len = 14 +
5974                 strlen(clp->cl_rpcclient->cl_nodename) +
5975                 1 +
5976                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5977                 1;
5978         rcu_read_unlock();
5979         if (nfs4_client_id_uniquifier[0] != '\0')
5980                 len += strlen(nfs4_client_id_uniquifier) + 1;
5981         if (len > NFS4_OPAQUE_LIMIT + 1)
5982                 return -EINVAL;
5983
5984         /*
5985          * Since this string is allocated at mount time, and held until the
5986          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5987          * about a memory-reclaim deadlock.
5988          */
5989         str = kmalloc(len, GFP_KERNEL);
5990         if (!str)
5991                 return -ENOMEM;
5992
5993         rcu_read_lock();
5994         if (nfs4_client_id_uniquifier[0] != '\0')
5995                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5996                           clp->cl_rpcclient->cl_nodename,
5997                           nfs4_client_id_uniquifier,
5998                           rpc_peeraddr2str(clp->cl_rpcclient,
5999                                            RPC_DISPLAY_ADDR));
6000         else
6001                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6002                           clp->cl_rpcclient->cl_nodename,
6003                           rpc_peeraddr2str(clp->cl_rpcclient,
6004                                            RPC_DISPLAY_ADDR));
6005         rcu_read_unlock();
6006
6007         clp->cl_owner_id = str;
6008         return 0;
6009 }
6010
6011 static int
6012 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
6013 {
6014         size_t len;
6015         char *str;
6016
6017         len = 10 + 10 + 1 + 10 + 1 +
6018                 strlen(nfs4_client_id_uniquifier) + 1 +
6019                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6020
6021         if (len > NFS4_OPAQUE_LIMIT + 1)
6022                 return -EINVAL;
6023
6024         /*
6025          * Since this string is allocated at mount time, and held until the
6026          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6027          * about a memory-reclaim deadlock.
6028          */
6029         str = kmalloc(len, GFP_KERNEL);
6030         if (!str)
6031                 return -ENOMEM;
6032
6033         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6034                         clp->rpc_ops->version, clp->cl_minorversion,
6035                         nfs4_client_id_uniquifier,
6036                         clp->cl_rpcclient->cl_nodename);
6037         clp->cl_owner_id = str;
6038         return 0;
6039 }
6040
6041 static int
6042 nfs4_init_uniform_client_string(struct nfs_client *clp)
6043 {
6044         size_t len;
6045         char *str;
6046
6047         if (clp->cl_owner_id != NULL)
6048                 return 0;
6049
6050         if (nfs4_client_id_uniquifier[0] != '\0')
6051                 return nfs4_init_uniquifier_client_string(clp);
6052
6053         len = 10 + 10 + 1 + 10 + 1 +
6054                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6055
6056         if (len > NFS4_OPAQUE_LIMIT + 1)
6057                 return -EINVAL;
6058
6059         /*
6060          * Since this string is allocated at mount time, and held until the
6061          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6062          * about a memory-reclaim deadlock.
6063          */
6064         str = kmalloc(len, GFP_KERNEL);
6065         if (!str)
6066                 return -ENOMEM;
6067
6068         scnprintf(str, len, "Linux NFSv%u.%u %s",
6069                         clp->rpc_ops->version, clp->cl_minorversion,
6070                         clp->cl_rpcclient->cl_nodename);
6071         clp->cl_owner_id = str;
6072         return 0;
6073 }
6074
6075 /*
6076  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6077  * services.  Advertise one based on the address family of the
6078  * clientaddr.
6079  */
6080 static unsigned int
6081 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6082 {
6083         if (strchr(clp->cl_ipaddr, ':') != NULL)
6084                 return scnprintf(buf, len, "tcp6");
6085         else
6086                 return scnprintf(buf, len, "tcp");
6087 }
6088
6089 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6090 {
6091         struct nfs4_setclientid *sc = calldata;
6092
6093         if (task->tk_status == 0)
6094                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6095 }
6096
6097 static const struct rpc_call_ops nfs4_setclientid_ops = {
6098         .rpc_call_done = nfs4_setclientid_done,
6099 };
6100
6101 /**
6102  * nfs4_proc_setclientid - Negotiate client ID
6103  * @clp: state data structure
6104  * @program: RPC program for NFSv4 callback service
6105  * @port: IP port number for NFS4 callback service
6106  * @cred: credential to use for this call
6107  * @res: where to place the result
6108  *
6109  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6110  */
6111 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6112                 unsigned short port, const struct cred *cred,
6113                 struct nfs4_setclientid_res *res)
6114 {
6115         nfs4_verifier sc_verifier;
6116         struct nfs4_setclientid setclientid = {
6117                 .sc_verifier = &sc_verifier,
6118                 .sc_prog = program,
6119                 .sc_clnt = clp,
6120         };
6121         struct rpc_message msg = {
6122                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6123                 .rpc_argp = &setclientid,
6124                 .rpc_resp = res,
6125                 .rpc_cred = cred,
6126         };
6127         struct rpc_task_setup task_setup_data = {
6128                 .rpc_client = clp->cl_rpcclient,
6129                 .rpc_message = &msg,
6130                 .callback_ops = &nfs4_setclientid_ops,
6131                 .callback_data = &setclientid,
6132                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6133         };
6134         unsigned long now = jiffies;
6135         int status;
6136
6137         /* nfs_client_id4 */
6138         nfs4_init_boot_verifier(clp, &sc_verifier);
6139
6140         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6141                 status = nfs4_init_uniform_client_string(clp);
6142         else
6143                 status = nfs4_init_nonuniform_client_string(clp);
6144
6145         if (status)
6146                 goto out;
6147
6148         /* cb_client4 */
6149         setclientid.sc_netid_len =
6150                                 nfs4_init_callback_netid(clp,
6151                                                 setclientid.sc_netid,
6152                                                 sizeof(setclientid.sc_netid));
6153         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6154                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6155                                 clp->cl_ipaddr, port >> 8, port & 255);
6156
6157         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6158                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6159                 clp->cl_owner_id);
6160
6161         status = nfs4_call_sync_custom(&task_setup_data);
6162         if (setclientid.sc_cred) {
6163                 kfree(clp->cl_acceptor);
6164                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6165                 put_rpccred(setclientid.sc_cred);
6166         }
6167
6168         if (status == 0)
6169                 do_renew_lease(clp, now);
6170 out:
6171         trace_nfs4_setclientid(clp, status);
6172         dprintk("NFS reply setclientid: %d\n", status);
6173         return status;
6174 }
6175
6176 /**
6177  * nfs4_proc_setclientid_confirm - Confirm client ID
6178  * @clp: state data structure
6179  * @arg: result of a previous SETCLIENTID
6180  * @cred: credential to use for this call
6181  *
6182  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6183  */
6184 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6185                 struct nfs4_setclientid_res *arg,
6186                 const struct cred *cred)
6187 {
6188         struct rpc_message msg = {
6189                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6190                 .rpc_argp = arg,
6191                 .rpc_cred = cred,
6192         };
6193         int status;
6194
6195         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6196                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6197                 clp->cl_clientid);
6198         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6199                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6200         trace_nfs4_setclientid_confirm(clp, status);
6201         dprintk("NFS reply setclientid_confirm: %d\n", status);
6202         return status;
6203 }
6204
6205 struct nfs4_delegreturndata {
6206         struct nfs4_delegreturnargs args;
6207         struct nfs4_delegreturnres res;
6208         struct nfs_fh fh;
6209         nfs4_stateid stateid;
6210         unsigned long timestamp;
6211         struct {
6212                 struct nfs4_layoutreturn_args arg;
6213                 struct nfs4_layoutreturn_res res;
6214                 struct nfs4_xdr_opaque_data ld_private;
6215                 u32 roc_barrier;
6216                 bool roc;
6217         } lr;
6218         struct nfs_fattr fattr;
6219         int rpc_status;
6220         struct inode *inode;
6221 };
6222
6223 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6224 {
6225         struct nfs4_delegreturndata *data = calldata;
6226         struct nfs4_exception exception = {
6227                 .inode = data->inode,
6228                 .stateid = &data->stateid,
6229         };
6230
6231         if (!nfs4_sequence_done(task, &data->res.seq_res))
6232                 return;
6233
6234         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6235
6236         /* Handle Layoutreturn errors */
6237         if (pnfs_roc_done(task, data->inode,
6238                                 &data->args.lr_args,
6239                                 &data->res.lr_res,
6240                                 &data->res.lr_ret) == -EAGAIN)
6241                 goto out_restart;
6242
6243         switch (task->tk_status) {
6244         case 0:
6245                 renew_lease(data->res.server, data->timestamp);
6246                 break;
6247         case -NFS4ERR_ADMIN_REVOKED:
6248         case -NFS4ERR_DELEG_REVOKED:
6249         case -NFS4ERR_EXPIRED:
6250                 nfs4_free_revoked_stateid(data->res.server,
6251                                 data->args.stateid,
6252                                 task->tk_msg.rpc_cred);
6253                 /* Fallthrough */
6254         case -NFS4ERR_BAD_STATEID:
6255         case -NFS4ERR_STALE_STATEID:
6256                 task->tk_status = 0;
6257                 break;
6258         case -NFS4ERR_OLD_STATEID:
6259                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6260                         nfs4_stateid_seqid_inc(&data->stateid);
6261                 if (data->args.bitmask) {
6262                         data->args.bitmask = NULL;
6263                         data->res.fattr = NULL;
6264                 }
6265                 goto out_restart;
6266         case -NFS4ERR_ACCESS:
6267                 if (data->args.bitmask) {
6268                         data->args.bitmask = NULL;
6269                         data->res.fattr = NULL;
6270                         goto out_restart;
6271                 }
6272                 /* Fallthrough */
6273         default:
6274                 task->tk_status = nfs4_async_handle_exception(task,
6275                                 data->res.server, task->tk_status,
6276                                 &exception);
6277                 if (exception.retry)
6278                         goto out_restart;
6279         }
6280         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6281         data->rpc_status = task->tk_status;
6282         return;
6283 out_restart:
6284         task->tk_status = 0;
6285         rpc_restart_call_prepare(task);
6286 }
6287
6288 static void nfs4_delegreturn_release(void *calldata)
6289 {
6290         struct nfs4_delegreturndata *data = calldata;
6291         struct inode *inode = data->inode;
6292
6293         if (inode) {
6294                 if (data->lr.roc)
6295                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6296                                         data->res.lr_ret);
6297                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6298                 nfs_iput_and_deactive(inode);
6299         }
6300         kfree(calldata);
6301 }
6302
6303 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6304 {
6305         struct nfs4_delegreturndata *d_data;
6306         struct pnfs_layout_hdr *lo;
6307
6308         d_data = (struct nfs4_delegreturndata *)data;
6309
6310         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6311                 nfs4_sequence_done(task, &d_data->res.seq_res);
6312                 return;
6313         }
6314
6315         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6316         if (lo && !pnfs_layout_is_valid(lo)) {
6317                 d_data->args.lr_args = NULL;
6318                 d_data->res.lr_res = NULL;
6319         }
6320
6321         nfs4_setup_sequence(d_data->res.server->nfs_client,
6322                         &d_data->args.seq_args,
6323                         &d_data->res.seq_res,
6324                         task);
6325 }
6326
6327 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6328         .rpc_call_prepare = nfs4_delegreturn_prepare,
6329         .rpc_call_done = nfs4_delegreturn_done,
6330         .rpc_release = nfs4_delegreturn_release,
6331 };
6332
6333 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6334 {
6335         struct nfs4_delegreturndata *data;
6336         struct nfs_server *server = NFS_SERVER(inode);
6337         struct rpc_task *task;
6338         struct rpc_message msg = {
6339                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6340                 .rpc_cred = cred,
6341         };
6342         struct rpc_task_setup task_setup_data = {
6343                 .rpc_client = server->client,
6344                 .rpc_message = &msg,
6345                 .callback_ops = &nfs4_delegreturn_ops,
6346                 .flags = RPC_TASK_ASYNC,
6347         };
6348         int status = 0;
6349
6350         data = kzalloc(sizeof(*data), GFP_NOFS);
6351         if (data == NULL)
6352                 return -ENOMEM;
6353         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6354
6355         nfs4_state_protect(server->nfs_client,
6356                         NFS_SP4_MACH_CRED_CLEANUP,
6357                         &task_setup_data.rpc_client, &msg);
6358
6359         data->args.fhandle = &data->fh;
6360         data->args.stateid = &data->stateid;
6361         data->args.bitmask = server->cache_consistency_bitmask;
6362         nfs_copy_fh(&data->fh, NFS_FH(inode));
6363         nfs4_stateid_copy(&data->stateid, stateid);
6364         data->res.fattr = &data->fattr;
6365         data->res.server = server;
6366         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6367         data->lr.arg.ld_private = &data->lr.ld_private;
6368         nfs_fattr_init(data->res.fattr);
6369         data->timestamp = jiffies;
6370         data->rpc_status = 0;
6371         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6372         data->inode = nfs_igrab_and_active(inode);
6373         if (data->inode) {
6374                 if (data->lr.roc) {
6375                         data->args.lr_args = &data->lr.arg;
6376                         data->res.lr_res = &data->lr.res;
6377                 }
6378         } else if (data->lr.roc) {
6379                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6380                 data->lr.roc = false;
6381         }
6382
6383         task_setup_data.callback_data = data;
6384         msg.rpc_argp = &data->args;
6385         msg.rpc_resp = &data->res;
6386         task = rpc_run_task(&task_setup_data);
6387         if (IS_ERR(task))
6388                 return PTR_ERR(task);
6389         if (!issync)
6390                 goto out;
6391         status = rpc_wait_for_completion_task(task);
6392         if (status != 0)
6393                 goto out;
6394         status = data->rpc_status;
6395 out:
6396         rpc_put_task(task);
6397         return status;
6398 }
6399
6400 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6401 {
6402         struct nfs_server *server = NFS_SERVER(inode);
6403         struct nfs4_exception exception = { };
6404         int err;
6405         do {
6406                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6407                 trace_nfs4_delegreturn(inode, stateid, err);
6408                 switch (err) {
6409                         case -NFS4ERR_STALE_STATEID:
6410                         case -NFS4ERR_EXPIRED:
6411                         case 0:
6412                                 return 0;
6413                 }
6414                 err = nfs4_handle_exception(server, err, &exception);
6415         } while (exception.retry);
6416         return err;
6417 }
6418
6419 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6420 {
6421         struct inode *inode = state->inode;
6422         struct nfs_server *server = NFS_SERVER(inode);
6423         struct nfs_client *clp = server->nfs_client;
6424         struct nfs_lockt_args arg = {
6425                 .fh = NFS_FH(inode),
6426                 .fl = request,
6427         };
6428         struct nfs_lockt_res res = {
6429                 .denied = request,
6430         };
6431         struct rpc_message msg = {
6432                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6433                 .rpc_argp       = &arg,
6434                 .rpc_resp       = &res,
6435                 .rpc_cred       = state->owner->so_cred,
6436         };
6437         struct nfs4_lock_state *lsp;
6438         int status;
6439
6440         arg.lock_owner.clientid = clp->cl_clientid;
6441         status = nfs4_set_lock_state(state, request);
6442         if (status != 0)
6443                 goto out;
6444         lsp = request->fl_u.nfs4_fl.owner;
6445         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6446         arg.lock_owner.s_dev = server->s_dev;
6447         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6448         switch (status) {
6449                 case 0:
6450                         request->fl_type = F_UNLCK;
6451                         break;
6452                 case -NFS4ERR_DENIED:
6453                         status = 0;
6454         }
6455         request->fl_ops->fl_release_private(request);
6456         request->fl_ops = NULL;
6457 out:
6458         return status;
6459 }
6460
6461 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6462 {
6463         struct nfs4_exception exception = {
6464                 .interruptible = true,
6465         };
6466         int err;
6467
6468         do {
6469                 err = _nfs4_proc_getlk(state, cmd, request);
6470                 trace_nfs4_get_lock(request, state, cmd, err);
6471                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6472                                 &exception);
6473         } while (exception.retry);
6474         return err;
6475 }
6476
6477 /*
6478  * Update the seqid of a lock stateid after receiving
6479  * NFS4ERR_OLD_STATEID
6480  */
6481 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6482                 struct nfs4_lock_state *lsp)
6483 {
6484         struct nfs4_state *state = lsp->ls_state;
6485         bool ret = false;
6486
6487         spin_lock(&state->state_lock);
6488         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6489                 goto out;
