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 = NULL;
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         label = nfs4_label_alloc(server, GFP_KERNEL);
4014         if (IS_ERR(label))
4015                 return PTR_ERR(label);
4016
4017         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4018         if (error < 0) {
4019                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4020                 goto err_free_label;
4021         }
4022
4023         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4024             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4025                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4026
4027 err_free_label:
4028         nfs4_label_free(label);
4029
4030         return error;
4031 }
4032
4033 /*
4034  * Get locations and (maybe) other attributes of a referral.
4035  * Note that we'll actually follow the referral later when
4036  * we detect fsid mismatch in inode revalidation
4037  */
4038 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4039                              const struct qstr *name, struct nfs_fattr *fattr,
4040                              struct nfs_fh *fhandle)
4041 {
4042         int status = -ENOMEM;
4043         struct page *page = NULL;
4044         struct nfs4_fs_locations *locations = NULL;
4045
4046         page = alloc_page(GFP_KERNEL);
4047         if (page == NULL)
4048                 goto out;
4049         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4050         if (locations == NULL)
4051                 goto out;
4052
4053         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4054         if (status != 0)
4055                 goto out;
4056
4057         /*
4058          * If the fsid didn't change, this is a migration event, not a
4059          * referral.  Cause us to drop into the exception handler, which
4060          * will kick off migration recovery.
4061          */
4062         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4063                 dprintk("%s: server did not return a different fsid for"
4064                         " a referral at %s\n", __func__, name->name);
4065                 status = -NFS4ERR_MOVED;
4066                 goto out;
4067         }
4068         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4069         nfs_fixup_referral_attributes(&locations->fattr);
4070
4071         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4072         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4073         memset(fhandle, 0, sizeof(struct nfs_fh));
4074 out:
4075         if (page)
4076                 __free_page(page);
4077         kfree(locations);
4078         return status;
4079 }
4080
4081 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4082                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4083                                 struct inode *inode)
4084 {
4085         __u32 bitmask[NFS4_BITMASK_SZ];
4086         struct nfs4_getattr_arg args = {
4087                 .fh = fhandle,
4088                 .bitmask = bitmask,
4089         };
4090         struct nfs4_getattr_res res = {
4091                 .fattr = fattr,
4092                 .label = label,
4093                 .server = server,
4094         };
4095         struct rpc_message msg = {
4096                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4097                 .rpc_argp = &args,
4098                 .rpc_resp = &res,
4099         };
4100         unsigned short task_flags = 0;
4101
4102         /* Is this is an attribute revalidation, subject to softreval? */
4103         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4104                 task_flags |= RPC_TASK_TIMEOUT;
4105
4106         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4107
4108         nfs_fattr_init(fattr);
4109         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4110         return nfs4_do_call_sync(server->client, server, &msg,
4111                         &args.seq_args, &res.seq_res, task_flags);
4112 }
4113
4114 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4115                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4116                                 struct inode *inode)
4117 {
4118         struct nfs4_exception exception = {
4119                 .interruptible = true,
4120         };
4121         int err;
4122         do {
4123                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4124                 trace_nfs4_getattr(server, fhandle, fattr, err);
4125                 err = nfs4_handle_exception(server, err,
4126                                 &exception);
4127         } while (exception.retry);
4128         return err;
4129 }
4130
4131 /*
4132  * The file is not closed if it is opened due to the a request to change
4133  * the size of the file. The open call will not be needed once the
4134  * VFS layer lookup-intents are implemented.
4135  *
4136  * Close is called when the inode is destroyed.
4137  * If we haven't opened the file for O_WRONLY, we
4138  * need to in the size_change case to obtain a stateid.
4139  *
4140  * Got race?
4141  * Because OPEN is always done by name in nfsv4, it is
4142  * possible that we opened a different file by the same
4143  * name.  We can recognize this race condition, but we
4144  * can't do anything about it besides returning an error.
4145  *
4146  * This will be fixed with VFS changes (lookup-intent).
4147  */
4148 static int
4149 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4150                   struct iattr *sattr)
4151 {
4152         struct inode *inode = d_inode(dentry);
4153         const struct cred *cred = NULL;
4154         struct nfs_open_context *ctx = NULL;
4155         struct nfs4_label *label = NULL;
4156         int status;
4157
4158         if (pnfs_ld_layoutret_on_setattr(inode) &&
4159             sattr->ia_valid & ATTR_SIZE &&
4160             sattr->ia_size < i_size_read(inode))
4161                 pnfs_commit_and_return_layout(inode);
4162
4163         nfs_fattr_init(fattr);
4164
4165         /* Deal with open(O_TRUNC) */
4166         if (sattr->ia_valid & ATTR_OPEN)
4167                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4168
4169         /* Optimization: if the end result is no change, don't RPC */
4170         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4171                 return 0;
4172
4173         /* Search for an existing open(O_WRITE) file */
4174         if (sattr->ia_valid & ATTR_FILE) {
4175
4176                 ctx = nfs_file_open_context(sattr->ia_file);
4177                 if (ctx)
4178                         cred = ctx->cred;
4179         }
4180
4181         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4182         if (IS_ERR(label))
4183                 return PTR_ERR(label);
4184
4185         /* Return any delegations if we're going to change ACLs */
4186         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4187                 nfs4_inode_make_writeable(inode);
4188
4189         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4190         if (status == 0) {
4191                 nfs_setattr_update_inode(inode, sattr, fattr);
4192                 nfs_setsecurity(inode, fattr, label);
4193         }
4194         nfs4_label_free(label);
4195         return status;
4196 }
4197
4198 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4199                 struct dentry *dentry, struct nfs_fh *fhandle,
4200                 struct nfs_fattr *fattr, struct nfs4_label *label)
4201 {
4202         struct nfs_server *server = NFS_SERVER(dir);
4203         int                    status;
4204         struct nfs4_lookup_arg args = {
4205                 .bitmask = server->attr_bitmask,
4206                 .dir_fh = NFS_FH(dir),
4207                 .name = &dentry->d_name,
4208         };
4209         struct nfs4_lookup_res res = {
4210                 .server = server,
4211                 .fattr = fattr,
4212                 .label = label,
4213                 .fh = fhandle,
4214         };
4215         struct rpc_message msg = {
4216                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4217                 .rpc_argp = &args,
4218                 .rpc_resp = &res,
4219         };
4220         unsigned short task_flags = 0;
4221
4222         /* Is this is an attribute revalidation, subject to softreval? */
4223         if (nfs_lookup_is_soft_revalidate(dentry))
4224                 task_flags |= RPC_TASK_TIMEOUT;
4225
4226         args.bitmask = nfs4_bitmask(server, label);
4227
4228         nfs_fattr_init(fattr);
4229
4230         dprintk("NFS call  lookup %pd2\n", dentry);
4231         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4232         status = nfs4_do_call_sync(clnt, server, &msg,
4233                         &args.seq_args, &res.seq_res, task_flags);
4234         dprintk("NFS reply lookup: %d\n", status);
4235         return status;
4236 }
4237
4238 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4239 {
4240         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4241                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4242         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4243         fattr->nlink = 2;
4244 }
4245
4246 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4247                                    struct dentry *dentry, struct nfs_fh *fhandle,
4248                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4249 {
4250         struct nfs4_exception exception = {
4251                 .interruptible = true,
4252         };
4253         struct rpc_clnt *client = *clnt;
4254         const struct qstr *name = &dentry->d_name;
4255         int err;
4256         do {
4257                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4258                 trace_nfs4_lookup(dir, name, err);
4259                 switch (err) {
4260                 case -NFS4ERR_BADNAME:
4261                         err = -ENOENT;
4262                         goto out;
4263                 case -NFS4ERR_MOVED:
4264                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4265                         if (err == -NFS4ERR_MOVED)
4266                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4267                         goto out;
4268                 case -NFS4ERR_WRONGSEC:
4269                         err = -EPERM;
4270                         if (client != *clnt)
4271                                 goto out;
4272                         client = nfs4_negotiate_security(client, dir, name);
4273                         if (IS_ERR(client))
4274                                 return PTR_ERR(client);
4275
4276                         exception.retry = 1;
4277                         break;
4278                 default:
4279                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4280                 }
4281         } while (exception.retry);
4282
4283 out:
4284         if (err == 0)
4285                 *clnt = client;
4286         else if (client != *clnt)
4287                 rpc_shutdown_client(client);
4288
4289         return err;
4290 }
4291
4292 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4293                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4294                             struct nfs4_label *label)
4295 {
4296         int status;
4297         struct rpc_clnt *client = NFS_CLIENT(dir);
4298
4299         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4300         if (client != NFS_CLIENT(dir)) {
4301                 rpc_shutdown_client(client);
4302                 nfs_fixup_secinfo_attributes(fattr);
4303         }
4304         return status;
4305 }
4306
4307 struct rpc_clnt *
4308 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4309                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4310 {
4311         struct rpc_clnt *client = NFS_CLIENT(dir);
4312         int status;
4313
4314         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4315         if (status < 0)
4316                 return ERR_PTR(status);
4317         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4318 }
4319
4320 static int _nfs4_proc_lookupp(struct inode *inode,
4321                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4322                 struct nfs4_label *label)
4323 {
4324         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4325         struct nfs_server *server = NFS_SERVER(inode);
4326         int                    status;
4327         struct nfs4_lookupp_arg args = {
4328                 .bitmask = server->attr_bitmask,
4329                 .fh = NFS_FH(inode),
4330         };
4331         struct nfs4_lookupp_res res = {
4332                 .server = server,
4333                 .fattr = fattr,
4334                 .label = label,
4335                 .fh = fhandle,
4336         };
4337         struct rpc_message msg = {
4338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4339                 .rpc_argp = &args,
4340                 .rpc_resp = &res,
4341         };
4342
4343         args.bitmask = nfs4_bitmask(server, label);
4344
4345         nfs_fattr_init(fattr);
4346
4347         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4348         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4349                                 &res.seq_res, 0);
4350         dprintk("NFS reply lookupp: %d\n", status);
4351         return status;
4352 }
4353
4354 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4355                              struct nfs_fattr *fattr, struct nfs4_label *label)
4356 {
4357         struct nfs4_exception exception = {
4358                 .interruptible = true,
4359         };
4360         int err;
4361         do {
4362                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4363                 trace_nfs4_lookupp(inode, err);
4364                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4365                                 &exception);
4366         } while (exception.retry);
4367         return err;
4368 }
4369
4370 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4371 {
4372         struct nfs_server *server = NFS_SERVER(inode);
4373         struct nfs4_accessargs args = {
4374                 .fh = NFS_FH(inode),
4375                 .access = entry->mask,
4376         };
4377         struct nfs4_accessres res = {
4378                 .server = server,
4379         };
4380         struct rpc_message msg = {
4381                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4382                 .rpc_argp = &args,
4383                 .rpc_resp = &res,
4384                 .rpc_cred = entry->cred,
4385         };
4386         int status = 0;
4387
4388         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4389                 res.fattr = nfs_alloc_fattr();
4390                 if (res.fattr == NULL)
4391                         return -ENOMEM;
4392                 args.bitmask = server->cache_consistency_bitmask;
4393         }
4394         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4395         if (!status) {
4396                 nfs_access_set_mask(entry, res.access);
4397                 if (res.fattr)
4398                         nfs_refresh_inode(inode, res.fattr);
4399         }
4400         nfs_free_fattr(res.fattr);
4401         return status;
4402 }
4403
4404 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4405 {
4406         struct nfs4_exception exception = {
4407                 .interruptible = true,
4408         };
4409         int err;
4410         do {
4411                 err = _nfs4_proc_access(inode, entry);
4412                 trace_nfs4_access(inode, err);
4413                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4414                                 &exception);
4415         } while (exception.retry);
4416         return err;
4417 }
4418
4419 /*
4420  * TODO: For the time being, we don't try to get any attributes
4421  * along with any of the zero-copy operations READ, READDIR,
4422  * READLINK, WRITE.
4423  *
4424  * In the case of the first three, we want to put the GETATTR
4425  * after the read-type operation -- this is because it is hard
4426  * to predict the length of a GETATTR response in v4, and thus
4427  * align the READ data correctly.  This means that the GETATTR
4428  * may end up partially falling into the page cache, and we should
4429  * shift it into the 'tail' of the xdr_buf before processing.
4430  * To do this efficiently, we need to know the total length
4431  * of data received, which doesn't seem to be available outside
4432  * of the RPC layer.
4433  *
4434  * In the case of WRITE, we also want to put the GETATTR after
4435  * the operation -- in this case because we want to make sure
4436  * we get the post-operation mtime and size.
4437  *
4438  * Both of these changes to the XDR layer would in fact be quite
4439  * minor, but I decided to leave them for a subsequent patch.
4440  */
4441 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4442                 unsigned int pgbase, unsigned int pglen)
4443 {
4444         struct nfs4_readlink args = {
4445                 .fh       = NFS_FH(inode),
4446                 .pgbase   = pgbase,
4447                 .pglen    = pglen,
4448                 .pages    = &page,
4449         };
4450         struct nfs4_readlink_res res;
4451         struct rpc_message msg = {
4452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4453                 .rpc_argp = &args,
4454                 .rpc_resp = &res,
4455         };
4456
4457         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4458 }
4459
4460 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4461                 unsigned int pgbase, unsigned int pglen)
4462 {
4463         struct nfs4_exception exception = {
4464                 .interruptible = true,
4465         };
4466         int err;
4467         do {
4468                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4469                 trace_nfs4_readlink(inode, err);
4470                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4471                                 &exception);
4472         } while (exception.retry);
4473         return err;
4474 }
4475
4476 /*
4477  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4478  */
4479 static int
4480 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4481                  int flags)
4482 {
4483         struct nfs_server *server = NFS_SERVER(dir);
4484         struct nfs4_label l, *ilabel = NULL;
4485         struct nfs_open_context *ctx;
4486         struct nfs4_state *state;
4487         int status = 0;
4488
4489         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4490         if (IS_ERR(ctx))
4491                 return PTR_ERR(ctx);
4492
4493         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4494
4495         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4496                 sattr->ia_mode &= ~current_umask();
4497         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4498         if (IS_ERR(state)) {
4499                 status = PTR_ERR(state);
4500                 goto out;
4501         }
4502 out:
4503         nfs4_label_release_security(ilabel);
4504         put_nfs_open_context(ctx);
4505         return status;
4506 }
4507
4508 static int
4509 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4510 {
4511         struct nfs_server *server = NFS_SERVER(dir);
4512         struct nfs_removeargs args = {
4513                 .fh = NFS_FH(dir),
4514                 .name = *name,
4515         };
4516         struct nfs_removeres res = {
4517                 .server = server,
4518         };
4519         struct rpc_message msg = {
4520                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4521                 .rpc_argp = &args,
4522                 .rpc_resp = &res,
4523         };
4524         unsigned long timestamp = jiffies;
4525         int status;
4526
4527         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4528         if (status == 0) {
4529                 spin_lock(&dir->i_lock);
4530                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4531                 /* Removing a directory decrements nlink in the parent */
4532                 if (ftype == NF4DIR && dir->i_nlink > 2)
4533                         nfs4_dec_nlink_locked(dir);
4534                 spin_unlock(&dir->i_lock);
4535         }
4536         return status;
4537 }
4538
4539 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4540 {
4541         struct nfs4_exception exception = {
4542                 .interruptible = true,
4543         };
4544         struct inode *inode = d_inode(dentry);
4545         int err;
4546
4547         if (inode) {
4548                 if (inode->i_nlink == 1)
4549                         nfs4_inode_return_delegation(inode);
4550                 else
4551                         nfs4_inode_make_writeable(inode);
4552         }
4553         do {
4554                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4555                 trace_nfs4_remove(dir, &dentry->d_name, err);
4556                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4557                                 &exception);
4558         } while (exception.retry);
4559         return err;
4560 }
4561
4562 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4563 {
4564         struct nfs4_exception exception = {
4565                 .interruptible = true,
4566         };
4567         int err;
4568
4569         do {
4570                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4571                 trace_nfs4_remove(dir, name, err);
4572                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4573                                 &exception);
4574         } while (exception.retry);
4575         return err;
4576 }
4577
4578 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4579                 struct dentry *dentry,
4580                 struct inode *inode)
4581 {
4582         struct nfs_removeargs *args = msg->rpc_argp;
4583         struct nfs_removeres *res = msg->rpc_resp;
4584
4585         res->server = NFS_SB(dentry->d_sb);
4586         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4587         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4588
4589         nfs_fattr_init(res->dir_attr);
4590
4591         if (inode)
4592                 nfs4_inode_return_delegation(inode);
4593 }
4594
4595 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4596 {
4597         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4598                         &data->args.seq_args,
4599                         &data->res.seq_res,
4600                         task);
4601 }
4602
4603 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4604 {
4605         struct nfs_unlinkdata *data = task->tk_calldata;
4606         struct nfs_removeres *res = &data->res;
4607
4608         if (!nfs4_sequence_done(task, &res->seq_res))
4609                 return 0;
4610         if (nfs4_async_handle_error(task, res->server, NULL,
4611                                     &data->timeout) == -EAGAIN)
4612                 return 0;
4613         if (task->tk_status == 0)
4614                 update_changeattr(dir, &res->cinfo,
4615                                 res->dir_attr->time_start, 0);
4616         return 1;
4617 }
4618
4619 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4620                 struct dentry *old_dentry,
4621                 struct dentry *new_dentry)
4622 {
4623         struct nfs_renameargs *arg = msg->rpc_argp;
4624         struct nfs_renameres *res = msg->rpc_resp;
4625         struct inode *old_inode = d_inode(old_dentry);
4626         struct inode *new_inode = d_inode(new_dentry);
4627
4628         if (old_inode)
4629                 nfs4_inode_make_writeable(old_inode);
4630         if (new_inode)
4631                 nfs4_inode_return_delegation(new_inode);
4632         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4633         res->server = NFS_SB(old_dentry->d_sb);
4634         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4635 }
4636
4637 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4638 {
4639         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4640                         &data->args.seq_args,
4641                         &data->res.seq_res,
4642                         task);
4643 }
4644
4645 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4646                                  struct inode *new_dir)
4647 {
4648         struct nfs_renamedata *data = task->tk_calldata;
4649         struct nfs_renameres *res = &data->res;
4650
4651         if (!nfs4_sequence_done(task, &res->seq_res))
4652                 return 0;
4653         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4654                 return 0;
4655
4656         if (task->tk_status == 0) {
4657                 if (new_dir != old_dir) {
4658                         /* Note: If we moved a directory, nlink will change */
4659                         update_changeattr(old_dir, &res->old_cinfo,
4660                                         res->old_fattr->time_start,
4661                                         NFS_INO_INVALID_OTHER);
4662                         update_changeattr(new_dir, &res->new_cinfo,
4663                                         res->new_fattr->time_start,
4664                                         NFS_INO_INVALID_OTHER);
4665                 } else
4666                         update_changeattr(old_dir, &res->old_cinfo,
4667                                         res->old_fattr->time_start,
4668                                         0);
4669         }
4670         return 1;
4671 }
4672
4673 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4674 {
4675         struct nfs_server *server = NFS_SERVER(inode);
4676         __u32 bitmask[NFS4_BITMASK_SZ];
4677         struct nfs4_link_arg arg = {
4678                 .fh     = NFS_FH(inode),
4679                 .dir_fh = NFS_FH(dir),
4680                 .name   = name,
4681                 .bitmask = bitmask,
4682         };
4683         struct nfs4_link_res res = {
4684                 .server = server,
4685                 .label = NULL,
4686         };
4687         struct rpc_message msg = {
4688                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4689                 .rpc_argp = &arg,
4690                 .rpc_resp = &res,
4691         };
4692         int status = -ENOMEM;
4693
4694         res.fattr = nfs_alloc_fattr();
4695         if (res.fattr == NULL)
4696                 goto out;
4697
4698         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4699         if (IS_ERR(res.label)) {
4700                 status = PTR_ERR(res.label);
4701                 goto out;
4702         }
4703
4704         nfs4_inode_make_writeable(inode);
4705         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4706
4707         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4708         if (!status) {
4709                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4710                 status = nfs_post_op_update_inode(inode, res.fattr);
4711                 if (!status)
4712                         nfs_setsecurity(inode, res.fattr, res.label);
4713         }
4714
4715
4716         nfs4_label_free(res.label);
4717
4718 out:
4719         nfs_free_fattr(res.fattr);
4720         return status;
4721 }
4722
4723 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4724 {
4725         struct nfs4_exception exception = {
4726                 .interruptible = true,
4727         };
4728         int err;
4729         do {
4730                 err = nfs4_handle_exception(NFS_SERVER(inode),
4731                                 _nfs4_proc_link(inode, dir, name),
4732                                 &exception);
4733         } while (exception.retry);
4734         return err;
4735 }
4736
4737 struct nfs4_createdata {
4738         struct rpc_message msg;
4739         struct nfs4_create_arg arg;
4740         struct nfs4_create_res res;
4741         struct nfs_fh fh;
4742         struct nfs_fattr fattr;
4743         struct nfs4_label *label;
4744 };
4745
4746 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4747                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4748 {
4749         struct nfs4_createdata *data;
4750
4751         data = kzalloc(sizeof(*data), GFP_KERNEL);
4752         if (data != NULL) {
4753                 struct nfs_server *server = NFS_SERVER(dir);
4754
4755                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4756                 if (IS_ERR(data->label))
4757                         goto out_free;
4758
4759                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4760                 data->msg.rpc_argp = &data->arg;
4761                 data->msg.rpc_resp = &data->res;
4762                 data->arg.dir_fh = NFS_FH(dir);
4763                 data->arg.server = server;
4764                 data->arg.name = name;
4765                 data->arg.attrs = sattr;
4766                 data->arg.ftype = ftype;
4767                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4768                 data->arg.umask = current_umask();
4769                 data->res.server = server;
4770                 data->res.fh = &data->fh;
4771                 data->res.fattr = &data->fattr;
4772                 data->res.label = data->label;
4773                 nfs_fattr_init(data->res.fattr);
4774         }
4775         return data;
4776 out_free:
4777         kfree(data);
4778         return NULL;
4779 }
4780
4781 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4782 {
4783         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4784                                     &data->arg.seq_args, &data->res.seq_res, 1);
4785         if (status == 0) {
4786                 spin_lock(&dir->i_lock);
4787                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4788                                 data->res.fattr->time_start, 0);
4789                 /* Creating a directory bumps nlink in the parent */
4790                 if (data->arg.ftype == NF4DIR)