6490         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6491                 nfs4_stateid_seqid_inc(dst);
6492         else
6493                 dst->seqid = lsp->ls_stateid.seqid;
6494         ret = true;
6495 out:
6496         spin_unlock(&state->state_lock);
6497         return ret;
6498 }
6499
6500 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6501                 struct nfs4_lock_state *lsp)
6502 {
6503         struct nfs4_state *state = lsp->ls_state;
6504         bool ret;
6505
6506         spin_lock(&state->state_lock);
6507         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6508         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6509         spin_unlock(&state->state_lock);
6510         return ret;
6511 }
6512
6513 struct nfs4_unlockdata {
6514         struct nfs_locku_args arg;
6515         struct nfs_locku_res res;
6516         struct nfs4_lock_state *lsp;
6517         struct nfs_open_context *ctx;
6518         struct nfs_lock_context *l_ctx;
6519         struct file_lock fl;
6520         struct nfs_server *server;
6521         unsigned long timestamp;
6522 };
6523
6524 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6525                 struct nfs_open_context *ctx,
6526                 struct nfs4_lock_state *lsp,
6527                 struct nfs_seqid *seqid)
6528 {
6529         struct nfs4_unlockdata *p;
6530         struct nfs4_state *state = lsp->ls_state;
6531         struct inode *inode = state->inode;
6532
6533         p = kzalloc(sizeof(*p), GFP_NOFS);
6534         if (p == NULL)
6535                 return NULL;
6536         p->arg.fh = NFS_FH(inode);
6537         p->arg.fl = &p->fl;
6538         p->arg.seqid = seqid;
6539         p->res.seqid = seqid;
6540         p->lsp = lsp;
6541         /* Ensure we don't close file until we're done freeing locks! */
6542         p->ctx = get_nfs_open_context(ctx);
6543         p->l_ctx = nfs_get_lock_context(ctx);
6544         locks_init_lock(&p->fl);
6545         locks_copy_lock(&p->fl, fl);
6546         p->server = NFS_SERVER(inode);
6547         spin_lock(&state->state_lock);
6548         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6549         spin_unlock(&state->state_lock);
6550         return p;
6551 }
6552
6553 static void nfs4_locku_release_calldata(void *data)
6554 {
6555         struct nfs4_unlockdata *calldata = data;
6556         nfs_free_seqid(calldata->arg.seqid);
6557         nfs4_put_lock_state(calldata->lsp);
6558         nfs_put_lock_context(calldata->l_ctx);
6559         put_nfs_open_context(calldata->ctx);
6560         kfree(calldata);
6561 }
6562
6563 static void nfs4_locku_done(struct rpc_task *task, void *data)
6564 {
6565         struct nfs4_unlockdata *calldata = data;
6566         struct nfs4_exception exception = {
6567                 .inode = calldata->lsp->ls_state->inode,
6568                 .stateid = &calldata->arg.stateid,
6569         };
6570
6571         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6572                 return;
6573         switch (task->tk_status) {
6574                 case 0:
6575                         renew_lease(calldata->server, calldata->timestamp);
6576                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6577                         if (nfs4_update_lock_stateid(calldata->lsp,
6578                                         &calldata->res.stateid))
6579                                 break;
6580                         /* Fall through */
6581                 case -NFS4ERR_ADMIN_REVOKED:
6582                 case -NFS4ERR_EXPIRED:
6583                         nfs4_free_revoked_stateid(calldata->server,
6584                                         &calldata->arg.stateid,
6585                                         task->tk_msg.rpc_cred);
6586                         /* Fall through */
6587                 case -NFS4ERR_BAD_STATEID:
6588                 case -NFS4ERR_STALE_STATEID:
6589                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6590                                                 calldata->lsp))
6591                                 rpc_restart_call_prepare(task);
6592                         break;
6593                 case -NFS4ERR_OLD_STATEID:
6594                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6595                                                 calldata->lsp))
6596                                 rpc_restart_call_prepare(task);
6597                         break;
6598                 default:
6599                         task->tk_status = nfs4_async_handle_exception(task,
6600                                         calldata->server, task->tk_status,
6601                                         &exception);
6602                         if (exception.retry)
6603                                 rpc_restart_call_prepare(task);
6604         }
6605         nfs_release_seqid(calldata->arg.seqid);
6606 }
6607
6608 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6609 {
6610         struct nfs4_unlockdata *calldata = data;
6611
6612         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6613                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6614                 return;
6615
6616         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6617                 goto out_wait;
6618         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6619                 /* Note: exit _without_ running nfs4_locku_done */
6620                 goto out_no_action;
6621         }
6622         calldata->timestamp = jiffies;
6623         if (nfs4_setup_sequence(calldata->server->nfs_client,
6624                                 &calldata->arg.seq_args,
6625                                 &calldata->res.seq_res,
6626                                 task) != 0)
6627                 nfs_release_seqid(calldata->arg.seqid);
6628         return;
6629 out_no_action:
6630         task->tk_action = NULL;
6631 out_wait:
6632         nfs4_sequence_done(task, &calldata->res.seq_res);
6633 }
6634
6635 static const struct rpc_call_ops nfs4_locku_ops = {
6636         .rpc_call_prepare = nfs4_locku_prepare,
6637         .rpc_call_done = nfs4_locku_done,
6638         .rpc_release = nfs4_locku_release_calldata,
6639 };
6640
6641 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6642                 struct nfs_open_context *ctx,
6643                 struct nfs4_lock_state *lsp,
6644                 struct nfs_seqid *seqid)
6645 {
6646         struct nfs4_unlockdata *data;
6647         struct rpc_message msg = {
6648                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6649                 .rpc_cred = ctx->cred,
6650         };
6651         struct rpc_task_setup task_setup_data = {
6652                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6653                 .rpc_message = &msg,
6654                 .callback_ops = &nfs4_locku_ops,
6655                 .workqueue = nfsiod_workqueue,
6656                 .flags = RPC_TASK_ASYNC,
6657         };
6658
6659         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6660                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6661
6662         /* Ensure this is an unlock - when canceling a lock, the
6663          * canceled lock is passed in, and it won't be an unlock.
6664          */
6665         fl->fl_type = F_UNLCK;
6666         if (fl->fl_flags & FL_CLOSE)
6667                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6668
6669         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6670         if (data == NULL) {
6671                 nfs_free_seqid(seqid);
6672                 return ERR_PTR(-ENOMEM);
6673         }
6674
6675         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6676         msg.rpc_argp = &data->arg;
6677         msg.rpc_resp = &data->res;
6678         task_setup_data.callback_data = data;
6679         return rpc_run_task(&task_setup_data);
6680 }
6681
6682 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6683 {
6684         struct inode *inode = state->inode;
6685         struct nfs4_state_owner *sp = state->owner;
6686         struct nfs_inode *nfsi = NFS_I(inode);
6687         struct nfs_seqid *seqid;
6688         struct nfs4_lock_state *lsp;
6689         struct rpc_task *task;
6690         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6691         int status = 0;
6692         unsigned char fl_flags = request->fl_flags;
6693
6694         status = nfs4_set_lock_state(state, request);
6695         /* Unlock _before_ we do the RPC call */
6696         request->fl_flags |= FL_EXISTS;
6697         /* Exclude nfs_delegation_claim_locks() */
6698         mutex_lock(&sp->so_delegreturn_mutex);
6699         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6700         down_read(&nfsi->rwsem);
6701         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6702                 up_read(&nfsi->rwsem);
6703                 mutex_unlock(&sp->so_delegreturn_mutex);
6704                 goto out;
6705         }
6706         up_read(&nfsi->rwsem);
6707         mutex_unlock(&sp->so_delegreturn_mutex);
6708         if (status != 0)
6709                 goto out;
6710         /* Is this a delegated lock? */
6711         lsp = request->fl_u.nfs4_fl.owner;
6712         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6713                 goto out;
6714         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6715         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6716         status = -ENOMEM;
6717         if (IS_ERR(seqid))
6718                 goto out;
6719         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6720         status = PTR_ERR(task);
6721         if (IS_ERR(task))
6722                 goto out;
6723         status = rpc_wait_for_completion_task(task);
6724         rpc_put_task(task);
6725 out:
6726         request->fl_flags = fl_flags;
6727         trace_nfs4_unlock(request, state, F_SETLK, status);
6728         return status;
6729 }
6730
6731 struct nfs4_lockdata {
6732         struct nfs_lock_args arg;
6733         struct nfs_lock_res res;
6734         struct nfs4_lock_state *lsp;
6735         struct nfs_open_context *ctx;
6736         struct file_lock fl;
6737         unsigned long timestamp;
6738         int rpc_status;
6739         int cancelled;
6740         struct nfs_server *server;
6741 };
6742
6743 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6744                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6745                 gfp_t gfp_mask)
6746 {
6747         struct nfs4_lockdata *p;
6748         struct inode *inode = lsp->ls_state->inode;
6749         struct nfs_server *server = NFS_SERVER(inode);
6750         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6751
6752         p = kzalloc(sizeof(*p), gfp_mask);
6753         if (p == NULL)
6754                 return NULL;
6755
6756         p->arg.fh = NFS_FH(inode);
6757         p->arg.fl = &p->fl;
6758         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6759         if (IS_ERR(p->arg.open_seqid))
6760                 goto out_free;
6761         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6762         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6763         if (IS_ERR(p->arg.lock_seqid))
6764                 goto out_free_seqid;
6765         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6766         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6767         p->arg.lock_owner.s_dev = server->s_dev;
6768         p->res.lock_seqid = p->arg.lock_seqid;
6769         p->lsp = lsp;
6770         p->server = server;
6771         p->ctx = get_nfs_open_context(ctx);
6772         locks_init_lock(&p->fl);
6773         locks_copy_lock(&p->fl, fl);
6774         return p;
6775 out_free_seqid:
6776         nfs_free_seqid(p->arg.open_seqid);
6777 out_free:
6778         kfree(p);
6779         return NULL;
6780 }
6781
6782 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6783 {
6784         struct nfs4_lockdata *data = calldata;
6785         struct nfs4_state *state = data->lsp->ls_state;
6786
6787         dprintk("%s: begin!\n", __func__);
6788         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6789                 goto out_wait;
6790         /* Do we need to do an open_to_lock_owner? */
6791         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6792                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6793                         goto out_release_lock_seqid;
6794                 }
6795                 nfs4_stateid_copy(&data->arg.open_stateid,
6796                                 &state->open_stateid);
6797                 data->arg.new_lock_owner = 1;
6798                 data->res.open_seqid = data->arg.open_seqid;
6799         } else {
6800                 data->arg.new_lock_owner = 0;
6801                 nfs4_stateid_copy(&data->arg.lock_stateid,
6802                                 &data->lsp->ls_stateid);
6803         }
6804         if (!nfs4_valid_open_stateid(state)) {
6805                 data->rpc_status = -EBADF;
6806                 task->tk_action = NULL;
6807                 goto out_release_open_seqid;
6808         }
6809         data->timestamp = jiffies;
6810         if (nfs4_setup_sequence(data->server->nfs_client,
6811                                 &data->arg.seq_args,
6812                                 &data->res.seq_res,
6813                                 task) == 0)
6814                 return;
6815 out_release_open_seqid:
6816         nfs_release_seqid(data->arg.open_seqid);
6817 out_release_lock_seqid:
6818         nfs_release_seqid(data->arg.lock_seqid);
6819 out_wait:
6820         nfs4_sequence_done(task, &data->res.seq_res);
6821         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6822 }
6823
6824 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6825 {
6826         struct nfs4_lockdata *data = calldata;
6827         struct nfs4_lock_state *lsp = data->lsp;
6828
6829         dprintk("%s: begin!\n", __func__);
6830
6831         if (!nfs4_sequence_done(task, &data->res.seq_res))
6832                 return;
6833
6834         data->rpc_status = task->tk_status;
6835         switch (task->tk_status) {
6836         case 0:
6837                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6838                                 data->timestamp);
6839                 if (data->arg.new_lock && !data->cancelled) {
6840                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6841                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6842                                 goto out_restart;
6843                 }
6844                 if (data->arg.new_lock_owner != 0) {
6845                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6846                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6847                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6848                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6849                         goto out_restart;
6850                 break;
6851         case -NFS4ERR_BAD_STATEID:
6852         case -NFS4ERR_OLD_STATEID:
6853         case -NFS4ERR_STALE_STATEID:
6854         case -NFS4ERR_EXPIRED:
6855                 if (data->arg.new_lock_owner != 0) {
6856                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6857                                                 &lsp->ls_state->open_stateid))
6858                                 goto out_restart;
6859                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6860                                                 &lsp->ls_stateid))
6861                                 goto out_restart;
6862         }
6863 out_done:
6864         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6865         return;
6866 out_restart:
6867         if (!data->cancelled)
6868                 rpc_restart_call_prepare(task);
6869         goto out_done;
6870 }
6871
6872 static void nfs4_lock_release(void *calldata)
6873 {
6874         struct nfs4_lockdata *data = calldata;
6875
6876         dprintk("%s: begin!\n", __func__);
6877         nfs_free_seqid(data->arg.open_seqid);
6878         if (data->cancelled && data->rpc_status == 0) {
6879                 struct rpc_task *task;
6880                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6881                                 data->arg.lock_seqid);
6882                 if (!IS_ERR(task))
6883                         rpc_put_task_async(task);
6884                 dprintk("%s: cancelling lock!\n", __func__);
6885         } else
6886                 nfs_free_seqid(data->arg.lock_seqid);
6887         nfs4_put_lock_state(data->lsp);
6888         put_nfs_open_context(data->ctx);
6889         kfree(data);
6890         dprintk("%s: done!\n", __func__);
6891 }
6892
6893 static const struct rpc_call_ops nfs4_lock_ops = {
6894         .rpc_call_prepare = nfs4_lock_prepare,
6895         .rpc_call_done = nfs4_lock_done,
6896         .rpc_release = nfs4_lock_release,
6897 };
6898
6899 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6900 {
6901         switch (error) {
6902         case -NFS4ERR_ADMIN_REVOKED:
6903         case -NFS4ERR_EXPIRED:
6904         case -NFS4ERR_BAD_STATEID:
6905                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6906                 if (new_lock_owner != 0 ||
6907                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6908                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6909                 break;
6910         case -NFS4ERR_STALE_STATEID:
6911                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6912                 nfs4_schedule_lease_recovery(server->nfs_client);
6913         }
6914 }
6915
6916 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6917 {
6918         struct nfs4_lockdata *data;
6919         struct rpc_task *task;
6920         struct rpc_message msg = {
6921                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6922                 .rpc_cred = state->owner->so_cred,
6923         };
6924         struct rpc_task_setup task_setup_data = {
6925                 .rpc_client = NFS_CLIENT(state->inode),
6926                 .rpc_message = &msg,
6927                 .callback_ops = &nfs4_lock_ops,
6928                 .workqueue = nfsiod_workqueue,
6929                 .flags = RPC_TASK_ASYNC,
6930         };
6931         int ret;
6932
6933         dprintk("%s: begin!\n", __func__);
6934         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6935                         fl->fl_u.nfs4_fl.owner,
6936                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6937         if (data == NULL)
6938                 return -ENOMEM;
6939         if (IS_SETLKW(cmd))
6940                 data->arg.block = 1;
6941         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6942                                 recovery_type > NFS_LOCK_NEW);
6943         msg.rpc_argp = &data->arg;
6944         msg.rpc_resp = &data->res;
6945         task_setup_data.callback_data = data;
6946         if (recovery_type > NFS_LOCK_NEW) {
6947                 if (recovery_type == NFS_LOCK_RECLAIM)
6948                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6949         } else
6950                 data->arg.new_lock = 1;
6951         task = rpc_run_task(&task_setup_data);
6952         if (IS_ERR(task))
6953                 return PTR_ERR(task);
6954         ret = rpc_wait_for_completion_task(task);
6955         if (ret == 0) {
6956                 ret = data->rpc_status;
6957                 if (ret)
6958                         nfs4_handle_setlk_error(data->server, data->lsp,
6959                                         data->arg.new_lock_owner, ret);
6960         } else
6961                 data->cancelled = true;
6962         rpc_put_task(task);
6963         dprintk("%s: done, ret = %d!\n", __func__, ret);
6964         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6965         return ret;
6966 }
6967
6968 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6969 {
6970         struct nfs_server *server = NFS_SERVER(state->inode);
6971         struct nfs4_exception exception = {
6972                 .inode = state->inode,
6973         };
6974         int err;
6975
6976         do {
6977                 /* Cache the lock if possible... */
6978                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6979                         return 0;
6980                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6981                 if (err != -NFS4ERR_DELAY)
6982                         break;
6983                 nfs4_handle_exception(server, err, &exception);
6984         } while (exception.retry);
6985         return err;
6986 }
6987
6988 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6989 {
6990         struct nfs_server *server = NFS_SERVER(state->inode);
6991         struct nfs4_exception exception = {
6992                 .inode = state->inode,
6993         };
6994         int err;
6995
6996         err = nfs4_set_lock_state(state, request);
6997         if (err != 0)
6998                 return err;
6999         if (!recover_lost_locks) {
7000                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7001                 return 0;
7002         }
7003         do {
7004                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7005                         return 0;
7006                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7007                 switch (err) {
7008                 default:
7009                         goto out;
7010                 case -NFS4ERR_GRACE:
7011                 case -NFS4ERR_DELAY:
7012                         nfs4_handle_exception(server, err, &exception);
7013                         err = 0;
7014                 }
7015         } while (exception.retry);
7016 out:
7017         return err;
7018 }
7019
7020 #if defined(CONFIG_NFS_V4_1)
7021 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7022 {
7023         struct nfs4_lock_state *lsp;
7024         int status;
7025
7026         status = nfs4_set_lock_state(state, request);
7027         if (status != 0)
7028                 return status;
7029         lsp = request->fl_u.nfs4_fl.owner;
7030         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7031             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7032                 return 0;
7033         return nfs4_lock_expired(state, request);
7034 }
7035 #endif
7036
7037 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7038 {
7039         struct nfs_inode *nfsi = NFS_I(state->inode);
7040         struct nfs4_state_owner *sp = state->owner;
7041         unsigned char fl_flags = request->fl_flags;
7042         int status;
7043
7044         request->fl_flags |= FL_ACCESS;
7045         status = locks_lock_inode_wait(state->inode, request);
7046         if (status < 0)
7047                 goto out;
7048         mutex_lock(&sp->so_delegreturn_mutex);
7049         down_read(&nfsi->rwsem);
7050         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7051                 /* Yes: cache locks! */
7052                 /* ...but avoid races with delegation recall... */
7053                 request->fl_flags = fl_flags & ~FL_SLEEP;
7054                 status = locks_lock_inode_wait(state->inode, request);
7055                 up_read(&nfsi->rwsem);
7056                 mutex_unlock(&sp->so_delegreturn_mutex);
7057                 goto out;
7058         }
7059         up_read(&nfsi->rwsem);
7060         mutex_unlock(&sp->so_delegreturn_mutex);
7061         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7062 out:
7063         request->fl_flags = fl_flags;
7064         return status;
7065 }
7066
7067 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7068 {
7069         struct nfs4_exception exception = {
7070                 .state = state,
7071                 .inode = state->inode,
7072                 .interruptible = true,
7073         };
7074         int err;
7075
7076         do {
7077                 err = _nfs4_proc_setlk(state, cmd, request);
7078                 if (err == -NFS4ERR_DENIED)
7079                         err = -EAGAIN;
7080                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7081                                 err, &exception);
7082         } while (exception.retry);
7083         return err;
7084 }
7085
7086 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7087 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7088
7089 static int
7090 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7091                         struct file_lock *request)
7092 {
7093         int             status = -ERESTARTSYS;
7094         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7095
7096         while(!signalled()) {
7097                 status = nfs4_proc_setlk(state, cmd, request);
7098                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7099                         break;
7100                 freezable_schedule_timeout_interruptible(timeout);
7101                 timeout *= 2;
7102                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7103                 status = -ERESTARTSYS;
7104         }
7105         return status;
7106 }
7107
7108 #ifdef CONFIG_NFS_V4_1
7109 struct nfs4_lock_waiter {
7110         struct task_struct      *task;
7111         struct inode            *inode;
7112         struct nfs_lowner       *owner;
7113 };
7114
7115 static int
7116 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7117 {
7118         int ret;
7119         struct nfs4_lock_waiter *waiter = wait->private;
7120
7121         /* NULL key means to wake up everyone */
7122         if (key) {
7123                 struct cb_notify_lock_args      *cbnl = key;
7124                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7125                                                 *wowner = waiter->owner;
7126
7127                 /* Only wake if the callback was for the same owner. */
7128                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7129                         return 0;
7130
7131                 /* Make sure it's for the right inode */
7132                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7133                         return 0;
7134         }
7135
7136         /* override "private" so we can use default_wake_function */
7137         wait->private = waiter->task;
7138         ret = woken_wake_function(wait, mode, flags, key);
7139         if (ret)
7140                 list_del_init(&wait->entry);
7141         wait->private = waiter;
7142         return ret;
7143 }
7144
7145 static int
7146 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7147 {
7148         int status = -ERESTARTSYS;
7149         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7150         struct nfs_server *server = NFS_SERVER(state->inode);
7151         struct nfs_client *clp = server->nfs_client;
7152         wait_queue_head_t *q = &clp->cl_lock_waitq;
7153         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7154                                     .id = lsp->ls_seqid.owner_id,
7155                                     .s_dev = server->s_dev };
7156         struct nfs4_lock_waiter waiter = { .task  = current,
7157                                            .inode = state->inode,
7158                                            .owner = &owner};
7159         wait_queue_entry_t wait;
7160
7161         /* Don't bother with waitqueue if we don't expect a callback */
7162         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7163                 return nfs4_retry_setlk_simple(state, cmd, request);
7164
7165         init_wait(&wait);
7166         wait.private = &waiter;
7167         wait.func = nfs4_wake_lock_waiter;
7168
7169         while(!signalled()) {
7170                 add_wait_queue(q, &wait);
7171                 status = nfs4_proc_setlk(state, cmd, request);
7172                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7173                         finish_wait(q, &wait);
7174                         break;
7175                 }
7176
7177                 status = -ERESTARTSYS;
7178                 freezer_do_not_count();
7179                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7180                 freezer_count();
7181                 finish_wait(q, &wait);
7182         }
7183
7184         return status;
7185 }
7186 #else /* !CONFIG_NFS_V4_1 */
7187 static inline int
7188 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7189 {
7190         return nfs4_retry_setlk_simple(state, cmd, request);
7191 }
7192 #endif
7193
7194 static int
7195 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7196 {
7197         struct nfs_open_context *ctx;
7198         struct nfs4_state *state;
7199         int status;
7200
7201         /* verify open state */
7202         ctx = nfs_file_open_context(filp);
7203         state = ctx->state;
7204
7205         if (IS_GETLK(cmd)) {
7206                 if (state != NULL)
7207                         return nfs4_proc_getlk(state, F_GETLK, request);
7208                 return 0;
7209         }
7210
7211         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7212                 return -EINVAL;
7213
7214         if (request->fl_type == F_UNLCK) {
7215                 if (state != NULL)
7216                         return nfs4_proc_unlck(state, cmd, request);
7217                 return 0;
7218         }
7219
7220         if (state == NULL)
7221                 return -ENOLCK;
7222
7223         if ((request->fl_flags & FL_POSIX) &&
7224             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7225                 return -ENOLCK;
7226
7227         /*
7228          * Don't rely on the VFS having checked the file open mode,
7229          * since it won't do this for flock() locks.
7230          */
7231         switch (request->fl_type) {
7232         case F_RDLCK:
7233                 if (!(filp->f_mode & FMODE_READ))
7234                         return -EBADF;
7235                 break;
7236         case F_WRLCK:
7237                 if (!(filp->f_mode & FMODE_WRITE))
7238                         return -EBADF;
7239         }
7240
7241         status = nfs4_set_lock_state(state, request);
7242         if (status != 0)
7243                 return status;
7244
7245         return nfs4_retry_setlk(state, cmd, request);
7246 }
7247
7248 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7249 {
7250         struct nfs_server *server = NFS_SERVER(state->inode);
7251         int err;
7252
7253         err = nfs4_set_lock_state(state, fl);
7254         if (err != 0)
7255                 return err;
7256         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7257         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7258 }
7259
7260 struct nfs_release_lockowner_data {
7261         struct nfs4_lock_state *lsp;
7262         struct nfs_server *server;
7263         struct nfs_release_lockowner_args args;
7264         struct nfs_release_lockowner_res res;
7265         unsigned long timestamp;
7266 };
7267
7268 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7269 {
7270         struct nfs_release_lockowner_data *data = calldata;
7271         struct nfs_server *server = data->server;
7272         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7273                            &data->res.seq_res, task);
7274         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7275         data->timestamp = jiffies;
7276 }
7277
7278 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7279 {
7280         struct nfs_release_lockowner_data *data = calldata;
7281         struct nfs_server *server = data->server;
7282
7283         nfs40_sequence_done(task, &data->res.seq_res);
7284
7285         switch (task->tk_status) {
7286         case 0:
7287                 renew_lease(server, data->timestamp);
7288                 break;
7289         case -NFS4ERR_STALE_CLIENTID:
7290         case -NFS4ERR_EXPIRED:
7291                 nfs4_schedule_lease_recovery(server->nfs_client);
7292                 break;
7293         case -NFS4ERR_LEASE_MOVED:
7294         case -NFS4ERR_DELAY:
7295                 if (nfs4_async_handle_error(task, server,
7296                                             NULL, NULL) == -EAGAIN)
7297                         rpc_restart_call_prepare(task);
7298         }
7299 }
7300
7301 static void nfs4_release_lockowner_release(void *calldata)
7302 {
7303         struct nfs_release_lockowner_data *data = calldata;
7304         nfs4_free_lock_state(data->server, data->lsp);
7305         kfree(calldata);
7306 }
7307
7308 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7309         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7310         .rpc_call_done = nfs4_release_lockowner_done,
7311         .rpc_release = nfs4_release_lockowner_release,
7312 };
7313
7314 static void
7315 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7316 {
7317         struct nfs_release_lockowner_data *data;
7318         struct rpc_message msg = {
7319                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7320         };
7321
7322         if (server->nfs_client->cl_mvops->minor_version != 0)
7323                 return;
7324
7325         data = kmalloc(sizeof(*data), GFP_NOFS);
7326         if (!data)
7327                 return;
7328         data->lsp = lsp;
7329         data->server = server;
7330         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7331         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7332         data->args.lock_owner.s_dev = server->s_dev;
7333
7334         msg.rpc_argp = &data->args;
7335         msg.rpc_resp = &data->res;
7336         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7337         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7338 }
7339
7340 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7341
7342 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7343                                    struct dentry *unused, struct inode *inode,
7344                                    const char *key, const void *buf,
7345                                    size_t buflen, int flags)
7346 {
7347         return nfs4_proc_set_acl(inode, buf, buflen);
7348 }
7349
7350 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7351                                    struct dentry *unused, struct inode *inode,
7352                                    const char *key, void *buf, size_t buflen)
7353 {
7354         return nfs4_proc_get_acl(inode, buf, buflen);
7355 }
7356
7357 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7358 {
7359         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7360 }
7361
7362 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7363
7364 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7365                                      struct dentry *unused, struct inode *inode,
7366                                      const char *key, const void *buf,
7367                                      size_t buflen, int flags)
7368 {
7369         if (security_ismaclabel(key))
7370                 return nfs4_set_security_label(inode, buf, buflen);
7371
7372         return -EOPNOTSUPP;
7373 }
7374
7375 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7376                                      struct dentry *unused, struct inode *inode,
7377                                      const char *key, void *buf, size_t buflen)
7378 {
7379         if (security_ismaclabel(key))
7380                 return nfs4_get_security_label(inode, buf, buflen);
7381         return -EOPNOTSUPP;
7382 }
7383
7384 static ssize_t
7385 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7386 {
7387         int len = 0;
7388
7389         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7390                 len = security_inode_listsecurity(inode, list, list_len);
7391                 if (list_len && len > list_len)
7392                         return -ERANGE;
7393         }
7394         return len;
7395 }
7396
7397 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7398         .prefix = XATTR_SECURITY_PREFIX,
7399         .get    = nfs4_xattr_get_nfs4_label,
7400         .set    = nfs4_xattr_set_nfs4_label,
7401 };
7402
7403 #else
7404
7405 static ssize_t
7406 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7407 {
7408         return 0;
7409 }
7410
7411 #endif
7412
7413 /*
7414  * nfs_fhget will use either the mounted_on_fileid or the fileid
7415  */
7416 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7417 {
7418         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7419                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7420               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7421               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7422                 return;
7423
7424         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7425                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7426         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7427         fattr->nlink = 2;
7428 }
7429
7430 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7431                                    const struct qstr *name,
7432                                    struct nfs4_fs_locations *fs_locations,
7433                                    struct page *page)
7434 {
7435         struct nfs_server *server = NFS_SERVER(dir);
7436         u32 bitmask[3];
7437         struct nfs4_fs_locations_arg args = {
7438                 .dir_fh = NFS_FH(dir),
7439                 .name = name,
7440                 .page = page,
7441                 .bitmask = bitmask,
7442         };
7443         struct nfs4_fs_locations_res res = {
7444                 .fs_locations = fs_locations,
7445         };
7446         struct rpc_message msg = {
7447                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7448                 .rpc_argp = &args,
7449                 .rpc_resp = &res,
7450         };
7451         int status;
7452
7453         dprintk("%s: start\n", __func__);
7454
7455         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7456         bitmask[1] = nfs4_fattr_bitmap[1];
7457
7458         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7459          * is not supported */
7460         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7461                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7462         else
7463                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7464
7465         nfs_fattr_init(&fs_locations->fattr);
7466         fs_locations->server = server;
7467         fs_locations->nlocations = 0;
7468         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7469         dprintk("%s: returned status = %d\n", __func__, status);
7470         return status;
7471 }
7472
7473 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7474                            const struct qstr *name,
7475                            struct nfs4_fs_locations *fs_locations,
7476                            struct page *page)
7477 {
7478         struct nfs4_exception exception = {
7479                 .interruptible = true,
7480         };
7481         int err;
7482         do {
7483                 err = _nfs4_proc_fs_locations(client, dir, name,
7484                                 fs_locations, page);
7485                 trace_nfs4_get_fs_locations(dir, name, err);
7486                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7487                                 &exception);
7488         } while (exception.retry);
7489         return err;
7490 }
7491
7492 /*
7493  * This operation also signals the server that this client is
7494  * performing migration recovery.  The server can stop returning
7495  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7496  * appended to this compound to identify the client ID which is
7497  * performing recovery.