4791                         nfs4_inc_nlink_locked(dir);
4792                 spin_unlock(&dir->i_lock);
4793                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4794         }
4795         return status;
4796 }
4797
4798 static void nfs4_free_createdata(struct nfs4_createdata *data)
4799 {
4800         nfs4_label_free(data->label);
4801         kfree(data);
4802 }
4803
4804 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4805                 struct page *page, unsigned int len, struct iattr *sattr,
4806                 struct nfs4_label *label)
4807 {
4808         struct nfs4_createdata *data;
4809         int status = -ENAMETOOLONG;
4810
4811         if (len > NFS4_MAXPATHLEN)
4812                 goto out;
4813
4814         status = -ENOMEM;
4815         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4816         if (data == NULL)
4817                 goto out;
4818
4819         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4820         data->arg.u.symlink.pages = &page;
4821         data->arg.u.symlink.len = len;
4822         data->arg.label = label;
4823
4824         status = nfs4_do_create(dir, dentry, data);
4825
4826         nfs4_free_createdata(data);
4827 out:
4828         return status;
4829 }
4830
4831 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4832                 struct page *page, unsigned int len, struct iattr *sattr)
4833 {
4834         struct nfs4_exception exception = {
4835                 .interruptible = true,
4836         };
4837         struct nfs4_label l, *label = NULL;
4838         int err;
4839
4840         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4841
4842         do {
4843                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4844                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4845                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4846                                 &exception);
4847         } while (exception.retry);
4848
4849         nfs4_label_release_security(label);
4850         return err;
4851 }
4852
4853 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4854                 struct iattr *sattr, struct nfs4_label *label)
4855 {
4856         struct nfs4_createdata *data;
4857         int status = -ENOMEM;
4858
4859         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4860         if (data == NULL)
4861                 goto out;
4862
4863         data->arg.label = label;
4864         status = nfs4_do_create(dir, dentry, data);
4865
4866         nfs4_free_createdata(data);
4867 out:
4868         return status;
4869 }
4870
4871 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4872                 struct iattr *sattr)
4873 {
4874         struct nfs_server *server = NFS_SERVER(dir);
4875         struct nfs4_exception exception = {
4876                 .interruptible = true,
4877         };
4878         struct nfs4_label l, *label = NULL;
4879         int err;
4880
4881         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4882
4883         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4884                 sattr->ia_mode &= ~current_umask();
4885         do {
4886                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4887                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4888                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4889                                 &exception);
4890         } while (exception.retry);
4891         nfs4_label_release_security(label);
4892
4893         return err;
4894 }
4895
4896 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4897                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4898 {
4899         struct inode            *dir = d_inode(dentry);
4900         struct nfs4_readdir_arg args = {
4901                 .fh = NFS_FH(dir),
4902                 .pages = pages,
4903                 .pgbase = 0,
4904                 .count = count,
4905                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4906                 .plus = plus,
4907         };
4908         struct nfs4_readdir_res res;
4909         struct rpc_message msg = {
4910                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4911                 .rpc_argp = &args,
4912                 .rpc_resp = &res,
4913                 .rpc_cred = cred,
4914         };
4915         int                     status;
4916
4917         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4918                         dentry,
4919                         (unsigned long long)cookie);
4920         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4921         res.pgbase = args.pgbase;
4922         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4923         if (status >= 0) {
4924                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4925                 status += args.pgbase;
4926         }
4927
4928         nfs_invalidate_atime(dir);
4929
4930         dprintk("%s: returns %d\n", __func__, status);
4931         return status;
4932 }
4933
4934 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4935                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4936 {
4937         struct nfs4_exception exception = {
4938                 .interruptible = true,
4939         };
4940         int err;
4941         do {
4942                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4943                                 pages, count, plus);
4944                 trace_nfs4_readdir(d_inode(dentry), err);
4945                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4946                                 &exception);
4947         } while (exception.retry);
4948         return err;
4949 }
4950
4951 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4952                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4953 {
4954         struct nfs4_createdata *data;
4955         int mode = sattr->ia_mode;
4956         int status = -ENOMEM;
4957
4958         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4959         if (data == NULL)
4960                 goto out;
4961
4962         if (S_ISFIFO(mode))
4963                 data->arg.ftype = NF4FIFO;
4964         else if (S_ISBLK(mode)) {
4965                 data->arg.ftype = NF4BLK;
4966                 data->arg.u.device.specdata1 = MAJOR(rdev);
4967                 data->arg.u.device.specdata2 = MINOR(rdev);
4968         }
4969         else if (S_ISCHR(mode)) {
4970                 data->arg.ftype = NF4CHR;
4971                 data->arg.u.device.specdata1 = MAJOR(rdev);
4972                 data->arg.u.device.specdata2 = MINOR(rdev);
4973         } else if (!S_ISSOCK(mode)) {
4974                 status = -EINVAL;
4975                 goto out_free;
4976         }
4977
4978         data->arg.label = label;
4979         status = nfs4_do_create(dir, dentry, data);
4980 out_free:
4981         nfs4_free_createdata(data);
4982 out:
4983         return status;
4984 }
4985
4986 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4987                 struct iattr *sattr, dev_t rdev)
4988 {
4989         struct nfs_server *server = NFS_SERVER(dir);
4990         struct nfs4_exception exception = {
4991                 .interruptible = true,
4992         };
4993         struct nfs4_label l, *label = NULL;
4994         int err;
4995
4996         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4997
4998         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4999                 sattr->ia_mode &= ~current_umask();
5000         do {
5001                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5002                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5003                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5004                                 &exception);
5005         } while (exception.retry);
5006
5007         nfs4_label_release_security(label);
5008
5009         return err;
5010 }
5011
5012 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5013                  struct nfs_fsstat *fsstat)
5014 {
5015         struct nfs4_statfs_arg args = {
5016                 .fh = fhandle,
5017                 .bitmask = server->attr_bitmask,
5018         };
5019         struct nfs4_statfs_res res = {
5020                 .fsstat = fsstat,
5021         };
5022         struct rpc_message msg = {
5023                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5024                 .rpc_argp = &args,
5025                 .rpc_resp = &res,
5026         };
5027
5028         nfs_fattr_init(fsstat->fattr);
5029         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5030 }
5031
5032 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5033 {
5034         struct nfs4_exception exception = {
5035                 .interruptible = true,
5036         };
5037         int err;
5038         do {
5039                 err = nfs4_handle_exception(server,
5040                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5041                                 &exception);
5042         } while (exception.retry);
5043         return err;
5044 }
5045
5046 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5047                 struct nfs_fsinfo *fsinfo)
5048 {
5049         struct nfs4_fsinfo_arg args = {
5050                 .fh = fhandle,
5051                 .bitmask = server->attr_bitmask,
5052         };
5053         struct nfs4_fsinfo_res res = {
5054                 .fsinfo = fsinfo,
5055         };
5056         struct rpc_message msg = {
5057                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5058                 .rpc_argp = &args,
5059                 .rpc_resp = &res,
5060         };
5061
5062         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5063 }
5064
5065 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5066 {
5067         struct nfs4_exception exception = {
5068                 .interruptible = true,
5069         };
5070         int err;
5071
5072         do {
5073                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5074                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5075                 if (err == 0) {
5076                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5077                         break;
5078                 }
5079                 err = nfs4_handle_exception(server, err, &exception);
5080         } while (exception.retry);
5081         return err;
5082 }
5083
5084 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5085 {
5086         int error;
5087
5088         nfs_fattr_init(fsinfo->fattr);
5089         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5090         if (error == 0) {
5091                 /* block layout checks this! */
5092                 server->pnfs_blksize = fsinfo->blksize;
5093                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5094         }
5095
5096         return error;
5097 }
5098
5099 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5100                 struct nfs_pathconf *pathconf)
5101 {
5102         struct nfs4_pathconf_arg args = {
5103                 .fh = fhandle,
5104                 .bitmask = server->attr_bitmask,
5105         };
5106         struct nfs4_pathconf_res res = {
5107                 .pathconf = pathconf,
5108         };
5109         struct rpc_message msg = {
5110                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5111                 .rpc_argp = &args,
5112                 .rpc_resp = &res,
5113         };
5114
5115         /* None of the pathconf attributes are mandatory to implement */
5116         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5117                 memset(pathconf, 0, sizeof(*pathconf));
5118                 return 0;
5119         }
5120
5121         nfs_fattr_init(pathconf->fattr);
5122         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5123 }
5124
5125 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5126                 struct nfs_pathconf *pathconf)
5127 {
5128         struct nfs4_exception exception = {
5129                 .interruptible = true,
5130         };
5131         int err;
5132
5133         do {
5134                 err = nfs4_handle_exception(server,
5135                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5136                                 &exception);
5137         } while (exception.retry);
5138         return err;
5139 }
5140
5141 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5142                 const struct nfs_open_context *ctx,
5143                 const struct nfs_lock_context *l_ctx,
5144                 fmode_t fmode)
5145 {
5146         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5147 }
5148 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5149
5150 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5151                 const struct nfs_open_context *ctx,
5152                 const struct nfs_lock_context *l_ctx,
5153                 fmode_t fmode)
5154 {
5155         nfs4_stateid _current_stateid;
5156
5157         /* If the current stateid represents a lost lock, then exit */
5158         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5159                 return true;
5160         return nfs4_stateid_match(stateid, &_current_stateid);
5161 }
5162
5163 static bool nfs4_error_stateid_expired(int err)
5164 {
5165         switch (err) {
5166         case -NFS4ERR_DELEG_REVOKED:
5167         case -NFS4ERR_ADMIN_REVOKED:
5168         case -NFS4ERR_BAD_STATEID:
5169         case -NFS4ERR_STALE_STATEID:
5170         case -NFS4ERR_OLD_STATEID:
5171         case -NFS4ERR_OPENMODE:
5172         case -NFS4ERR_EXPIRED:
5173                 return true;
5174         }
5175         return false;
5176 }
5177
5178 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5179 {
5180         struct nfs_server *server = NFS_SERVER(hdr->inode);
5181
5182         trace_nfs4_read(hdr, task->tk_status);
5183         if (task->tk_status < 0) {
5184                 struct nfs4_exception exception = {
5185                         .inode = hdr->inode,
5186                         .state = hdr->args.context->state,
5187                         .stateid = &hdr->args.stateid,
5188                 };
5189                 task->tk_status = nfs4_async_handle_exception(task,
5190                                 server, task->tk_status, &exception);
5191                 if (exception.retry) {
5192                         rpc_restart_call_prepare(task);
5193                         return -EAGAIN;
5194                 }
5195         }
5196
5197         if (task->tk_status > 0)
5198                 renew_lease(server, hdr->timestamp);
5199         return 0;
5200 }
5201
5202 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5203                 struct nfs_pgio_args *args)
5204 {
5205
5206         if (!nfs4_error_stateid_expired(task->tk_status) ||
5207                 nfs4_stateid_is_current(&args->stateid,
5208                                 args->context,
5209                                 args->lock_context,
5210                                 FMODE_READ))
5211                 return false;
5212         rpc_restart_call_prepare(task);
5213         return true;
5214 }
5215
5216 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5217 {
5218
5219         dprintk("--> %s\n", __func__);
5220
5221         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5222                 return -EAGAIN;
5223         if (nfs4_read_stateid_changed(task, &hdr->args))
5224                 return -EAGAIN;
5225         if (task->tk_status > 0)
5226                 nfs_invalidate_atime(hdr->inode);
5227         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5228                                     nfs4_read_done_cb(task, hdr);
5229 }
5230
5231 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5232                                  struct rpc_message *msg)
5233 {
5234         hdr->timestamp   = jiffies;
5235         if (!hdr->pgio_done_cb)
5236                 hdr->pgio_done_cb = nfs4_read_done_cb;
5237         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5238         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5239 }
5240
5241 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5242                                       struct nfs_pgio_header *hdr)
5243 {
5244         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5245                         &hdr->args.seq_args,
5246                         &hdr->res.seq_res,
5247                         task))
5248                 return 0;
5249         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5250                                 hdr->args.lock_context,
5251                                 hdr->rw_mode) == -EIO)
5252                 return -EIO;
5253         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5254                 return -EIO;
5255         return 0;
5256 }
5257
5258 static int nfs4_write_done_cb(struct rpc_task *task,
5259                               struct nfs_pgio_header *hdr)
5260 {
5261         struct inode *inode = hdr->inode;
5262
5263         trace_nfs4_write(hdr, task->tk_status);
5264         if (task->tk_status < 0) {
5265                 struct nfs4_exception exception = {
5266                         .inode = hdr->inode,
5267                         .state = hdr->args.context->state,
5268                         .stateid = &hdr->args.stateid,
5269                 };
5270                 task->tk_status = nfs4_async_handle_exception(task,
5271                                 NFS_SERVER(inode), task->tk_status,
5272                                 &exception);
5273                 if (exception.retry) {
5274                         rpc_restart_call_prepare(task);
5275                         return -EAGAIN;
5276                 }
5277         }
5278         if (task->tk_status >= 0) {
5279                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5280                 nfs_writeback_update_inode(hdr);
5281         }
5282         return 0;
5283 }
5284
5285 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5286                 struct nfs_pgio_args *args)
5287 {
5288
5289         if (!nfs4_error_stateid_expired(task->tk_status) ||
5290                 nfs4_stateid_is_current(&args->stateid,
5291                                 args->context,
5292                                 args->lock_context,
5293                                 FMODE_WRITE))
5294                 return false;
5295         rpc_restart_call_prepare(task);
5296         return true;
5297 }
5298
5299 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5300 {
5301         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5302                 return -EAGAIN;
5303         if (nfs4_write_stateid_changed(task, &hdr->args))
5304                 return -EAGAIN;
5305         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5306                 nfs4_write_done_cb(task, hdr);
5307 }
5308
5309 static
5310 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5311 {
5312         /* Don't request attributes for pNFS or O_DIRECT writes */
5313         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5314                 return false;
5315         /* Otherwise, request attributes if and only if we don't hold
5316          * a delegation
5317          */
5318         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5319 }
5320
5321 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5322                                   struct rpc_message *msg,
5323                                   struct rpc_clnt **clnt)
5324 {
5325         struct nfs_server *server = NFS_SERVER(hdr->inode);
5326
5327         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5328                 hdr->args.bitmask = NULL;
5329                 hdr->res.fattr = NULL;
5330         } else
5331                 hdr->args.bitmask = server->cache_consistency_bitmask;
5332
5333         if (!hdr->pgio_done_cb)
5334                 hdr->pgio_done_cb = nfs4_write_done_cb;
5335         hdr->res.server = server;
5336         hdr->timestamp   = jiffies;
5337
5338         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5339         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5340         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5341 }
5342
5343 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5344 {
5345         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5346                         &data->args.seq_args,
5347                         &data->res.seq_res,
5348                         task);
5349 }
5350
5351 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5352 {
5353         struct inode *inode = data->inode;
5354
5355         trace_nfs4_commit(data, task->tk_status);
5356         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5357                                     NULL, NULL) == -EAGAIN) {
5358                 rpc_restart_call_prepare(task);
5359                 return -EAGAIN;
5360         }
5361         return 0;
5362 }
5363
5364 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5365 {
5366         if (!nfs4_sequence_done(task, &data->res.seq_res))
5367                 return -EAGAIN;
5368         return data->commit_done_cb(task, data);
5369 }
5370
5371 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5372                                    struct rpc_clnt **clnt)
5373 {
5374         struct nfs_server *server = NFS_SERVER(data->inode);
5375
5376         if (data->commit_done_cb == NULL)
5377                 data->commit_done_cb = nfs4_commit_done_cb;
5378         data->res.server = server;
5379         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5380         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5381         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5382 }
5383
5384 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5385                                 struct nfs_commitres *res)
5386 {
5387         struct inode *dst_inode = file_inode(dst);
5388         struct nfs_server *server = NFS_SERVER(dst_inode);
5389         struct rpc_message msg = {
5390                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5391                 .rpc_argp = args,
5392                 .rpc_resp = res,
5393         };
5394
5395         args->fh = NFS_FH(dst_inode);
5396         return nfs4_call_sync(server->client, server, &msg,
5397                         &args->seq_args, &res->seq_res, 1);
5398 }
5399
5400 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5401 {
5402         struct nfs_commitargs args = {
5403                 .offset = offset,
5404                 .count = count,
5405         };
5406         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5407         struct nfs4_exception exception = { };
5408         int status;
5409
5410         do {
5411                 status = _nfs4_proc_commit(dst, &args, res);
5412                 status = nfs4_handle_exception(dst_server, status, &exception);
5413         } while (exception.retry);
5414
5415         return status;
5416 }
5417
5418 struct nfs4_renewdata {
5419         struct nfs_client       *client;
5420         unsigned long           timestamp;
5421 };
5422
5423 /*
5424  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5425  * standalone procedure for queueing an asynchronous RENEW.
5426  */
5427 static void nfs4_renew_release(void *calldata)
5428 {
5429         struct nfs4_renewdata *data = calldata;
5430         struct nfs_client *clp = data->client;
5431
5432         if (refcount_read(&clp->cl_count) > 1)
5433                 nfs4_schedule_state_renewal(clp);
5434         nfs_put_client(clp);
5435         kfree(data);
5436 }
5437
5438 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5439 {
5440         struct nfs4_renewdata *data = calldata;
5441         struct nfs_client *clp = data->client;
5442         unsigned long timestamp = data->timestamp;
5443
5444         trace_nfs4_renew_async(clp, task->tk_status);
5445         switch (task->tk_status) {
5446         case 0:
5447                 break;
5448         case -NFS4ERR_LEASE_MOVED:
5449                 nfs4_schedule_lease_moved_recovery(clp);
5450                 break;
5451         default:
5452                 /* Unless we're shutting down, schedule state recovery! */
5453                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5454                         return;
5455                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5456                         nfs4_schedule_lease_recovery(clp);
5457                         return;
5458                 }
5459                 nfs4_schedule_path_down_recovery(clp);
5460         }
5461         do_renew_lease(clp, timestamp);
5462 }
5463
5464 static const struct rpc_call_ops nfs4_renew_ops = {
5465         .rpc_call_done = nfs4_renew_done,
5466         .rpc_release = nfs4_renew_release,
5467 };
5468
5469 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5470 {
5471         struct rpc_message msg = {
5472                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5473                 .rpc_argp       = clp,
5474                 .rpc_cred       = cred,
5475         };
5476         struct nfs4_renewdata *data;
5477
5478         if (renew_flags == 0)
5479                 return 0;
5480         if (!refcount_inc_not_zero(&clp->cl_count))
5481                 return -EIO;
5482         data = kmalloc(sizeof(*data), GFP_NOFS);
5483         if (data == NULL) {
5484                 nfs_put_client(clp);
5485                 return -ENOMEM;
5486         }
5487         data->client = clp;
5488         data->timestamp = jiffies;
5489         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5490                         &nfs4_renew_ops, data);
5491 }
5492
5493 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5494 {
5495         struct rpc_message msg = {
5496                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5497                 .rpc_argp       = clp,
5498                 .rpc_cred       = cred,
5499         };
5500         unsigned long now = jiffies;
5501         int status;
5502
5503         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5504         if (status < 0)
5505                 return status;
5506         do_renew_lease(clp, now);
5507         return 0;
5508 }
5509
5510 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5511 {
5512         return server->caps & NFS_CAP_ACLS;
5513 }
5514
5515 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5516  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5517  * the stack.
5518  */
5519 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5520
5521 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5522                 struct page **pages)
5523 {
5524         struct page *newpage, **spages;
5525         int rc = 0;
5526         size_t len;
5527         spages = pages;
5528
5529         do {
5530                 len = min_t(size_t, PAGE_SIZE, buflen);
5531                 newpage = alloc_page(GFP_KERNEL);
5532
5533                 if (newpage == NULL)
5534                         goto unwind;
5535                 memcpy(page_address(newpage), buf, len);
5536                 buf += len;
5537                 buflen -= len;
5538                 *pages++ = newpage;
5539                 rc++;
5540         } while (buflen != 0);
5541
5542         return rc;
5543
5544 unwind:
5545         for(; rc > 0; rc--)
5546                 __free_page(spages[rc-1]);
5547         return -ENOMEM;
5548 }
5549
5550 struct nfs4_cached_acl {
5551         int cached;
5552         size_t len;
5553         char data[0];
5554 };
5555
5556 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5557 {
5558         struct nfs_inode *nfsi = NFS_I(inode);
5559
5560         spin_lock(&inode->i_lock);
5561         kfree(nfsi->nfs4_acl);
5562         nfsi->nfs4_acl = acl;
5563         spin_unlock(&inode->i_lock);
5564 }
5565
5566 static void nfs4_zap_acl_attr(struct inode *inode)
5567 {
5568         nfs4_set_cached_acl(inode, NULL);
5569 }
5570
5571 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5572 {
5573         struct nfs_inode *nfsi = NFS_I(inode);
5574         struct nfs4_cached_acl *acl;
5575         int ret = -ENOENT;
5576
5577         spin_lock(&inode->i_lock);
5578         acl = nfsi->nfs4_acl;
5579         if (acl == NULL)
5580                 goto out;
5581         if (buf == NULL) /* user is just asking for length */
5582                 goto out_len;
5583         if (acl->cached == 0)
5584                 goto out;
5585         ret = -ERANGE; /* see getxattr(2) man page */
5586         if (acl->len > buflen)
5587                 goto out;
5588         memcpy(buf, acl->data, acl->len);
5589 out_len:
5590         ret = acl->len;
5591 out:
5592         spin_unlock(&inode->i_lock);
5593         return ret;
5594 }
5595
5596 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5597 {
5598         struct nfs4_cached_acl *acl;
5599         size_t buflen = sizeof(*acl) + acl_len;
5600
5601         if (buflen <= PAGE_SIZE) {
5602                 acl = kmalloc(buflen, GFP_KERNEL);
5603                 if (acl == NULL)
5604                         goto out;
5605                 acl->cached = 1;
5606                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5607         } else {
5608                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5609                 if (acl == NULL)
5610                         goto out;
5611                 acl->cached = 0;
5612         }
5613         acl->len = acl_len;
5614 out:
5615         nfs4_set_cached_acl(inode, acl);
5616 }
5617
5618 /*
5619  * The getxattr API returns the required buffer length when called with a
5620  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5621  * the required buf.  On a NULL buf, we send a page of data to the server
5622  * guessing that the ACL request can be serviced by a page. If so, we cache
5623  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5624  * the cache. If not so, we throw away the page, and cache the required
5625  * length. The next getxattr call will then produce another round trip to
5626  * the server, this time with the input buf of the required size.