7498  */
7499 static int _nfs40_proc_get_locations(struct inode *inode,
7500                                      struct nfs4_fs_locations *locations,
7501                                      struct page *page, const struct cred *cred)
7502 {
7503         struct nfs_server *server = NFS_SERVER(inode);
7504         struct rpc_clnt *clnt = server->client;
7505         u32 bitmask[2] = {
7506                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7507         };
7508         struct nfs4_fs_locations_arg args = {
7509                 .clientid       = server->nfs_client->cl_clientid,
7510                 .fh             = NFS_FH(inode),
7511                 .page           = page,
7512                 .bitmask        = bitmask,
7513                 .migration      = 1,            /* skip LOOKUP */
7514                 .renew          = 1,            /* append RENEW */
7515         };
7516         struct nfs4_fs_locations_res res = {
7517                 .fs_locations   = locations,
7518                 .migration      = 1,
7519                 .renew          = 1,
7520         };
7521         struct rpc_message msg = {
7522                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7523                 .rpc_argp       = &args,
7524                 .rpc_resp       = &res,
7525                 .rpc_cred       = cred,
7526         };
7527         unsigned long now = jiffies;
7528         int status;
7529
7530         nfs_fattr_init(&locations->fattr);
7531         locations->server = server;
7532         locations->nlocations = 0;
7533
7534         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7535         status = nfs4_call_sync_sequence(clnt, server, &msg,
7536                                         &args.seq_args, &res.seq_res);
7537         if (status)
7538                 return status;
7539
7540         renew_lease(server, now);
7541         return 0;
7542 }
7543
7544 #ifdef CONFIG_NFS_V4_1
7545
7546 /*
7547  * This operation also signals the server that this client is
7548  * performing migration recovery.  The server can stop asserting
7549  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7550  * performing this operation is identified in the SEQUENCE
7551  * operation in this compound.
7552  *
7553  * When the client supports GETATTR(fs_locations_info), it can
7554  * be plumbed in here.
7555  */
7556 static int _nfs41_proc_get_locations(struct inode *inode,
7557                                      struct nfs4_fs_locations *locations,
7558                                      struct page *page, const struct cred *cred)
7559 {
7560         struct nfs_server *server = NFS_SERVER(inode);
7561         struct rpc_clnt *clnt = server->client;
7562         u32 bitmask[2] = {
7563                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7564         };
7565         struct nfs4_fs_locations_arg args = {
7566                 .fh             = NFS_FH(inode),
7567                 .page           = page,
7568                 .bitmask        = bitmask,
7569                 .migration      = 1,            /* skip LOOKUP */
7570         };
7571         struct nfs4_fs_locations_res res = {
7572                 .fs_locations   = locations,
7573                 .migration      = 1,
7574         };
7575         struct rpc_message msg = {
7576                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7577                 .rpc_argp       = &args,
7578                 .rpc_resp       = &res,
7579                 .rpc_cred       = cred,
7580         };
7581         int status;
7582
7583         nfs_fattr_init(&locations->fattr);
7584         locations->server = server;
7585         locations->nlocations = 0;
7586
7587         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7588         status = nfs4_call_sync_sequence(clnt, server, &msg,
7589                                         &args.seq_args, &res.seq_res);
7590         if (status == NFS4_OK &&
7591             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7592                 status = -NFS4ERR_LEASE_MOVED;
7593         return status;
7594 }
7595
7596 #endif  /* CONFIG_NFS_V4_1 */
7597
7598 /**
7599  * nfs4_proc_get_locations - discover locations for a migrated FSID
7600  * @inode: inode on FSID that is migrating
7601  * @locations: result of query
7602  * @page: buffer
7603  * @cred: credential to use for this operation
7604  *
7605  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7606  * operation failed, or a negative errno if a local error occurred.
7607  *
7608  * On success, "locations" is filled in, but if the server has
7609  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7610  * asserted.
7611  *
7612  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7613  * from this client that require migration recovery.
7614  */
7615 int nfs4_proc_get_locations(struct inode *inode,
7616                             struct nfs4_fs_locations *locations,
7617                             struct page *page, const struct cred *cred)
7618 {
7619         struct nfs_server *server = NFS_SERVER(inode);
7620         struct nfs_client *clp = server->nfs_client;
7621         const struct nfs4_mig_recovery_ops *ops =
7622                                         clp->cl_mvops->mig_recovery_ops;
7623         struct nfs4_exception exception = {
7624                 .interruptible = true,
7625         };
7626         int status;
7627
7628         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7629                 (unsigned long long)server->fsid.major,
7630                 (unsigned long long)server->fsid.minor,
7631                 clp->cl_hostname);
7632         nfs_display_fhandle(NFS_FH(inode), __func__);
7633
7634         do {
7635                 status = ops->get_locations(inode, locations, page, cred);
7636                 if (status != -NFS4ERR_DELAY)
7637                         break;
7638                 nfs4_handle_exception(server, status, &exception);
7639         } while (exception.retry);
7640         return status;
7641 }
7642
7643 /*
7644  * This operation also signals the server that this client is
7645  * performing "lease moved" recovery.  The server can stop
7646  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7647  * is appended to this compound to identify the client ID which is
7648  * performing recovery.
7649  */
7650 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7651 {
7652         struct nfs_server *server = NFS_SERVER(inode);
7653         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7654         struct rpc_clnt *clnt = server->client;
7655         struct nfs4_fsid_present_arg args = {
7656                 .fh             = NFS_FH(inode),
7657                 .clientid       = clp->cl_clientid,
7658                 .renew          = 1,            /* append RENEW */
7659         };
7660         struct nfs4_fsid_present_res res = {
7661                 .renew          = 1,
7662         };
7663         struct rpc_message msg = {
7664                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7665                 .rpc_argp       = &args,
7666                 .rpc_resp       = &res,
7667                 .rpc_cred       = cred,
7668         };
7669         unsigned long now = jiffies;
7670         int status;
7671
7672         res.fh = nfs_alloc_fhandle();
7673         if (res.fh == NULL)
7674                 return -ENOMEM;
7675
7676         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7677         status = nfs4_call_sync_sequence(clnt, server, &msg,
7678                                                 &args.seq_args, &res.seq_res);
7679         nfs_free_fhandle(res.fh);
7680         if (status)
7681                 return status;
7682
7683         do_renew_lease(clp, now);
7684         return 0;
7685 }
7686
7687 #ifdef CONFIG_NFS_V4_1
7688
7689 /*
7690  * This operation also signals the server that this client is
7691  * performing "lease moved" recovery.  The server can stop asserting
7692  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7693  * this operation is identified in the SEQUENCE operation in this
7694  * compound.
7695  */
7696 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7697 {
7698         struct nfs_server *server = NFS_SERVER(inode);
7699         struct rpc_clnt *clnt = server->client;
7700         struct nfs4_fsid_present_arg args = {
7701                 .fh             = NFS_FH(inode),
7702         };
7703         struct nfs4_fsid_present_res res = {
7704         };
7705         struct rpc_message msg = {
7706                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7707                 .rpc_argp       = &args,
7708                 .rpc_resp       = &res,
7709                 .rpc_cred       = cred,
7710         };
7711         int status;
7712
7713         res.fh = nfs_alloc_fhandle();
7714         if (res.fh == NULL)
7715                 return -ENOMEM;
7716
7717         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7718         status = nfs4_call_sync_sequence(clnt, server, &msg,
7719                                                 &args.seq_args, &res.seq_res);
7720         nfs_free_fhandle(res.fh);
7721         if (status == NFS4_OK &&
7722             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7723                 status = -NFS4ERR_LEASE_MOVED;
7724         return status;
7725 }
7726
7727 #endif  /* CONFIG_NFS_V4_1 */
7728
7729 /**
7730  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7731  * @inode: inode on FSID to check
7732  * @cred: credential to use for this operation
7733  *
7734  * Server indicates whether the FSID is present, moved, or not
7735  * recognized.  This operation is necessary to clear a LEASE_MOVED
7736  * condition for this client ID.
7737  *
7738  * Returns NFS4_OK if the FSID is present on this server,
7739  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7740  *  NFS4ERR code if some error occurred on the server, or a
7741  *  negative errno if a local failure occurred.
7742  */
7743 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7744 {
7745         struct nfs_server *server = NFS_SERVER(inode);
7746         struct nfs_client *clp = server->nfs_client;
7747         const struct nfs4_mig_recovery_ops *ops =
7748                                         clp->cl_mvops->mig_recovery_ops;
7749         struct nfs4_exception exception = {
7750                 .interruptible = true,
7751         };
7752         int status;
7753
7754         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7755                 (unsigned long long)server->fsid.major,
7756                 (unsigned long long)server->fsid.minor,
7757                 clp->cl_hostname);
7758         nfs_display_fhandle(NFS_FH(inode), __func__);
7759
7760         do {
7761                 status = ops->fsid_present(inode, cred);
7762                 if (status != -NFS4ERR_DELAY)
7763                         break;
7764                 nfs4_handle_exception(server, status, &exception);
7765         } while (exception.retry);
7766         return status;
7767 }
7768
7769 /*
7770  * If 'use_integrity' is true and the state managment nfs_client
7771  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7772  * and the machine credential as per RFC3530bis and RFC5661 Security
7773  * Considerations sections. Otherwise, just use the user cred with the
7774  * filesystem's rpc_client.
7775  */
7776 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7777 {
7778         int status;
7779         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7780         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7781         struct nfs4_secinfo_arg args = {
7782                 .dir_fh = NFS_FH(dir),
7783                 .name   = name,
7784         };
7785         struct nfs4_secinfo_res res = {
7786                 .flavors     = flavors,
7787         };
7788         struct rpc_message msg = {
7789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7790                 .rpc_argp = &args,
7791                 .rpc_resp = &res,
7792         };
7793         struct nfs4_call_sync_data data = {
7794                 .seq_server = NFS_SERVER(dir),
7795                 .seq_args = &args.seq_args,
7796                 .seq_res = &res.seq_res,
7797         };
7798         struct rpc_task_setup task_setup = {
7799                 .rpc_client = clnt,
7800                 .rpc_message = &msg,
7801                 .callback_ops = clp->cl_mvops->call_sync_ops,
7802                 .callback_data = &data,
7803                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7804         };
7805         const struct cred *cred = NULL;
7806
7807         if (use_integrity) {
7808                 clnt = clp->cl_rpcclient;
7809                 task_setup.rpc_client = clnt;
7810
7811                 cred = nfs4_get_clid_cred(clp);
7812                 msg.rpc_cred = cred;
7813         }
7814
7815         dprintk("NFS call  secinfo %s\n", name->name);
7816
7817         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7818         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7819         status = nfs4_call_sync_custom(&task_setup);
7820
7821         dprintk("NFS reply  secinfo: %d\n", status);
7822
7823         put_cred(cred);
7824         return status;
7825 }
7826
7827 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7828                       struct nfs4_secinfo_flavors *flavors)
7829 {
7830         struct nfs4_exception exception = {
7831                 .interruptible = true,
7832         };
7833         int err;
7834         do {
7835                 err = -NFS4ERR_WRONGSEC;
7836
7837                 /* try to use integrity protection with machine cred */
7838                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7839                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7840
7841                 /*
7842                  * if unable to use integrity protection, or SECINFO with
7843                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7844                  * disallowed by spec, but exists in deployed servers) use
7845                  * the current filesystem's rpc_client and the user cred.
7846                  */
7847                 if (err == -NFS4ERR_WRONGSEC)
7848                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7849
7850                 trace_nfs4_secinfo(dir, name, err);
7851                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7852                                 &exception);
7853         } while (exception.retry);
7854         return err;
7855 }
7856
7857 #ifdef CONFIG_NFS_V4_1
7858 /*
7859  * Check the exchange flags returned by the server for invalid flags, having
7860  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7861  * DS flags set.
7862  */
7863 static int nfs4_check_cl_exchange_flags(u32 flags)
7864 {
7865         if (flags & ~EXCHGID4_FLAG_MASK_R)
7866                 goto out_inval;
7867         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7868             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7869                 goto out_inval;
7870         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7871                 goto out_inval;
7872         return NFS_OK;
7873 out_inval:
7874         return -NFS4ERR_INVAL;
7875 }
7876
7877 static bool
7878 nfs41_same_server_scope(struct nfs41_server_scope *a,
7879                         struct nfs41_server_scope *b)
7880 {
7881         if (a->server_scope_sz != b->server_scope_sz)
7882                 return false;
7883         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7884 }
7885
7886 static void
7887 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7888 {
7889         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
7890         struct nfs_client *clp = args->client;
7891
7892         switch (task->tk_status) {
7893         case -NFS4ERR_BADSESSION:
7894         case -NFS4ERR_DEADSESSION:
7895                 nfs4_schedule_session_recovery(clp->cl_session,
7896                                 task->tk_status);
7897         }
7898 }
7899
7900 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7901         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7902 };
7903
7904 /*
7905  * nfs4_proc_bind_one_conn_to_session()
7906  *
7907  * The 4.1 client currently uses the same TCP connection for the
7908  * fore and backchannel.
7909  */
7910 static
7911 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7912                 struct rpc_xprt *xprt,
7913                 struct nfs_client *clp,
7914                 const struct cred *cred)
7915 {
7916         int status;
7917         struct nfs41_bind_conn_to_session_args args = {
7918                 .client = clp,
7919                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7920         };
7921         struct nfs41_bind_conn_to_session_res res;
7922         struct rpc_message msg = {
7923                 .rpc_proc =
7924                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7925                 .rpc_argp = &args,
7926                 .rpc_resp = &res,
7927                 .rpc_cred = cred,
7928         };
7929         struct rpc_task_setup task_setup_data = {
7930                 .rpc_client = clnt,
7931                 .rpc_xprt = xprt,
7932                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7933                 .rpc_message = &msg,
7934                 .flags = RPC_TASK_TIMEOUT,
7935         };
7936         struct rpc_task *task;
7937
7938         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7939         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7940                 args.dir = NFS4_CDFC4_FORE;
7941
7942         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7943         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7944                 args.dir = NFS4_CDFC4_FORE;
7945
7946         task = rpc_run_task(&task_setup_data);
7947         if (!IS_ERR(task)) {
7948                 status = task->tk_status;
7949                 rpc_put_task(task);
7950         } else
7951                 status = PTR_ERR(task);
7952         trace_nfs4_bind_conn_to_session(clp, status);
7953         if (status == 0) {
7954                 if (memcmp(res.sessionid.data,
7955                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7956                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7957                         return -EIO;
7958                 }
7959                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7960                         dprintk("NFS: %s: Unexpected direction from server\n",
7961                                 __func__);
7962                         return -EIO;
7963                 }
7964                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7965                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7966                                 __func__);
7967                         return -EIO;
7968                 }
7969         }
7970
7971         return status;
7972 }
7973
7974 struct rpc_bind_conn_calldata {
7975         struct nfs_client *clp;
7976         const struct cred *cred;
7977 };
7978
7979 static int
7980 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7981                 struct rpc_xprt *xprt,
7982                 void *calldata)
7983 {
7984         struct rpc_bind_conn_calldata *p = calldata;
7985
7986         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7987 }
7988
7989 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7990 {
7991         struct rpc_bind_conn_calldata data = {
7992                 .clp = clp,
7993                 .cred = cred,
7994         };
7995         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7996                         nfs4_proc_bind_conn_to_session_callback, &data);
7997 }
7998
7999 /*
8000  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8001  * and operations we'd like to see to enable certain features in the allow map
8002  */
8003 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8004         .how = SP4_MACH_CRED,
8005         .enforce.u.words = {
8006                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8007                       1 << (OP_EXCHANGE_ID - 32) |
8008                       1 << (OP_CREATE_SESSION - 32) |
8009                       1 << (OP_DESTROY_SESSION - 32) |
8010                       1 << (OP_DESTROY_CLIENTID - 32)
8011         },
8012         .allow.u.words = {
8013                 [0] = 1 << (OP_CLOSE) |
8014                       1 << (OP_OPEN_DOWNGRADE) |
8015                       1 << (OP_LOCKU) |
8016                       1 << (OP_DELEGRETURN) |
8017                       1 << (OP_COMMIT),
8018                 [1] = 1 << (OP_SECINFO - 32) |
8019                       1 << (OP_SECINFO_NO_NAME - 32) |
8020                       1 << (OP_LAYOUTRETURN - 32) |
8021                       1 << (OP_TEST_STATEID - 32) |
8022                       1 << (OP_FREE_STATEID - 32) |
8023                       1 << (OP_WRITE - 32)
8024         }
8025 };
8026
8027 /*
8028  * Select the state protection mode for client `clp' given the server results
8029  * from exchange_id in `sp'.
8030  *
8031  * Returns 0 on success, negative errno otherwise.
8032  */
8033 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8034                                  struct nfs41_state_protection *sp)
8035 {
8036         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8037                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8038                       1 << (OP_EXCHANGE_ID - 32) |
8039                       1 << (OP_CREATE_SESSION - 32) |
8040                       1 << (OP_DESTROY_SESSION - 32) |
8041                       1 << (OP_DESTROY_CLIENTID - 32)
8042         };
8043         unsigned long flags = 0;
8044         unsigned int i;
8045         int ret = 0;
8046
8047         if (sp->how == SP4_MACH_CRED) {
8048                 /* Print state protect result */
8049                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8050                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8051                         if (test_bit(i, sp->enforce.u.longs))
8052                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8053                         if (test_bit(i, sp->allow.u.longs))
8054                                 dfprintk(MOUNT, "  allow op %d\n", i);
8055                 }
8056
8057                 /* make sure nothing is on enforce list that isn't supported */
8058                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8059                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8060                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8061                                 ret = -EINVAL;
8062                                 goto out;
8063                         }
8064                 }
8065
8066                 /*
8067                  * Minimal mode - state operations are allowed to use machine
8068                  * credential.  Note this already happens by default, so the
8069                  * client doesn't have to do anything more than the negotiation.
8070                  *
8071                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8072                  *       we're already using the machine cred for exchange_id
8073                  *       and will never use a different cred.
8074                  */
8075                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8076                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8077                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8078                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8079                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8080                         dfprintk(MOUNT, "  minimal mode enabled\n");
8081                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8082                 } else {
8083                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8084                         ret = -EINVAL;
8085                         goto out;
8086                 }
8087
8088                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8089                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8090                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8091                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8092                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8093                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8094                 }
8095
8096                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8097                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8098                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8099                 }
8100
8101                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8102                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8103                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8104                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8105                 }
8106
8107                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8108                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8109                         dfprintk(MOUNT, "  stateid mode enabled\n");
8110                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8111                 }
8112
8113                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8114                         dfprintk(MOUNT, "  write mode enabled\n");
8115                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8116                 }
8117
8118                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8119                         dfprintk(MOUNT, "  commit mode enabled\n");
8120                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8121                 }
8122         }
8123 out:
8124         clp->cl_sp4_flags = flags;
8125         return ret;
8126 }
8127
8128 struct nfs41_exchange_id_data {
8129         struct nfs41_exchange_id_res res;
8130         struct nfs41_exchange_id_args args;
8131 };
8132
8133 static void nfs4_exchange_id_release(void *data)
8134 {
8135         struct nfs41_exchange_id_data *cdata =
8136                                         (struct nfs41_exchange_id_data *)data;
8137
8138         nfs_put_client(cdata->args.client);
8139         kfree(cdata->res.impl_id);
8140         kfree(cdata->res.server_scope);
8141         kfree(cdata->res.server_owner);
8142         kfree(cdata);
8143 }
8144
8145 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8146         .rpc_release = nfs4_exchange_id_release,
8147 };
8148
8149 /*
8150  * _nfs4_proc_exchange_id()
8151  *
8152  * Wrapper for EXCHANGE_ID operation.
8153  */
8154 static struct rpc_task *
8155 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8156                         u32 sp4_how, struct rpc_xprt *xprt)
8157 {
8158         struct rpc_message msg = {
8159                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8160                 .rpc_cred = cred,
8161         };
8162         struct rpc_task_setup task_setup_data = {
8163                 .rpc_client = clp->cl_rpcclient,
8164                 .callback_ops = &nfs4_exchange_id_call_ops,
8165                 .rpc_message = &msg,
8166                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8167         };
8168         struct nfs41_exchange_id_data *calldata;
8169         int status;
8170
8171         if (!refcount_inc_not_zero(&clp->cl_count))
8172                 return ERR_PTR(-EIO);
8173
8174         status = -ENOMEM;
8175         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8176         if (!calldata)
8177                 goto out;
8178
8179         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8180
8181         status = nfs4_init_uniform_client_string(clp);
8182         if (status)
8183                 goto out_calldata;
8184
8185         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8186                                                 GFP_NOFS);
8187         status = -ENOMEM;
8188         if (unlikely(calldata->res.server_owner == NULL))
8189                 goto out_calldata;
8190
8191         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8192                                         GFP_NOFS);
8193         if (unlikely(calldata->res.server_scope == NULL))
8194                 goto out_server_owner;
8195
8196         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8197         if (unlikely(calldata->res.impl_id == NULL))
8198                 goto out_server_scope;
8199
8200         switch (sp4_how) {
8201         case SP4_NONE:
8202                 calldata->args.state_protect.how = SP4_NONE;
8203                 break;
8204
8205         case SP4_MACH_CRED:
8206                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8207                 break;
8208
8209         default:
8210                 /* unsupported! */
8211                 WARN_ON_ONCE(1);
8212                 status = -EINVAL;
8213                 goto out_impl_id;
8214         }
8215         if (xprt) {
8216                 task_setup_data.rpc_xprt = xprt;
8217                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8218                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8219                                 sizeof(calldata->args.verifier.data));
8220         }
8221         calldata->args.client = clp;
8222         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8223         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8224 #ifdef CONFIG_NFS_V4_1_MIGRATION
8225         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8226 #endif
8227         msg.rpc_argp = &calldata->args;
8228         msg.rpc_resp = &calldata->res;
8229         task_setup_data.callback_data = calldata;
8230
8231         return rpc_run_task(&task_setup_data);
8232
8233 out_impl_id:
8234         kfree(calldata->res.impl_id);
8235 out_server_scope:
8236         kfree(calldata->res.server_scope);
8237 out_server_owner:
8238         kfree(calldata->res.server_owner);
8239 out_calldata:
8240         kfree(calldata);
8241 out:
8242         nfs_put_client(clp);
8243         return ERR_PTR(status);
8244 }
8245
8246 /*
8247  * _nfs4_proc_exchange_id()
8248  *
8249  * Wrapper for EXCHANGE_ID operation.
8250  */
8251 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8252                         u32 sp4_how)
8253 {
8254         struct rpc_task *task;
8255         struct nfs41_exchange_id_args *argp;
8256         struct nfs41_exchange_id_res *resp;
8257         unsigned long now = jiffies;
8258         int status;
8259
8260         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8261         if (IS_ERR(task))
8262                 return PTR_ERR(task);
8263
8264         argp = task->tk_msg.rpc_argp;
8265         resp = task->tk_msg.rpc_resp;
8266         status = task->tk_status;
8267         if (status  != 0)
8268                 goto out;
8269
8270         status = nfs4_check_cl_exchange_flags(resp->flags);
8271         if (status  != 0)
8272                 goto out;
8273
8274         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8275         if (status != 0)
8276                 goto out;
8277
8278         do_renew_lease(clp, now);
8279
8280         clp->cl_clientid = resp->clientid;
8281         clp->cl_exchange_flags = resp->flags;
8282         clp->cl_seqid = resp->seqid;
8283         /* Client ID is not confirmed */
8284         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8285                 clear_bit(NFS4_SESSION_ESTABLISHED,
8286                           &clp->cl_session->session_state);
8287
8288         if (clp->cl_serverscope != NULL &&
8289             !nfs41_same_server_scope(clp->cl_serverscope,
8290                                 resp->server_scope)) {
8291                 dprintk("%s: server_scope mismatch detected\n",
8292                         __func__);
8293                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8294         }
8295
8296         swap(clp->cl_serverowner, resp->server_owner);
8297         swap(clp->cl_serverscope, resp->server_scope);
8298         swap(clp->cl_implid, resp->impl_id);
8299
8300         /* Save the EXCHANGE_ID verifier session trunk tests */
8301         memcpy(clp->cl_confirm.data, argp->verifier.data,
8302                sizeof(clp->cl_confirm.data));
8303 out:
8304         trace_nfs4_exchange_id(clp, status);
8305         rpc_put_task(task);
8306         return status;
8307 }
8308
8309 /*
8310  * nfs4_proc_exchange_id()
8311  *
8312  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8313  *
8314  * Since the clientid has expired, all compounds using sessions
8315  * associated with the stale clientid will be returning
8316  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8317  * be in some phase of session reset.
8318  *
8319  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8320  */
8321 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8322 {
8323         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8324         int status;
8325
8326         /* try SP4_MACH_CRED if krb5i/p */
8327         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8328             authflavor == RPC_AUTH_GSS_KRB5P) {
8329                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8330                 if (!status)
8331                         return 0;
8332         }
8333
8334         /* try SP4_NONE */
8335         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8336 }
8337
8338 /**
8339  * nfs4_test_session_trunk
8340  *
8341  * This is an add_xprt_test() test function called from
8342  * rpc_clnt_setup_test_and_add_xprt.
8343  *
8344  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8345  * and is dereferrenced in nfs4_exchange_id_release
8346  *
8347  * Upon success, add the new transport to the rpc_clnt
8348  *
8349  * @clnt: struct rpc_clnt to get new transport
8350  * @xprt: the rpc_xprt to test
8351  * @data: call data for _nfs4_proc_exchange_id.
8352  */
8353 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8354                             void *data)
8355 {
8356         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8357         struct rpc_task *task;
8358         int status;
8359
8360         u32 sp4_how;
8361
8362         dprintk("--> %s try %s\n", __func__,
8363                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8364
8365         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8366
8367         /* Test connection for session trunking. Async exchange_id call */
8368         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8369         if (IS_ERR(task))
8370                 return;
8371
8372         status = task->tk_status;
8373         if (status == 0)
8374                 status = nfs4_detect_session_trunking(adata->clp,
8375                                 task->tk_msg.rpc_resp, xprt);
8376
8377         if (status == 0)
8378                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8379
8380         rpc_put_task(task);
8381 }
8382 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8383
8384 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8385                 const struct cred *cred)
8386 {
8387         struct rpc_message msg = {
8388                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8389                 .rpc_argp = clp,
8390                 .rpc_cred = cred,
8391         };
8392         int status;
8393
8394         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8395                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8396         trace_nfs4_destroy_clientid(clp, status);
8397         if (status)
8398                 dprintk("NFS: Got error %d from the server %s on "
8399                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8400         return status;
8401 }
8402
8403 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8404                 const struct cred *cred)
8405 {
8406         unsigned int loop;
8407         int ret;
8408
8409         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8410                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8411                 switch (ret) {
8412                 case -NFS4ERR_DELAY:
8413                 case -NFS4ERR_CLIENTID_BUSY:
8414                         ssleep(1);
8415                         break;
8416                 default:
8417                         return ret;
8418                 }
8419         }
8420         return 0;
8421 }
8422
8423 int nfs4_destroy_clientid(struct nfs_client *clp)
8424 {
8425         const struct cred *cred;
8426         int ret = 0;
8427
8428         if (clp->cl_mvops->minor_version < 1)
8429                 goto out;
8430         if (clp->cl_exchange_flags == 0)
8431                 goto out;
8432         if (clp->cl_preserve_clid)
8433                 goto out;
8434         cred = nfs4_get_clid_cred(clp);
8435         ret = nfs4_proc_destroy_clientid(clp, cred);
8436         put_cred(cred);
8437         switch (ret) {
8438         case 0:
8439         case -NFS4ERR_STALE_CLIENTID:
8440                 clp->cl_exchange_flags = 0;
8441         }
8442 out:
8443         return ret;
8444 }
8445
8446 #endif /* CONFIG_NFS_V4_1 */
8447
8448 struct nfs4_get_lease_time_data {
8449         struct nfs4_get_lease_time_args *args;
8450         struct nfs4_get_lease_time_res *res;
8451         struct nfs_client *clp;
8452 };
8453
8454 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8455                                         void *calldata)
8456 {
8457         struct nfs4_get_lease_time_data *data =
8458                         (struct nfs4_get_lease_time_data *)calldata;
8459
8460         dprintk("--> %s\n", __func__);
8461         /* just setup sequence, do not trigger session recovery
8462            since we're invoked within one */
8463         nfs4_setup_sequence(data->clp,
8464                         &data->args->la_seq_args,
8465                         &data->res->lr_seq_res,
8466                         task);
8467         dprintk("<-- %s\n", __func__);
8468 }
8469
8470 /*
8471  * Called from nfs4_state_manager thread for session setup, so don't recover
8472  * from sequence operation or clientid errors.
8473  */
8474 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8475 {
8476         struct nfs4_get_lease_time_data *data =
8477                         (struct nfs4_get_lease_time_data *)calldata;
8478
8479         dprintk("--> %s\n", __func__);
8480         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8481                 return;
8482         switch (task->tk_status) {
8483         case -NFS4ERR_DELAY:
8484         case -NFS4ERR_GRACE:
8485                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8486                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8487                 task->tk_status = 0;
8488                 /* fall through */
8489         case -NFS4ERR_RETRY_UNCACHED_REP:
8490                 rpc_restart_call_prepare(task);
8491                 return;
8492         }
8493         dprintk("<-- %s\n", __func__);
8494 }
8495
8496 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8497         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8498         .rpc_call_done = nfs4_get_lease_time_done,
8499 };
8500
8501 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8502 {
8503         struct nfs4_get_lease_time_args args;
8504         struct nfs4_get_lease_time_res res = {
8505                 .lr_fsinfo = fsinfo,
8506         };
8507         struct nfs4_get_lease_time_data data = {
8508                 .args = &args,
8509                 .res = &res,
8510                 .clp = clp,
8511         };
8512         struct rpc_message msg = {
8513                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8514                 .rpc_argp = &args,
8515                 .rpc_resp = &res,
8516         };
8517         struct rpc_task_setup task_setup = {
8518                 .rpc_client = clp->cl_rpcclient,
8519                 .rpc_message = &msg,
8520                 .callback_ops = &nfs4_get_lease_time_ops,
8521                 .callback_data = &data,
8522                 .flags = RPC_TASK_TIMEOUT,
8523         };
8524
8525         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8526         return nfs4_call_sync_custom(&task_setup);
8527 }
8528
8529 #ifdef CONFIG_NFS_V4_1
8530
8531 /*
8532  * Initialize the values to be used by the client in CREATE_SESSION
8533  * If nfs4_init_session set the fore channel request and response sizes,
8534  * use them.
8535  *
8536  * Set the back channel max_resp_sz_cached to zero to force the client to
8537  * always set csa_cachethis to FALSE because the current implementation
8538  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8539  */
8540 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8541                                     struct rpc_clnt *clnt)
8542 {
8543         unsigned int max_rqst_sz, max_resp_sz;
8544         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8545         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8546
8547         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8548         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8549
8550         /* Fore channel attributes */
8551         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8552         args->fc_attrs.max_resp_sz = max_resp_sz;
8553         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8554         args->fc_attrs.max_reqs = max_session_slots;
8555
8556         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8557                 "max_ops=%u max_reqs=%u\n",
8558                 __func__,
8559                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8560                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8561
8562         /* Back channel attributes */
8563         args->bc_attrs.max_rqst_sz = max_bc_payload;
8564         args->bc_attrs.max_resp_sz = max_bc_payload;
8565         args->bc_attrs.max_resp_sz_cached = 0;
8566         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8567         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8568         if (args->bc_attrs.max_reqs > max_bc_slots)
8569                 args->bc_attrs.max_reqs = max_bc_slots;
8570
8571         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8572                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8573                 __func__,
8574                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8575                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8576                 args->bc_attrs.max_reqs);
8577 }
8578
8579 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8580                 struct nfs41_create_session_res *res)
8581 {
8582         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8583         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8584
8585         if (rcvd->max_resp_sz > sent->max_resp_sz)
8586                 return -EINVAL;
8587         /*
8588          * Our requested max_ops is the minimum we need; we're not
8589          * prepared to break up compounds into smaller pieces than that.