5627  */
5628 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5629 {
5630         struct page **pages;
5631         struct nfs_getaclargs args = {
5632                 .fh = NFS_FH(inode),
5633                 .acl_len = buflen,
5634         };
5635         struct nfs_getaclres res = {
5636                 .acl_len = buflen,
5637         };
5638         struct rpc_message msg = {
5639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5640                 .rpc_argp = &args,
5641                 .rpc_resp = &res,
5642         };
5643         unsigned int npages;
5644         int ret = -ENOMEM, i;
5645         struct nfs_server *server = NFS_SERVER(inode);
5646
5647         if (buflen == 0)
5648                 buflen = server->rsize;
5649
5650         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5651         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5652         if (!pages)
5653                 return -ENOMEM;
5654
5655         args.acl_pages = pages;
5656
5657         for (i = 0; i < npages; i++) {
5658                 pages[i] = alloc_page(GFP_KERNEL);
5659                 if (!pages[i])
5660                         goto out_free;
5661         }
5662
5663         /* for decoding across pages */
5664         res.acl_scratch = alloc_page(GFP_KERNEL);
5665         if (!res.acl_scratch)
5666                 goto out_free;
5667
5668         args.acl_len = npages * PAGE_SIZE;
5669
5670         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5671                 __func__, buf, buflen, npages, args.acl_len);
5672         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5673                              &msg, &args.seq_args, &res.seq_res, 0);
5674         if (ret)
5675                 goto out_free;
5676
5677         /* Handle the case where the passed-in buffer is too short */
5678         if (res.acl_flags & NFS4_ACL_TRUNC) {
5679                 /* Did the user only issue a request for the acl length? */
5680                 if (buf == NULL)
5681                         goto out_ok;
5682                 ret = -ERANGE;
5683                 goto out_free;
5684         }
5685         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5686         if (buf) {
5687                 if (res.acl_len > buflen) {
5688                         ret = -ERANGE;
5689                         goto out_free;
5690                 }
5691                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5692         }
5693 out_ok:
5694         ret = res.acl_len;
5695 out_free:
5696         for (i = 0; i < npages; i++)
5697                 if (pages[i])
5698                         __free_page(pages[i]);
5699         if (res.acl_scratch)
5700                 __free_page(res.acl_scratch);
5701         kfree(pages);
5702         return ret;
5703 }
5704
5705 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5706 {
5707         struct nfs4_exception exception = {
5708                 .interruptible = true,
5709         };
5710         ssize_t ret;
5711         do {
5712                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5713                 trace_nfs4_get_acl(inode, ret);
5714                 if (ret >= 0)
5715                         break;
5716                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5717         } while (exception.retry);
5718         return ret;
5719 }
5720
5721 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5722 {
5723         struct nfs_server *server = NFS_SERVER(inode);
5724         int ret;
5725
5726         if (!nfs4_server_supports_acls(server))
5727                 return -EOPNOTSUPP;
5728         ret = nfs_revalidate_inode(server, inode);
5729         if (ret < 0)
5730                 return ret;
5731         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5732                 nfs_zap_acl_cache(inode);
5733         ret = nfs4_read_cached_acl(inode, buf, buflen);
5734         if (ret != -ENOENT)
5735                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5736                  * but no cached acl data, just the acl length */
5737                 return ret;
5738         return nfs4_get_acl_uncached(inode, buf, buflen);
5739 }
5740
5741 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5742 {
5743         struct nfs_server *server = NFS_SERVER(inode);
5744         struct page *pages[NFS4ACL_MAXPAGES];
5745         struct nfs_setaclargs arg = {
5746                 .fh             = NFS_FH(inode),
5747                 .acl_pages      = pages,
5748                 .acl_len        = buflen,
5749         };
5750         struct nfs_setaclres res;
5751         struct rpc_message msg = {
5752                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5753                 .rpc_argp       = &arg,
5754                 .rpc_resp       = &res,
5755         };
5756         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5757         int ret, i;
5758
5759         if (!nfs4_server_supports_acls(server))
5760                 return -EOPNOTSUPP;
5761         if (npages > ARRAY_SIZE(pages))
5762                 return -ERANGE;
5763         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5764         if (i < 0)
5765                 return i;
5766         nfs4_inode_make_writeable(inode);
5767         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5768
5769         /*
5770          * Free each page after tx, so the only ref left is
5771          * held by the network stack
5772          */
5773         for (; i > 0; i--)
5774                 put_page(pages[i-1]);
5775
5776         /*
5777          * Acl update can result in inode attribute update.
5778          * so mark the attribute cache invalid.
5779          */
5780         spin_lock(&inode->i_lock);
5781         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5782                 | NFS_INO_INVALID_CTIME
5783                 | NFS_INO_REVAL_FORCED;
5784         spin_unlock(&inode->i_lock);
5785         nfs_access_zap_cache(inode);
5786         nfs_zap_acl_cache(inode);
5787         return ret;
5788 }
5789
5790 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5791 {
5792         struct nfs4_exception exception = { };
5793         int err;
5794         do {
5795                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5796                 trace_nfs4_set_acl(inode, err);
5797                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5798                                 &exception);
5799         } while (exception.retry);
5800         return err;
5801 }
5802
5803 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5804 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5805                                         size_t buflen)
5806 {
5807         struct nfs_server *server = NFS_SERVER(inode);
5808         struct nfs_fattr fattr;
5809         struct nfs4_label label = {0, 0, buflen, buf};
5810
5811         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5812         struct nfs4_getattr_arg arg = {
5813                 .fh             = NFS_FH(inode),
5814                 .bitmask        = bitmask,
5815         };
5816         struct nfs4_getattr_res res = {
5817                 .fattr          = &fattr,
5818                 .label          = &label,
5819                 .server         = server,
5820         };
5821         struct rpc_message msg = {
5822                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5823                 .rpc_argp       = &arg,
5824                 .rpc_resp       = &res,
5825         };
5826         int ret;
5827
5828         nfs_fattr_init(&fattr);
5829
5830         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5831         if (ret)
5832                 return ret;
5833         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5834                 return -ENOENT;
5835         if (buflen < label.len)
5836                 return -ERANGE;
5837         return 0;
5838 }
5839
5840 static int nfs4_get_security_label(struct inode *inode, void *buf,
5841                                         size_t buflen)
5842 {
5843         struct nfs4_exception exception = {
5844                 .interruptible = true,
5845         };
5846         int err;
5847
5848         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5849                 return -EOPNOTSUPP;
5850
5851         do {
5852                 err = _nfs4_get_security_label(inode, buf, buflen);
5853                 trace_nfs4_get_security_label(inode, err);
5854                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5855                                 &exception);
5856         } while (exception.retry);
5857         return err;
5858 }
5859
5860 static int _nfs4_do_set_security_label(struct inode *inode,
5861                 struct nfs4_label *ilabel,
5862                 struct nfs_fattr *fattr,
5863                 struct nfs4_label *olabel)
5864 {
5865
5866         struct iattr sattr = {0};
5867         struct nfs_server *server = NFS_SERVER(inode);
5868         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5869         struct nfs_setattrargs arg = {
5870                 .fh             = NFS_FH(inode),
5871                 .iap            = &sattr,
5872                 .server         = server,
5873                 .bitmask        = bitmask,
5874                 .label          = ilabel,
5875         };
5876         struct nfs_setattrres res = {
5877                 .fattr          = fattr,
5878                 .label          = olabel,
5879                 .server         = server,
5880         };
5881         struct rpc_message msg = {
5882                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5883                 .rpc_argp       = &arg,
5884                 .rpc_resp       = &res,
5885         };
5886         int status;
5887
5888         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5889
5890         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5891         if (status)
5892                 dprintk("%s failed: %d\n", __func__, status);
5893
5894         return status;
5895 }
5896
5897 static int nfs4_do_set_security_label(struct inode *inode,
5898                 struct nfs4_label *ilabel,
5899                 struct nfs_fattr *fattr,
5900                 struct nfs4_label *olabel)
5901 {
5902         struct nfs4_exception exception = { };
5903         int err;
5904
5905         do {
5906                 err = _nfs4_do_set_security_label(inode, ilabel,
5907                                 fattr, olabel);
5908                 trace_nfs4_set_security_label(inode, err);
5909                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5910                                 &exception);
5911         } while (exception.retry);
5912         return err;
5913 }
5914
5915 static int
5916 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5917 {
5918         struct nfs4_label ilabel, *olabel = NULL;
5919         struct nfs_fattr fattr;
5920         int status;
5921
5922         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5923                 return -EOPNOTSUPP;
5924
5925         nfs_fattr_init(&fattr);
5926
5927         ilabel.pi = 0;
5928         ilabel.lfs = 0;
5929         ilabel.label = (char *)buf;
5930         ilabel.len = buflen;
5931
5932         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5933         if (IS_ERR(olabel)) {
5934                 status = -PTR_ERR(olabel);
5935                 goto out;
5936         }
5937
5938         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5939         if (status == 0)
5940                 nfs_setsecurity(inode, &fattr, olabel);
5941
5942         nfs4_label_free(olabel);
5943 out:
5944         return status;
5945 }
5946 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5947
5948
5949 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5950                                     nfs4_verifier *bootverf)
5951 {
5952         __be32 verf[2];
5953
5954         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5955                 /* An impossible timestamp guarantees this value
5956                  * will never match a generated boot time. */
5957                 verf[0] = cpu_to_be32(U32_MAX);
5958                 verf[1] = cpu_to_be32(U32_MAX);
5959         } else {
5960                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5961                 u64 ns = ktime_to_ns(nn->boot_time);
5962
5963                 verf[0] = cpu_to_be32(ns >> 32);
5964                 verf[1] = cpu_to_be32(ns);
5965         }
5966         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5967 }
5968
5969 static int
5970 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5971 {
5972         size_t len;
5973         char *str;
5974
5975         if (clp->cl_owner_id != NULL)
5976                 return 0;
5977
5978         rcu_read_lock();
5979         len = 14 +
5980                 strlen(clp->cl_rpcclient->cl_nodename) +
5981                 1 +
5982                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5983                 1;
5984         rcu_read_unlock();
5985         if (nfs4_client_id_uniquifier[0] != '\0')
5986                 len += strlen(nfs4_client_id_uniquifier) + 1;
5987         if (len > NFS4_OPAQUE_LIMIT + 1)
5988                 return -EINVAL;
5989
5990         /*
5991          * Since this string is allocated at mount time, and held until the
5992          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5993          * about a memory-reclaim deadlock.
5994          */
5995         str = kmalloc(len, GFP_KERNEL);
5996         if (!str)
5997                 return -ENOMEM;
5998
5999         rcu_read_lock();
6000         if (nfs4_client_id_uniquifier[0] != '\0')
6001                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6002                           clp->cl_rpcclient->cl_nodename,
6003                           nfs4_client_id_uniquifier,
6004                           rpc_peeraddr2str(clp->cl_rpcclient,
6005                                            RPC_DISPLAY_ADDR));
6006         else
6007                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6008                           clp->cl_rpcclient->cl_nodename,
6009                           rpc_peeraddr2str(clp->cl_rpcclient,
6010                                            RPC_DISPLAY_ADDR));
6011         rcu_read_unlock();
6012
6013         clp->cl_owner_id = str;
6014         return 0;
6015 }
6016
6017 static int
6018 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
6019 {
6020         size_t len;
6021         char *str;
6022
6023         len = 10 + 10 + 1 + 10 + 1 +
6024                 strlen(nfs4_client_id_uniquifier) + 1 +
6025                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6026
6027         if (len > NFS4_OPAQUE_LIMIT + 1)
6028                 return -EINVAL;
6029
6030         /*
6031          * Since this string is allocated at mount time, and held until the
6032          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6033          * about a memory-reclaim deadlock.
6034          */
6035         str = kmalloc(len, GFP_KERNEL);
6036         if (!str)
6037                 return -ENOMEM;
6038
6039         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6040                         clp->rpc_ops->version, clp->cl_minorversion,
6041                         nfs4_client_id_uniquifier,
6042                         clp->cl_rpcclient->cl_nodename);
6043         clp->cl_owner_id = str;
6044         return 0;
6045 }
6046
6047 static int
6048 nfs4_init_uniform_client_string(struct nfs_client *clp)
6049 {
6050         size_t len;
6051         char *str;
6052
6053         if (clp->cl_owner_id != NULL)
6054                 return 0;
6055
6056         if (nfs4_client_id_uniquifier[0] != '\0')
6057                 return nfs4_init_uniquifier_client_string(clp);
6058
6059         len = 10 + 10 + 1 + 10 + 1 +
6060                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6061
6062         if (len > NFS4_OPAQUE_LIMIT + 1)
6063                 return -EINVAL;
6064
6065         /*
6066          * Since this string is allocated at mount time, and held until the
6067          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6068          * about a memory-reclaim deadlock.
6069          */
6070         str = kmalloc(len, GFP_KERNEL);
6071         if (!str)
6072                 return -ENOMEM;
6073
6074         scnprintf(str, len, "Linux NFSv%u.%u %s",
6075                         clp->rpc_ops->version, clp->cl_minorversion,
6076                         clp->cl_rpcclient->cl_nodename);
6077         clp->cl_owner_id = str;
6078         return 0;
6079 }
6080
6081 /*
6082  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6083  * services.  Advertise one based on the address family of the
6084  * clientaddr.
6085  */
6086 static unsigned int
6087 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6088 {
6089         if (strchr(clp->cl_ipaddr, ':') != NULL)
6090                 return scnprintf(buf, len, "tcp6");
6091         else
6092                 return scnprintf(buf, len, "tcp");
6093 }
6094
6095 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6096 {
6097         struct nfs4_setclientid *sc = calldata;
6098
6099         if (task->tk_status == 0)
6100                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6101 }
6102
6103 static const struct rpc_call_ops nfs4_setclientid_ops = {
6104         .rpc_call_done = nfs4_setclientid_done,
6105 };
6106
6107 /**
6108  * nfs4_proc_setclientid - Negotiate client ID
6109  * @clp: state data structure
6110  * @program: RPC program for NFSv4 callback service
6111  * @port: IP port number for NFS4 callback service
6112  * @cred: credential to use for this call
6113  * @res: where to place the result
6114  *
6115  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6116  */
6117 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6118                 unsigned short port, const struct cred *cred,
6119                 struct nfs4_setclientid_res *res)
6120 {
6121         nfs4_verifier sc_verifier;
6122         struct nfs4_setclientid setclientid = {
6123                 .sc_verifier = &sc_verifier,
6124                 .sc_prog = program,
6125                 .sc_clnt = clp,
6126         };
6127         struct rpc_message msg = {
6128                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6129                 .rpc_argp = &setclientid,
6130                 .rpc_resp = res,
6131                 .rpc_cred = cred,
6132         };
6133         struct rpc_task_setup task_setup_data = {
6134                 .rpc_client = clp->cl_rpcclient,
6135                 .rpc_message = &msg,
6136                 .callback_ops = &nfs4_setclientid_ops,
6137                 .callback_data = &setclientid,
6138                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6139         };
6140         unsigned long now = jiffies;
6141         int status;
6142
6143         /* nfs_client_id4 */
6144         nfs4_init_boot_verifier(clp, &sc_verifier);
6145
6146         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6147                 status = nfs4_init_uniform_client_string(clp);
6148         else
6149                 status = nfs4_init_nonuniform_client_string(clp);
6150
6151         if (status)
6152                 goto out;
6153
6154         /* cb_client4 */
6155         setclientid.sc_netid_len =
6156                                 nfs4_init_callback_netid(clp,
6157                                                 setclientid.sc_netid,
6158                                                 sizeof(setclientid.sc_netid));
6159         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6160                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6161                                 clp->cl_ipaddr, port >> 8, port & 255);
6162
6163         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6164                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6165                 clp->cl_owner_id);
6166
6167         status = nfs4_call_sync_custom(&task_setup_data);
6168         if (setclientid.sc_cred) {
6169                 kfree(clp->cl_acceptor);
6170                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6171                 put_rpccred(setclientid.sc_cred);
6172         }
6173
6174         if (status == 0)
6175                 do_renew_lease(clp, now);
6176 out:
6177         trace_nfs4_setclientid(clp, status);
6178         dprintk("NFS reply setclientid: %d\n", status);
6179         return status;
6180 }
6181
6182 /**
6183  * nfs4_proc_setclientid_confirm - Confirm client ID
6184  * @clp: state data structure
6185  * @arg: result of a previous SETCLIENTID
6186  * @cred: credential to use for this call
6187  *
6188  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6189  */
6190 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6191                 struct nfs4_setclientid_res *arg,
6192                 const struct cred *cred)
6193 {
6194         struct rpc_message msg = {
6195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6196                 .rpc_argp = arg,
6197                 .rpc_cred = cred,
6198         };
6199         int status;
6200
6201         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6202                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6203                 clp->cl_clientid);
6204         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6205                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6206         trace_nfs4_setclientid_confirm(clp, status);
6207         dprintk("NFS reply setclientid_confirm: %d\n", status);
6208         return status;
6209 }
6210
6211 struct nfs4_delegreturndata {
6212         struct nfs4_delegreturnargs args;
6213         struct nfs4_delegreturnres res;
6214         struct nfs_fh fh;
6215         nfs4_stateid stateid;
6216         unsigned long timestamp;
6217         struct {
6218                 struct nfs4_layoutreturn_args arg;
6219                 struct nfs4_layoutreturn_res res;
6220                 struct nfs4_xdr_opaque_data ld_private;
6221                 u32 roc_barrier;
6222                 bool roc;
6223         } lr;
6224         struct nfs_fattr fattr;
6225         int rpc_status;
6226         struct inode *inode;
6227 };
6228
6229 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6230 {
6231         struct nfs4_delegreturndata *data = calldata;
6232         struct nfs4_exception exception = {
6233                 .inode = data->inode,
6234                 .stateid = &data->stateid,
6235         };
6236
6237         if (!nfs4_sequence_done(task, &data->res.seq_res))
6238                 return;
6239
6240         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6241
6242         /* Handle Layoutreturn errors */
6243         if (pnfs_roc_done(task, data->inode,
6244                                 &data->args.lr_args,
6245                                 &data->res.lr_res,
6246                                 &data->res.lr_ret) == -EAGAIN)
6247                 goto out_restart;
6248
6249         switch (task->tk_status) {
6250         case 0:
6251                 renew_lease(data->res.server, data->timestamp);
6252                 break;
6253         case -NFS4ERR_ADMIN_REVOKED:
6254         case -NFS4ERR_DELEG_REVOKED:
6255         case -NFS4ERR_EXPIRED:
6256                 nfs4_free_revoked_stateid(data->res.server,
6257                                 data->args.stateid,
6258                                 task->tk_msg.rpc_cred);
6259                 /* Fallthrough */
6260         case -NFS4ERR_BAD_STATEID:
6261         case -NFS4ERR_STALE_STATEID:
6262                 task->tk_status = 0;
6263                 break;
6264         case -NFS4ERR_OLD_STATEID:
6265                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6266                         nfs4_stateid_seqid_inc(&data->stateid);
6267                 if (data->args.bitmask) {
6268                         data->args.bitmask = NULL;
6269                         data->res.fattr = NULL;
6270                 }
6271                 goto out_restart;
6272         case -NFS4ERR_ACCESS:
6273                 if (data->args.bitmask) {
6274                         data->args.bitmask = NULL;
6275                         data->res.fattr = NULL;
6276                         goto out_restart;
6277                 }
6278                 /* Fallthrough */
6279         default:
6280                 task->tk_status = nfs4_async_handle_exception(task,
6281                                 data->res.server, task->tk_status,
6282                                 &exception);
6283                 if (exception.retry)
6284                         goto out_restart;
6285         }
6286         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6287         data->rpc_status = task->tk_status;
6288         return;
6289 out_restart:
6290         task->tk_status = 0;
6291         rpc_restart_call_prepare(task);
6292 }
6293
6294 static void nfs4_delegreturn_release(void *calldata)
6295 {
6296         struct nfs4_delegreturndata *data = calldata;
6297         struct inode *inode = data->inode;
6298
6299         if (inode) {
6300                 if (data->lr.roc)
6301                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6302                                         data->res.lr_ret);
6303                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6304                 nfs_iput_and_deactive(inode);
6305         }
6306         kfree(calldata);
6307 }
6308
6309 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6310 {
6311         struct nfs4_delegreturndata *d_data;
6312         struct pnfs_layout_hdr *lo;
6313
6314         d_data = (struct nfs4_delegreturndata *)data;
6315
6316         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6317                 nfs4_sequence_done(task, &d_data->res.seq_res);
6318                 return;
6319         }
6320
6321         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6322         if (lo && !pnfs_layout_is_valid(lo)) {
6323                 d_data->args.lr_args = NULL;
6324                 d_data->res.lr_res = NULL;
6325         }
6326
6327         nfs4_setup_sequence(d_data->res.server->nfs_client,
6328                         &d_data->args.seq_args,
6329                         &d_data->res.seq_res,
6330                         task);
6331 }
6332
6333 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6334         .rpc_call_prepare = nfs4_delegreturn_prepare,
6335         .rpc_call_done = nfs4_delegreturn_done,
6336         .rpc_release = nfs4_delegreturn_release,
6337 };
6338
6339 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6340 {
6341         struct nfs4_delegreturndata *data;
6342         struct nfs_server *server = NFS_SERVER(inode);
6343         struct rpc_task *task;
6344         struct rpc_message msg = {
6345                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6346                 .rpc_cred = cred,
6347         };
6348         struct rpc_task_setup task_setup_data = {
6349                 .rpc_client = server->client,
6350                 .rpc_message = &msg,
6351                 .callback_ops = &nfs4_delegreturn_ops,
6352                 .flags = RPC_TASK_ASYNC,
6353         };
6354         int status = 0;
6355
6356         data = kzalloc(sizeof(*data), GFP_NOFS);
6357         if (data == NULL)
6358                 return -ENOMEM;
6359         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6360
6361         nfs4_state_protect(server->nfs_client,
6362                         NFS_SP4_MACH_CRED_CLEANUP,
6363                         &task_setup_data.rpc_client, &msg);
6364
6365         data->args.fhandle = &data->fh;
6366         data->args.stateid = &data->stateid;
6367         data->args.bitmask = server->cache_consistency_bitmask;
6368         nfs_copy_fh(&data->fh, NFS_FH(inode));
6369         nfs4_stateid_copy(&data->stateid, stateid);
6370         data->res.fattr = &data->fattr;
6371         data->res.server = server;
6372         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6373         data->lr.arg.ld_private = &data->lr.ld_private;
6374         nfs_fattr_init(data->res.fattr);
6375         data->timestamp = jiffies;
6376         data->rpc_status = 0;
6377         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6378         data->inode = nfs_igrab_and_active(inode);
6379         if (data->inode) {
6380                 if (data->lr.roc) {
6381                         data->args.lr_args = &data->lr.arg;
6382                         data->res.lr_res = &data->lr.res;
6383                 }
6384         } else if (data->lr.roc) {
6385                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6386                 data->lr.roc = false;
6387         }
6388
6389         task_setup_data.callback_data = data;
6390         msg.rpc_argp = &data->args;
6391         msg.rpc_resp = &data->res;
6392         task = rpc_run_task(&task_setup_data);
6393         if (IS_ERR(task))
6394                 return PTR_ERR(task);
6395         if (!issync)
6396                 goto out;
6397         status = rpc_wait_for_completion_task(task);
6398         if (status != 0)
6399                 goto out;
6400         status = data->rpc_status;
6401 out:
6402         rpc_put_task(task);
6403         return status;
6404 }
6405
6406 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6407 {
6408         struct nfs_server *server = NFS_SERVER(inode);
6409         struct nfs4_exception exception = { };
6410         int err;
6411         do {
6412                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6413                 trace_nfs4_delegreturn(inode, stateid, err);
6414                 switch (err) {
6415                         case -NFS4ERR_STALE_STATEID:
6416                         case -NFS4ERR_EXPIRED:
6417                         case 0:
6418                                 return 0;
6419                 }
6420                 err = nfs4_handle_exception(server, err, &exception);
6421         } while (exception.retry);
6422         return err;
6423 }
6424
6425 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6426 {
6427         struct inode *inode = state->inode;
6428         struct nfs_server *server = NFS_SERVER(inode);
6429         struct nfs_client *clp = server->nfs_client;
6430         struct nfs_lockt_args arg = {
6431                 .fh = NFS_FH(inode),
6432                 .fl = request,
6433         };
6434         struct nfs_lockt_res res = {
6435                 .denied = request,
6436         };
6437         struct rpc_message msg = {
6438                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6439                 .rpc_argp       = &arg,
6440                 .rpc_resp       = &res,
6441                 .rpc_cred       = state->owner->so_cred,
6442         };
6443         struct nfs4_lock_state *lsp;
6444         int status;
6445
6446         arg.lock_owner.clientid = clp->cl_clientid;
6447         status = nfs4_set_lock_state(state, request);
6448         if (status != 0)
6449                 goto out;
6450         lsp = request->fl_u.nfs4_fl.owner;
6451         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6452         arg.lock_owner.s_dev = server->s_dev;