8590          * So, no point even trying to continue if the server won't
8591          * cooperate:
8592          */
8593         if (rcvd->max_ops < sent->max_ops)
8594                 return -EINVAL;
8595         if (rcvd->max_reqs == 0)
8596                 return -EINVAL;
8597         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8598                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8599         return 0;
8600 }
8601
8602 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8603                 struct nfs41_create_session_res *res)
8604 {
8605         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8606         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8607
8608         if (!(res->flags & SESSION4_BACK_CHAN))
8609                 goto out;
8610         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8611                 return -EINVAL;
8612         if (rcvd->max_resp_sz < sent->max_resp_sz)
8613                 return -EINVAL;
8614         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8615                 return -EINVAL;
8616         if (rcvd->max_ops > sent->max_ops)
8617                 return -EINVAL;
8618         if (rcvd->max_reqs > sent->max_reqs)
8619                 return -EINVAL;
8620 out:
8621         return 0;
8622 }
8623
8624 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8625                                      struct nfs41_create_session_res *res)
8626 {
8627         int ret;
8628
8629         ret = nfs4_verify_fore_channel_attrs(args, res);
8630         if (ret)
8631                 return ret;
8632         return nfs4_verify_back_channel_attrs(args, res);
8633 }
8634
8635 static void nfs4_update_session(struct nfs4_session *session,
8636                 struct nfs41_create_session_res *res)
8637 {
8638         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8639         /* Mark client id and session as being confirmed */
8640         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8641         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8642         session->flags = res->flags;
8643         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8644         if (res->flags & SESSION4_BACK_CHAN)
8645                 memcpy(&session->bc_attrs, &res->bc_attrs,
8646                                 sizeof(session->bc_attrs));
8647 }
8648
8649 static int _nfs4_proc_create_session(struct nfs_client *clp,
8650                 const struct cred *cred)
8651 {
8652         struct nfs4_session *session = clp->cl_session;
8653         struct nfs41_create_session_args args = {
8654                 .client = clp,
8655                 .clientid = clp->cl_clientid,
8656                 .seqid = clp->cl_seqid,
8657                 .cb_program = NFS4_CALLBACK,
8658         };
8659         struct nfs41_create_session_res res;
8660
8661         struct rpc_message msg = {
8662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8663                 .rpc_argp = &args,
8664                 .rpc_resp = &res,
8665                 .rpc_cred = cred,
8666         };
8667         int status;
8668
8669         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8670         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8671
8672         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8673                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8674         trace_nfs4_create_session(clp, status);
8675
8676         switch (status) {
8677         case -NFS4ERR_STALE_CLIENTID:
8678         case -NFS4ERR_DELAY:
8679         case -ETIMEDOUT:
8680         case -EACCES:
8681         case -EAGAIN:
8682                 goto out;
8683         }
8684
8685         clp->cl_seqid++;
8686         if (!status) {
8687                 /* Verify the session's negotiated channel_attrs values */
8688                 status = nfs4_verify_channel_attrs(&args, &res);
8689                 /* Increment the clientid slot sequence id */
8690                 if (status)
8691                         goto out;
8692                 nfs4_update_session(session, &res);
8693         }
8694 out:
8695         return status;
8696 }
8697
8698 /*
8699  * Issues a CREATE_SESSION operation to the server.
8700  * It is the responsibility of the caller to verify the session is
8701  * expired before calling this routine.
8702  */
8703 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8704 {
8705         int status;
8706         unsigned *ptr;
8707         struct nfs4_session *session = clp->cl_session;
8708
8709         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8710
8711         status = _nfs4_proc_create_session(clp, cred);
8712         if (status)
8713                 goto out;
8714
8715         /* Init or reset the session slot tables */
8716         status = nfs4_setup_session_slot_tables(session);
8717         dprintk("slot table setup returned %d\n", status);
8718         if (status)
8719                 goto out;
8720
8721         ptr = (unsigned *)&session->sess_id.data[0];
8722         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8723                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8724 out:
8725         dprintk("<-- %s\n", __func__);
8726         return status;
8727 }
8728
8729 /*
8730  * Issue the over-the-wire RPC DESTROY_SESSION.
8731  * The caller must serialize access to this routine.
8732  */
8733 int nfs4_proc_destroy_session(struct nfs4_session *session,
8734                 const struct cred *cred)
8735 {
8736         struct rpc_message msg = {
8737                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8738                 .rpc_argp = session,
8739                 .rpc_cred = cred,
8740         };
8741         int status = 0;
8742
8743         dprintk("--> nfs4_proc_destroy_session\n");
8744
8745         /* session is still being setup */
8746         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8747                 return 0;
8748
8749         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8750                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8751         trace_nfs4_destroy_session(session->clp, status);
8752
8753         if (status)
8754                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8755                         "Session has been destroyed regardless...\n", status);
8756
8757         dprintk("<-- nfs4_proc_destroy_session\n");
8758         return status;
8759 }
8760
8761 /*
8762  * Renew the cl_session lease.
8763  */
8764 struct nfs4_sequence_data {
8765         struct nfs_client *clp;
8766         struct nfs4_sequence_args args;
8767         struct nfs4_sequence_res res;
8768 };
8769
8770 static void nfs41_sequence_release(void *data)
8771 {
8772         struct nfs4_sequence_data *calldata = data;
8773         struct nfs_client *clp = calldata->clp;
8774
8775         if (refcount_read(&clp->cl_count) > 1)
8776                 nfs4_schedule_state_renewal(clp);
8777         nfs_put_client(clp);
8778         kfree(calldata);
8779 }
8780
8781 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8782 {
8783         switch(task->tk_status) {
8784         case -NFS4ERR_DELAY:
8785                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8786                 return -EAGAIN;
8787         default:
8788                 nfs4_schedule_lease_recovery(clp);
8789         }
8790         return 0;
8791 }
8792
8793 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8794 {
8795         struct nfs4_sequence_data *calldata = data;
8796         struct nfs_client *clp = calldata->clp;
8797
8798         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8799                 return;
8800
8801         trace_nfs4_sequence(clp, task->tk_status);
8802         if (task->tk_status < 0) {
8803                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8804                 if (refcount_read(&clp->cl_count) == 1)
8805                         goto out;
8806
8807                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8808                         rpc_restart_call_prepare(task);
8809                         return;
8810                 }
8811         }
8812         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8813 out:
8814         dprintk("<-- %s\n", __func__);
8815 }
8816
8817 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8818 {
8819         struct nfs4_sequence_data *calldata = data;
8820         struct nfs_client *clp = calldata->clp;
8821         struct nfs4_sequence_args *args;
8822         struct nfs4_sequence_res *res;
8823
8824         args = task->tk_msg.rpc_argp;
8825         res = task->tk_msg.rpc_resp;
8826
8827         nfs4_setup_sequence(clp, args, res, task);
8828 }
8829
8830 static const struct rpc_call_ops nfs41_sequence_ops = {
8831         .rpc_call_done = nfs41_sequence_call_done,
8832         .rpc_call_prepare = nfs41_sequence_prepare,
8833         .rpc_release = nfs41_sequence_release,
8834 };
8835
8836 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8837                 const struct cred *cred,
8838                 struct nfs4_slot *slot,
8839                 bool is_privileged)
8840 {
8841         struct nfs4_sequence_data *calldata;
8842         struct rpc_message msg = {
8843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8844                 .rpc_cred = cred,
8845         };
8846         struct rpc_task_setup task_setup_data = {
8847                 .rpc_client = clp->cl_rpcclient,
8848                 .rpc_message = &msg,
8849                 .callback_ops = &nfs41_sequence_ops,
8850                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8851         };
8852         struct rpc_task *ret;
8853
8854         ret = ERR_PTR(-EIO);
8855         if (!refcount_inc_not_zero(&clp->cl_count))
8856                 goto out_err;
8857
8858         ret = ERR_PTR(-ENOMEM);
8859         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8860         if (calldata == NULL)
8861                 goto out_put_clp;
8862         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8863         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8864         msg.rpc_argp = &calldata->args;
8865         msg.rpc_resp = &calldata->res;
8866         calldata->clp = clp;
8867         task_setup_data.callback_data = calldata;
8868
8869         ret = rpc_run_task(&task_setup_data);
8870         if (IS_ERR(ret))
8871                 goto out_err;
8872         return ret;
8873 out_put_clp:
8874         nfs_put_client(clp);
8875 out_err:
8876         nfs41_release_slot(slot);
8877         return ret;
8878 }
8879
8880 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8881 {
8882         struct rpc_task *task;
8883         int ret = 0;
8884
8885         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8886                 return -EAGAIN;
8887         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8888         if (IS_ERR(task))
8889                 ret = PTR_ERR(task);
8890         else
8891                 rpc_put_task_async(task);
8892         dprintk("<-- %s status=%d\n", __func__, ret);
8893         return ret;
8894 }
8895
8896 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8897 {
8898         struct rpc_task *task;
8899         int ret;
8900
8901         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8902         if (IS_ERR(task)) {
8903                 ret = PTR_ERR(task);
8904                 goto out;
8905         }
8906         ret = rpc_wait_for_completion_task(task);
8907         if (!ret)
8908                 ret = task->tk_status;
8909         rpc_put_task(task);
8910 out:
8911         dprintk("<-- %s status=%d\n", __func__, ret);
8912         return ret;
8913 }
8914
8915 struct nfs4_reclaim_complete_data {
8916         struct nfs_client *clp;
8917         struct nfs41_reclaim_complete_args arg;
8918         struct nfs41_reclaim_complete_res res;
8919 };
8920
8921 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8922 {
8923         struct nfs4_reclaim_complete_data *calldata = data;
8924
8925         nfs4_setup_sequence(calldata->clp,
8926                         &calldata->arg.seq_args,
8927                         &calldata->res.seq_res,
8928                         task);
8929 }
8930
8931 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8932 {
8933         switch(task->tk_status) {
8934         case 0:
8935                 wake_up_all(&clp->cl_lock_waitq);
8936                 /* Fallthrough */
8937         case -NFS4ERR_COMPLETE_ALREADY:
8938         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8939                 break;
8940         case -NFS4ERR_DELAY:
8941                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8942                 /* fall through */
8943         case -NFS4ERR_RETRY_UNCACHED_REP:
8944                 return -EAGAIN;
8945         case -NFS4ERR_BADSESSION:
8946         case -NFS4ERR_DEADSESSION:
8947         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8948                 break;
8949         default:
8950                 nfs4_schedule_lease_recovery(clp);
8951         }
8952         return 0;
8953 }
8954
8955 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8956 {
8957         struct nfs4_reclaim_complete_data *calldata = data;
8958         struct nfs_client *clp = calldata->clp;
8959         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8960
8961         dprintk("--> %s\n", __func__);
8962         if (!nfs41_sequence_done(task, res))
8963                 return;
8964
8965         trace_nfs4_reclaim_complete(clp, task->tk_status);
8966         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8967                 rpc_restart_call_prepare(task);
8968                 return;
8969         }
8970         dprintk("<-- %s\n", __func__);
8971 }
8972
8973 static void nfs4_free_reclaim_complete_data(void *data)
8974 {
8975         struct nfs4_reclaim_complete_data *calldata = data;
8976
8977         kfree(calldata);
8978 }
8979
8980 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8981         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8982         .rpc_call_done = nfs4_reclaim_complete_done,
8983         .rpc_release = nfs4_free_reclaim_complete_data,
8984 };
8985
8986 /*
8987  * Issue a global reclaim complete.
8988  */
8989 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8990                 const struct cred *cred)
8991 {
8992         struct nfs4_reclaim_complete_data *calldata;
8993         struct rpc_message msg = {
8994                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8995                 .rpc_cred = cred,
8996         };
8997         struct rpc_task_setup task_setup_data = {
8998                 .rpc_client = clp->cl_rpcclient,
8999                 .rpc_message = &msg,
9000                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9001                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9002         };
9003         int status = -ENOMEM;
9004
9005         dprintk("--> %s\n", __func__);
9006         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9007         if (calldata == NULL)
9008                 goto out;
9009         calldata->clp = clp;
9010         calldata->arg.one_fs = 0;
9011
9012         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9013         msg.rpc_argp = &calldata->arg;
9014         msg.rpc_resp = &calldata->res;
9015         task_setup_data.callback_data = calldata;
9016         status = nfs4_call_sync_custom(&task_setup_data);
9017 out:
9018         dprintk("<-- %s status=%d\n", __func__, status);
9019         return status;
9020 }
9021
9022 static void
9023 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9024 {
9025         struct nfs4_layoutget *lgp = calldata;
9026         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9027
9028         dprintk("--> %s\n", __func__);
9029         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9030                                 &lgp->res.seq_res, task);
9031         dprintk("<-- %s\n", __func__);
9032 }
9033
9034 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9035 {
9036         struct nfs4_layoutget *lgp = calldata;
9037
9038         dprintk("--> %s\n", __func__);
9039         nfs41_sequence_process(task, &lgp->res.seq_res);
9040         dprintk("<-- %s\n", __func__);
9041 }
9042
9043 static int
9044 nfs4_layoutget_handle_exception(struct rpc_task *task,
9045                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9046 {
9047         struct inode *inode = lgp->args.inode;
9048         struct nfs_server *server = NFS_SERVER(inode);
9049         struct pnfs_layout_hdr *lo;
9050         int nfs4err = task->tk_status;
9051         int err, status = 0;
9052         LIST_HEAD(head);
9053
9054         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9055
9056         nfs4_sequence_free_slot(&lgp->res.seq_res);
9057
9058         switch (nfs4err) {
9059         case 0:
9060                 goto out;
9061
9062         /*
9063          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9064          * on the file. set tk_status to -ENODATA to tell upper layer to
9065          * retry go inband.
9066          */
9067         case -NFS4ERR_LAYOUTUNAVAILABLE:
9068                 status = -ENODATA;
9069                 goto out;
9070         /*
9071          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9072          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9073          */
9074         case -NFS4ERR_BADLAYOUT:
9075                 status = -EOVERFLOW;
9076                 goto out;
9077         /*
9078          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9079          * (or clients) writing to the same RAID stripe except when
9080          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9081          *
9082          * Treat it like we would RECALLCONFLICT -- we retry for a little
9083          * while, and then eventually give up.
9084          */
9085         case -NFS4ERR_LAYOUTTRYLATER:
9086                 if (lgp->args.minlength == 0) {
9087                         status = -EOVERFLOW;
9088                         goto out;
9089                 }
9090                 status = -EBUSY;
9091                 break;
9092         case -NFS4ERR_RECALLCONFLICT:
9093                 status = -ERECALLCONFLICT;
9094                 break;
9095         case -NFS4ERR_DELEG_REVOKED:
9096         case -NFS4ERR_ADMIN_REVOKED:
9097         case -NFS4ERR_EXPIRED:
9098         case -NFS4ERR_BAD_STATEID:
9099                 exception->timeout = 0;
9100                 spin_lock(&inode->i_lock);
9101                 lo = NFS_I(inode)->layout;
9102                 /* If the open stateid was bad, then recover it. */
9103                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9104                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9105                         spin_unlock(&inode->i_lock);
9106                         exception->state = lgp->args.ctx->state;
9107                         exception->stateid = &lgp->args.stateid;
9108                         break;
9109                 }
9110
9111                 /*
9112                  * Mark the bad layout state as invalid, then retry
9113                  */
9114                 pnfs_mark_layout_stateid_invalid(lo, &head);
9115                 spin_unlock(&inode->i_lock);
9116                 nfs_commit_inode(inode, 0);
9117                 pnfs_free_lseg_list(&head);
9118                 status = -EAGAIN;
9119                 goto out;
9120         }
9121
9122         err = nfs4_handle_exception(server, nfs4err, exception);
9123         if (!status) {
9124                 if (exception->retry)
9125                         status = -EAGAIN;
9126                 else
9127                         status = err;
9128         }
9129 out:
9130         dprintk("<-- %s\n", __func__);
9131         return status;
9132 }
9133
9134 size_t max_response_pages(struct nfs_server *server)
9135 {
9136         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9137         return nfs_page_array_len(0, max_resp_sz);
9138 }
9139
9140 static void nfs4_layoutget_release(void *calldata)
9141 {
9142         struct nfs4_layoutget *lgp = calldata;
9143
9144         dprintk("--> %s\n", __func__);
9145         nfs4_sequence_free_slot(&lgp->res.seq_res);
9146         pnfs_layoutget_free(lgp);
9147         dprintk("<-- %s\n", __func__);
9148 }
9149
9150 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9151         .rpc_call_prepare = nfs4_layoutget_prepare,
9152         .rpc_call_done = nfs4_layoutget_done,
9153         .rpc_release = nfs4_layoutget_release,
9154 };
9155
9156 struct pnfs_layout_segment *
9157 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9158 {
9159         struct inode *inode = lgp->args.inode;
9160         struct nfs_server *server = NFS_SERVER(inode);
9161         struct rpc_task *task;
9162         struct rpc_message msg = {
9163                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9164                 .rpc_argp = &lgp->args,
9165                 .rpc_resp = &lgp->res,
9166                 .rpc_cred = lgp->cred,
9167         };
9168         struct rpc_task_setup task_setup_data = {
9169                 .rpc_client = server->client,
9170                 .rpc_message = &msg,
9171                 .callback_ops = &nfs4_layoutget_call_ops,
9172                 .callback_data = lgp,
9173                 .flags = RPC_TASK_ASYNC,
9174         };
9175         struct pnfs_layout_segment *lseg = NULL;
9176         struct nfs4_exception exception = {
9177                 .inode = inode,
9178                 .timeout = *timeout,
9179         };
9180         int status = 0;
9181
9182         dprintk("--> %s\n", __func__);
9183
9184         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9185         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9186
9187         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9188
9189         task = rpc_run_task(&task_setup_data);
9190         if (IS_ERR(task))
9191                 return ERR_CAST(task);
9192         status = rpc_wait_for_completion_task(task);
9193         if (status != 0)
9194                 goto out;
9195
9196         if (task->tk_status < 0) {
9197                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9198                 *timeout = exception.timeout;
9199         } else if (lgp->res.layoutp->len == 0) {
9200                 status = -EAGAIN;
9201                 *timeout = nfs4_update_delay(&exception.timeout);
9202         } else
9203                 lseg = pnfs_layout_process(lgp);
9204 out:
9205         trace_nfs4_layoutget(lgp->args.ctx,
9206                         &lgp->args.range,
9207                         &lgp->res.range,
9208                         &lgp->res.stateid,
9209                         status);
9210
9211         rpc_put_task(task);
9212         dprintk("<-- %s status=%d\n", __func__, status);
9213         if (status)
9214                 return ERR_PTR(status);
9215         return lseg;
9216 }
9217
9218 static void
9219 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9220 {
9221         struct nfs4_layoutreturn *lrp = calldata;
9222
9223         dprintk("--> %s\n", __func__);
9224         nfs4_setup_sequence(lrp->clp,
9225                         &lrp->args.seq_args,
9226                         &lrp->res.seq_res,
9227                         task);
9228         if (!pnfs_layout_is_valid(lrp->args.layout))
9229                 rpc_exit(task, 0);
9230 }
9231
9232 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9233 {
9234         struct nfs4_layoutreturn *lrp = calldata;
9235         struct nfs_server *server;
9236
9237         dprintk("--> %s\n", __func__);
9238
9239         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9240                 return;
9241
9242         /*
9243          * Was there an RPC level error? Assume the call succeeded,
9244          * and that we need to release the layout
9245          */
9246         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9247                 lrp->res.lrs_present = 0;
9248                 return;
9249         }
9250
9251         server = NFS_SERVER(lrp->args.inode);
9252         switch (task->tk_status) {
9253         case -NFS4ERR_OLD_STATEID:
9254                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9255                                         &lrp->args.range,
9256                                         lrp->args.inode))
9257                         goto out_restart;
9258                 /* Fallthrough */
9259         default:
9260                 task->tk_status = 0;
9261                 /* Fallthrough */
9262         case 0:
9263                 break;
9264         case -NFS4ERR_DELAY:
9265                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9266                         break;
9267                 goto out_restart;
9268         }
9269         dprintk("<-- %s\n", __func__);
9270         return;
9271 out_restart:
9272         task->tk_status = 0;
9273         nfs4_sequence_free_slot(&lrp->res.seq_res);
9274         rpc_restart_call_prepare(task);
9275 }
9276
9277 static void nfs4_layoutreturn_release(void *calldata)
9278 {
9279         struct nfs4_layoutreturn *lrp = calldata;
9280         struct pnfs_layout_hdr *lo = lrp->args.layout;
9281
9282         dprintk("--> %s\n", __func__);
9283         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9284                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9285         nfs4_sequence_free_slot(&lrp->res.seq_res);
9286         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9287                 lrp->ld_private.ops->free(&lrp->ld_private);
9288         pnfs_put_layout_hdr(lrp->args.layout);
9289         nfs_iput_and_deactive(lrp->inode);
9290         kfree(calldata);
9291         dprintk("<-- %s\n", __func__);
9292 }
9293
9294 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9295         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9296         .rpc_call_done = nfs4_layoutreturn_done,
9297         .rpc_release = nfs4_layoutreturn_release,
9298 };
9299
9300 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9301 {
9302         struct rpc_task *task;
9303         struct rpc_message msg = {
9304                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9305                 .rpc_argp = &lrp->args,
9306                 .rpc_resp = &lrp->res,
9307                 .rpc_cred = lrp->cred,
9308         };
9309         struct rpc_task_setup task_setup_data = {
9310                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9311                 .rpc_message = &msg,
9312                 .callback_ops = &nfs4_layoutreturn_call_ops,
9313                 .callback_data = lrp,
9314         };
9315         int status = 0;
9316
9317         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9318                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9319                         &task_setup_data.rpc_client, &msg);
9320
9321         dprintk("--> %s\n", __func__);
9322         if (!sync) {
9323                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9324                 if (!lrp->inode) {
9325                         nfs4_layoutreturn_release(lrp);
9326                         return -EAGAIN;
9327                 }
9328                 task_setup_data.flags |= RPC_TASK_ASYNC;
9329         }
9330         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9331         task = rpc_run_task(&task_setup_data);
9332         if (IS_ERR(task))
9333                 return PTR_ERR(task);
9334         if (sync)
9335                 status = task->tk_status;
9336         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9337         dprintk("<-- %s status=%d\n", __func__, status);
9338         rpc_put_task(task);
9339         return status;
9340 }
9341
9342 static int
9343 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9344                 struct pnfs_device *pdev,
9345                 const struct cred *cred)
9346 {
9347         struct nfs4_getdeviceinfo_args args = {
9348                 .pdev = pdev,
9349                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9350                         NOTIFY_DEVICEID4_DELETE,
9351         };
9352         struct nfs4_getdeviceinfo_res res = {
9353                 .pdev = pdev,
9354         };
9355         struct rpc_message msg = {
9356                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9357                 .rpc_argp = &args,
9358                 .rpc_resp = &res,
9359                 .rpc_cred = cred,
9360         };
9361         int status;
9362
9363         dprintk("--> %s\n", __func__);
9364         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9365         if (res.notification & ~args.notify_types)
9366                 dprintk("%s: unsupported notification\n", __func__);
9367         if (res.notification != args.notify_types)
9368                 pdev->nocache = 1;
9369
9370         dprintk("<-- %s status=%d\n", __func__, status);
9371
9372         return status;
9373 }
9374
9375 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9376                 struct pnfs_device *pdev,
9377                 const struct cred *cred)
9378 {
9379         struct nfs4_exception exception = { };
9380         int err;
9381
9382         do {
9383                 err = nfs4_handle_exception(server,
9384                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9385                                         &exception);
9386         } while (exception.retry);
9387         return err;
9388 }
9389 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9390
9391 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9392 {
9393         struct nfs4_layoutcommit_data *data = calldata;
9394         struct nfs_server *server = NFS_SERVER(data->args.inode);
9395
9396         nfs4_setup_sequence(server->nfs_client,
9397                         &data->args.seq_args,
9398                         &data->res.seq_res,
9399                         task);
9400 }
9401
9402 static void
9403 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9404 {
9405         struct nfs4_layoutcommit_data *data = calldata;
9406         struct nfs_server *server = NFS_SERVER(data->args.inode);
9407
9408         if (!nfs41_sequence_done(task, &data->res.seq_res))
9409                 return;
9410
9411         switch (task->tk_status) { /* Just ignore these failures */
9412         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9413         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9414         case -NFS4ERR_BADLAYOUT:     /* no layout */
9415         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9416                 task->tk_status = 0;
9417         case 0:
9418                 break;
9419         default:
9420                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9421                         rpc_restart_call_prepare(task);
9422                         return;
9423                 }
9424         }
9425 }
9426
9427 static void nfs4_layoutcommit_release(void *calldata)
9428 {
9429         struct nfs4_layoutcommit_data *data = calldata;
9430
9431         pnfs_cleanup_layoutcommit(data);
9432         nfs_post_op_update_inode_force_wcc(data->args.inode,
9433                                            data->res.fattr);
9434         put_cred(data->cred);
9435         nfs_iput_and_deactive(data->inode);
9436         kfree(data);
9437 }
9438
9439 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9440         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9441         .rpc_call_done = nfs4_layoutcommit_done,
9442         .rpc_release = nfs4_layoutcommit_release,
9443 };
9444
9445 int
9446 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9447 {
9448         struct rpc_message msg = {
9449                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9450                 .rpc_argp = &data->args,
9451                 .rpc_resp = &data->res,
9452                 .rpc_cred = data->cred,
9453         };
9454         struct rpc_task_setup task_setup_data = {
9455                 .task = &data->task,
9456                 .rpc_client = NFS_CLIENT(data->args.inode),
9457                 .rpc_message = &msg,
9458                 .callback_ops = &nfs4_layoutcommit_ops,
9459                 .callback_data = data,
9460         };
9461         struct rpc_task *task;
9462         int status = 0;
9463
9464         dprintk("NFS: initiating layoutcommit call. sync %d "
9465                 "lbw: %llu inode %lu\n", sync,
9466                 data->args.lastbytewritten,
9467                 data->args.inode->i_ino);
9468
9469         if (!sync) {
9470                 data->inode = nfs_igrab_and_active(data->args.inode);
9471                 if (data->inode == NULL) {
9472                         nfs4_layoutcommit_release(data);
9473                         return -EAGAIN;
9474                 }
9475                 task_setup_data.flags = RPC_TASK_ASYNC;
9476         }
9477         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9478         task = rpc_run_task(&task_setup_data);
9479         if (IS_ERR(task))
9480                 return PTR_ERR(task);
9481         if (sync)
9482                 status = task->tk_status;
9483         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9484         dprintk("%s: status %d\n", __func__, status);
9485         rpc_put_task(task);
9486         return status;
9487 }
9488
9489 /*
9490  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9491  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9492  */
9493 static int
9494 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9495                     struct nfs_fsinfo *info,
9496                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9497 {
9498         struct nfs41_secinfo_no_name_args args = {
9499                 .style = SECINFO_STYLE_CURRENT_FH,
9500         };
9501         struct nfs4_secinfo_res res = {
9502                 .flavors = flavors,
9503         };
9504         struct rpc_message msg = {
9505                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9506                 .rpc_argp = &args,
9507                 .rpc_resp = &res,
9508         };
9509         struct rpc_clnt *clnt = server->client;
9510         struct nfs4_call_sync_data data = {
9511                 .seq_server = server,
9512                 .seq_args = &args.seq_args,
9513                 .seq_res = &res.seq_res,
9514         };
9515         struct rpc_task_setup task_setup = {
9516                 .rpc_client = server->client,
9517                 .rpc_message = &msg,
9518                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9519                 .callback_data = &data,
9520                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9521         };
9522         const struct cred *cred = NULL;
9523         int status;
9524
9525         if (use_integrity) {
9526                 clnt = server->nfs_client->cl_rpcclient;
9527                 task_setup.rpc_client = clnt;
9528
9529                 cred = nfs4_get_clid_cred(server->nfs_client);
9530                 msg.rpc_cred = cred;
9531         }
9532
9533         dprintk("--> %s\n", __func__);
9534         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9535         status = nfs4_call_sync_custom(&task_setup);
9536         dprintk("<-- %s status=%d\n", __func__, status);
9537
9538         put_cred(cred);
9539
9540         return status;
9541 }
9542
9543 static int
9544 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9545                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9546 {
9547         struct nfs4_exception exception = {
9548                 .interruptible = true,
9549         };
9550         int err;
9551         do {
9552                 /* first try using integrity protection */
9553                 err = -NFS4ERR_WRONGSEC;
9554
9555                 /* try to use integrity protection with machine cred */
9556                 if (_nfs4_is_integrity_protected(server->nfs_client))
9557                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9558                                                           flavors, true);
9559
9560                 /*
9561                  * if unable to use integrity protection, or SECINFO with
9562                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9563                  * disallowed by spec, but exists in deployed servers) use
9564                  * the current filesystem's rpc_client and the user cred.
9565                  */
9566                 if (err == -NFS4ERR_WRONGSEC)
9567                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9568                                                           flavors, false);
9569
9570                 switch (err) {
9571                 case 0:
9572                 case -NFS4ERR_WRONGSEC:
9573                 case -ENOTSUPP:
9574                         goto out;
9575                 default:
9576                         err = nfs4_handle_exception(server, err, &exception);
9577                 }
9578         } while (exception.retry);
9579 out:
9580         return err;
9581 }
9582
9583 static int
9584 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9585                     struct nfs_fsinfo *info)
9586 {
9587         int err;
9588         struct page *page;
9589         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9590         struct nfs4_secinfo_flavors *flavors;
9591         struct nfs4_secinfo4 *secinfo;
9592         int i;
9593
9594         page = alloc_page(GFP_KERNEL);
9595         if (!page) {
9596                 err = -ENOMEM;
9597                 goto out;
9598         }
9599
9600         flavors = page_address(page);
9601         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9602
9603         /*
9604          * Fall back on "guess and check" method if
9605          * the server doesn't support SECINFO_NO_NAME
9606          */
9607         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9608                 err = nfs4_find_root_sec(server, fhandle, info);
9609                 goto out_freepage;
9610         }
9611         if (err)
9612                 goto out_freepage;
9613
9614         for (i = 0; i < flavors->num_flavors; i++) {
9615                 secinfo = &flavors->flavors[i];
9616
9617                 switch (secinfo->flavor) {
9618                 case RPC_AUTH_NULL:
9619                 case RPC_AUTH_UNIX:
9620                 case RPC_AUTH_GSS:
9621                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9622                                         &secinfo->flavor_info);
9623                         break;
9624                 default:
9625                         flavor = RPC_AUTH_MAXFLAVOR;
9626                         break;
9627                 }
9628
9629                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9630                         flavor = RPC_AUTH_MAXFLAVOR;
9631
9632                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9633                         err = nfs4_lookup_root_sec(server, fhandle,
9634                                                    info, flavor);
9635                         if (!err)
9636                                 break;
9637                 }
9638         }
9639
9640         if (flavor == RPC_AUTH_MAXFLAVOR)
9641                 err = -EPERM;
9642
9643 out_freepage:
9644         put_page(page);
9645         if (err == -EACCES)
9646                 return -EPERM;
9647 out:
9648         return err;
9649 }
9650
9651 static int _nfs41_test_stateid(struct nfs_server *server,
9652                 nfs4_stateid *stateid,
9653                 const struct cred *cred)
9654 {
9655         int status;
9656         struct nfs41_test_stateid_args args = {
9657                 .stateid = stateid,
9658         };
9659         struct nfs41_test_stateid_res res;
9660         struct rpc_message msg = {
9661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9662                 .rpc_argp = &args,
9663                 .rpc_resp = &res,
9664                 .rpc_cred = cred,
9665         };
9666         struct rpc_clnt *rpc_client = server->client;
9667
9668         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9669                 &rpc_client, &msg);
9670
9671         dprintk("NFS call  test_stateid %p\n", stateid);
9672         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9673         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9674                         &args.seq_args, &res.seq_res);
9675         if (status != NFS_OK) {
9676                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9677                 return status;
9678         }
9679         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9680         return -res.status;
9681 }
9682
9683 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9684                 int err, struct nfs4_exception *exception)
9685 {
9686         exception->retry = 0;
9687         switch(err) {
9688         case -NFS4ERR_DELAY:
9689         case -NFS4ERR_RETRY_UNCACHED_REP:
9690                 nfs4_handle_exception(server, err, exception);
9691                 break;
9692         case -NFS4ERR_BADSESSION:
9693         case -NFS4ERR_BADSLOT:
9694         case -NFS4ERR_BAD_HIGH_SLOT:
9695         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9696         case -NFS4ERR_DEADSESSION:
9697                 nfs4_do_handle_exception(server, err, exception);
9698         }
9699 }
9700
9701 /**
9702  * nfs41_test_stateid - perform a TEST_STATEID operation
9703  *
9704  * @server: server / transport on which to perform the operation
9705  * @stateid: state ID to test
9706  * @cred: credential
9707  *
9708  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9709  * Otherwise a negative NFS4ERR value is returned if the operation
9710  * failed or the state ID is not currently valid.