6453         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6454         switch (status) {
6455                 case 0:
6456                         request->fl_type = F_UNLCK;
6457                         break;
6458                 case -NFS4ERR_DENIED:
6459                         status = 0;
6460         }
6461         request->fl_ops->fl_release_private(request);
6462         request->fl_ops = NULL;
6463 out:
6464         return status;
6465 }
6466
6467 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6468 {
6469         struct nfs4_exception exception = {
6470                 .interruptible = true,
6471         };
6472         int err;
6473
6474         do {
6475                 err = _nfs4_proc_getlk(state, cmd, request);
6476                 trace_nfs4_get_lock(request, state, cmd, err);
6477                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6478                                 &exception);
6479         } while (exception.retry);
6480         return err;
6481 }
6482
6483 /*
6484  * Update the seqid of a lock stateid after receiving
6485  * NFS4ERR_OLD_STATEID
6486  */
6487 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6488                 struct nfs4_lock_state *lsp)
6489 {
6490         struct nfs4_state *state = lsp->ls_state;
6491         bool ret = false;
6492
6493         spin_lock(&state->state_lock);
6494         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6495                 goto out;
6496         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6497                 nfs4_stateid_seqid_inc(dst);
6498         else
6499                 dst->seqid = lsp->ls_stateid.seqid;
6500         ret = true;
6501 out:
6502         spin_unlock(&state->state_lock);
6503         return ret;
6504 }
6505
6506 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6507                 struct nfs4_lock_state *lsp)
6508 {
6509         struct nfs4_state *state = lsp->ls_state;
6510         bool ret;
6511
6512         spin_lock(&state->state_lock);
6513         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6514         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6515         spin_unlock(&state->state_lock);
6516         return ret;
6517 }
6518
6519 struct nfs4_unlockdata {
6520         struct nfs_locku_args arg;
6521         struct nfs_locku_res res;
6522         struct nfs4_lock_state *lsp;
6523         struct nfs_open_context *ctx;
6524         struct nfs_lock_context *l_ctx;
6525         struct file_lock fl;
6526         struct nfs_server *server;
6527         unsigned long timestamp;
6528 };
6529
6530 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6531                 struct nfs_open_context *ctx,
6532                 struct nfs4_lock_state *lsp,
6533                 struct nfs_seqid *seqid)
6534 {
6535         struct nfs4_unlockdata *p;
6536         struct nfs4_state *state = lsp->ls_state;
6537         struct inode *inode = state->inode;
6538
6539         p = kzalloc(sizeof(*p), GFP_NOFS);
6540         if (p == NULL)
6541                 return NULL;
6542         p->arg.fh = NFS_FH(inode);
6543         p->arg.fl = &p->fl;
6544         p->arg.seqid = seqid;
6545         p->res.seqid = seqid;
6546         p->lsp = lsp;
6547         /* Ensure we don't close file until we're done freeing locks! */
6548         p->ctx = get_nfs_open_context(ctx);
6549         p->l_ctx = nfs_get_lock_context(ctx);
6550         locks_init_lock(&p->fl);
6551         locks_copy_lock(&p->fl, fl);
6552         p->server = NFS_SERVER(inode);
6553         spin_lock(&state->state_lock);
6554         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6555         spin_unlock(&state->state_lock);
6556         return p;
6557 }
6558
6559 static void nfs4_locku_release_calldata(void *data)
6560 {
6561         struct nfs4_unlockdata *calldata = data;
6562         nfs_free_seqid(calldata->arg.seqid);
6563         nfs4_put_lock_state(calldata->lsp);
6564         nfs_put_lock_context(calldata->l_ctx);
6565         put_nfs_open_context(calldata->ctx);
6566         kfree(calldata);
6567 }
6568
6569 static void nfs4_locku_done(struct rpc_task *task, void *data)
6570 {
6571         struct nfs4_unlockdata *calldata = data;
6572         struct nfs4_exception exception = {
6573                 .inode = calldata->lsp->ls_state->inode,
6574                 .stateid = &calldata->arg.stateid,
6575         };
6576
6577         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6578                 return;
6579         switch (task->tk_status) {
6580                 case 0:
6581                         renew_lease(calldata->server, calldata->timestamp);
6582                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6583                         if (nfs4_update_lock_stateid(calldata->lsp,
6584                                         &calldata->res.stateid))
6585                                 break;
6586                         /* Fall through */
6587                 case -NFS4ERR_ADMIN_REVOKED:
6588                 case -NFS4ERR_EXPIRED:
6589                         nfs4_free_revoked_stateid(calldata->server,
6590                                         &calldata->arg.stateid,
6591                                         task->tk_msg.rpc_cred);
6592                         /* Fall through */
6593                 case -NFS4ERR_BAD_STATEID:
6594                 case -NFS4ERR_STALE_STATEID:
6595                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6596                                                 calldata->lsp))
6597                                 rpc_restart_call_prepare(task);
6598                         break;
6599                 case -NFS4ERR_OLD_STATEID:
6600                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6601                                                 calldata->lsp))
6602                                 rpc_restart_call_prepare(task);
6603                         break;
6604                 default:
6605                         task->tk_status = nfs4_async_handle_exception(task,
6606                                         calldata->server, task->tk_status,
6607                                         &exception);
6608                         if (exception.retry)
6609                                 rpc_restart_call_prepare(task);
6610         }
6611         nfs_release_seqid(calldata->arg.seqid);
6612 }
6613
6614 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6615 {
6616         struct nfs4_unlockdata *calldata = data;
6617
6618         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6619                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6620                 return;
6621
6622         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6623                 goto out_wait;
6624         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6625                 /* Note: exit _without_ running nfs4_locku_done */
6626                 goto out_no_action;
6627         }
6628         calldata->timestamp = jiffies;
6629         if (nfs4_setup_sequence(calldata->server->nfs_client,
6630                                 &calldata->arg.seq_args,
6631                                 &calldata->res.seq_res,
6632                                 task) != 0)
6633                 nfs_release_seqid(calldata->arg.seqid);
6634         return;
6635 out_no_action:
6636         task->tk_action = NULL;
6637 out_wait:
6638         nfs4_sequence_done(task, &calldata->res.seq_res);
6639 }
6640
6641 static const struct rpc_call_ops nfs4_locku_ops = {
6642         .rpc_call_prepare = nfs4_locku_prepare,
6643         .rpc_call_done = nfs4_locku_done,
6644         .rpc_release = nfs4_locku_release_calldata,
6645 };
6646
6647 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6648                 struct nfs_open_context *ctx,
6649                 struct nfs4_lock_state *lsp,
6650                 struct nfs_seqid *seqid)
6651 {
6652         struct nfs4_unlockdata *data;
6653         struct rpc_message msg = {
6654                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6655                 .rpc_cred = ctx->cred,
6656         };
6657         struct rpc_task_setup task_setup_data = {
6658                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6659                 .rpc_message = &msg,
6660                 .callback_ops = &nfs4_locku_ops,
6661                 .workqueue = nfsiod_workqueue,
6662                 .flags = RPC_TASK_ASYNC,
6663         };
6664
6665         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6666                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6667
6668         /* Ensure this is an unlock - when canceling a lock, the
6669          * canceled lock is passed in, and it won't be an unlock.
6670          */
6671         fl->fl_type = F_UNLCK;
6672         if (fl->fl_flags & FL_CLOSE)
6673                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6674
6675         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6676         if (data == NULL) {
6677                 nfs_free_seqid(seqid);
6678                 return ERR_PTR(-ENOMEM);
6679         }
6680
6681         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6682         msg.rpc_argp = &data->arg;
6683         msg.rpc_resp = &data->res;
6684         task_setup_data.callback_data = data;
6685         return rpc_run_task(&task_setup_data);
6686 }
6687
6688 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6689 {
6690         struct inode *inode = state->inode;
6691         struct nfs4_state_owner *sp = state->owner;
6692         struct nfs_inode *nfsi = NFS_I(inode);
6693         struct nfs_seqid *seqid;
6694         struct nfs4_lock_state *lsp;
6695         struct rpc_task *task;
6696         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6697         int status = 0;
6698         unsigned char fl_flags = request->fl_flags;
6699
6700         status = nfs4_set_lock_state(state, request);
6701         /* Unlock _before_ we do the RPC call */
6702         request->fl_flags |= FL_EXISTS;
6703         /* Exclude nfs_delegation_claim_locks() */
6704         mutex_lock(&sp->so_delegreturn_mutex);
6705         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6706         down_read(&nfsi->rwsem);
6707         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6708                 up_read(&nfsi->rwsem);
6709                 mutex_unlock(&sp->so_delegreturn_mutex);
6710                 goto out;
6711         }
6712         up_read(&nfsi->rwsem);
6713         mutex_unlock(&sp->so_delegreturn_mutex);
6714         if (status != 0)
6715                 goto out;
6716         /* Is this a delegated lock? */
6717         lsp = request->fl_u.nfs4_fl.owner;
6718         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6719                 goto out;
6720         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6721         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6722         status = -ENOMEM;
6723         if (IS_ERR(seqid))
6724                 goto out;
6725         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6726         status = PTR_ERR(task);
6727         if (IS_ERR(task))
6728                 goto out;
6729         status = rpc_wait_for_completion_task(task);
6730         rpc_put_task(task);
6731 out:
6732         request->fl_flags = fl_flags;
6733         trace_nfs4_unlock(request, state, F_SETLK, status);
6734         return status;
6735 }
6736
6737 struct nfs4_lockdata {
6738         struct nfs_lock_args arg;
6739         struct nfs_lock_res res;
6740         struct nfs4_lock_state *lsp;
6741         struct nfs_open_context *ctx;
6742         struct file_lock fl;
6743         unsigned long timestamp;
6744         int rpc_status;
6745         int cancelled;
6746         struct nfs_server *server;
6747 };
6748
6749 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6750                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6751                 gfp_t gfp_mask)
6752 {
6753         struct nfs4_lockdata *p;
6754         struct inode *inode = lsp->ls_state->inode;
6755         struct nfs_server *server = NFS_SERVER(inode);
6756         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6757
6758         p = kzalloc(sizeof(*p), gfp_mask);
6759         if (p == NULL)
6760                 return NULL;
6761
6762         p->arg.fh = NFS_FH(inode);
6763         p->arg.fl = &p->fl;
6764         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6765         if (IS_ERR(p->arg.open_seqid))
6766                 goto out_free;
6767         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6768         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6769         if (IS_ERR(p->arg.lock_seqid))
6770                 goto out_free_seqid;
6771         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6772         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6773         p->arg.lock_owner.s_dev = server->s_dev;
6774         p->res.lock_seqid = p->arg.lock_seqid;
6775         p->lsp = lsp;
6776         p->server = server;
6777         p->ctx = get_nfs_open_context(ctx);
6778         locks_init_lock(&p->fl);
6779         locks_copy_lock(&p->fl, fl);
6780         return p;
6781 out_free_seqid:
6782         nfs_free_seqid(p->arg.open_seqid);
6783 out_free:
6784         kfree(p);
6785         return NULL;
6786 }
6787
6788 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6789 {
6790         struct nfs4_lockdata *data = calldata;
6791         struct nfs4_state *state = data->lsp->ls_state;
6792
6793         dprintk("%s: begin!\n", __func__);
6794         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6795                 goto out_wait;
6796         /* Do we need to do an open_to_lock_owner? */
6797         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6798                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6799                         goto out_release_lock_seqid;
6800                 }
6801                 nfs4_stateid_copy(&data->arg.open_stateid,
6802                                 &state->open_stateid);
6803                 data->arg.new_lock_owner = 1;
6804                 data->res.open_seqid = data->arg.open_seqid;
6805         } else {
6806                 data->arg.new_lock_owner = 0;
6807                 nfs4_stateid_copy(&data->arg.lock_stateid,
6808                                 &data->lsp->ls_stateid);
6809         }
6810         if (!nfs4_valid_open_stateid(state)) {
6811                 data->rpc_status = -EBADF;
6812                 task->tk_action = NULL;
6813                 goto out_release_open_seqid;
6814         }
6815         data->timestamp = jiffies;
6816         if (nfs4_setup_sequence(data->server->nfs_client,
6817                                 &data->arg.seq_args,
6818                                 &data->res.seq_res,
6819                                 task) == 0)
6820                 return;
6821 out_release_open_seqid:
6822         nfs_release_seqid(data->arg.open_seqid);
6823 out_release_lock_seqid:
6824         nfs_release_seqid(data->arg.lock_seqid);
6825 out_wait:
6826         nfs4_sequence_done(task, &data->res.seq_res);
6827         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6828 }
6829
6830 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6831 {
6832         struct nfs4_lockdata *data = calldata;
6833         struct nfs4_lock_state *lsp = data->lsp;
6834
6835         dprintk("%s: begin!\n", __func__);
6836
6837         if (!nfs4_sequence_done(task, &data->res.seq_res))
6838                 return;
6839
6840         data->rpc_status = task->tk_status;
6841         switch (task->tk_status) {
6842         case 0:
6843                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6844                                 data->timestamp);
6845                 if (data->arg.new_lock && !data->cancelled) {
6846                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6847                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6848                                 goto out_restart;
6849                 }
6850                 if (data->arg.new_lock_owner != 0) {
6851                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6852                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6853                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6854                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6855                         goto out_restart;
6856                 break;
6857         case -NFS4ERR_BAD_STATEID:
6858         case -NFS4ERR_OLD_STATEID:
6859         case -NFS4ERR_STALE_STATEID:
6860         case -NFS4ERR_EXPIRED:
6861                 if (data->arg.new_lock_owner != 0) {
6862                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6863                                                 &lsp->ls_state->open_stateid))
6864                                 goto out_restart;
6865                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6866                                                 &lsp->ls_stateid))
6867                                 goto out_restart;
6868         }
6869 out_done:
6870         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6871         return;
6872 out_restart:
6873         if (!data->cancelled)
6874                 rpc_restart_call_prepare(task);
6875         goto out_done;
6876 }
6877
6878 static void nfs4_lock_release(void *calldata)
6879 {
6880         struct nfs4_lockdata *data = calldata;
6881
6882         dprintk("%s: begin!\n", __func__);
6883         nfs_free_seqid(data->arg.open_seqid);
6884         if (data->cancelled && data->rpc_status == 0) {
6885                 struct rpc_task *task;
6886                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6887                                 data->arg.lock_seqid);
6888                 if (!IS_ERR(task))
6889                         rpc_put_task_async(task);
6890                 dprintk("%s: cancelling lock!\n", __func__);
6891         } else
6892                 nfs_free_seqid(data->arg.lock_seqid);
6893         nfs4_put_lock_state(data->lsp);
6894         put_nfs_open_context(data->ctx);
6895         kfree(data);
6896         dprintk("%s: done!\n", __func__);
6897 }
6898
6899 static const struct rpc_call_ops nfs4_lock_ops = {
6900         .rpc_call_prepare = nfs4_lock_prepare,
6901         .rpc_call_done = nfs4_lock_done,
6902         .rpc_release = nfs4_lock_release,
6903 };
6904
6905 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6906 {
6907         switch (error) {
6908         case -NFS4ERR_ADMIN_REVOKED:
6909         case -NFS4ERR_EXPIRED:
6910         case -NFS4ERR_BAD_STATEID:
6911                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6912                 if (new_lock_owner != 0 ||
6913                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6914                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6915                 break;
6916         case -NFS4ERR_STALE_STATEID:
6917                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6918                 nfs4_schedule_lease_recovery(server->nfs_client);
6919         }
6920 }
6921
6922 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6923 {
6924         struct nfs4_lockdata *data;
6925         struct rpc_task *task;
6926         struct rpc_message msg = {
6927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6928                 .rpc_cred = state->owner->so_cred,
6929         };
6930         struct rpc_task_setup task_setup_data = {
6931                 .rpc_client = NFS_CLIENT(state->inode),
6932                 .rpc_message = &msg,
6933                 .callback_ops = &nfs4_lock_ops,
6934                 .workqueue = nfsiod_workqueue,
6935                 .flags = RPC_TASK_ASYNC,
6936         };
6937         int ret;
6938
6939         dprintk("%s: begin!\n", __func__);
6940         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6941                         fl->fl_u.nfs4_fl.owner,
6942                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6943         if (data == NULL)
6944                 return -ENOMEM;
6945         if (IS_SETLKW(cmd))
6946                 data->arg.block = 1;
6947         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6948                                 recovery_type > NFS_LOCK_NEW);
6949         msg.rpc_argp = &data->arg;
6950         msg.rpc_resp = &data->res;
6951         task_setup_data.callback_data = data;
6952         if (recovery_type > NFS_LOCK_NEW) {
6953                 if (recovery_type == NFS_LOCK_RECLAIM)
6954                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6955         } else
6956                 data->arg.new_lock = 1;
6957         task = rpc_run_task(&task_setup_data);
6958         if (IS_ERR(task))
6959                 return PTR_ERR(task);
6960         ret = rpc_wait_for_completion_task(task);
6961         if (ret == 0) {
6962                 ret = data->rpc_status;
6963                 if (ret)
6964                         nfs4_handle_setlk_error(data->server, data->lsp,
6965                                         data->arg.new_lock_owner, ret);
6966         } else
6967                 data->cancelled = true;
6968         rpc_put_task(task);
6969         dprintk("%s: done, ret = %d!\n", __func__, ret);
6970         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6971         return ret;
6972 }
6973
6974 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6975 {
6976         struct nfs_server *server = NFS_SERVER(state->inode);
6977         struct nfs4_exception exception = {
6978                 .inode = state->inode,
6979         };
6980         int err;
6981
6982         do {
6983                 /* Cache the lock if possible... */
6984                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6985                         return 0;
6986                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6987                 if (err != -NFS4ERR_DELAY)
6988                         break;
6989                 nfs4_handle_exception(server, err, &exception);
6990         } while (exception.retry);
6991         return err;
6992 }
6993
6994 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6995 {
6996         struct nfs_server *server = NFS_SERVER(state->inode);
6997         struct nfs4_exception exception = {
6998                 .inode = state->inode,
6999         };
7000         int err;
7001
7002         err = nfs4_set_lock_state(state, request);
7003         if (err != 0)
7004                 return err;
7005         if (!recover_lost_locks) {
7006                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7007                 return 0;
7008         }
7009         do {
7010                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7011                         return 0;
7012                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7013                 switch (err) {
7014                 default:
7015                         goto out;
7016                 case -NFS4ERR_GRACE:
7017                 case -NFS4ERR_DELAY:
7018                         nfs4_handle_exception(server, err, &exception);
7019                         err = 0;
7020                 }
7021         } while (exception.retry);
7022 out:
7023         return err;
7024 }
7025
7026 #if defined(CONFIG_NFS_V4_1)
7027 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7028 {
7029         struct nfs4_lock_state *lsp;
7030         int status;
7031
7032         status = nfs4_set_lock_state(state, request);
7033         if (status != 0)
7034                 return status;
7035         lsp = request->fl_u.nfs4_fl.owner;
7036         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7037             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7038                 return 0;
7039         return nfs4_lock_expired(state, request);
7040 }
7041 #endif
7042
7043 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7044 {
7045         struct nfs_inode *nfsi = NFS_I(state->inode);
7046         struct nfs4_state_owner *sp = state->owner;
7047         unsigned char fl_flags = request->fl_flags;
7048         int status;
7049
7050         request->fl_flags |= FL_ACCESS;
7051         status = locks_lock_inode_wait(state->inode, request);
7052         if (status < 0)
7053                 goto out;
7054         mutex_lock(&sp->so_delegreturn_mutex);
7055         down_read(&nfsi->rwsem);
7056         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7057                 /* Yes: cache locks! */
7058                 /* ...but avoid races with delegation recall... */
7059                 request->fl_flags = fl_flags & ~FL_SLEEP;
7060                 status = locks_lock_inode_wait(state->inode, request);
7061                 up_read(&nfsi->rwsem);
7062                 mutex_unlock(&sp->so_delegreturn_mutex);
7063                 goto out;
7064         }
7065         up_read(&nfsi->rwsem);
7066         mutex_unlock(&sp->so_delegreturn_mutex);
7067         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7068 out:
7069         request->fl_flags = fl_flags;
7070         return status;
7071 }
7072
7073 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7074 {
7075         struct nfs4_exception exception = {
7076                 .state = state,
7077                 .inode = state->inode,
7078                 .interruptible = true,
7079         };
7080         int err;
7081
7082         do {
7083                 err = _nfs4_proc_setlk(state, cmd, request);
7084                 if (err == -NFS4ERR_DENIED)
7085                         err = -EAGAIN;
7086                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7087                                 err, &exception);
7088         } while (exception.retry);
7089         return err;
7090 }
7091
7092 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7093 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7094
7095 static int
7096 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7097                         struct file_lock *request)
7098 {
7099         int             status = -ERESTARTSYS;
7100         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7101
7102         while(!signalled()) {
7103                 status = nfs4_proc_setlk(state, cmd, request);
7104                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7105                         break;
7106                 freezable_schedule_timeout_interruptible(timeout);
7107                 timeout *= 2;
7108                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7109                 status = -ERESTARTSYS;
7110         }
7111         return status;
7112 }
7113
7114 #ifdef CONFIG_NFS_V4_1
7115 struct nfs4_lock_waiter {
7116         struct task_struct      *task;
7117         struct inode            *inode;
7118         struct nfs_lowner       *owner;
7119 };
7120
7121 static int
7122 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7123 {
7124         int ret;
7125         struct nfs4_lock_waiter *waiter = wait->private;
7126
7127         /* NULL key means to wake up everyone */
7128         if (key) {
7129                 struct cb_notify_lock_args      *cbnl = key;
7130                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7131                                                 *wowner = waiter->owner;
7132
7133                 /* Only wake if the callback was for the same owner. */
7134                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7135                         return 0;
7136
7137                 /* Make sure it's for the right inode */
7138                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7139                         return 0;
7140         }
7141
7142         /* override "private" so we can use default_wake_function */
7143         wait->private = waiter->task;
7144         ret = woken_wake_function(wait, mode, flags, key);
7145         if (ret)
7146                 list_del_init(&wait->entry);
7147         wait->private = waiter;
7148         return ret;
7149 }
7150
7151 static int
7152 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7153 {
7154         int status = -ERESTARTSYS;
7155         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7156         struct nfs_server *server = NFS_SERVER(state->inode);
7157         struct nfs_client *clp = server->nfs_client;
7158         wait_queue_head_t *q = &clp->cl_lock_waitq;
7159         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7160                                     .id = lsp->ls_seqid.owner_id,
7161                                     .s_dev = server->s_dev };
7162         struct nfs4_lock_waiter waiter = { .task  = current,
7163                                            .inode = state->inode,
7164                                            .owner = &owner};
7165         wait_queue_entry_t wait;
7166
7167         /* Don't bother with waitqueue if we don't expect a callback */
7168         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7169                 return nfs4_retry_setlk_simple(state, cmd, request);
7170
7171         init_wait(&wait);
7172         wait.private = &waiter;
7173         wait.func = nfs4_wake_lock_waiter;
7174
7175         while(!signalled()) {
7176                 add_wait_queue(q, &wait);
7177                 status = nfs4_proc_setlk(state, cmd, request);
7178                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7179                         finish_wait(q, &wait);
7180                         break;
7181                 }
7182
7183                 status = -ERESTARTSYS;
7184                 freezer_do_not_count();
7185                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7186                 freezer_count();
7187                 finish_wait(q, &wait);
7188         }
7189
7190         return status;
7191 }
7192 #else /* !CONFIG_NFS_V4_1 */
7193 static inline int
7194 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7195 {
7196         return nfs4_retry_setlk_simple(state, cmd, request);
7197 }
7198 #endif
7199
7200 static int
7201 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7202 {
7203         struct nfs_open_context *ctx;
7204         struct nfs4_state *state;
7205         int status;
7206
7207         /* verify open state */
7208         ctx = nfs_file_open_context(filp);
7209         state = ctx->state;
7210
7211         if (IS_GETLK(cmd)) {
7212                 if (state != NULL)
7213                         return nfs4_proc_getlk(state, F_GETLK, request);
7214                 return 0;
7215         }
7216
7217         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7218                 return -EINVAL;
7219
7220         if (request->fl_type == F_UNLCK) {
7221                 if (state != NULL)
7222                         return nfs4_proc_unlck(state, cmd, request);
7223                 return 0;
7224         }
7225
7226         if (state == NULL)
7227                 return -ENOLCK;
7228
7229         if ((request->fl_flags & FL_POSIX) &&
7230             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7231                 return -ENOLCK;
7232
7233         /*
7234          * Don't rely on the VFS having checked the file open mode,
7235          * since it won't do this for flock() locks.
7236          */
7237         switch (request->fl_type) {
7238         case F_RDLCK:
7239                 if (!(filp->f_mode & FMODE_READ))
7240                         return -EBADF;
7241                 break;
7242         case F_WRLCK:
7243                 if (!(filp->f_mode & FMODE_WRITE))
7244                         return -EBADF;
7245         }
7246
7247         status = nfs4_set_lock_state(state, request);
7248         if (status != 0)
7249                 return status;
7250
7251         return nfs4_retry_setlk(state, cmd, request);
7252 }
7253
7254 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7255 {
7256         struct nfs_server *server = NFS_SERVER(state->inode);
7257         int err;
7258
7259         err = nfs4_set_lock_state(state, fl);
7260         if (err != 0)
7261                 return err;
7262         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7263         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7264 }
7265
7266 struct nfs_release_lockowner_data {
7267         struct nfs4_lock_state *lsp;
7268         struct nfs_server *server;
7269         struct nfs_release_lockowner_args args;
7270         struct nfs_release_lockowner_res res;
7271         unsigned long timestamp;
7272 };
7273
7274 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7275 {
7276         struct nfs_release_lockowner_data *data = calldata;
7277         struct nfs_server *server = data->server;
7278         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7279                            &data->res.seq_res, task);
7280         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7281         data->timestamp = jiffies;
7282 }
7283
7284 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7285 {
7286         struct nfs_release_lockowner_data *data = calldata;
7287         struct nfs_server *server = data->server;
7288
7289         nfs40_sequence_done(task, &data->res.seq_res);
7290
7291         switch (task->tk_status) {
7292         case 0:
7293                 renew_lease(server, data->timestamp);
7294                 break;
7295         case -NFS4ERR_STALE_CLIENTID:
7296         case -NFS4ERR_EXPIRED:
7297                 nfs4_schedule_lease_recovery(server->nfs_client);
7298                 break;
7299         case -NFS4ERR_LEASE_MOVED:
7300         case -NFS4ERR_DELAY:
7301                 if (nfs4_async_handle_error(task, server,
7302                                             NULL, NULL) == -EAGAIN)
7303                         rpc_restart_call_prepare(task);
7304         }
7305 }
7306
7307 static void nfs4_release_lockowner_release(void *calldata)
7308 {
7309         struct nfs_release_lockowner_data *data = calldata;
7310         nfs4_free_lock_state(data->server, data->lsp);
7311         kfree(calldata);
7312 }
7313
7314 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7315         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7316         .rpc_call_done = nfs4_release_lockowner_done,
7317         .rpc_release = nfs4_release_lockowner_release,
7318 };
7319
7320 static void
7321 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7322 {
7323         struct nfs_release_lockowner_data *data;
7324         struct rpc_message msg = {
7325                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7326         };
7327
7328         if (server->nfs_client->cl_mvops->minor_version != 0)
7329                 return;
7330
7331         data = kmalloc(sizeof(*data), GFP_NOFS);
7332         if (!data)
7333                 return;
7334         data->lsp = lsp;
7335         data->server = server;
7336         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7337         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7338         data->args.lock_owner.s_dev = server->s_dev;
7339
7340         msg.rpc_argp = &data->args;
7341         msg.rpc_resp = &data->res;
7342         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7343         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7344 }
7345
7346 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7347
7348 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7349                                    struct dentry *unused, struct inode *inode,
7350                                    const char *key, const void *buf,
7351                                    size_t buflen, int flags)
7352 {
7353         return nfs4_proc_set_acl(inode, buf, buflen);
7354 }
7355
7356 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7357                                    struct dentry *unused, struct inode *inode,
7358                                    const char *key, void *buf, size_t buflen)
7359 {
7360         return nfs4_proc_get_acl(inode, buf, buflen);
7361 }
7362
7363 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7364 {
7365         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7366 }
7367
7368 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7369
7370 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7371                                      struct dentry *unused, struct inode *inode,
7372                                      const char *key, const void *buf,
7373                                      size_t buflen, int flags)
7374 {
7375         if (security_ismaclabel(key))
7376                 return nfs4_set_security_label(inode, buf, buflen);
7377
7378         return -EOPNOTSUPP;
7379 }
7380
7381 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7382                                      struct dentry *unused, struct inode *inode,
7383                                      const char *key, void *buf, size_t buflen)
7384 {
7385         if (security_ismaclabel(key))
7386                 return nfs4_get_security_label(inode, buf, buflen);
7387         return -EOPNOTSUPP;
7388 }
7389
7390 static ssize_t
7391 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7392 {
7393         int len = 0;
7394
7395         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7396                 len = security_inode_listsecurity(inode, list, list_len);
7397                 if (list_len && len > list_len)
7398                         return -ERANGE;
7399         }
7400         return len;
7401 }
7402
7403 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7404         .prefix = XATTR_SECURITY_PREFIX,
7405         .get    = nfs4_xattr_get_nfs4_label,
7406         .set    = nfs4_xattr_set_nfs4_label,
7407 };
7408
7409 #else
7410
7411 static ssize_t
7412 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7413 {
7414         return 0;
7415 }
7416
7417 #endif
7418
7419 /*
7420  * nfs_fhget will use either the mounted_on_fileid or the fileid
7421  */
7422 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7423 {
7424         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7425                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7426               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7427               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7428                 return;
7429
7430         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7431                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7432         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7433         fattr->nlink = 2;
7434 }
7435
7436 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7437                                    const struct qstr *name,
7438                                    struct nfs4_fs_locations *fs_locations,
7439                                    struct page *page)
7440 {
7441         struct nfs_server *server = NFS_SERVER(dir);
7442         u32 bitmask[3];
7443         struct nfs4_fs_locations_arg args = {
7444                 .dir_fh = NFS_FH(dir),
7445                 .name = name,
7446                 .page = page,
7447                 .bitmask = bitmask,
7448         };
7449         struct nfs4_fs_locations_res res = {
7450                 .fs_locations = fs_locations,
7451         };
7452         struct rpc_message msg = {
7453                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7454                 .rpc_argp = &args,
7455                 .rpc_resp = &res,
7456         };
7457         int status;
7458
7459         dprintk("%s: start\n", __func__);
7460
7461         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7462         bitmask[1] = nfs4_fattr_bitmap[1];
7463
7464         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7465          * is not supported */
7466         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7467                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7468         else
7469                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7470
7471         nfs_fattr_init(&fs_locations->fattr);
7472         fs_locations->server = server;
7473         fs_locations->nlocations = 0;
7474         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7475         dprintk("%s: returned status = %d\n", __func__, status);
7476         return status;
7477 }
7478
7479 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7480                            const struct qstr *name,
7481                            struct nfs4_fs_locations *fs_locations,
7482                            struct page *page)
7483 {
7484         struct nfs4_exception exception = {
7485                 .interruptible = true,
7486         };
7487         int err;
7488         do {
7489                 err = _nfs4_proc_fs_locations(client, dir, name,
7490                                 fs_locations, page);
7491                 trace_nfs4_get_fs_locations(dir, name, err);
7492                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7493                                 &exception);
7494         } while (exception.retry);
7495         return err;
7496 }
7497
7498 /*
7499  * This operation also signals the server that this client is
7500  * performing migration recovery.  The server can stop returning
7501  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7502  * appended to this compound to identify the client ID which is
7503  * performing recovery.
7504  */
7505 static int _nfs40_proc_get_locations(struct inode *inode,
7506                                      struct nfs4_fs_locations *locations,
7507                                      struct page *page, const struct cred *cred)
7508 {
7509         struct nfs_server *server = NFS_SERVER(inode);
7510         struct rpc_clnt *clnt = server->client;
7511         u32 bitmask[2] = {
7512                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7513         };
7514         struct nfs4_fs_locations_arg args = {
7515                 .clientid       = server->nfs_client->cl_clientid,
7516                 .fh             = NFS_FH(inode),
7517                 .page           = page,
7518                 .bitmask        = bitmask,
7519                 .migration      = 1,            /* skip LOOKUP */
7520                 .renew          = 1,            /* append RENEW */
7521         };
7522         struct nfs4_fs_locations_res res = {
7523                 .fs_locations   = locations,
7524                 .migration      = 1,
7525                 .renew          = 1,
7526         };
7527         struct rpc_message msg = {
7528                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7529                 .rpc_argp       = &args,
7530                 .rpc_resp       = &res,
7531                 .rpc_cred       = cred,
7532         };
7533         unsigned long now = jiffies;
7534         int status;
7535
7536         nfs_fattr_init(&locations->fattr);
7537         locations->server = server;
7538         locations->nlocations = 0;
7539
7540         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7541         status = nfs4_call_sync_sequence(clnt, server, &msg,
7542                                         &args.seq_args, &res.seq_res);
7543         if (status)
7544                 return status;
7545
7546         renew_lease(server, now);
7547         return 0;
7548 }
7549
7550 #ifdef CONFIG_NFS_V4_1
7551
7552 /*
7553  * This operation also signals the server that this client is
7554  * performing migration recovery.  The server can stop asserting
7555  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7556  * performing this operation is identified in the SEQUENCE
7557  * operation in this compound.
7558  *
7559  * When the client supports GETATTR(fs_locations_info), it can
7560  * be plumbed in here.
7561  */
7562 static int _nfs41_proc_get_locations(struct inode *inode,
7563                                      struct nfs4_fs_locations *locations,
7564                                      struct page *page, const struct cred *cred)
7565 {
7566         struct nfs_server *server = NFS_SERVER(inode);
7567         struct rpc_clnt *clnt = server->client;
7568         u32 bitmask[2] = {
7569                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7570         };
7571         struct nfs4_fs_locations_arg args = {
7572                 .fh             = NFS_FH(inode),
7573                 .page           = page,
7574                 .bitmask        = bitmask,
7575                 .migration      = 1,            /* skip LOOKUP */
7576         };
7577         struct nfs4_fs_locations_res res = {
7578                 .fs_locations   = locations,
7579                 .migration      = 1,
7580         };
7581         struct rpc_message msg = {
7582                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7583                 .rpc_argp       = &args,
7584                 .rpc_resp       = &res,
7585                 .rpc_cred       = cred,
7586         };
7587         int status;
7588
7589         nfs_fattr_init(&locations->fattr);
7590         locations->server = server;
7591         locations->nlocations = 0;
7592
7593         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7594         status = nfs4_call_sync_sequence(clnt, server, &msg,
7595                                         &args.seq_args, &res.seq_res);
7596         if (status == NFS4_OK &&
7597             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7598                 status = -NFS4ERR_LEASE_MOVED;
7599         return status;
7600 }
7601
7602 #endif  /* CONFIG_NFS_V4_1 */
7603
7604 /**
7605  * nfs4_proc_get_locations - discover locations for a migrated FSID
7606  * @inode: inode on FSID that is migrating
7607  * @locations: result of query
7608  * @page: buffer
7609  * @cred: credential to use for this operation
7610  *
7611  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7612  * operation failed, or a negative errno if a local error occurred.
7613  *
7614  * On success, "locations" is filled in, but if the server has
7615  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7616  * asserted.
7617  *
7618  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7619  * from this client that require migration recovery.
7620  */
7621 int nfs4_proc_get_locations(struct inode *inode,
7622                             struct nfs4_fs_locations *locations,
7623                             struct page *page, const struct cred *cred)
7624 {
7625         struct nfs_server *server = NFS_SERVER(inode);
7626         struct nfs_client *clp = server->nfs_client;
7627         const struct nfs4_mig_recovery_ops *ops =
7628                                         clp->cl_mvops->mig_recovery_ops;
7629         struct nfs4_exception exception = {
7630                 .interruptible = true,
7631         };
7632         int status;
7633
7634         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7635                 (unsigned long long)server->fsid.major,
7636                 (unsigned long long)server->fsid.minor,
7637                 clp->cl_hostname);
7638         nfs_display_fhandle(NFS_FH(inode), __func__);
7639
7640         do {
7641                 status = ops->get_locations(inode, locations, page, cred);
7642                 if (status != -NFS4ERR_DELAY)
7643                         break;
7644                 nfs4_handle_exception(server, status, &exception);
7645         } while (exception.retry);
7646         return status;
7647 }
7648
7649 /*
7650  * This operation also signals the server that this client is
7651  * performing "lease moved" recovery.  The server can stop
7652  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7653  * is appended to this compound to identify the client ID which is
7654  * performing recovery.
7655  */
7656 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7657 {
7658         struct nfs_server *server = NFS_SERVER(inode);
7659         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7660         struct rpc_clnt *clnt = server->client;
7661         struct nfs4_fsid_present_arg args = {
7662                 .fh             = NFS_FH(inode),
7663                 .clientid       = clp->cl_clientid,
7664                 .renew          = 1,            /* append RENEW */
7665         };
7666         struct nfs4_fsid_present_res res = {
7667                 .renew          = 1,
7668         };
7669         struct rpc_message msg = {
7670                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7671                 .rpc_argp       = &args,
7672                 .rpc_resp       = &res,
7673                 .rpc_cred       = cred,
7674         };
7675         unsigned long now = jiffies;
7676         int status;
7677
7678         res.fh = nfs_alloc_fhandle();
7679         if (res.fh == NULL)
7680                 return -ENOMEM;
7681
7682         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7683         status = nfs4_call_sync_sequence(clnt, server, &msg,
7684                                                 &args.seq_args, &res.seq_res);
7685         nfs_free_fhandle(res.fh);
7686         if (status)
7687                 return status;
7688
7689         do_renew_lease(clp, now);
7690         return 0;
7691 }
7692
7693 #ifdef CONFIG_NFS_V4_1
7694
7695 /*
7696  * This operation also signals the server that this client is
7697  * performing "lease moved" recovery.  The server can stop asserting
7698  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7699  * this operation is identified in the SEQUENCE operation in this
7700  * compound.
7701  */
7702 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7703 {
7704         struct nfs_server *server = NFS_SERVER(inode);
7705         struct rpc_clnt *clnt = server->client;
7706         struct nfs4_fsid_present_arg args = {
7707                 .fh             = NFS_FH(inode),
7708         };
7709         struct nfs4_fsid_present_res res = {
7710         };
7711         struct rpc_message msg = {
7712                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7713                 .rpc_argp       = &args,
7714                 .rpc_resp       = &res,
7715                 .rpc_cred       = cred,
7716         };
7717         int status;
7718
7719         res.fh = nfs_alloc_fhandle();
7720         if (res.fh == NULL)
7721                 return -ENOMEM;
7722
7723         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7724         status = nfs4_call_sync_sequence(clnt, server, &msg,
7725                                                 &args.seq_args, &res.seq_res);
7726         nfs_free_fhandle(res.fh);
7727         if (status == NFS4_OK &&
7728             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7729                 status = -NFS4ERR_LEASE_MOVED;
7730         return status;
7731 }
7732
7733 #endif  /* CONFIG_NFS_V4_1 */
7734
7735 /**
7736  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7737  * @inode: inode on FSID to check
7738  * @cred: credential to use for this operation
7739  *
7740  * Server indicates whether the FSID is present, moved, or not
7741  * recognized.  This operation is necessary to clear a LEASE_MOVED
7742  * condition for this client ID.
7743  *
7744  * Returns NFS4_OK if the FSID is present on this server,
7745  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7746  *  NFS4ERR code if some error occurred on the server, or a
7747  *  negative errno if a local failure occurred.
7748  */
7749 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7750 {
7751         struct nfs_server *server = NFS_SERVER(inode);
7752         struct nfs_client *clp = server->nfs_client;
7753         const struct nfs4_mig_recovery_ops *ops =
7754                                         clp->cl_mvops->mig_recovery_ops;
7755         struct nfs4_exception exception = {
7756                 .interruptible = true,
7757         };
7758         int status;
7759
7760         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7761                 (unsigned long long)server->fsid.major,
7762                 (unsigned long long)server->fsid.minor,
7763                 clp->cl_hostname);
7764         nfs_display_fhandle(NFS_FH(inode), __func__);
7765
7766         do {
7767                 status = ops->fsid_present(inode, cred);
7768                 if (status != -NFS4ERR_DELAY)
7769                         break;
7770                 nfs4_handle_exception(server, status, &exception);
7771         } while (exception.retry);
7772         return status;
7773 }
7774
7775 /*
7776  * If 'use_integrity' is true and the state managment nfs_client
7777  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7778  * and the machine credential as per RFC3530bis and RFC5661 Security
7779  * Considerations sections. Otherwise, just use the user cred with the
7780  * filesystem's rpc_client.
7781  */
7782 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7783 {
7784         int status;
7785         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7786         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7787         struct nfs4_secinfo_arg args = {
7788                 .dir_fh = NFS_FH(dir),
7789                 .name   = name,
7790         };
7791         struct nfs4_secinfo_res res = {
7792                 .flavors     = flavors,
7793         };
7794         struct rpc_message msg = {
7795                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7796                 .rpc_argp = &args,
7797                 .rpc_resp = &res,
7798         };
7799         struct nfs4_call_sync_data data = {
7800                 .seq_server = NFS_SERVER(dir),
7801                 .seq_args = &args.seq_args,
7802                 .seq_res = &res.seq_res,
7803         };
7804         struct rpc_task_setup task_setup = {
7805                 .rpc_client = clnt,
7806                 .rpc_message = &msg,
7807                 .callback_ops = clp->cl_mvops->call_sync_ops,
7808                 .callback_data = &data,
7809                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7810         };
7811         const struct cred *cred = NULL;
7812
7813         if (use_integrity) {
7814                 clnt = clp->cl_rpcclient;
7815                 task_setup.rpc_client = clnt;
7816
7817                 cred = nfs4_get_clid_cred(clp);
7818                 msg.rpc_cred = cred;
7819         }
7820
7821         dprintk("NFS call  secinfo %s\n", name->name);
7822
7823         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7824         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7825         status = nfs4_call_sync_custom(&task_setup);
7826
7827         dprintk("NFS reply  secinfo: %d\n", status);
7828
7829         put_cred(cred);
7830         return status;
7831 }
7832
7833 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7834                       struct nfs4_secinfo_flavors *flavors)
7835 {
7836         struct nfs4_exception exception = {
7837                 .interruptible = true,
7838         };
7839         int err;
7840         do {
7841                 err = -NFS4ERR_WRONGSEC;
7842
7843                 /* try to use integrity protection with machine cred */
7844                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7845                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7846
7847                 /*
7848                  * if unable to use integrity protection, or SECINFO with
7849                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7850                  * disallowed by spec, but exists in deployed servers) use
7851                  * the current filesystem's rpc_client and the user cred.
7852                  */
7853                 if (err == -NFS4ERR_WRONGSEC)
7854                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7855
7856                 trace_nfs4_secinfo(dir, name, err);
7857                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7858                                 &exception);
7859         } while (exception.retry);
7860         return err;
7861 }
7862
7863 #ifdef CONFIG_NFS_V4_1
7864 /*
7865  * Check the exchange flags returned by the server for invalid flags, having
7866  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7867  * DS flags set.