9711  */
9712 static int nfs41_test_stateid(struct nfs_server *server,
9713                 nfs4_stateid *stateid,
9714                 const struct cred *cred)
9715 {
9716         struct nfs4_exception exception = {
9717                 .interruptible = true,
9718         };
9719         int err;
9720         do {
9721                 err = _nfs41_test_stateid(server, stateid, cred);
9722                 nfs4_handle_delay_or_session_error(server, err, &exception);
9723         } while (exception.retry);
9724         return err;
9725 }
9726
9727 struct nfs_free_stateid_data {
9728         struct nfs_server *server;
9729         struct nfs41_free_stateid_args args;
9730         struct nfs41_free_stateid_res res;
9731 };
9732
9733 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9734 {
9735         struct nfs_free_stateid_data *data = calldata;
9736         nfs4_setup_sequence(data->server->nfs_client,
9737                         &data->args.seq_args,
9738                         &data->res.seq_res,
9739                         task);
9740 }
9741
9742 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9743 {
9744         struct nfs_free_stateid_data *data = calldata;
9745
9746         nfs41_sequence_done(task, &data->res.seq_res);
9747
9748         switch (task->tk_status) {
9749         case -NFS4ERR_DELAY:
9750                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9751                         rpc_restart_call_prepare(task);
9752         }
9753 }
9754
9755 static void nfs41_free_stateid_release(void *calldata)
9756 {
9757         kfree(calldata);
9758 }
9759
9760 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9761         .rpc_call_prepare = nfs41_free_stateid_prepare,
9762         .rpc_call_done = nfs41_free_stateid_done,
9763         .rpc_release = nfs41_free_stateid_release,
9764 };
9765
9766 /**
9767  * nfs41_free_stateid - perform a FREE_STATEID operation
9768  *
9769  * @server: server / transport on which to perform the operation
9770  * @stateid: state ID to release
9771  * @cred: credential
9772  * @privileged: set to true if this call needs to be privileged
9773  *
9774  * Note: this function is always asynchronous.
9775  */
9776 static int nfs41_free_stateid(struct nfs_server *server,
9777                 const nfs4_stateid *stateid,
9778                 const struct cred *cred,
9779                 bool privileged)
9780 {
9781         struct rpc_message msg = {
9782                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9783                 .rpc_cred = cred,
9784         };
9785         struct rpc_task_setup task_setup = {
9786                 .rpc_client = server->client,
9787                 .rpc_message = &msg,
9788                 .callback_ops = &nfs41_free_stateid_ops,
9789                 .flags = RPC_TASK_ASYNC,
9790         };
9791         struct nfs_free_stateid_data *data;
9792         struct rpc_task *task;
9793
9794         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9795                 &task_setup.rpc_client, &msg);
9796
9797         dprintk("NFS call  free_stateid %p\n", stateid);
9798         data = kmalloc(sizeof(*data), GFP_NOFS);
9799         if (!data)
9800                 return -ENOMEM;
9801         data->server = server;
9802         nfs4_stateid_copy(&data->args.stateid, stateid);
9803
9804         task_setup.callback_data = data;
9805
9806         msg.rpc_argp = &data->args;
9807         msg.rpc_resp = &data->res;
9808         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9809         task = rpc_run_task(&task_setup);
9810         if (IS_ERR(task))
9811                 return PTR_ERR(task);
9812         rpc_put_task(task);
9813         return 0;
9814 }
9815
9816 static void
9817 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9818 {
9819         const struct cred *cred = lsp->ls_state->owner->so_cred;
9820
9821         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9822         nfs4_free_lock_state(server, lsp);
9823 }
9824
9825 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9826                 const nfs4_stateid *s2)
9827 {
9828         if (s1->type != s2->type)
9829                 return false;
9830
9831         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9832                 return false;
9833
9834         if (s1->seqid == s2->seqid)
9835                 return true;
9836
9837         return s1->seqid == 0 || s2->seqid == 0;
9838 }
9839
9840 #endif /* CONFIG_NFS_V4_1 */
9841
9842 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9843                 const nfs4_stateid *s2)
9844 {
9845         return nfs4_stateid_match(s1, s2);
9846 }
9847
9848
9849 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9850         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9851         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9852         .recover_open   = nfs4_open_reclaim,
9853         .recover_lock   = nfs4_lock_reclaim,
9854         .establish_clid = nfs4_init_clientid,
9855         .detect_trunking = nfs40_discover_server_trunking,
9856 };
9857
9858 #if defined(CONFIG_NFS_V4_1)
9859 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9860         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9861         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9862         .recover_open   = nfs4_open_reclaim,
9863         .recover_lock   = nfs4_lock_reclaim,
9864         .establish_clid = nfs41_init_clientid,
9865         .reclaim_complete = nfs41_proc_reclaim_complete,
9866         .detect_trunking = nfs41_discover_server_trunking,
9867 };
9868 #endif /* CONFIG_NFS_V4_1 */
9869
9870 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9871         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9872         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9873         .recover_open   = nfs40_open_expired,
9874         .recover_lock   = nfs4_lock_expired,
9875         .establish_clid = nfs4_init_clientid,
9876 };
9877
9878 #if defined(CONFIG_NFS_V4_1)
9879 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9880         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9881         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9882         .recover_open   = nfs41_open_expired,
9883         .recover_lock   = nfs41_lock_expired,
9884         .establish_clid = nfs41_init_clientid,
9885 };
9886 #endif /* CONFIG_NFS_V4_1 */
9887
9888 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9889         .sched_state_renewal = nfs4_proc_async_renew,
9890         .get_state_renewal_cred = nfs4_get_renew_cred,
9891         .renew_lease = nfs4_proc_renew,
9892 };
9893
9894 #if defined(CONFIG_NFS_V4_1)
9895 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9896         .sched_state_renewal = nfs41_proc_async_sequence,
9897         .get_state_renewal_cred = nfs4_get_machine_cred,
9898         .renew_lease = nfs4_proc_sequence,
9899 };
9900 #endif
9901
9902 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9903         .get_locations = _nfs40_proc_get_locations,
9904         .fsid_present = _nfs40_proc_fsid_present,
9905 };
9906
9907 #if defined(CONFIG_NFS_V4_1)
9908 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9909         .get_locations = _nfs41_proc_get_locations,
9910         .fsid_present = _nfs41_proc_fsid_present,
9911 };
9912 #endif  /* CONFIG_NFS_V4_1 */
9913
9914 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9915         .minor_version = 0,
9916         .init_caps = NFS_CAP_READDIRPLUS
9917                 | NFS_CAP_ATOMIC_OPEN
9918                 | NFS_CAP_POSIX_LOCK,
9919         .init_client = nfs40_init_client,
9920         .shutdown_client = nfs40_shutdown_client,
9921         .match_stateid = nfs4_match_stateid,
9922         .find_root_sec = nfs4_find_root_sec,
9923         .free_lock_state = nfs4_release_lockowner,
9924         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9925         .alloc_seqid = nfs_alloc_seqid,
9926         .call_sync_ops = &nfs40_call_sync_ops,
9927         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9928         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9929         .state_renewal_ops = &nfs40_state_renewal_ops,
9930         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9931 };
9932
9933 #if defined(CONFIG_NFS_V4_1)
9934 static struct nfs_seqid *
9935 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9936 {
9937         return NULL;
9938 }
9939
9940 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9941         .minor_version = 1,
9942         .init_caps = NFS_CAP_READDIRPLUS
9943                 | NFS_CAP_ATOMIC_OPEN
9944                 | NFS_CAP_POSIX_LOCK
9945                 | NFS_CAP_STATEID_NFSV41
9946                 | NFS_CAP_ATOMIC_OPEN_V1
9947                 | NFS_CAP_LGOPEN,
9948         .init_client = nfs41_init_client,
9949         .shutdown_client = nfs41_shutdown_client,
9950         .match_stateid = nfs41_match_stateid,
9951         .find_root_sec = nfs41_find_root_sec,
9952         .free_lock_state = nfs41_free_lock_state,
9953         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9954         .alloc_seqid = nfs_alloc_no_seqid,
9955         .session_trunk = nfs4_test_session_trunk,
9956         .call_sync_ops = &nfs41_call_sync_ops,
9957         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9958         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9959         .state_renewal_ops = &nfs41_state_renewal_ops,
9960         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9961 };
9962 #endif
9963
9964 #if defined(CONFIG_NFS_V4_2)
9965 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9966         .minor_version = 2,
9967         .init_caps = NFS_CAP_READDIRPLUS
9968                 | NFS_CAP_ATOMIC_OPEN
9969                 | NFS_CAP_POSIX_LOCK
9970                 | NFS_CAP_STATEID_NFSV41
9971                 | NFS_CAP_ATOMIC_OPEN_V1
9972                 | NFS_CAP_LGOPEN
9973                 | NFS_CAP_ALLOCATE
9974                 | NFS_CAP_COPY
9975                 | NFS_CAP_OFFLOAD_CANCEL
9976                 | NFS_CAP_COPY_NOTIFY
9977                 | NFS_CAP_DEALLOCATE
9978                 | NFS_CAP_SEEK
9979                 | NFS_CAP_LAYOUTSTATS
9980                 | NFS_CAP_CLONE
9981                 | NFS_CAP_LAYOUTERROR,
9982         .init_client = nfs41_init_client,
9983         .shutdown_client = nfs41_shutdown_client,
9984         .match_stateid = nfs41_match_stateid,
9985         .find_root_sec = nfs41_find_root_sec,
9986         .free_lock_state = nfs41_free_lock_state,
9987         .call_sync_ops = &nfs41_call_sync_ops,
9988         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9989         .alloc_seqid = nfs_alloc_no_seqid,
9990         .session_trunk = nfs4_test_session_trunk,
9991         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9992         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9993         .state_renewal_ops = &nfs41_state_renewal_ops,
9994         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9995 };
9996 #endif
9997
9998 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9999         [0] = &nfs_v4_0_minor_ops,
10000 #if defined(CONFIG_NFS_V4_1)
10001         [1] = &nfs_v4_1_minor_ops,
10002 #endif
10003 #if defined(CONFIG_NFS_V4_2)
10004         [2] = &nfs_v4_2_minor_ops,
10005 #endif
10006 };
10007
10008 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10009 {
10010         ssize_t error, error2;
10011
10012         error = generic_listxattr(dentry, list, size);
10013         if (error < 0)
10014                 return error;
10015         if (list) {
10016                 list += error;
10017                 size -= error;
10018         }
10019
10020         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10021         if (error2 < 0)
10022                 return error2;
10023         return error + error2;
10024 }
10025
10026 static const struct inode_operations nfs4_dir_inode_operations = {
10027         .create         = nfs_create,
10028         .lookup         = nfs_lookup,
10029         .atomic_open    = nfs_atomic_open,
10030         .link           = nfs_link,
10031         .unlink         = nfs_unlink,
10032         .symlink        = nfs_symlink,
10033         .mkdir          = nfs_mkdir,
10034         .rmdir          = nfs_rmdir,
10035         .mknod          = nfs_mknod,
10036         .rename         = nfs_rename,
10037         .permission     = nfs_permission,
10038         .getattr        = nfs_getattr,
10039         .setattr        = nfs_setattr,
10040         .listxattr      = nfs4_listxattr,
10041 };
10042
10043 static const struct inode_operations nfs4_file_inode_operations = {
10044         .permission     = nfs_permission,
10045         .getattr        = nfs_getattr,
10046         .setattr        = nfs_setattr,
10047         .listxattr      = nfs4_listxattr,
10048 };
10049
10050 const struct nfs_rpc_ops nfs_v4_clientops = {
10051         .version        = 4,                    /* protocol version */
10052         .dentry_ops     = &nfs4_dentry_operations,
10053         .dir_inode_ops  = &nfs4_dir_inode_operations,
10054         .file_inode_ops = &nfs4_file_inode_operations,
10055         .file_ops       = &nfs4_file_operations,
10056         .getroot        = nfs4_proc_get_root,
10057         .submount       = nfs4_submount,
10058         .try_get_tree   = nfs4_try_get_tree,
10059         .getattr        = nfs4_proc_getattr,
10060         .setattr        = nfs4_proc_setattr,
10061         .lookup         = nfs4_proc_lookup,
10062         .lookupp        = nfs4_proc_lookupp,
10063         .access         = nfs4_proc_access,
10064         .readlink       = nfs4_proc_readlink,
10065         .create         = nfs4_proc_create,
10066         .remove         = nfs4_proc_remove,
10067         .unlink_setup   = nfs4_proc_unlink_setup,
10068         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10069         .unlink_done    = nfs4_proc_unlink_done,
10070         .rename_setup   = nfs4_proc_rename_setup,
10071         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10072         .rename_done    = nfs4_proc_rename_done,
10073         .link           = nfs4_proc_link,
10074         .symlink        = nfs4_proc_symlink,
10075         .mkdir          = nfs4_proc_mkdir,
10076         .rmdir          = nfs4_proc_rmdir,
10077         .readdir        = nfs4_proc_readdir,
10078         .mknod          = nfs4_proc_mknod,
10079         .statfs         = nfs4_proc_statfs,
10080         .fsinfo         = nfs4_proc_fsinfo,
10081         .pathconf       = nfs4_proc_pathconf,
10082         .set_capabilities = nfs4_server_capabilities,
10083         .decode_dirent  = nfs4_decode_dirent,
10084         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10085         .read_setup     = nfs4_proc_read_setup,
10086         .read_done      = nfs4_read_done,
10087         .write_setup    = nfs4_proc_write_setup,
10088         .write_done     = nfs4_write_done,
10089         .commit_setup   = nfs4_proc_commit_setup,
10090         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10091         .commit_done    = nfs4_commit_done,
10092         .lock           = nfs4_proc_lock,
10093         .clear_acl_cache = nfs4_zap_acl_attr,
10094         .close_context  = nfs4_close_context,
10095         .open_context   = nfs4_atomic_open,
10096         .have_delegation = nfs4_have_delegation,
10097         .alloc_client   = nfs4_alloc_client,
10098         .init_client    = nfs4_init_client,
10099         .free_client    = nfs4_free_client,
10100         .create_server  = nfs4_create_server,
10101         .clone_server   = nfs_clone_server,
10102 };
10103
10104 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10105         .name   = XATTR_NAME_NFSV4_ACL,
10106         .list   = nfs4_xattr_list_nfs4_acl,
10107         .get    = nfs4_xattr_get_nfs4_acl,
10108         .set    = nfs4_xattr_set_nfs4_acl,
10109 };
10110
10111 const struct xattr_handler *nfs4_xattr_handlers[] = {
10112         &nfs4_xattr_nfs4_acl_handler,
10113 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10114         &nfs4_xattr_nfs4_label_handler,
10115 #endif
10116         NULL
10117 };
10118
10119 /*
10120  * Local variables:
10121  *  c-basic-offset: 8
10122  * End:
10123  */