7868  */
7869 static int nfs4_check_cl_exchange_flags(u32 flags)
7870 {
7871         if (flags & ~EXCHGID4_FLAG_MASK_R)
7872                 goto out_inval;
7873         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7874             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7875                 goto out_inval;
7876         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7877                 goto out_inval;
7878         return NFS_OK;
7879 out_inval:
7880         return -NFS4ERR_INVAL;
7881 }
7882
7883 static bool
7884 nfs41_same_server_scope(struct nfs41_server_scope *a,
7885                         struct nfs41_server_scope *b)
7886 {
7887         if (a->server_scope_sz != b->server_scope_sz)
7888                 return false;
7889         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7890 }
7891
7892 static void
7893 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7894 {
7895         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
7896         struct nfs_client *clp = args->client;
7897
7898         switch (task->tk_status) {
7899         case -NFS4ERR_BADSESSION:
7900         case -NFS4ERR_DEADSESSION:
7901                 nfs4_schedule_session_recovery(clp->cl_session,
7902                                 task->tk_status);
7903         }
7904 }
7905
7906 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7907         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7908 };
7909
7910 /*
7911  * nfs4_proc_bind_one_conn_to_session()
7912  *
7913  * The 4.1 client currently uses the same TCP connection for the
7914  * fore and backchannel.
7915  */
7916 static
7917 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7918                 struct rpc_xprt *xprt,
7919                 struct nfs_client *clp,
7920                 const struct cred *cred)
7921 {
7922         int status;
7923         struct nfs41_bind_conn_to_session_args args = {
7924                 .client = clp,
7925                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7926         };
7927         struct nfs41_bind_conn_to_session_res res;
7928         struct rpc_message msg = {
7929                 .rpc_proc =
7930                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7931                 .rpc_argp = &args,
7932                 .rpc_resp = &res,
7933                 .rpc_cred = cred,
7934         };
7935         struct rpc_task_setup task_setup_data = {
7936                 .rpc_client = clnt,
7937                 .rpc_xprt = xprt,
7938                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7939                 .rpc_message = &msg,
7940                 .flags = RPC_TASK_TIMEOUT,
7941         };
7942         struct rpc_task *task;
7943
7944         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7945         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7946                 args.dir = NFS4_CDFC4_FORE;
7947
7948         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7949         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7950                 args.dir = NFS4_CDFC4_FORE;
7951
7952         task = rpc_run_task(&task_setup_data);
7953         if (!IS_ERR(task)) {
7954                 status = task->tk_status;
7955                 rpc_put_task(task);
7956         } else
7957                 status = PTR_ERR(task);
7958         trace_nfs4_bind_conn_to_session(clp, status);
7959         if (status == 0) {
7960                 if (memcmp(res.sessionid.data,
7961                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7962                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7963                         return -EIO;
7964                 }
7965                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7966                         dprintk("NFS: %s: Unexpected direction from server\n",
7967                                 __func__);
7968                         return -EIO;
7969                 }
7970                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7971                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7972                                 __func__);
7973                         return -EIO;
7974                 }
7975         }
7976
7977         return status;
7978 }
7979
7980 struct rpc_bind_conn_calldata {
7981         struct nfs_client *clp;
7982         const struct cred *cred;
7983 };
7984
7985 static int
7986 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7987                 struct rpc_xprt *xprt,
7988                 void *calldata)
7989 {
7990         struct rpc_bind_conn_calldata *p = calldata;
7991
7992         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7993 }
7994
7995 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7996 {
7997         struct rpc_bind_conn_calldata data = {
7998                 .clp = clp,
7999                 .cred = cred,
8000         };
8001         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8002                         nfs4_proc_bind_conn_to_session_callback, &data);
8003 }
8004
8005 /*
8006  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8007  * and operations we'd like to see to enable certain features in the allow map
8008  */
8009 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8010         .how = SP4_MACH_CRED,
8011         .enforce.u.words = {
8012                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8013                       1 << (OP_EXCHANGE_ID - 32) |
8014                       1 << (OP_CREATE_SESSION - 32) |
8015                       1 << (OP_DESTROY_SESSION - 32) |
8016                       1 << (OP_DESTROY_CLIENTID - 32)
8017         },
8018         .allow.u.words = {
8019                 [0] = 1 << (OP_CLOSE) |
8020                       1 << (OP_OPEN_DOWNGRADE) |
8021                       1 << (OP_LOCKU) |
8022                       1 << (OP_DELEGRETURN) |
8023                       1 << (OP_COMMIT),
8024                 [1] = 1 << (OP_SECINFO - 32) |
8025                       1 << (OP_SECINFO_NO_NAME - 32) |
8026                       1 << (OP_LAYOUTRETURN - 32) |
8027                       1 << (OP_TEST_STATEID - 32) |
8028                       1 << (OP_FREE_STATEID - 32) |
8029                       1 << (OP_WRITE - 32)
8030         }
8031 };
8032
8033 /*
8034  * Select the state protection mode for client `clp' given the server results
8035  * from exchange_id in `sp'.
8036  *
8037  * Returns 0 on success, negative errno otherwise.
8038  */
8039 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8040                                  struct nfs41_state_protection *sp)
8041 {
8042         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8043                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8044                       1 << (OP_EXCHANGE_ID - 32) |
8045                       1 << (OP_CREATE_SESSION - 32) |
8046                       1 << (OP_DESTROY_SESSION - 32) |
8047                       1 << (OP_DESTROY_CLIENTID - 32)
8048         };
8049         unsigned long flags = 0;
8050         unsigned int i;
8051         int ret = 0;
8052
8053         if (sp->how == SP4_MACH_CRED) {
8054                 /* Print state protect result */
8055                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8056                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8057                         if (test_bit(i, sp->enforce.u.longs))
8058                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8059                         if (test_bit(i, sp->allow.u.longs))
8060                                 dfprintk(MOUNT, "  allow op %d\n", i);
8061                 }
8062
8063                 /* make sure nothing is on enforce list that isn't supported */
8064                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8065                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8066                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8067                                 ret = -EINVAL;
8068                                 goto out;
8069                         }
8070                 }
8071
8072                 /*
8073                  * Minimal mode - state operations are allowed to use machine
8074                  * credential.  Note this already happens by default, so the
8075                  * client doesn't have to do anything more than the negotiation.
8076                  *
8077                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8078                  *       we're already using the machine cred for exchange_id
8079                  *       and will never use a different cred.
8080                  */
8081                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8082                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8083                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8084                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8085                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8086                         dfprintk(MOUNT, "  minimal mode enabled\n");
8087                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8088                 } else {
8089                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8090                         ret = -EINVAL;
8091                         goto out;
8092                 }
8093
8094                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8095                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8096                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8097                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8098                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8099                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8100                 }
8101
8102                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8103                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8104                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8105                 }
8106
8107                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8108                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8109                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8110                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8111                 }
8112
8113                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8114                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8115                         dfprintk(MOUNT, "  stateid mode enabled\n");
8116                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8117                 }
8118
8119                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8120                         dfprintk(MOUNT, "  write mode enabled\n");
8121                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8122                 }
8123
8124                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8125                         dfprintk(MOUNT, "  commit mode enabled\n");
8126                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8127                 }
8128         }
8129 out:
8130         clp->cl_sp4_flags = flags;
8131         return ret;
8132 }
8133
8134 struct nfs41_exchange_id_data {
8135         struct nfs41_exchange_id_res res;
8136         struct nfs41_exchange_id_args args;
8137 };
8138
8139 static void nfs4_exchange_id_release(void *data)
8140 {
8141         struct nfs41_exchange_id_data *cdata =
8142                                         (struct nfs41_exchange_id_data *)data;
8143
8144         nfs_put_client(cdata->args.client);
8145         kfree(cdata->res.impl_id);
8146         kfree(cdata->res.server_scope);
8147         kfree(cdata->res.server_owner);
8148         kfree(cdata);
8149 }
8150
8151 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8152         .rpc_release = nfs4_exchange_id_release,
8153 };
8154
8155 /*
8156  * _nfs4_proc_exchange_id()
8157  *
8158  * Wrapper for EXCHANGE_ID operation.
8159  */
8160 static struct rpc_task *
8161 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8162                         u32 sp4_how, struct rpc_xprt *xprt)
8163 {
8164         struct rpc_message msg = {
8165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8166                 .rpc_cred = cred,
8167         };
8168         struct rpc_task_setup task_setup_data = {
8169                 .rpc_client = clp->cl_rpcclient,
8170                 .callback_ops = &nfs4_exchange_id_call_ops,
8171                 .rpc_message = &msg,
8172                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8173         };
8174         struct nfs41_exchange_id_data *calldata;
8175         int status;
8176
8177         if (!refcount_inc_not_zero(&clp->cl_count))
8178                 return ERR_PTR(-EIO);
8179
8180         status = -ENOMEM;
8181         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8182         if (!calldata)
8183                 goto out;
8184
8185         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8186
8187         status = nfs4_init_uniform_client_string(clp);
8188         if (status)
8189                 goto out_calldata;
8190
8191         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8192                                                 GFP_NOFS);
8193         status = -ENOMEM;
8194         if (unlikely(calldata->res.server_owner == NULL))
8195                 goto out_calldata;
8196
8197         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8198                                         GFP_NOFS);
8199         if (unlikely(calldata->res.server_scope == NULL))
8200                 goto out_server_owner;
8201
8202         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8203         if (unlikely(calldata->res.impl_id == NULL))
8204                 goto out_server_scope;
8205
8206         switch (sp4_how) {
8207         case SP4_NONE:
8208                 calldata->args.state_protect.how = SP4_NONE;
8209                 break;
8210
8211         case SP4_MACH_CRED:
8212                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8213                 break;
8214
8215         default:
8216                 /* unsupported! */
8217                 WARN_ON_ONCE(1);
8218                 status = -EINVAL;
8219                 goto out_impl_id;
8220         }
8221         if (xprt) {
8222                 task_setup_data.rpc_xprt = xprt;
8223                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8224                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8225                                 sizeof(calldata->args.verifier.data));
8226         }
8227         calldata->args.client = clp;
8228         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8229         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8230 #ifdef CONFIG_NFS_V4_1_MIGRATION
8231         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8232 #endif
8233         msg.rpc_argp = &calldata->args;
8234         msg.rpc_resp = &calldata->res;
8235         task_setup_data.callback_data = calldata;
8236
8237         return rpc_run_task(&task_setup_data);
8238
8239 out_impl_id:
8240         kfree(calldata->res.impl_id);
8241 out_server_scope:
8242         kfree(calldata->res.server_scope);
8243 out_server_owner:
8244         kfree(calldata->res.server_owner);
8245 out_calldata:
8246         kfree(calldata);
8247 out:
8248         nfs_put_client(clp);
8249         return ERR_PTR(status);
8250 }
8251
8252 /*
8253  * _nfs4_proc_exchange_id()
8254  *
8255  * Wrapper for EXCHANGE_ID operation.
8256  */
8257 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8258                         u32 sp4_how)
8259 {
8260         struct rpc_task *task;
8261         struct nfs41_exchange_id_args *argp;
8262         struct nfs41_exchange_id_res *resp;
8263         unsigned long now = jiffies;
8264         int status;
8265
8266         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8267         if (IS_ERR(task))
8268                 return PTR_ERR(task);
8269
8270         argp = task->tk_msg.rpc_argp;
8271         resp = task->tk_msg.rpc_resp;
8272         status = task->tk_status;
8273         if (status  != 0)
8274                 goto out;
8275
8276         status = nfs4_check_cl_exchange_flags(resp->flags);
8277         if (status  != 0)
8278                 goto out;
8279
8280         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8281         if (status != 0)
8282                 goto out;
8283
8284         do_renew_lease(clp, now);
8285
8286         clp->cl_clientid = resp->clientid;
8287         clp->cl_exchange_flags = resp->flags;
8288         clp->cl_seqid = resp->seqid;
8289         /* Client ID is not confirmed */
8290         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8291                 clear_bit(NFS4_SESSION_ESTABLISHED,
8292                           &clp->cl_session->session_state);
8293
8294         if (clp->cl_serverscope != NULL &&
8295             !nfs41_same_server_scope(clp->cl_serverscope,
8296                                 resp->server_scope)) {
8297                 dprintk("%s: server_scope mismatch detected\n",
8298                         __func__);
8299                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8300         }
8301
8302         swap(clp->cl_serverowner, resp->server_owner);
8303         swap(clp->cl_serverscope, resp->server_scope);
8304         swap(clp->cl_implid, resp->impl_id);
8305
8306         /* Save the EXCHANGE_ID verifier session trunk tests */
8307         memcpy(clp->cl_confirm.data, argp->verifier.data,
8308                sizeof(clp->cl_confirm.data));
8309 out:
8310         trace_nfs4_exchange_id(clp, status);
8311         rpc_put_task(task);
8312         return status;
8313 }
8314
8315 /*
8316  * nfs4_proc_exchange_id()
8317  *
8318  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8319  *
8320  * Since the clientid has expired, all compounds using sessions
8321  * associated with the stale clientid will be returning
8322  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8323  * be in some phase of session reset.
8324  *
8325  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8326  */
8327 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8328 {
8329         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8330         int status;
8331
8332         /* try SP4_MACH_CRED if krb5i/p */
8333         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8334             authflavor == RPC_AUTH_GSS_KRB5P) {
8335                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8336                 if (!status)
8337                         return 0;
8338         }
8339
8340         /* try SP4_NONE */
8341         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8342 }
8343
8344 /**
8345  * nfs4_test_session_trunk
8346  *
8347  * This is an add_xprt_test() test function called from
8348  * rpc_clnt_setup_test_and_add_xprt.
8349  *
8350  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8351  * and is dereferrenced in nfs4_exchange_id_release
8352  *
8353  * Upon success, add the new transport to the rpc_clnt
8354  *
8355  * @clnt: struct rpc_clnt to get new transport
8356  * @xprt: the rpc_xprt to test
8357  * @data: call data for _nfs4_proc_exchange_id.
8358  */
8359 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8360                             void *data)
8361 {
8362         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8363         struct rpc_task *task;
8364         int status;
8365
8366         u32 sp4_how;
8367
8368         dprintk("--> %s try %s\n", __func__,
8369                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8370
8371         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8372
8373         /* Test connection for session trunking. Async exchange_id call */
8374         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8375         if (IS_ERR(task))
8376                 return;
8377
8378         status = task->tk_status;
8379         if (status == 0)
8380                 status = nfs4_detect_session_trunking(adata->clp,
8381                                 task->tk_msg.rpc_resp, xprt);
8382
8383         if (status == 0)
8384                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8385
8386         rpc_put_task(task);
8387 }
8388 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8389
8390 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8391                 const struct cred *cred)
8392 {
8393         struct rpc_message msg = {
8394                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8395                 .rpc_argp = clp,
8396                 .rpc_cred = cred,
8397         };
8398         int status;
8399
8400         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8401                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8402         trace_nfs4_destroy_clientid(clp, status);
8403         if (status)
8404                 dprintk("NFS: Got error %d from the server %s on "
8405                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8406         return status;
8407 }
8408
8409 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8410                 const struct cred *cred)
8411 {
8412         unsigned int loop;
8413         int ret;
8414
8415         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8416                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8417                 switch (ret) {
8418                 case -NFS4ERR_DELAY:
8419                 case -NFS4ERR_CLIENTID_BUSY:
8420                         ssleep(1);
8421                         break;
8422                 default:
8423                         return ret;
8424                 }
8425         }
8426         return 0;
8427 }
8428
8429 int nfs4_destroy_clientid(struct nfs_client *clp)
8430 {
8431         const struct cred *cred;
8432         int ret = 0;
8433
8434         if (clp->cl_mvops->minor_version < 1)
8435                 goto out;
8436         if (clp->cl_exchange_flags == 0)
8437                 goto out;
8438         if (clp->cl_preserve_clid)
8439                 goto out;
8440         cred = nfs4_get_clid_cred(clp);
8441         ret = nfs4_proc_destroy_clientid(clp, cred);
8442         put_cred(cred);
8443         switch (ret) {
8444         case 0:
8445         case -NFS4ERR_STALE_CLIENTID:
8446                 clp->cl_exchange_flags = 0;
8447         }
8448 out:
8449         return ret;
8450 }
8451
8452 #endif /* CONFIG_NFS_V4_1 */
8453
8454 struct nfs4_get_lease_time_data {
8455         struct nfs4_get_lease_time_args *args;
8456         struct nfs4_get_lease_time_res *res;
8457         struct nfs_client *clp;
8458 };
8459
8460 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8461                                         void *calldata)
8462 {
8463         struct nfs4_get_lease_time_data *data =
8464                         (struct nfs4_get_lease_time_data *)calldata;
8465
8466         dprintk("--> %s\n", __func__);
8467         /* just setup sequence, do not trigger session recovery
8468            since we're invoked within one */
8469         nfs4_setup_sequence(data->clp,
8470                         &data->args->la_seq_args,
8471                         &data->res->lr_seq_res,
8472                         task);
8473         dprintk("<-- %s\n", __func__);
8474 }
8475
8476 /*
8477  * Called from nfs4_state_manager thread for session setup, so don't recover
8478  * from sequence operation or clientid errors.
8479  */
8480 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8481 {
8482         struct nfs4_get_lease_time_data *data =
8483                         (struct nfs4_get_lease_time_data *)calldata;
8484
8485         dprintk("--> %s\n", __func__);
8486         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8487                 return;
8488         switch (task->tk_status) {
8489         case -NFS4ERR_DELAY:
8490         case -NFS4ERR_GRACE:
8491                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8492                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8493                 task->tk_status = 0;
8494                 /* fall through */
8495         case -NFS4ERR_RETRY_UNCACHED_REP:
8496                 rpc_restart_call_prepare(task);
8497                 return;
8498         }
8499         dprintk("<-- %s\n", __func__);
8500 }
8501
8502 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8503         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8504         .rpc_call_done = nfs4_get_lease_time_done,
8505 };
8506
8507 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8508 {
8509         struct nfs4_get_lease_time_args args;
8510         struct nfs4_get_lease_time_res res = {
8511                 .lr_fsinfo = fsinfo,
8512         };
8513         struct nfs4_get_lease_time_data data = {
8514                 .args = &args,
8515                 .res = &res,
8516                 .clp = clp,
8517         };
8518         struct rpc_message msg = {
8519                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8520                 .rpc_argp = &args,
8521                 .rpc_resp = &res,
8522         };
8523         struct rpc_task_setup task_setup = {
8524                 .rpc_client = clp->cl_rpcclient,
8525                 .rpc_message = &msg,
8526                 .callback_ops = &nfs4_get_lease_time_ops,
8527                 .callback_data = &data,
8528                 .flags = RPC_TASK_TIMEOUT,
8529         };
8530
8531         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8532         return nfs4_call_sync_custom(&task_setup);
8533 }
8534
8535 #ifdef CONFIG_NFS_V4_1
8536
8537 /*
8538  * Initialize the values to be used by the client in CREATE_SESSION
8539  * If nfs4_init_session set the fore channel request and response sizes,
8540  * use them.
8541  *
8542  * Set the back channel max_resp_sz_cached to zero to force the client to
8543  * always set csa_cachethis to FALSE because the current implementation
8544  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8545  */
8546 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8547                                     struct rpc_clnt *clnt)
8548 {
8549         unsigned int max_rqst_sz, max_resp_sz;
8550         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8551         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8552
8553         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8554         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8555
8556         /* Fore channel attributes */
8557         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8558         args->fc_attrs.max_resp_sz = max_resp_sz;
8559         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8560         args->fc_attrs.max_reqs = max_session_slots;
8561
8562         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8563                 "max_ops=%u max_reqs=%u\n",
8564                 __func__,
8565                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8566                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8567
8568         /* Back channel attributes */
8569         args->bc_attrs.max_rqst_sz = max_bc_payload;
8570         args->bc_attrs.max_resp_sz = max_bc_payload;
8571         args->bc_attrs.max_resp_sz_cached = 0;
8572         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8573         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8574         if (args->bc_attrs.max_reqs > max_bc_slots)
8575                 args->bc_attrs.max_reqs = max_bc_slots;
8576
8577         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8578                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8579                 __func__,
8580                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8581                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8582                 args->bc_attrs.max_reqs);
8583 }
8584
8585 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8586                 struct nfs41_create_session_res *res)
8587 {
8588         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8589         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8590
8591         if (rcvd->max_resp_sz > sent->max_resp_sz)
8592                 return -EINVAL;
8593         /*
8594          * Our requested max_ops is the minimum we need; we're not
8595          * prepared to break up compounds into smaller pieces than that.
8596          * So, no point even trying to continue if the server won't
8597          * cooperate:
8598          */
8599         if (rcvd->max_ops < sent->max_ops)
8600                 return -EINVAL;
8601         if (rcvd->max_reqs == 0)
8602                 return -EINVAL;
8603         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8604                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8605         return 0;
8606 }
8607
8608 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8609                 struct nfs41_create_session_res *res)
8610 {
8611         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8612         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8613
8614         if (!(res->flags & SESSION4_BACK_CHAN))
8615                 goto out;
8616         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8617                 return -EINVAL;
8618         if (rcvd->max_resp_sz < sent->max_resp_sz)
8619                 return -EINVAL;
8620         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8621                 return -EINVAL;
8622         if (rcvd->max_ops > sent->max_ops)
8623                 return -EINVAL;
8624         if (rcvd->max_reqs > sent->max_reqs)
8625                 return -EINVAL;
8626 out:
8627         return 0;
8628 }
8629
8630 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8631                                      struct nfs41_create_session_res *res)
8632 {
8633         int ret;
8634
8635         ret = nfs4_verify_fore_channel_attrs(args, res);
8636         if (ret)
8637                 return ret;
8638         return nfs4_verify_back_channel_attrs(args, res);
8639 }
8640
8641 static void nfs4_update_session(struct nfs4_session *session,
8642                 struct nfs41_create_session_res *res)
8643 {
8644         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8645         /* Mark client id and session as being confirmed */
8646         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8647         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8648         session->flags = res->flags;
8649         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8650         if (res->flags & SESSION4_BACK_CHAN)
8651                 memcpy(&session->bc_attrs, &res->bc_attrs,
8652                                 sizeof(session->bc_attrs));
8653 }
8654
8655 static int _nfs4_proc_create_session(struct nfs_client *clp,
8656                 const struct cred *cred)
8657 {
8658         struct nfs4_session *session = clp->cl_session;
8659         struct nfs41_create_session_args args = {
8660                 .client = clp,
8661                 .clientid = clp->cl_clientid,
8662                 .seqid = clp->cl_seqid,
8663                 .cb_program = NFS4_CALLBACK,
8664         };
8665         struct nfs41_create_session_res res;
8666
8667         struct rpc_message msg = {
8668                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8669                 .rpc_argp = &args,
8670                 .rpc_resp = &res,
8671                 .rpc_cred = cred,
8672         };
8673         int status;
8674
8675         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8676         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8677
8678         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8679                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8680         trace_nfs4_create_session(clp, status);
8681
8682         switch (status) {
8683         case -NFS4ERR_STALE_CLIENTID:
8684         case -NFS4ERR_DELAY:
8685         case -ETIMEDOUT:
8686         case -EACCES:
8687         case -EAGAIN:
8688                 goto out;
8689         }
8690
8691         clp->cl_seqid++;
8692         if (!status) {
8693                 /* Verify the session's negotiated channel_attrs values */
8694                 status = nfs4_verify_channel_attrs(&args, &res);
8695                 /* Increment the clientid slot sequence id */
8696                 if (status)
8697                         goto out;
8698                 nfs4_update_session(session, &res);
8699         }
8700 out:
8701         return status;
8702 }
8703
8704 /*
8705  * Issues a CREATE_SESSION operation to the server.
8706  * It is the responsibility of the caller to verify the session is
8707  * expired before calling this routine.
8708  */
8709 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8710 {
8711         int status;
8712         unsigned *ptr;
8713         struct nfs4_session *session = clp->cl_session;
8714
8715         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8716
8717         status = _nfs4_proc_create_session(clp, cred);
8718         if (status)
8719                 goto out;
8720
8721         /* Init or reset the session slot tables */
8722         status = nfs4_setup_session_slot_tables(session);
8723         dprintk("slot table setup returned %d\n", status);
8724         if (status)
8725                 goto out;
8726
8727         ptr = (unsigned *)&session->sess_id.data[0];
8728         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8729                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8730 out:
8731         dprintk("<-- %s\n", __func__);
8732         return status;
8733 }
8734
8735 /*
8736  * Issue the over-the-wire RPC DESTROY_SESSION.
8737  * The caller must serialize access to this routine.
8738  */
8739 int nfs4_proc_destroy_session(struct nfs4_session *session,
8740                 const struct cred *cred)
8741 {
8742         struct rpc_message msg = {
8743                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8744                 .rpc_argp = session,
8745                 .rpc_cred = cred,
8746         };
8747         int status = 0;
8748
8749         dprintk("--> nfs4_proc_destroy_session\n");
8750
8751         /* session is still being setup */
8752         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8753                 return 0;
8754
8755         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8756                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8757         trace_nfs4_destroy_session(session->clp, status);
8758
8759         if (status)
8760                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8761                         "Session has been destroyed regardless...\n", status);
8762
8763         dprintk("<-- nfs4_proc_destroy_session\n");
8764         return status;
8765 }
8766
8767 /*
8768  * Renew the cl_session lease.
8769  */
8770 struct nfs4_sequence_data {
8771         struct nfs_client *clp;
8772         struct nfs4_sequence_args args;
8773         struct nfs4_sequence_res res;
8774 };
8775
8776 static void nfs41_sequence_release(void *data)
8777 {
8778         struct nfs4_sequence_data *calldata = data;
8779         struct nfs_client *clp = calldata->clp;
8780
8781         if (refcount_read(&clp->cl_count) > 1)
8782                 nfs4_schedule_state_renewal(clp);
8783         nfs_put_client(clp);
8784         kfree(calldata);
8785 }
8786
8787 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8788 {
8789         switch(task->tk_status) {
8790         case -NFS4ERR_DELAY:
8791                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8792                 return -EAGAIN;
8793         default:
8794                 nfs4_schedule_lease_recovery(clp);
8795         }
8796         return 0;
8797 }
8798
8799 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8800 {
8801         struct nfs4_sequence_data *calldata = data;
8802         struct nfs_client *clp = calldata->clp;
8803
8804         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8805                 return;
8806
8807         trace_nfs4_sequence(clp, task->tk_status);
8808         if (task->tk_status < 0) {
8809                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8810                 if (refcount_read(&clp->cl_count) == 1)
8811                         goto out;
8812
8813                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8814                         rpc_restart_call_prepare(task);
8815                         return;
8816                 }
8817         }
8818         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8819 out:
8820         dprintk("<-- %s\n", __func__);
8821 }
8822
8823 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8824 {
8825         struct nfs4_sequence_data *calldata = data;
8826         struct nfs_client *clp = calldata->clp;
8827         struct nfs4_sequence_args *args;
8828         struct nfs4_sequence_res *res;
8829
8830         args = task->tk_msg.rpc_argp;
8831         res = task->tk_msg.rpc_resp;
8832
8833         nfs4_setup_sequence(clp, args, res, task);
8834 }
8835
8836 static const struct rpc_call_ops nfs41_sequence_ops = {
8837         .rpc_call_done = nfs41_sequence_call_done,
8838         .rpc_call_prepare = nfs41_sequence_prepare,
8839         .rpc_release = nfs41_sequence_release,
8840 };
8841
8842 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8843                 const struct cred *cred,
8844                 struct nfs4_slot *slot,
8845                 bool is_privileged)
8846 {
8847         struct nfs4_sequence_data *calldata;
8848         struct rpc_message msg = {
8849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8850                 .rpc_cred = cred,
8851         };
8852         struct rpc_task_setup task_setup_data = {
8853                 .rpc_client = clp->cl_rpcclient,
8854                 .rpc_message = &msg,
8855                 .callback_ops = &nfs41_sequence_ops,
8856                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8857         };
8858         struct rpc_task *ret;
8859
8860         ret = ERR_PTR(-EIO);
8861         if (!refcount_inc_not_zero(&clp->cl_count))
8862                 goto out_err;
8863
8864         ret = ERR_PTR(-ENOMEM);
8865         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8866         if (calldata == NULL)
8867                 goto out_put_clp;
8868         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8869         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8870         msg.rpc_argp = &calldata->args;
8871         msg.rpc_resp = &calldata->res;
8872         calldata->clp = clp;
8873         task_setup_data.callback_data = calldata;
8874
8875         ret = rpc_run_task(&task_setup_data);
8876         if (IS_ERR(ret))
8877                 goto out_err;
8878         return ret;
8879 out_put_clp:
8880         nfs_put_client(clp);
8881 out_err:
8882         nfs41_release_slot(slot);
8883         return ret;
8884 }
8885
8886 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8887 {
8888         struct rpc_task *task;
8889         int ret = 0;
8890
8891         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8892                 return -EAGAIN;
8893         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8894         if (IS_ERR(task))
8895                 ret = PTR_ERR(task);
8896         else
8897                 rpc_put_task_async(task);
8898         dprintk("<-- %s status=%d\n", __func__, ret);
8899         return ret;
8900 }
8901
8902 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8903 {
8904         struct rpc_task *task;
8905         int ret;
8906
8907         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8908         if (IS_ERR(task)) {
8909                 ret = PTR_ERR(task);
8910                 goto out;
8911         }
8912         ret = rpc_wait_for_completion_task(task);
8913         if (!ret)
8914                 ret = task->tk_status;
8915         rpc_put_task(task);
8916 out:
8917         dprintk("<-- %s status=%d\n", __func__, ret);
8918         return ret;
8919 }
8920
8921 struct nfs4_reclaim_complete_data {
8922         struct nfs_client *clp;
8923         struct nfs41_reclaim_complete_args arg;
8924         struct nfs41_reclaim_complete_res res;
8925 };
8926
8927 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8928 {
8929         struct nfs4_reclaim_complete_data *calldata = data;
8930
8931         nfs4_setup_sequence(calldata->clp,
8932                         &calldata->arg.seq_args,
8933                         &calldata->res.seq_res,
8934                         task);
8935 }
8936
8937 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8938 {
8939         switch(task->tk_status) {
8940         case 0:
8941                 wake_up_all(&clp->cl_lock_waitq);
8942                 /* Fallthrough */
8943         case -NFS4ERR_COMPLETE_ALREADY:
8944         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8945                 break;
8946         case -NFS4ERR_DELAY:
8947                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8948                 /* fall through */
8949         case -NFS4ERR_RETRY_UNCACHED_REP:
8950                 return -EAGAIN;
8951         case -NFS4ERR_BADSESSION:
8952         case -NFS4ERR_DEADSESSION:
8953         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8954                 break;
8955         default:
8956                 nfs4_schedule_lease_recovery(clp);
8957         }
8958         return 0;
8959 }
8960
8961 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8962 {
8963         struct nfs4_reclaim_complete_data *calldata = data;
8964         struct nfs_client *clp = calldata->clp;
8965         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8966
8967         dprintk("--> %s\n", __func__);
8968         if (!nfs41_sequence_done(task, res))
8969                 return;
8970
8971         trace_nfs4_reclaim_complete(clp, task->tk_status);
8972         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8973                 rpc_restart_call_prepare(task);
8974                 return;
8975         }
8976         dprintk("<-- %s\n", __func__);
8977 }
8978
8979 static void nfs4_free_reclaim_complete_data(void *data)
8980 {
8981         struct nfs4_reclaim_complete_data *calldata = data;
8982
8983         kfree(calldata);
8984 }
8985
8986 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8987         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8988         .rpc_call_done = nfs4_reclaim_complete_done,
8989         .rpc_release = nfs4_free_reclaim_complete_data,
8990 };
8991
8992 /*
8993  * Issue a global reclaim complete.
8994  */
8995 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8996                 const struct cred *cred)
8997 {
8998         struct nfs4_reclaim_complete_data *calldata;
8999         struct rpc_message msg = {
9000                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9001                 .rpc_cred = cred,
9002         };
9003         struct rpc_task_setup task_setup_data = {
9004                 .rpc_client = clp->cl_rpcclient,
9005                 .rpc_message = &msg,
9006                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9007                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9008         };
9009         int status = -ENOMEM;
9010
9011         dprintk("--> %s\n", __func__);
9012         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9013         if (calldata == NULL)
9014                 goto out;
9015         calldata->clp = clp;
9016         calldata->arg.one_fs = 0;
9017
9018         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9019         msg.rpc_argp = &calldata->arg;
9020         msg.rpc_resp = &calldata->res;
9021         task_setup_data.callback_data = calldata;
9022         status = nfs4_call_sync_custom(&task_setup_data);
9023 out:
9024         dprintk("<-- %s status=%d\n", __func__, status);
9025         return status;
9026 }
9027
9028 static void
9029 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9030 {
9031         struct nfs4_layoutget *lgp = calldata;
9032         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9033
9034         dprintk("--> %s\n", __func__);
9035         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9036                                 &lgp->res.seq_res, task);
9037         dprintk("<-- %s\n", __func__);
9038 }
9039
9040 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9041 {
9042         struct nfs4_layoutget *lgp = calldata;
9043
9044         dprintk("--> %s\n", __func__);
9045         nfs41_sequence_process(task, &lgp->res.seq_res);
9046         dprintk("<-- %s\n", __func__);
9047 }
9048
9049 static int
9050 nfs4_layoutget_handle_exception(struct rpc_task *task,
9051                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9052 {
9053         struct inode *inode = lgp->args.inode;
9054         struct nfs_server *server = NFS_SERVER(inode);
9055         struct pnfs_layout_hdr *lo;
9056         int nfs4err = task->tk_status;
9057         int err, status = 0;
9058         LIST_HEAD(head);
9059
9060         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9061
9062         nfs4_sequence_free_slot(&lgp->res.seq_res);
9063
9064         switch (nfs4err) {
9065         case 0:
9066                 goto out;
9067
9068         /*
9069          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9070          * on the file. set tk_status to -ENODATA to tell upper layer to
9071          * retry go inband.
9072          */
9073         case -NFS4ERR_LAYOUTUNAVAILABLE:
9074                 status = -ENODATA;
9075                 goto out;
9076         /*
9077          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9078          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9079          */
9080         case -NFS4ERR_BADLAYOUT:
9081                 status = -EOVERFLOW;
9082                 goto out;
9083         /*
9084          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9085          * (or clients) writing to the same RAID stripe except when
9086          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9087          *
9088          * Treat it like we would RECALLCONFLICT -- we retry for a little
9089          * while, and then eventually give up.
9090          */
9091         case -NFS4ERR_LAYOUTTRYLATER:
9092                 if (lgp->args.minlength == 0) {
9093                         status = -EOVERFLOW;
9094                         goto out;
9095                 }
9096                 status = -EBUSY;
9097                 break;
9098         case -NFS4ERR_RECALLCONFLICT:
9099                 status = -ERECALLCONFLICT;
9100                 break;
9101         case -NFS4ERR_DELEG_REVOKED:
9102         case -NFS4ERR_ADMIN_REVOKED:
9103         case -NFS4ERR_EXPIRED:
9104         case -NFS4ERR_BAD_STATEID:
9105                 exception->timeout = 0;
9106                 spin_lock(&inode->i_lock);
9107                 lo = NFS_I(inode)->layout;
9108                 /* If the open stateid was bad, then recover it. */
9109                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9110                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9111                         spin_unlock(&inode->i_lock);
9112                         exception->state = lgp->args.ctx->state;
9113                         exception->stateid = &lgp->args.stateid;
9114                         break;
9115                 }
9116
9117                 /*
9118                  * Mark the bad layout state as invalid, then retry
9119                  */
9120                 pnfs_mark_layout_stateid_invalid(lo, &head);
9121                 spin_unlock(&inode->i_lock);
9122                 nfs_commit_inode(inode, 0);
9123                 pnfs_free_lseg_list(&head);
9124                 status = -EAGAIN;
9125                 goto out;
9126         }
9127
9128         err = nfs4_handle_exception(server, nfs4err, exception);
9129         if (!status) {
9130                 if (exception->retry)
9131                         status = -EAGAIN;
9132                 else
9133                         status = err;
9134         }
9135 out:
9136         dprintk("<-- %s\n", __func__);
9137         return status;
9138 }
9139
9140 size_t max_response_pages(struct nfs_server *server)
9141 {
9142         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9143         return nfs_page_array_len(0, max_resp_sz);
9144 }
9145
9146 static void nfs4_layoutget_release(void *calldata)
9147 {
9148         struct nfs4_layoutget *lgp = calldata;
9149
9150         dprintk("--> %s\n", __func__);
9151         nfs4_sequence_free_slot(&lgp->res.seq_res);
9152         pnfs_layoutget_free(lgp);
9153         dprintk("<-- %s\n", __func__);
9154 }
9155
9156 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9157         .rpc_call_prepare = nfs4_layoutget_prepare,
9158         .rpc_call_done = nfs4_layoutget_done,
9159         .rpc_release = nfs4_layoutget_release,
9160 };
9161
9162 struct pnfs_layout_segment *
9163 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9164 {
9165         struct inode *inode = lgp->args.inode;
9166         struct nfs_server *server = NFS_SERVER(inode);
9167         struct rpc_task *task;
9168         struct rpc_message msg = {
9169                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9170                 .rpc_argp = &lgp->args,
9171                 .rpc_resp = &lgp->res,
9172                 .rpc_cred = lgp->cred,
9173         };
9174         struct rpc_task_setup task_setup_data = {
9175                 .rpc_client = server->client,
9176                 .rpc_message = &msg,
9177                 .callback_ops = &nfs4_layoutget_call_ops,
9178                 .callback_data = lgp,
9179                 .flags = RPC_TASK_ASYNC,
9180         };
9181         struct pnfs_layout_segment *lseg = NULL;
9182         struct nfs4_exception exception = {
9183                 .inode = inode,
9184                 .timeout = *timeout,
9185         };
9186         int status = 0;
9187
9188         dprintk("--> %s\n", __func__);
9189
9190         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9191         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9192
9193         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9194
9195         task = rpc_run_task(&task_setup_data);
9196         if (IS_ERR(task))
9197                 return ERR_CAST(task);
9198         status = rpc_wait_for_completion_task(task);
9199         if (status != 0)
9200                 goto out;
9201
9202         if (task->tk_status < 0) {
9203                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9204                 *timeout = exception.timeout;
9205         } else if (lgp->res.layoutp->len == 0) {
9206                 status = -EAGAIN;
9207                 *timeout = nfs4_update_delay(&exception.timeout);
9208         } else
9209                 lseg = pnfs_layout_process(lgp);
9210 out:
9211         trace_nfs4_layoutget(lgp->args.ctx,
9212                         &lgp->args.range,
9213                         &lgp->res.range,
9214                         &lgp->res.stateid,
9215                         status);
9216
9217         rpc_put_task(task);
9218         dprintk("<-- %s status=%d\n", __func__, status);
9219         if (status)
9220                 return ERR_PTR(status);
9221         return lseg;
9222 }
9223
9224 static void
9225 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9226 {
9227         struct nfs4_layoutreturn *lrp = calldata;
9228
9229         dprintk("--> %s\n", __func__);
9230         nfs4_setup_sequence(lrp->clp,
9231                         &lrp->args.seq_args,
9232                         &lrp->res.seq_res,
9233                         task);
9234         if (!pnfs_layout_is_valid(lrp->args.layout))
9235                 rpc_exit(task, 0);
9236 }
9237
9238 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9239 {
9240         struct nfs4_layoutreturn *lrp = calldata;
9241         struct nfs_server *server;
9242
9243         dprintk("--> %s\n", __func__);
9244
9245         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9246                 return;
9247
9248         /*
9249          * Was there an RPC level error? Assume the call succeeded,
9250          * and that we need to release the layout
9251          */
9252         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9253                 lrp->res.lrs_present = 0;
9254                 return;
9255         }
9256
9257         server = NFS_SERVER(lrp->args.inode);
9258         switch (task->tk_status) {
9259         case -NFS4ERR_OLD_STATEID:
9260                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9261                                         &lrp->args.range,
9262                                         lrp->args.inode))
9263                         goto out_restart;
9264                 /* Fallthrough */
9265         default:
9266                 task->tk_status = 0;
9267                 /* Fallthrough */
9268         case 0:
9269                 break;
9270         case -NFS4ERR_DELAY:
9271                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9272                         break;
9273                 goto out_restart;
9274         }
9275         dprintk("<-- %s\n", __func__);
9276         return;
9277 out_restart:
9278         task->tk_status = 0;
9279         nfs4_sequence_free_slot(&lrp->res.seq_res);
9280         rpc_restart_call_prepare(task);
9281 }
9282
9283 static void nfs4_layoutreturn_release(void *calldata)
9284 {
9285         struct nfs4_layoutreturn *lrp = calldata;
9286         struct pnfs_layout_hdr *lo = lrp->args.layout;
9287
9288         dprintk("--> %s\n", __func__);
9289         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9290                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9291         nfs4_sequence_free_slot(&lrp->res.seq_res);
9292         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9293                 lrp->ld_private.ops->free(&lrp->ld_private);
9294         pnfs_put_layout_hdr(lrp->args.layout);
9295         nfs_iput_and_deactive(lrp->inode);
9296         kfree(calldata);
9297         dprintk("<-- %s\n", __func__);
9298 }
9299
9300 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9301         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9302         .rpc_call_done = nfs4_layoutreturn_done,
9303         .rpc_release = nfs4_layoutreturn_release,
9304 };
9305
9306 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9307 {
9308         struct rpc_task *task;
9309         struct rpc_message msg = {
9310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9311                 .rpc_argp = &lrp->args,
9312                 .rpc_resp = &lrp->res,
9313                 .rpc_cred = lrp->cred,
9314         };
9315         struct rpc_task_setup task_setup_data = {
9316                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9317                 .rpc_message = &msg,
9318                 .callback_ops = &nfs4_layoutreturn_call_ops,
9319                 .callback_data = lrp,
9320         };
9321         int status = 0;
9322
9323         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9324                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9325                         &task_setup_data.rpc_client, &msg);
9326
9327         dprintk("--> %s\n", __func__);
9328         if (!sync) {
9329                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9330                 if (!lrp->inode) {
9331                         nfs4_layoutreturn_release(lrp);
9332                         return -EAGAIN;
9333                 }
9334                 task_setup_data.flags |= RPC_TASK_ASYNC;
9335         }
9336         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9337         task = rpc_run_task(&task_setup_data);
9338         if (IS_ERR(task))
9339                 return PTR_ERR(task);
9340         if (sync)
9341                 status = task->tk_status;
9342         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9343         dprintk("<-- %s status=%d\n", __func__, status);
9344         rpc_put_task(task);
9345         return status;
9346 }
9347
9348 static int
9349 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9350                 struct pnfs_device *pdev,
9351                 const struct cred *cred)
9352 {
9353         struct nfs4_getdeviceinfo_args args = {
9354                 .pdev = pdev,
9355                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9356                         NOTIFY_DEVICEID4_DELETE,
9357         };
9358         struct nfs4_getdeviceinfo_res res = {
9359                 .pdev = pdev,
9360         };
9361         struct rpc_message msg = {
9362                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9363                 .rpc_argp = &args,
9364                 .rpc_resp = &res,
9365                 .rpc_cred = cred,
9366         };
9367         int status;
9368
9369         dprintk("--> %s\n", __func__);
9370         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9371         if (res.notification & ~args.notify_types)
9372                 dprintk("%s: unsupported notification\n", __func__);
9373         if (res.notification != args.notify_types)
9374                 pdev->nocache = 1;
9375
9376         dprintk("<-- %s status=%d\n", __func__, status);
9377
9378         return status;
9379 }
9380
9381 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9382                 struct pnfs_device *pdev,
9383                 const struct cred *cred)
9384 {
9385         struct nfs4_exception exception = { };
9386         int err;
9387
9388         do {
9389                 err = nfs4_handle_exception(server,
9390                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9391                                         &exception);
9392         } while (exception.retry);
9393         return err;
9394 }
9395 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9396
9397 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9398 {
9399         struct nfs4_layoutcommit_data *data = calldata;
9400         struct nfs_server *server = NFS_SERVER(data->args.inode);
9401
9402         nfs4_setup_sequence(server->nfs_client,
9403                         &data->args.seq_args,
9404                         &data->res.seq_res,
9405                         task);
9406 }
9407
9408 static void
9409 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9410 {
9411         struct nfs4_layoutcommit_data *data = calldata;
9412         struct nfs_server *server = NFS_SERVER(data->args.inode);
9413
9414         if (!nfs41_sequence_done(task, &data->res.seq_res))
9415                 return;
9416
9417         switch (task->tk_status) { /* Just ignore these failures */
9418         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9419         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9420         case -NFS4ERR_BADLAYOUT:     /* no layout */
9421         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9422                 task->tk_status = 0;
9423         case 0:
9424                 break;
9425         default:
9426                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9427                         rpc_restart_call_prepare(task);
9428                         return;
9429                 }
9430         }
9431 }
9432
9433 static void nfs4_layoutcommit_release(void *calldata)
9434 {
9435         struct nfs4_layoutcommit_data *data = calldata;
9436
9437         pnfs_cleanup_layoutcommit(data);
9438         nfs_post_op_update_inode_force_wcc(data->args.inode,
9439                                            data->res.fattr);
9440         put_cred(data->cred);
9441         nfs_iput_and_deactive(data->inode);
9442         kfree(data);
9443 }
9444
9445 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9446         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9447         .rpc_call_done = nfs4_layoutcommit_done,
9448         .rpc_release = nfs4_layoutcommit_release,
9449 };
9450
9451 int
9452 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9453 {
9454         struct rpc_message msg = {
9455                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9456                 .rpc_argp = &data->args,
9457                 .rpc_resp = &data->res,
9458                 .rpc_cred = data->cred,
9459         };
9460         struct rpc_task_setup task_setup_data = {
9461                 .task = &data->task,
9462                 .rpc_client = NFS_CLIENT(data->args.inode),
9463                 .rpc_message = &msg,
9464                 .callback_ops = &nfs4_layoutcommit_ops,
9465                 .callback_data = data,
9466         };
9467         struct rpc_task *task;
9468         int status = 0;
9469
9470         dprintk("NFS: initiating layoutcommit call. sync %d "
9471                 "lbw: %llu inode %lu\n", sync,
9472                 data->args.lastbytewritten,
9473                 data->args.inode->i_ino);
9474
9475         if (!sync) {
9476                 data->inode = nfs_igrab_and_active(data->args.inode);
9477                 if (data->inode == NULL) {
9478                         nfs4_layoutcommit_release(data);
9479                         return -EAGAIN;
9480                 }
9481                 task_setup_data.flags = RPC_TASK_ASYNC;
9482         }
9483         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9484         task = rpc_run_task(&task_setup_data);
9485         if (IS_ERR(task))
9486                 return PTR_ERR(task);
9487         if (sync)
9488                 status = task->tk_status;
9489         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9490         dprintk("%s: status %d\n", __func__, status);
9491         rpc_put_task(task);
9492         return status;
9493 }
9494
9495 /*
9496  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9497  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9498  */
9499 static int
9500 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9501                     struct nfs_fsinfo *info,
9502                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9503 {
9504         struct nfs41_secinfo_no_name_args args = {
9505                 .style = SECINFO_STYLE_CURRENT_FH,
9506         };
9507         struct nfs4_secinfo_res res = {
9508                 .flavors = flavors,
9509         };
9510         struct rpc_message msg = {
9511                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9512                 .rpc_argp = &args,
9513                 .rpc_resp = &res,
9514         };
9515         struct rpc_clnt *clnt = server->client;
9516         struct nfs4_call_sync_data data = {
9517                 .seq_server = server,
9518                 .seq_args = &args.seq_args,
9519                 .seq_res = &res.seq_res,
9520         };
9521         struct rpc_task_setup task_setup = {
9522                 .rpc_client = server->client,
9523                 .rpc_message = &msg,
9524                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9525                 .callback_data = &data,
9526                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9527         };
9528         const struct cred *cred = NULL;
9529         int status;
9530
9531         if (use_integrity) {
9532                 clnt = server->nfs_client->cl_rpcclient;
9533                 task_setup.rpc_client = clnt;
9534
9535                 cred = nfs4_get_clid_cred(server->nfs_client);
9536                 msg.rpc_cred = cred;
9537         }
9538
9539         dprintk("--> %s\n", __func__);
9540         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9541         status = nfs4_call_sync_custom(&task_setup);
9542         dprintk("<-- %s status=%d\n", __func__, status);
9543
9544         put_cred(cred);
9545
9546         return status;
9547 }
9548
9549 static int
9550 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9551                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9552 {
9553         struct nfs4_exception exception = {
9554                 .interruptible = true,
9555         };
9556         int err;
9557         do {
9558                 /* first try using integrity protection */
9559                 err = -NFS4ERR_WRONGSEC;
9560
9561                 /* try to use integrity protection with machine cred */
9562                 if (_nfs4_is_integrity_protected(server->nfs_client))
9563                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9564                                                           flavors, true);
9565
9566                 /*
9567                  * if unable to use integrity protection, or SECINFO with
9568                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9569                  * disallowed by spec, but exists in deployed servers) use
9570                  * the current filesystem's rpc_client and the user cred.
9571                  */
9572                 if (err == -NFS4ERR_WRONGSEC)
9573                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9574                                                           flavors, false);
9575
9576                 switch (err) {
9577                 case 0:
9578                 case -NFS4ERR_WRONGSEC:
9579                 case -ENOTSUPP:
9580                         goto out;
9581                 default:
9582                         err = nfs4_handle_exception(server, err, &exception);
9583                 }
9584         } while (exception.retry);
9585 out:
9586         return err;
9587 }
9588
9589 static int
9590 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9591                     struct nfs_fsinfo *info)
9592 {
9593         int err;
9594         struct page *page;
9595         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9596         struct nfs4_secinfo_flavors *flavors;
9597         struct nfs4_secinfo4 *secinfo;
9598         int i;
9599
9600         page = alloc_page(GFP_KERNEL);
9601         if (!page) {
9602                 err = -ENOMEM;
9603                 goto out;
9604         }
9605
9606         flavors = page_address(page);
9607         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9608
9609         /*
9610          * Fall back on "guess and check" method if
9611          * the server doesn't support SECINFO_NO_NAME
9612          */
9613         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9614                 err = nfs4_find_root_sec(server, fhandle, info);
9615                 goto out_freepage;
9616         }
9617         if (err)
9618                 goto out_freepage;
9619
9620         for (i = 0; i < flavors->num_flavors; i++) {
9621                 secinfo = &flavors->flavors[i];
9622
9623                 switch (secinfo->flavor) {
9624                 case RPC_AUTH_NULL:
9625                 case RPC_AUTH_UNIX:
9626                 case RPC_AUTH_GSS:
9627                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9628                                         &secinfo->flavor_info);
9629                         break;
9630                 default:
9631                         flavor = RPC_AUTH_MAXFLAVOR;
9632                         break;
9633                 }
9634
9635                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9636                         flavor = RPC_AUTH_MAXFLAVOR;
9637
9638                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9639                         err = nfs4_lookup_root_sec(server, fhandle,
9640                                                    info, flavor);
9641                         if (!err)
9642                                 break;
9643                 }
9644         }
9645
9646         if (flavor == RPC_AUTH_MAXFLAVOR)
9647                 err = -EPERM;
9648
9649 out_freepage:
9650         put_page(page);
9651         if (err == -EACCES)
9652                 return -EPERM;
9653 out:
9654         return err;
9655 }
9656
9657 static int _nfs41_test_stateid(struct nfs_server *server,
9658                 nfs4_stateid *stateid,
9659                 const struct cred *cred)
9660 {
9661         int status;
9662         struct nfs41_test_stateid_args args = {
9663                 .stateid = stateid,
9664         };
9665         struct nfs41_test_stateid_res res;
9666         struct rpc_message msg = {
9667                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9668                 .rpc_argp = &args,
9669                 .rpc_resp = &res,
9670                 .rpc_cred = cred,
9671         };
9672         struct rpc_clnt *rpc_client = server->client;
9673
9674         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9675                 &rpc_client, &msg);
9676
9677         dprintk("NFS call  test_stateid %p\n", stateid);
9678         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9679         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9680                         &args.seq_args, &res.seq_res);
9681         if (status != NFS_OK) {
9682                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9683                 return status;
9684         }
9685         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9686         return -res.status;
9687 }
9688
9689 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9690                 int err, struct nfs4_exception *exception)
9691 {
9692         exception->retry = 0;
9693         switch(err) {
9694         case -NFS4ERR_DELAY:
9695         case -NFS4ERR_RETRY_UNCACHED_REP:
9696                 nfs4_handle_exception(server, err, exception);
9697                 break;
9698         case -NFS4ERR_BADSESSION:
9699         case -NFS4ERR_BADSLOT:
9700         case -NFS4ERR_BAD_HIGH_SLOT:
9701         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9702         case -NFS4ERR_DEADSESSION:
9703                 nfs4_do_handle_exception(server, err, exception);
9704         }
9705 }
9706
9707 /**
9708  * nfs41_test_stateid - perform a TEST_STATEID operation
9709  *
9710  * @server: server / transport on which to perform the operation
9711  * @stateid: state ID to test
9712  * @cred: credential
9713  *
9714  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9715  * Otherwise a negative NFS4ERR value is returned if the operation
9716  * failed or the state ID is not currently valid.
9717  */
9718 static int nfs41_test_stateid(struct nfs_server *server,
9719                 nfs4_stateid *stateid,
9720                 const struct cred *cred)
9721 {
9722         struct nfs4_exception exception = {
9723                 .interruptible = true,
9724         };
9725         int err;
9726         do {
9727                 err = _nfs41_test_stateid(server, stateid, cred);
9728                 nfs4_handle_delay_or_session_error(server, err, &exception);
9729         } while (exception.retry);
9730         return err;
9731 }
9732
9733 struct nfs_free_stateid_data {
9734         struct nfs_server *server;
9735         struct nfs41_free_stateid_args args;
9736         struct nfs41_free_stateid_res res;
9737 };
9738
9739 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9740 {
9741         struct nfs_free_stateid_data *data = calldata;
9742         nfs4_setup_sequence(data->server->nfs_client,
9743                         &data->args.seq_args,
9744                         &data->res.seq_res,
9745                         task);
9746 }
9747
9748 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9749 {
9750         struct nfs_free_stateid_data *data = calldata;
9751
9752         nfs41_sequence_done(task, &data->res.seq_res);
9753
9754         switch (task->tk_status) {
9755         case -NFS4ERR_DELAY:
9756                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9757                         rpc_restart_call_prepare(task);
9758         }
9759 }
9760
9761 static void nfs41_free_stateid_release(void *calldata)
9762 {
9763         kfree(calldata);
9764 }
9765
9766 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9767         .rpc_call_prepare = nfs41_free_stateid_prepare,
9768         .rpc_call_done = nfs41_free_stateid_done,
9769         .rpc_release = nfs41_free_stateid_release,
9770 };
9771
9772 /**
9773  * nfs41_free_stateid - perform a FREE_STATEID operation
9774  *
9775  * @server: server / transport on which to perform the operation
9776  * @stateid: state ID to release
9777  * @cred: credential
9778  * @privileged: set to true if this call needs to be privileged
9779  *
9780  * Note: this function is always asynchronous.
9781  */
9782 static int nfs41_free_stateid(struct nfs_server *server,
9783                 const nfs4_stateid *stateid,
9784                 const struct cred *cred,
9785                 bool privileged)
9786 {
9787         struct rpc_message msg = {
9788                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9789                 .rpc_cred = cred,
9790         };
9791         struct rpc_task_setup task_setup = {
9792                 .rpc_client = server->client,
9793                 .rpc_message = &msg,
9794                 .callback_ops = &nfs41_free_stateid_ops,
9795                 .flags = RPC_TASK_ASYNC,
9796         };
9797         struct nfs_free_stateid_data *data;
9798         struct rpc_task *task;
9799
9800         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9801                 &task_setup.rpc_client, &msg);
9802
9803         dprintk("NFS call  free_stateid %p\n", stateid);
9804         data = kmalloc(sizeof(*data), GFP_NOFS);
9805         if (!data)
9806                 return -ENOMEM;
9807         data->server = server;
9808         nfs4_stateid_copy(&data->args.stateid, stateid);
9809
9810         task_setup.callback_data = data;
9811
9812         msg.rpc_argp = &data->args;
9813         msg.rpc_resp = &data->res;
9814         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9815         task = rpc_run_task(&task_setup);
9816         if (IS_ERR(task))
9817                 return PTR_ERR(task);
9818         rpc_put_task(task);
9819         return 0;
9820 }
9821
9822 static void
9823 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9824 {
9825         const struct cred *cred = lsp->ls_state->owner->so_cred;
9826
9827         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9828         nfs4_free_lock_state(server, lsp);
9829 }
9830
9831 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9832                 const nfs4_stateid *s2)
9833 {
9834         if (s1->type != s2->type)
9835                 return false;
9836
9837         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9838                 return false;
9839
9840         if (s1->seqid == s2->seqid)
9841                 return true;
9842
9843         return s1->seqid == 0 || s2->seqid == 0;
9844 }
9845
9846 #endif /* CONFIG_NFS_V4_1 */
9847
9848 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9849                 const nfs4_stateid *s2)
9850 {
9851         return nfs4_stateid_match(s1, s2);
9852 }
9853
9854
9855 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9856         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9857         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9858         .recover_open   = nfs4_open_reclaim,
9859         .recover_lock   = nfs4_lock_reclaim,
9860         .establish_clid = nfs4_init_clientid,
9861         .detect_trunking = nfs40_discover_server_trunking,
9862 };
9863
9864 #if defined(CONFIG_NFS_V4_1)
9865 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9866         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9867         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9868         .recover_open   = nfs4_open_reclaim,
9869         .recover_lock   = nfs4_lock_reclaim,
9870         .establish_clid = nfs41_init_clientid,
9871         .reclaim_complete = nfs41_proc_reclaim_complete,
9872         .detect_trunking = nfs41_discover_server_trunking,
9873 };
9874 #endif /* CONFIG_NFS_V4_1 */
9875
9876 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9877         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9878         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9879         .recover_open   = nfs40_open_expired,
9880         .recover_lock   = nfs4_lock_expired,
9881         .establish_clid = nfs4_init_clientid,
9882 };
9883
9884 #if defined(CONFIG_NFS_V4_1)
9885 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9886         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9887         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9888         .recover_open   = nfs41_open_expired,
9889         .recover_lock   = nfs41_lock_expired,
9890         .establish_clid = nfs41_init_clientid,
9891 };
9892 #endif /* CONFIG_NFS_V4_1 */
9893
9894 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9895         .sched_state_renewal = nfs4_proc_async_renew,
9896         .get_state_renewal_cred = nfs4_get_renew_cred,
9897         .renew_lease = nfs4_proc_renew,
9898 };
9899
9900 #if defined(CONFIG_NFS_V4_1)
9901 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9902         .sched_state_renewal = nfs41_proc_async_sequence,
9903         .get_state_renewal_cred = nfs4_get_machine_cred,
9904         .renew_lease = nfs4_proc_sequence,
9905 };
9906 #endif
9907
9908 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9909         .get_locations = _nfs40_proc_get_locations,
9910         .fsid_present = _nfs40_proc_fsid_present,
9911 };
9912
9913 #if defined(CONFIG_NFS_V4_1)
9914 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9915         .get_locations = _nfs41_proc_get_locations,
9916         .fsid_present = _nfs41_proc_fsid_present,
9917 };
9918 #endif  /* CONFIG_NFS_V4_1 */
9919
9920 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9921         .minor_version = 0,
9922         .init_caps = NFS_CAP_READDIRPLUS
9923                 | NFS_CAP_ATOMIC_OPEN
9924                 | NFS_CAP_POSIX_LOCK,
9925         .init_client = nfs40_init_client,
9926         .shutdown_client = nfs40_shutdown_client,
9927         .match_stateid = nfs4_match_stateid,
9928         .find_root_sec = nfs4_find_root_sec,
9929         .free_lock_state = nfs4_release_lockowner,
9930         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9931         .alloc_seqid = nfs_alloc_seqid,
9932         .call_sync_ops = &nfs40_call_sync_ops,
9933         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9934         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9935         .state_renewal_ops = &nfs40_state_renewal_ops,
9936         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9937 };
9938
9939 #if defined(CONFIG_NFS_V4_1)
9940 static struct nfs_seqid *
9941 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9942 {
9943         return NULL;
9944 }
9945
9946 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9947         .minor_version = 1,
9948         .init_caps = NFS_CAP_READDIRPLUS
9949                 | NFS_CAP_ATOMIC_OPEN
9950                 | NFS_CAP_POSIX_LOCK
9951                 | NFS_CAP_STATEID_NFSV41
9952                 | NFS_CAP_ATOMIC_OPEN_V1
9953                 | NFS_CAP_LGOPEN,
9954         .init_client = nfs41_init_client,
9955         .shutdown_client = nfs41_shutdown_client,
9956         .match_stateid = nfs41_match_stateid,
9957         .find_root_sec = nfs41_find_root_sec,
9958         .free_lock_state = nfs41_free_lock_state,
9959         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9960         .alloc_seqid = nfs_alloc_no_seqid,
9961         .session_trunk = nfs4_test_session_trunk,
9962         .call_sync_ops = &nfs41_call_sync_ops,
9963         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9964         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9965         .state_renewal_ops = &nfs41_state_renewal_ops,
9966         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9967 };
9968 #endif
9969
9970 #if defined(CONFIG_NFS_V4_2)
9971 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9972         .minor_version = 2,
9973         .init_caps = NFS_CAP_READDIRPLUS
9974                 | NFS_CAP_ATOMIC_OPEN
9975                 | NFS_CAP_POSIX_LOCK
9976                 | NFS_CAP_STATEID_NFSV41
9977                 | NFS_CAP_ATOMIC_OPEN_V1
9978                 | NFS_CAP_LGOPEN
9979                 | NFS_CAP_ALLOCATE
9980                 | NFS_CAP_COPY
9981                 | NFS_CAP_OFFLOAD_CANCEL
9982                 | NFS_CAP_COPY_NOTIFY
9983                 | NFS_CAP_DEALLOCATE
9984                 | NFS_CAP_SEEK
9985                 | NFS_CAP_LAYOUTSTATS
9986                 | NFS_CAP_CLONE
9987                 | NFS_CAP_LAYOUTERROR,
9988         .init_client = nfs41_init_client,
9989         .shutdown_client = nfs41_shutdown_client,
9990         .match_stateid = nfs41_match_stateid,
9991         .find_root_sec = nfs41_find_root_sec,
9992         .free_lock_state = nfs41_free_lock_state,
9993         .call_sync_ops = &nfs41_call_sync_ops,
9994         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9995         .alloc_seqid = nfs_alloc_no_seqid,
9996         .session_trunk = nfs4_test_session_trunk,
9997         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9998         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9999         .state_renewal_ops = &nfs41_state_renewal_ops,
10000         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10001 };
10002 #endif
10003
10004 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10005         [0] = &nfs_v4_0_minor_ops,
10006 #if defined(CONFIG_NFS_V4_1)
10007         [1] = &nfs_v4_1_minor_ops,
10008 #endif
10009 #if defined(CONFIG_NFS_V4_2)
10010         [2] = &nfs_v4_2_minor_ops,
10011 #endif
10012 };
10013
10014 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10015 {
10016         ssize_t error, error2;
10017
10018         error = generic_listxattr(dentry, list, size);
10019         if (error < 0)
10020                 return error;
10021         if (list) {
10022                 list += error;
10023                 size -= error;
10024         }
10025
10026         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10027         if (error2 < 0)
10028                 return error2;
10029         return error + error2;
10030 }
10031
10032 static const struct inode_operations nfs4_dir_inode_operations = {
10033         .create         = nfs_create,
10034         .lookup         = nfs_lookup,
10035         .atomic_open    = nfs_atomic_open,
10036         .link           = nfs_link,
10037         .unlink         = nfs_unlink,
10038         .symlink        = nfs_symlink,
10039         .mkdir          = nfs_mkdir,
10040         .rmdir          = nfs_rmdir,
10041         .mknod          = nfs_mknod,
10042         .rename         = nfs_rename,
10043         .permission     = nfs_permission,
10044         .getattr        = nfs_getattr,
10045         .setattr        = nfs_setattr,
10046         .listxattr      = nfs4_listxattr,
10047 };
10048
10049 static const struct inode_operations nfs4_file_inode_operations = {
10050         .permission     = nfs_permission,
10051         .getattr        = nfs_getattr,
10052         .setattr        = nfs_setattr,
10053         .listxattr      = nfs4_listxattr,
10054 };
10055
10056 const struct nfs_rpc_ops nfs_v4_clientops = {
10057         .version        = 4,                    /* protocol version */
10058         .dentry_ops     = &nfs4_dentry_operations,
10059         .dir_inode_ops  = &nfs4_dir_inode_operations,
10060         .file_inode_ops = &nfs4_file_inode_operations,
10061         .file_ops       = &nfs4_file_operations,
10062         .getroot        = nfs4_proc_get_root,
10063         .submount       = nfs4_submount,
10064         .try_get_tree   = nfs4_try_get_tree,
10065         .getattr        = nfs4_proc_getattr,
10066         .setattr        = nfs4_proc_setattr,
10067         .lookup         = nfs4_proc_lookup,
10068         .lookupp        = nfs4_proc_lookupp,
10069         .access         = nfs4_proc_access,
10070         .readlink       = nfs4_proc_readlink,
10071         .create         = nfs4_proc_create,
10072         .remove         = nfs4_proc_remove,
10073         .unlink_setup   = nfs4_proc_unlink_setup,
10074         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10075         .unlink_done    = nfs4_proc_unlink_done,
10076         .rename_setup   = nfs4_proc_rename_setup,
10077         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10078         .rename_done    = nfs4_proc_rename_done,
10079         .link           = nfs4_proc_link,
10080         .symlink        = nfs4_proc_symlink,
10081         .mkdir          = nfs4_proc_mkdir,
10082         .rmdir          = nfs4_proc_rmdir,
10083         .readdir        = nfs4_proc_readdir,
10084         .mknod          = nfs4_proc_mknod,
10085         .statfs         = nfs4_proc_statfs,
10086         .fsinfo         = nfs4_proc_fsinfo,
10087         .pathconf       = nfs4_proc_pathconf,
10088         .set_capabilities = nfs4_server_capabilities,
10089         .decode_dirent  = nfs4_decode_dirent,
10090         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10091         .read_setup     = nfs4_proc_read_setup,
10092         .read_done      = nfs4_read_done,
10093         .write_setup    = nfs4_proc_write_setup,
10094         .write_done     = nfs4_write_done,
10095         .commit_setup   = nfs4_proc_commit_setup,
10096         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10097         .commit_done    = nfs4_commit_done,
10098         .lock           = nfs4_proc_lock,
10099         .clear_acl_cache = nfs4_zap_acl_attr,
10100         .close_context  = nfs4_close_context,
10101         .open_context   = nfs4_atomic_open,
10102         .have_delegation = nfs4_have_delegation,
10103         .alloc_client   = nfs4_alloc_client,
10104         .init_client    = nfs4_init_client,
10105         .free_client    = nfs4_free_client,
10106         .create_server  = nfs4_create_server,
10107         .clone_server   = nfs_clone_server,
10108 };
10109
10110 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10111         .name   = XATTR_NAME_NFSV4_ACL,
10112         .list   = nfs4_xattr_list_nfs4_acl,
10113         .get    = nfs4_xattr_get_nfs4_acl,
10114         .set    = nfs4_xattr_set_nfs4_acl,
10115 };
10116
10117 const struct xattr_handler *nfs4_xattr_handlers[] = {
10118         &nfs4_xattr_nfs4_acl_handler,
10119 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10120         &nfs4_xattr_nfs4_label_handler,
10121 #endif
10122         NULL
10123 };
10124
10125 /*
10126  * Local variables:
10127  *  c-basic-offset: 8
10128  * End:
10129  */